VIEWS: 33 PAGES: 230 CATEGORY: Legal POSTED ON: 3/15/2010
ALLIANCE FOR CERVICAL CANCER PREVENTION International Agency for Research in Cancer (IARC), Lyon John Hopkins Program for International Education of Gynecologists Scientific Institute and Obstetricians (JHPIEGO Corporation), Baltimore Program for Appropriate Technology in Health (PATH), Seattle of Public Health Funded by the Bill & Melinda Gates Foundation Assessment of Innovative Approaches to Cervical Cancer Screening, Follow-Up and Treatment of Screen Detected Cervical Lesions in Developing Countries A Pooled Analysis of ACCP Study Results Marc ARBYN1, Lynne GAFFIKIN2, Rengaswamy SANKARANARAYANAN3, Richard MUWONGE3, Namory KEITA4, Ramani WESLEY5, Vivien TSU6 1. Scientific Institute of Public Health, Brussels, Belgium 2. JHPIEGO Corporation, Baltimore, USA 3. International Agency for Research in Cancer, Lyon, France 4. Service de Gynécologie/Obstétrique, Université de Conakry, Guinée-Conakry 5. Medicine, Regional Cancer Center, Trivandrum, Kerala State, India 6. PATH, Seattle, USA IPH/EPI REPORTS Nr. 2006–002 1 2 Epidemiology, January 2006; Brussels (Belgium) Scientific Institute of Public Heatlh, IPH/EPI REPORTS 2006 - 002 Depotnumber: D/2006/2505/002 File: ACCP_Meta9a.doc. Funded by the Bill & Melinda Gates Foundation ALLIANCE FOR CERVICAL CANCER PREVENTION IARC, Lyon (France) JHPIEGO Corporation, Baltimore (USA) PATH, Seattle (USA) Assessment of Innovative Approaches to Cervical Cancer Screening, Follow-Up and Treatment of Screen Detected Cervical Lesions in Developing Countries A Pooled Analysis of ACCP Study Results Marc ARBYN1, Lynne GAFFIKIN2, Rengaswamy SANKARANARAYANAN3, Richard MUWONGE3, Namory KEITA4, Ramani WESLEY5, Vivien TSU6 1. Scientific Institute of Public Health, Brussels, Belgium 2. JHPIEGO Corporation, Baltimore, USA 3. International Agency for Research in Cancer, Lyon, France 4. Service de Gynécologie/Obstétrique, Université de Conakry, Guinée-Conakry 5. Medicine, Regional Cancer Center, Trivandrum, Kerala State, India 6. PATH, Seattle, USA Scientific Institute of Public Health J. Wytsmanstr. 14 B1050 Brussels Belgium e-mail: marc.arbyn@iph.fgov.be http://www.iph.fgov.be/epidemio/epien/PROG2.HTM IPH/EPI REPORTS Nr. 2006 – 002 3 4 1. Table of contents 1. TABLE OF CONTENTS.................................................................................5 2. EXECUTIVE SUMMARY..............................................................................8 2.1. BACKGROUND ................................................................................................8 2.2. MATERIAL AND METHODS ..............................................................................8 2.2.1. Conducted studies .....................................................................................8 2.2.2. Statistical analyses ....................................................................................9 2.3. RESULTS .......................................................................................................10 2.3.1. Prevalence of cervical disease ................................................................10 2.3.2. Test characteristics of the evaluated screening tests...............................10 2.3.3. Influence of study characteristics on test accuracy.................................12 2.3.4. Triage of VIA-positive cases ..................................................................13 2.3.5. Cryotherapy: compliance, acceptability and safety ................................13 2.4. CONCLUSIONS ..............................................................................................13 3. INTRODUCTION..........................................................................................15 4. MATERIAL AND METHODS.....................................................................17 4.1. CONDUCTED STUDIES ...................................................................................17 4.1.1. Cross-sectional multi-center study of the test accuracy of VIA, VIAM, VILI, cytology and the HC2 assay..........................................................18 4.1.2. Cross-sectional study on the test accuracy of VIA, cytology and the HC2 assay in Zimbabwe..................................................................................19 4.1.3. Randomized trial: VIA screening versus no screening in rural South India ........................................................................................................20 4.1.4. Western Kenya Cervical Cancer Prevention Project ..............................21 4.1.5. Safety, acceptability, and feasibility of a single-visit approach to cervical cancer prevention in rural Thailand ........................................................23 4.1.6. Safety, acceptability, and feasibility of a single-visit approach to cervical-cancer prevention in Ghana ......................................................23 4.2. METHODS .....................................................................................................25 4.3. STATISTICAL ANALYSES ...............................................................................27 5. RESULTS OF THE SEPARATE STUDIES ...............................................29 5.1. CROSS-SECTIONAL MULTI-CENTER STUDY ON THE TEST ACCURACY OF VIA, VIAM, VILI, CYTOLOGY AND THE HC2 ASSAY ..................................29 5.1.1. Test accuracy of VIA and VILI for CIN2+ ............................................29 5.1.2. Pooled relative accuracy of the screening tests.......................................32 5.2. CROSS-SECTIONAL STUDY ON THE TEST ACCURACY OF VIA, CYTOLOGY AND THE HC2 ASSAY IN ZIMBABWE .............................................................33 5.3. RANDOMIZED TRIAL OF VIA SCREENING IN RURAL SOUTH INDIA ................35 5.4. WESTERN KENYA CERVICAL CANCER PREVENTION PROJECT (WKCCPP)..37 5.4.1. Positive predictive value and approximated specificity of VIA .............37 5.4.2. Performance of triage methods ...............................................................38 5.4.3. Combination of VIA screening with subsequent triage ..........................41 5.4.4. Experience of women in the 1-3 month period after cryotherapy ..........42 5.4.5. Follow-up 1 year after cryotherapy.........................................................46 5 5.5. SAFETY, ACCEPTABILITY, AND FEASIBILITY OF A SINGLE-VISIT APPROACH TO CERVICAL-CANCER PREVENTION IN RURAL THAILAND ............................48 5.6. SAFETY, ACCEPTABILITY, AND FEASIBILITY OF A SINGLE-VISIT APPROACH TO CERVICAL CANCER PREVENTION IN GHANA .............................................54 6. POOLED ANALYSES...................................................................................59 6.1. META-ANALYSIS OF TEST CHARACTERISTICS OF APPLIED SCREENING METHODS ......................................................................................................59 6.1.1. Test characteristics of VIA .....................................................................59 6.1.2. Test characteristics of VILI ....................................................................70 6.1.3. Test characteristics of VIAM..................................................................80 6.1.4. Test characteristics of HPV DNA detection by HC2..............................90 6.1.5. Test characteristics of cervical cytology...............................................100 6.1.6. Summary of the meta-analyses of the accuracy for CIN1+..................127 6.1.7. Summary of the meta-analyses of the accuracy for CIN2+..................129 6.1.8. Summary of the meta-analyses of the accuracy for CIN3+..................131 6.1.9. Summary of the meta-analyses of the accuracy for cancer...................133 6.2. META-ANALYSES OF THE RELATIVE SENSITIVITY AND SPECIFICITY OF CERVICAL CANCER SCREENING TESTS .........................................................135 6.2.1. Relative sensitivity and specificity of VILI compared to VIA .............135 6.2.2. Relative sensitivity and specificity of VIAM compared to VIA ..........138 6.2.3. Relative sensitivity and specificity of HC2 compared to VIA .............140 6.2.4. Relative sensitivity and specificity of cervical cytology (at ASCUS+) compared to VIA...................................................................................142 6.2.5. Relative sensitivity and specificity of cervical cytology (at LSIL+) compared to VIA...................................................................................144 6.2.6. Relative sensitivity and specificity of cervical cytology (at HSIL+) compared to VIA...................................................................................146 6.2.7. Relative sensitivity and specificity of VIAM compared to VILI .........148 6.2.8. Relative sensitivity and specificity of HC2 compared to VILI.............150 6.2.9. Relative sensitivity and specificity of cytology (at ASCUS+) compared to VILI ..................................................................................................152 6.2.10. Relative sensitivity and specificity of cytology (at LSIL+) compared to VILI.......................................................................................................154 6.2.11. Relative sensitivity and specificity of cytology (at HSIL+) compared to VILI.......................................................................................................156 6.2.12. Relative sensitivity and specificity of HC2 compared to VIAM ..........158 6.2.13. Relative sensitivity and specificity of cytology at (ASCUS+) compared to VIAM................................................................................................160 6.2.14. Relative sensitivity and specificity of cytology (at LSIL+) compared to VIAM ....................................................................................................162 6.2.15. Relative sensitivity and specificity of cytology (at HSIL+) compared to VIAM ....................................................................................................164 6.2.16. Relative sensitivity and specificity of cytology (at ASCUS+) compared to HC2...................................................................................................166 6.2.17. Relative sensitivity and specificity of cytology (at LSIL+) compared to HC2 .......................................................................................................168 6.2.18. Relative sensitivity and specificity of cytology (at HSIL+) compared to HC2 .......................................................................................................170 6 6.2.19. Summary of the meta-analyses of the relative accuracy of screening tests ...............................................................................................................172 6.3. SUMMARY ROC CURVES ............................................................................174 6.4. INFLUENCE OF STUDY CHARACTERISTICS ON THE PREVALENCE OF DISEASE ......................................................................................................178 6.5. INFLUENCE OF STUDY COVARIATES ON TEST CHARACTERISTICS ............180 6.5.1. Variation of the test positivity rate by age group..................................180 6.5.2. Variation of the test positivity rate by time period ...............................183 6.5.3. Influence of age, period and study setting on the PPV .........................185 6.5.4. Influence of study characteristics on the sensitivity and specificity of screen tests ............................................................................................189 6.5.5. Influence of study characteristics on the diagnostic odds ratio using multi-variate SROC-regression analysis...............................................192 6.6. CORRELATIONS BETWEEN TESTS ................................................................194 6.7. PPV OF COLPOSCOPY TRIAGE .....................................................................194 6.8. TREATMENT OF VIA-POSITIVE CASES.........................................................198 7. DISCUSSION ...............................................................................................202 8. ACKNOWLEDGEMENTS.........................................................................206 9. ABBREVIATIONS ......................................................................................208 10. REFERENCES.............................................................................................209 11. ANNEXES.....................................................................................................213 11.1. DATA FILES AND STATISTICAL SYNTAX FILES .............................................213 11.2. ABSTRACTS FROM THE INCLUDED SEPARATE PUBLISHED STUDIES..............214 11.2.1. Cross-sectional multi-country study (5 publications) ...........................214 11.2.2. Zimbabwe study (3 publications)..........................................................216 11.2.3. Ambilikai RCT (2 publications) ...........................................................218 11.2.4. Kenya study ..........................................................................................219 11.2.5. Thailand demonstration project ............................................................219 11.2.6. Ghana demonstration project ................................................................220 11.3. COMPREHENSIVE TABLE WITH CHARACTERISTICS OF THE INCLUDED STUDIES ......................................................................................................221 7 2. Executive summary 2.1. Background Cervical cancer is the most common female cancer in developing counties. Cervical cancer has a long detectable preclinical phase and this makes it particularly preventable by screening. The incidence of and mortality from this disease has fallen dramatically in western countries where the use of the Pap smear was well established and quality could be assured. By lack of financial and technical resources it is unrealistic to introduce similar screening policies in developing countries. Therefore new screening methods, applicable in resource-poor settings, are currently being investigated. The Bill & Melinda Gates Foundation has made it possible for the Alliance for Cervical Cancer Prevention to set up large-scale studies, which in the long run may result in the development of appropriate health policies using effective and affordable technologies to prevent cervical cancer. 2.2. Material and methods 2.2.1. Conducted studies Five operational partners in the Alliance for Cervical Cancer Prevention (ACCP)-- PATH (Seattle), IARC (Lyon), JHPIEGO (Baltimore), EngenderHealth (New York) and PAHO (Washington)--conducted several trials in Africa, Asia and Latin America between 1999 and 2004. At an intermediate evaluation that took place in Geneva in April 2004, experts suggested that a pooled analysis of available data be performed. Data from six different studies were obtained. One research project was conducted by the International Agency for Research in Cancer (IARC, Lyon). It was a cross-sectional multi-center study set up in 11 locations, in five African countries and India, where in total 58 000 women were screened with naked-eye visual inspection after application of 3 to 5% acetic acid (VIA). In several settings other screening tests were performed as well. In three locations, the cervix was inspected with a magnifying device (VIAM); in 10 locations Lugol’s iodine was used to identify suspect yellow non-iodine uptake areas. At five places a Pap smear was taken and in four settings the Hybrid Capture II (HC2) assay was used to detect presence of high-risk HPV DNA. The gold standard verification consisted of a colposcopic exploration of the cervix, offered to all women, irrespective of the screen test result and performed during the same visit. Biopsies were taken from colposcopically suspect areas. The histological result or, if histology was absent, the negative colposcopy result was considered as the outcome. Women with CIN or more serious cervical disease received adequate treatment. A similar study was conducted by JHPIEGO (Baltimore) in collaboration with the University of Zimbabwe. Here more than 2000 enrolled women were tested with VIA, a Pap smear and the Hybrid Capture II assay. The gold standard verification was as described for the first study. 8 A third study was a cluster-randomized controlled trial conducted by IARC in Ambilikai in the Dindigul District (India). In study arm A, VIA screening was offered to voluntary participating women whereas in the second group no screening took place. VIA-positive women, in group A, were verified with colposcopy and biopsies from colposcopically suspect areas. Women with colposcopically localized cervical intra-epithelial neoplasia were treated on the spot with cryotherapy. Women with more advanced disease were referred for appropriate treatment. Cancer and mortality surveillance is organized similarly in both groups, by the Ambilikai Cancer Register and mortality registries. The cumulative incidence of and mortality from cervical cancer over ten years will be compared for the two study arms. The VIA results and colposcopy and histology results of VIA-positive women enrolled in arm A are already available and included in this meta-analysis. The Ambilikai study is the first randomized trial ever conducted which will document effectiveness of cervical cancer screening directly. A fourth study was conducted by PATH (Seattle) in the Busia district, within the Western Kenya Cervical Cancer Prevention Project. More than 2000 women were screened at the local health center with VIA, and screen-positive women were referred to the district hospital where two alternative triage methods were tested in parallel: cytology and visual inspection with acetic acid using the Aviscope for magnification (VIAM). All women presenting for triage were colposcoped as well. Women with a colposcopy-based diagnosis of localized CIN were treated with cryotherapy. Treated women were followed at 1-3 months and at 12 months after treatment to verify secondary effects of cryotherapy and occurrence of residual/recurrent CIN. Outcomes of the study were: positive predictive value of the screen test, performance of VIAM and cytology to triage VIA-positive women; compliance of treated women with follow-up procedures and the safety of VIA treatment. At the end of the study, VILI was evaluated as well in a small subset of women as a screening test (at the health center) and as a triage test (at the referral hospital). The fifth and sixth studies were demonstration projects, conducted by JHPIEGO, which were purposefully set up not to measure test accuracy parameters but safety, acceptability and feasibility of cryotherapy offered to VIA-positive eligible women in a single visit. Almost 6000 and 4000 women from Roi-et Province (Thailand) and from Accra (Ghana), respectively, were screened with VIA. Non eligible women were referred for further examination. VIA-positive women having received treatment were interviewed about their experience and possible side-effects at and after cryotherapy and were followed at 3 and 12 months after treatment. We obtained the individual anonymized data files of all six studies described above. 2.2.2. Statistical analyses The main test accuracy parameters – sensitivity, specificity, positive and predictive value, prevalence of disease, detection rate of disease – could be evaluated for the first two studies where all screened women were verified with a gold standard, assuming that colposcopy followed by the histological examination of colposcopically oriented biopsies provided perfect ascertainment of the cervical status. 9 From the Ambilikai RCT and the Busia study, we could derive the positive predictive value of VIA, the detection rate of disease and an approximated specificity. From all six studies the test-positivity rate for the VIA-test could be evaluated. Four levels of disease outcome were discerned: CIN I or worse (CIN1+), CIN2+, CIN3+ and invasive cervical cancer. Compliance with treatment recommendations could be addressed in the last four studies and the occurrence of side-effects and the acceptability of cryotherapy can be analyzed from the studies conducted in Kenya, Thailand and Ghana. In first set of meta-analyses, data documenting the test characteristics of each separate test were pooled and analyzed. Subsequently we computed the relative sensitivity and specificity of each test compared to the other study tests. Random effect models were used for pooling and forest plots were produced for all meta-analyses of test parameters of all tests and outcomes. The simultaneous variation of sensitivity and specificity was addressed by constructing summary ROC curves (sROC) and computing the area under this sROC curve. The influence of study covariates on test characteristics was studied graphically and by logistic regression using the individual patient data and using age group, study phase and study location as co-variates. Finally the multi-variate variation of the overall accuracy was analyzed using sROC- regression analyses. 2.3. Results 2.3.1. Prevalence of cervical disease The prevalence of cervical intra-epithelial neoplasia (CIN) could be assessed in the first two studies. The prevalence of high-grade cervical disease (CIN2+) was exceptionally high in Zimbabwe: 9.7% for CIN2+, whereas in the cross-sectional multi-center study the prevalence of CIN2+ ranged only from 0.8% to 3.3%. In general the prevalence increased slightly until the age of 50-54 and increased with a steeper slope thereafter. 2.3.2. Test characteristics of the evaluated screening tests Accuracy of VIA The pooled sensitivity of VIA to detect presence of CIN2+ and CIN3+ was respectively 79% (CI: 74% to 84%) and 83% (CI: 77% to 89%). Its specificity was 83% (CI: 79% to 87%) for CIN2+ and 84% (CI: 80% to 88%) for CIN3+. The pooled predictive value of a positive VIA test was 12% (CI: 9% to 15%) if CIN2+ was the considered outcome and 8% (CI: 5% to 9%)) if CIN3+ was the considered outcome. The pooled predictive value of a negative VIA result was higher than 99% for both outcomes. Important statistically significant inter-study variation was observed. Visual inspection after application of Lugol’s iodine solution was 11% (CI: 5% to 17%) more sensitive to detect CIN2+ than after application of diluted acetic acid. For CIN3+, VILI was 7% (CI: 4% to 11%) more sensitive than VIA. VIA had a similar sensitivity and specificity as inspection assisted with a magnifying device (VIAM). In these studies, VIA showed a higher sensitivity overall than the HC2 for CIN2+ and CIN3+, but the difference was not statistically significant. On 10 the other hand, the HC2 was 6% (CI: 3% to 9%) more specific for excluding CIN2+ and 8% (CI: 5% to 10%) more specific for excluding CIN3+ than VIA. It must be remarked that these findings were dominated by results observed in India. In Zimbabwe, HC2 showed an almost significantly higher sensitivity and lower specificity for CIN2+ than VIA. VIA was always more sensitive than the conventional Pap smear but this latter test, in turn, was more specific. Accuracy differences between VIA and cytology (higher sensitivity and lower specificity of VIA) were significant, with the exception of the sensitivity when the cytological cutoff was ASCUS+ and the outcome was CIN3+. Accuracy of VILI The sensitivity and specificity of VILI as a screening test was only assessed by the IARC. Of all evaluated tests, VILI showed the highest sensitivity: the pooled value was 91% (CI: 88% to 95%) for CIN2+ and 94% (91% to 97%) for CIN3+. The specificity of VILI was 85% (CI: 81% to 88%) when CIN2+ was the outcome and 84% (CI: 81% to 87%) when CIN3+ was the outcome. The PPV and NPV for the same outcomes were respectively 13% (CI: 8% to 18%) and 99.8% (CI: 99.7% to 99.9%). As mentioned above, VILI is more sensitive but equally specific compared to VIA. VILI was more sensitive than the HC2 (ratio = 1.20; CI: 1.06-1.35) for CIN2+. There was no statistically significant difference between VILI and HC2 when CIN3+ was considered in the three Indian settings where both screening tests were performed. The specificity of VILI was significantly lower than that of HC2 for both CIN2+ and CIN3+. VILI always was more sensitive but less specific than cytology at whatever cytological cutoff. Accuracy of VIAM As mentioned VIAM showed accuracy values which did not differ from those of VIA. Accuracy of the Hybrid Capture II assay The pooled sensitivity for high-grade CIN was unexpectedly low compared to previously published studies [summarized by Lorincz, 2003 and Franco, 2003]. It was 66% (CI: 62% to 77%) for CIN2+ and 68% (CI: 62% to 75%) for CIN3+. The sensitivity for CIN2+ was homogeneously low in the four Indian settings where the HC2 was evaluated (62%; CI: 56% to 68%), whereas, in Harare, it was 80% (CI: 74.1% to.85.4%). The specificity of HC2 for CIN2+, pooled for the four Indian settings, was 94% (CI: 92% to 95%). In Harare, it was only 61% (CI: 59% to 63%). The PPV of HC2 varied between minimally 9% (in Kolkata 1) and maximally 18% (in Zimbabwe). The range of the NPV was situated between 96.7% and 99.7%. HC2 showed a higher pooled sensitivity for CIN2+ than cytology, even at cutoff ASCUS+. In the four Indian settings this difference was insignificant, but in Harare it was significant. HC2 showed a significantly higher pooled specificity than cytology. In Zimbabwe, HC2 showed a significantly lower specificity. The accuracy of the Pap smear The pooled sensitivity of the Pap smear changed by cytological cut-off: at ASCUS+ it was 58% (CI: 44% to 73%) for CIN2+ and 63% (CI: 38% to 88%) for CIN3+; at LSIL it was 50% (CI: 33% to 67%) for CIN2+ and 56% (CI: 33% to 80%) for CIN3+, at HSIL it was substantially lower: 38% (CI: 22% to 54%) for CIN2+ and 52% (CI: 32% to 71%). Due to large interstudy heterogeneity, these differences were 11 insignificant. On the contrary, the specificity increased significantly by cytological cut-off: at ASCUS+ it was 90% (CI: 86% to 94%) and at HSIL it was 99.1% (98.6% to 99.5%) for the outcome of CIN2 or worse. The specificity for CIN3+ did not change by cytological cut-off. The pooled positive predictive value of the Pap smear for the outcome CIN2+ using the cut-off HSIL (56%, 35% to 77%) was higher than when the cut-off ASCUS (21%, CI: 13% to 29%) was used. The NPV always was higher than 98% in spite of the rather low sensitivity. Among all evaluated screening tests, cytology had the lowest sensitivity for all degrees of CIN. Only for the comparison with HC2 assay was the difference not significant. On the other hand, cytology was the most specific test. Only when ASCUS was the cut-off, and when compared with the HC2 test, was the difference in specificity not significant. Test positivity rate The range of variation in VIA test positivity rate was wide: from 6% in Amasaman (Ghana) to 40% in Harare (Zimbabwe). The VILI test positivity rate ranged from 9% to 29%. The HPV test positivity rate varied between 6% and 9% in the Indian settings but was 43% in Harare. The test positivity rate of all visual tests varied considerably by age group, study phase and setting. The variation in VIA test positivity by age group was inconsistent: it increased significantly in certain settings, whereas in others it decreased or remained rather constant, or changed following a "V" or inversed-"V" shape curve. Three periods were defined using tertiles of the screening dates. Important period effects were observed for the visual inspection methods, in particular for VIA and for cytology. Remarkable was the absence of a period effect in the HPV- test positivity rate. Summary ROC curve (sROC) An overall summary measure of test accuracy, the area under the sROC curve, was not different for VIA and VILI. However, given the restricted range of VILI points, the sROC curve could not describe VILI accuracy appropriately. No sROC curves could be constructed for the other tests because of insufficient points. 2.3.3. Influence of study characteristics on test accuracy By logistic regression the multivariate effects of age group, study period and setting on the PPV (14 locations) and on sensitivity and specificity (12 settings) for CIN2+ were analyzed. In general, the PPV increased with age and period and varied further by location. The PPV of HC2 did not vary by period. Age did not influence the sensitivity for CIN2+ of the screening tests, with the exception of the Pap smear. Period had an effect only on the sensitivity of VIA. The sensitivity always varied by study setting. The sensitivity of VILI was lower in the Indian settings, compared to the African settings. Age, period and setting always influenced the specificity for CIN2+, with the exception of the HC2 assay, which was not influenced by period. For all other tests, the specificity increased by period. The influence of study characteristics on the diagnostic odds ratio, an accuracy measure that integrates sensitivity and specificity, was assessed by sROC regression. The S term always was significantly different from zero, indicating change of accuracy by degree of positivity of the screen test. Age never was influential. The 12 diagnostic odds ratio (DOR) always was higher in the second or third period compared with the first, with the exception of the HC2 test. High DOR-values for VIA (considering CIN2+ as outcome) were observed in Conakry and Bamako, where at the same time high sensitivities and specificities were found. Visual screening methods (VIA, VIAM, VILI) always had higher DOR values in Kolkata 2 and Trivandrum 2 compared to Kolkata 1 and Trivandrum 1. 2.3.4. Triage of VIA-positive cases One of the purposes of colposcopy triage is to reduce the number of women needing treatment. The low-grade colposcopy rate among VIA-positive women varied widely from 40% to 93%. Its pooled value was 65% (CI: 54% to 76%). This means that, on average, the number of treated women could be decreased by 35% in a "see-triage- treat" strategy compared with a "see & treat" strategy. The pooled PPV of VIA testing could be increased by adding colposcopy triage by a factor 1.59 (CI: 1.48- 1.64) when CIN2+ was the outcome, but only by a factor of 1.13 (CI: 1.03 to 1.18) when CIN3+ was the outcome. Only in Busia (Kenya) were alternative triage methods evaluated. Cytology appeared to be unacceptable, because it showed, at cutoff ASCUS+, only a moderate sensitivity (72%) and low specificity (49%) for the outcome CIN2+. VIAM had a higher sensitivity (92%) but a lower specificity (40%). VILI showed the best triage results (100% sensitivity and 76% specificity), but this was evaluated only on a small group. Triage was offered only at the referral hospital, which yielded a drop out of 36% of VIA+ cases. 2.3.5. Cryotherapy: compliance, acceptability and safety More than 80% of women with known CIN2+ status in Ambilikai and Busia got treatment. In Thailand and Ghana more than 90% of VIA-positive women were treated. Ninety-one percent of treated women in Ambilikai were treated with cryotherapy. Among women receiving cryotherapy, between 71% and 82% were treated on the same day. Cryotherapy offered to VIA-positive women immediately after screening is highly accepted and quite safe. Pain during or just after cryotherapy was the most noticed secondary effect (40% in Thailand, 23% in Ghana), which was most often mild (79% in Roi-Et and 86% in Accra), and only in 2% (Roi-Et) and 3% (Accra) was it severe. In 2% of treated women some minor blood loss occurred. In Thailand, 83% and 93% of treated women presented at the follow-up visits at respectively 3 and 12 month after treatment. In Ghana these proportions were 77% and 55%. 2.4. Conclusions An enormous quantity of new information has now become available from the ACCP research projects analyzed in this report, which provides insight into two important aspects of cervical cancer screening in developing countries: the cross-sectional test- accuracy of five screening tests and the feasibility, acceptability and safety of immediate treatment of women who are positive at visual cervical inspection. Meta- analytical pooling allowed us to formulate the following conclusions: 13 • Direct visual inspection of the cervix after application of acetic acid solution or Lugol’s iodine allows detection of cervical cancer precursors with an accuracy that is as good as or even better than the standard Pap smear or testing for presence of DNA from high-risk human papillomaviruses with the Hybrid Capture II assay. • Direct visual inspection of the cervix after application of Lugol’s-iodine (VILI) is 10 % more sensitive but equally specific in detecting high-grade cervical intra-epithelial lesions as application of dilute acetic acid (VIA). The test characteristics varied widely between the settings and between the start and end phase of the studies, which might reflect low reproducibility and which underlines the need for training and continuous supervision. • The use of a magnifying device does not improve the accuracy of the visual interpretation. • The Hybrid Capture II assay (HC2) showed an unexpectedly low sensitivity, which is in conflict with previously published results. Nevertheless, the variability of the test results between the study locations and the study periods was limited, suggesting high reproducibility. The reason for its low sensitivity must be explored before conclusions can be drawn concerning the performance of the HC2 test. Currently existing HPV DNA detection methods are nowadays still expensive and require laboratory infrastructure and appropriate logistics. • Among all evaluated tests, cytology showed the lowest sensitivity, which is in agreement with the documented very wide variation of the accuracy of this test. Moreover, cytology is not highly reproducible, it does not allow an instantaneous diagnosis, and it requires important resources for training and logistics. • By colposcopy or other visual triage, the number of women needing treatment can be reduced and targeted biopsies can be taken for a definitive diagnosis. Nevertheless, throughout the ACCP studies, the triaging performance varied widely. When triage is only available at distance from the screening place, an important drop out of patients can be expected. • Therefore "see and treatment" options are important to maximize treatment compliance. The experiences from the demonstration projects in Thailand and Ghana indicate that immediate cryotherapy of VIA-positive women is accepted by women and their partners and only causes limited side effects. 14 3. Introduction Cervical cancer is the second most common cancer in women worldwide and is the most common female cancer in Central America, Sub-Saharan Africa, South-Central Asia and Melanesia. It is estimated that, in 2002, approximately 493,000 women developed cervical cancer and that 273,000 died from the disease [Ferlay, 2004]. 83% of all cervical cancers occur in developing countries. The disease primarily affects younger women with the majority of cases appearing between the ages of 30 and 50, an age when many women are actively involved in their careers or caring for their families [Yang, 2004]. Nevertheless, among all malignant tumors, cervical cancer is the one that is most easily preventable by screening. The detection of cytological abnormalities by microscopic examination of Pap smears, and the subsequent treatment of women where cytological abnormalities are high-grade, avoids development of cancer [Miller, 1993]. Through well-organized cytological screening at the population level, every three to five years, the incidence can be reduced up to 80% [Expert team cervical cancer screening, 2005]. In industrial countries, incidence of and mortality from cervical cancer dropped dramatically, most probably as a consequence of cytological screening [Bray, 2005; Devesa, 1987]. However, in only a few of them-such as the Nordic countries, the United Kingdom and the Netherlands--is cytological screening well organized [Anttila, 2004]. In most other countries, screening is opportunistic, depending on the initiative of the individual woman or her doctor. Such opportunistic screening is most often characterized by a high coverage in selected parts of the population with screening done too frequently, coexisting with a low coverage in lower socio-economic groups and heterogeneous quality, which results in poor cost-effectiveness [van Ballegooijen, 2000; van den Akker-van Marle, 2002; Miller 2002]. It is nowadays widely recognized that policies calling for repetitive cytological screening and prolonged surveillance of screen-positive women are difficult to realize in developing countries, because of financial, technical and human constraints. Therefore alternative test methods are being investigated which might perform as well as or even better than the Pap smear but be more affordable, applicable and acceptable in wide ranges of settings in the low-resource countries [Miller, 2000]. Swabbing the cervix with 3 to 5% acetic acid (VIA) turns cervical intra-epithelial abnormalities white, and detection of such aceto-white lesions by visual inspection might be a useful alternative screening test. It is cheap and can be applied by local health workers after a few weeks of training. Visual inspection with Lugol’s iodine (VILI) can be used as well [Sankaranarayanan, 2003c]. In this case yellow non-iodine uptake areas are indicative of cervical lesions. The cervix can be inspected with the naked eye or using a low-level magnifying device (VIAM). The establishment of the evidence that oncogenic types of the human papillomavirus are a necessary but insufficient condition for the development of virtually all cervical cancers has resulted in the development of HPV DNA detection methods [Bosch, 2002; 2003]. The Hybrid Capture II assay (Digene Corp., Gaithersburg, MD, USA) is, nowadays, the only USFDA approved commercially available HPV DNA detection test applicable in cervical cancer screening [Lörincz, 1997]. In order to avoid development of cancer, screen-detected intra-epithelial lesions with potential of progression must be treated. Several excisional or ablative techniques exist, and all are believed to be equally effective in treatment of localized CIN 15 [Mitchell, 1998; Martin-Hirsh, 2000]. Among all, cryotherapy appears the most appropriate method for treatment of localized cervical lesions in low-resource settings, since it requires only limited equipment and is easy to handle by mid-level health workers. In 1999, the Bill & Melinda Gates Foundation awarded US$50 million to a group of international research organizations through the Alliance for Cervical Cancer Prevention (ACCP). A series of large-scale studies were set up between 1999 and 2004 to assess new screening and treatment approaches and technologies appropriate to low-resource settingsa. The current report pools the study results obtained from three ACCP partners: IARC, JHPIEGO and PATH. The main aims of the current pooled analysis are the estimation of the test accuracy of five cervical cancer screening techniques: the visual inspection methods (VIA, VILI and VIAM), the cytological examination of the conventional Pap smear and HPV DNA detection using the Hybrid Capture II (HC2) test; the evaluation of alternative triage options for management of screen-positive women; and finally the assessment of the feasibility, acceptability and safety of immediate cryotherapy offered to women with a positive VIA test. a ACCP (Alliance for Cervical Cancer Prevention): see http://www.alliance-cxca.org/ 16 4. Material and methods 4.1. Conducted studies We obtained data files including the individual records of more than 106,000 women, from six studies conducted at 19 settings in 10 countries in Asia and Africa (see map in Table 1 and Figure 1). Table 1. Study size, location of settings and coordinating agency of the six studies included in the pooled analysis. Study Location Country N Coordinating agency 1. Multi-center cross- 11 urban settings Burkina-Faso, 58,679 IARC sectional study Congo, Guinea, Mali, Niger, India 2. Zimbabwe cross- Chitungwiza & Zimbabwe 2,206 JHPEIGO sectional study greater Harare area 3. Ambilikai RCT Rural area in India 31,317 IARC Dindigul district 4. WKCCPPa 5 rural divisions Kenya 2,295 PATH in Busia district 5. Thai demonstration 4 rural districts in Thailand 5,999 JHPEIGO project Roi-Et Province 6. Ghanaian Accra Ghana 3,665 JHPEIGO demonstration project Metropolitan area 7. Amansaman Amansaman, rural Ghana 3,206 JHPEIGO demonstration projectb Total 10 countries 106,626 Throughout the meta-analysis we will rank the study settings by study and within the study by alphabetic order of the setting, as following: • Multi-center study: (1) Bamako, (2) Brazzaville, (3) Conakry, (4) Jaipur, (5) Kolkata-1, (6) Kolkata-2, (7) Mumbai, (8) Niamey, (9) Ouagadougou, (10) Trivandrum 1, (11) Trivandrum 2 • Zimbabwe study: (12) Harare • Ambilikai RCT: (13) Ambilikai • WKCCSP: (14) Busia • Thai demonstration project: (15) Chaturapakpiman ,(16) Kasetwisai, (17) Pathumrat, (18) Phanomphrai • Ghanaian demonstration project: 19) Accra. The six studies are described separately in the following subchapters. a WKCCPP: Western Kenya Cervical Cancer Prevention Project. b The Amansanan data file will be included a posteriori once the completed cleaned data file will become available. 17 Jaipur Calculta Bamako Niamey Dindigul Roy Et Ouagadougou Trivandrum Conakry Accra Busia Brazzaville Harare Figure 1. Study locations in 10 countries where the ACCP studies were conducted included in this pooled analysis. In red: studies coordinated by IARC: Bamako (Mali), Brazzaville (Congo), Conakry (Guinea), Niamey (Niger), Ouagadougou (Burkina Faso), Calcutta, Dindigul [Ambilikai], Jaipur and Trivandrum (India); in green: study conducted by PATH: Busia (Kenya); in blue: studies conducted by JHPIEGO: Harare (Zimbabwe), Accra (Ghana) and Roi-Et (Thailand). 4.1.1. Cross-sectional multi-center study of the test accuracy of VIA, VIAM, VILI, cytology and the HC2 assay Study design: A cross-sectional study was conducted in 11 centers and consisted of concomitant testing with VIA, VILI, VIAM, the Pap smear and the Hybrid Capture II Assay. VIA was evaluated in all centers, VILI in 10 centers, VIAM in 3, conventional cytology in 5 and the HC2 assay in 4 centers. Screening tests were followed by colposcopy in all women. Punch biopsies were taken from colposcopically suspect areas. The histological result of the biopsy or, if absent, the colposcopic interpretation was considered as a combined gold standard (assuming 100% accurate). Different health workers, blinded to the results of the other tests, performed all tests. Since the gold standard is applied to all study subjects, the accuracy for CIN or cancer can be evaluated without verification bias. 18 Sample size: Table 2. Tests evaluated and sample size by setting involved in the multi-center cross-sectional study. Number of Center Country Tests evaluated women included Bamako Mali VIA, VILI 5552 Brazzaville Congo VIA, VILI 6935 Calcutta 1a India VIA, VIAM, Cytology, HC2 5894 Calcutta 2 India VIA, VILI, VIAM, HC2 8080 Conakry Guinea VIA, VILI 8627 Jaipur India VIA, VILI, Cytology 5786 Mumbai India VIA, VILI, VIAM, Cytology, HC2 4004 Niamey Niger VIA, VILI 2534 Ouagadougou Burkina Faso VIA, VILI 2051 Trivandrum 1 India VIA, VILI, Cytology 4457 Trivandrum 2 India VIA, VILI, Cytology, HC2 4759 Total 58,679 Time period: Between 1999 and 2003. Setting and characteristics of the study population: Unscreened population from urban areas in India and five African countries (see Table 2). Women were between 25 and 65 years old. For more details on the study population, the health workers performing the tests and the technical aspects of the tests, we refer to a series of previous publications [Sankaranarayanan 2003b, 2004a, 2004c, 2004d, 2004e]. 4.1.2. Cross-sectional study on the test accuracy of VIA, cytology and the HC2 assay in Zimbabwe Study design: Participants in this cross-sectional study were invited to a health education talk on cervical cancer and were screened with three tests: VIA, the Pap test and the Hybrid Capture II assay. In Phase II of the study all women were examined by colposcopy and biopsies were taken in colposcopically suspect areas, so that sensitivity and specificity could be evaluated without verification bias, assuming 100% accuracy for the applied gold standard. The outcome thresholds were CIN1+ and CIN2+. Data from phase I of the study were not included in this meta-analysis, because of the occurrence of verification bias. In phase I, diagnosis was supposed to be conducted for all screen-positives and a random fraction of screen-negatives, but due to circumstances the fraction of screen-negative women being verified was not random. a Throughout the study we use the Indian name for Calcutta: "Kolkata". 19 Sample size: 2205 women in phase II of the study. Time period: The study took place between October 1995 and August 1997. Setting and characteristics of the study population: The participants were women, 25-55 years old, consulting at one of 15 primary health care clinics in Chitungwiza and greater Harare area. Pregnant women, women having a history of cervical cancer and hysterectomized women were excluded. More details can be found in four publications from the University of Zimbabwe/JHPIEGO [1999], Blumenthal [2001] and Womack [2000a; 2000b]. 4.1.3. Randomized trial: VIA screening versus no screening in rural South India Study design: cluster-randomized screening trial Villages were randomized into two arms: women in villages of group A were offered VIA testing, followed by verification of VIA+ cases with colposcopy and punch biopsies if colposcopically suspect. Women with localized CIN were offered cryotherapya, and women with more extended disease were referred to the hospital for treatment with LEEP or cold knife conization; specific oncological treatment was provided in case of cancer. No screening was given to women in villages from group B. Incidence of invasive cervical cancer and mortality from cervical cancer are the final outcomes of the study. Cancer and mortality surveillance is organized similarly in both groups, by the Ambilikai Cancer Register and mortality registers. VIA screening, colposcopy and biopsies were performed by nurses trained to perform the examinations under the supervision of physicians. Histological examination of the punch biopsies was performed by pathologists. Process indicators in group A: - Participation rate among the eligible population - Follow-up compliance of screen positives - Treatment compliance among women referred for treatment. - Compliance with post-treatment follow-up at 1 year after treatment. Intermediate outcomes in group A are: - Test positivity rate - Detection rate of confirmed CIN or cancer at screening a Local CIN: <3/4 of the TZ involved; not extending to endocervical canal or vaginal wall, whole lesion could be covered by the nipple of the cryoprobe. 20 - Program sensitivity (proportion of cancers detected at screening among all cancers detected in the whole follow-up period). Intermediate outcomes in group A and B are: - Stage distribution at diagnosis of cancer. - Survival from cancer. Initial available results concern the screen test results, the colposcopic and histological verification of VIA+ cases, and the incidence of cervical cancer in the short-term follow-up period. This trial will be the first randomized study ever conducted that will yield a direct estimate of the reduction in incidence of and mortality from cervical cancer as a consequence of screening and management of screen-positive women. Sample size: Group A: 57 clusters, 48 225 women included among whom 30 577 were screened. Group B: 56 clusters, 30 167 women Time period: Recruitment: 1999-2003. Follow-up: 2003-2012. Characteristics of the study population: Age: 30-59 years. Location: Ambilikai area in Dindigul District (South India). Rural never screened population. For more details, we refer to a publication in the International Journal of Cancer [Sankaranarayanan, 2004c]. 4.1.4. Western Kenya Cervical Cancer Prevention Project Study design: Women were screened at the health center with VIAa. Screen-positive women were referred to the district hospital, where they received 3 triage tests: VIAM with the AviScope, a Pap test and colposcopy. Different providers performed VIAM and colposcopy. The providers were family planning nurses or clinical officers previously trained in both visual cervical inspection techniques. Punch biopsies were taken, when there was a colposcopic impression of CIN. Endocervical curettage was performed when it was difficult to see the upper border of the cervical lesion. Immediate cryotherapy was offered to colpo-positive cases. Pap smears were processed in the district hospital and along with the biopsies were sent initially to the provincial hospital of Kisumu for interpretation by cytotechnicians and pathologists and later to a private pathologist in Nairobi. The pathologists were blinded to screening tests results. The Bethesda System (version 1991) was used for reporting the Pap smear results and the CIN classification was used for reporting the a At the end of the study, a group of women was also screened with VILI and the VILI results were used to determine referral to the district hospital. 21 histological results. Women treated with cryotherapy were invited to return for follow-up at 1-3 months (when cytology and biopsy results were available) and for a final visit at 12 months. The main outcomes were: • Accuracy of VIA for different degrees of CIN in a small subset of 133 women receiving all expert colposcopy (by Dr. E. Kauffman) irrespective of screen test results. • PPV of the screen test for different degrees of CIN. • Differences in accuracy of triage methods, using the biopsy result and, if colposcopy was negative, the colposcopy result as gold standard.a • Compliance of screen positives with referral and follow-up • Safety of cryotherapy and outcome of treated lesions. Sample size: 2295 were screened with VIA and 137 among them also with VILI. 116 women received cryotherapy. Time period: Screening took place between November 2000 and March 2004. Before Spring 2005 all treated women could be expected to have undergone follow-up visit. Setting and characteristics of the study population: Women from a rural area, in the Busia District (Western Kenya), in the age range 29- 40 years. a Expert colposcopy results were used as gold standard if available. 22 4.1.5. Safety, acceptability, and feasibility of a single-visit approach to cervical cancer prevention in rural Thailand Study design: Enrolled women were tested by VIA. VIA-positive women were offered immediate cryotherapy, if eligible, after counseling about the benefits, potential risks, and probable side-effects. Treated women were followed at 3 and 12 months after treatment to assess safety, acceptability of cryotherapy. Compliance with hygiene and follow-up instructions was an additional outcome measure. The test positivity rate and the proportion of test-positives that are not eligible for treatment are also documented. Sample Size: 5999 women were tested by VIA. 756 women received cryotherapy. Time period: Women were enrolled in the period February – November 2000; treated women were followed at 12 months. Setting and characteristics of the study population: Twelve trained nurses provided services in mobile (village health center-baseda) and fixed (hospital-based) teams in four rural districts of Roi-et Province (Thailand) Women were in the age range 30-45 years. 4.1.6. Safety, acceptability, and feasibility of a single-visit approach to cervical-cancer prevention in Ghana Study design: The study enrolled women, consulting at the hospital. They were tested with VIA by trained health workers. Many came to the facility specifically for testing based on word of mouth information from family members or friends. VIA-positive cases were offered immediate cryotherapy or referral if indicated. The study design was similar to that of the Thai demonstration project (see 4.1.5) and is further explicated in the flow chart in Figure 2 [Gaffikin, 2004]. Contra-indications for immediate treatment were: suspicion of cancer, large lesions extending onto the vaginal wall or occupying >3/4 of the cervix, extending 2mm or more beyond the cryoprobe, cervicitis, polyps. Women with cervicitis were reassessed after antibiotic treatment. Women suspect for having cervical cancer or with large lesions were referred for evaluation with biopsies and were offered appropriate treatment. Women, receiving cryotherapy were amply informed about expected side effects and alarm signals requiring further care. They were counseled to abstain from sexual relations for 1 month. If sexual abstinence was not possible, condom use was recommended and a supply of condoms was offered. a Mobile teams: services provided by a trained team from the fixed site. 23 Figure 2. Flow chart diagram applied in the SAFE project (extracted from Gaffikin, 2004). Sample Size: 3665 women were tested with VIA.a a A second similar study was conducted in the rural Health Centre of Amansaman. Three thousand two hundred and six were tested with VIA. The Amasaman data files could not be analysed as the data had not been fully cleaned at the time of this analysis. 24 Time period: Women were enrolled during an 18-month period (March 2001-August 2002); treated women were followed at 12 months. Setting and characteristics of the study population: The target population was restricted to women in the age range 25-45 years. Enrolment took place in the family planning clinic in the Ridge Hospital, an urban regional hospital located in the Accra Metropolitan Area of the Greater Accra Region in Ghana. Total hysterectomy, history of cervical cancer and pregnancy (based on clinical examination) were exclusion criteria for participation. 4.2. Methods Addressed study questions What is the accuracy of five cervical cancer screening tests to detect different degrees of cervical cancer or cervical cancer precursors? This question was the focus of the first two studies. The evaluated tests were: VIA (naked eye visual inspection after application of 3-5% acetic acid); VIAM (visual inspection after application of acetic acid solution using a magnifying instrument); VILI (naked eye visual inspection of the cervix after application of Lugol’s iodine); cytological examination of a Pap smear; high-risk HPV DNA detection using the Hybrid Capture II assay [Lörincz, 1997]. What is the test positivity rate and does this vary through the different studies? This question could be addressed in all included studies. How accurate are triage tests applied to VIA-positive women? The evaluated triage methods were: cytological examination of a Pap smear, VIAM, VILI, and colposcopy. What was the treatment compliance of women referred for treatment? What is the safety and acceptability of cryotherapy applied to VIA-positive women or women with colposcopic confirmation of CIN? Definition of disease Four levels of prevalent disease are distinguished throughout this pooled analysis: cervical intra-epithelial neoplasia grade I (CIN 1), CIN 2, CIN 3, and invasive cervical cancer [Richart, 1973]. High-grade CIN is used for the lumping of CIN2 and CIN3. Definition of test positivity VIA or VIAM positivity is defined as: presence of opaque, dull, well-defined confluent acetowhite lesions touching the squamo-columnar junction (SCJ) or close to the external ostium uteri or surrounding the os; presence of warts and leukoplakia turning acetowhite. Dense, opaque aceto-white visible ulceroproliferative growth indicates lesions that are compatible with suspicion of cervical cancer. VILI positivity was defined as: presence of well-defined, dense, yellow non-iodine uptake areas in the transformation zone touching the SCJ, or, circumorificial area occupying a large portion of the cervix, or ulceroproliferatrive growth on the cervix turning yellow after 25 application of Lugol’s iodine [Sankarananarayan, 2003b; Sankarananarayan, 2004c; Sankaranarayanan, 2000]. For cytology, the three-tier classification of Bethesda (1991) was applied: atypical squamous cells of unspecified significance (ASCUS), low-grade squamous intra- epithelial lesion (LSIL) and high-grade squamous intra-epithelial lesion (HSIL) [Lundberg, 1989; Luff, 1992]. For the detection of DNA of 13 high-risk human papillomaviruses, the Hybrid Capture II assay was used. Positivity in all included studies was defined as RLU (relative light units)>1, which is the standard criterion proposed by the manufacturer. This cut-off corresponds with presence of more than 1pg of HPV DNA/pg. We always dichotomized test results in our statistical analysis. For instance: "LSIL+" means: a cytological result compatible with LSIL, HSIL or cancer. Evaluated test parameters We have assumed that colposcopy followed by the histological examination of punch biopsies provided complete ascertainment of the true cervical status. We did not deal with potential misclassification due to potential imperfect gold standard verification. Under this assumption we were able to compute the following test characteristics for the two first studies (the cross-sectional multi-center and the Zimbabwean trials): sensitivity, specificity, the negative and predictive values. We refer to previous work [Arbyn, 2002] and standard epidemiology handbooks for the definition of these parameters. Further, for these studies, we can compute: the test positivity rate, the prevalence of disease and the detection rate of disease. The detection rate of disease is the proportion of the true positives among all tested subjects, which equals the product of the test sensitivity and the prevalence of disease. An overall test accuracy parameter, integrating sensitivity and specificity, is the diagnostic odds ratio (DOR). OddsTPR Sensi (1 − Sensi ) Sensi * Speci DOR = = = . OddsFPR (1 − Speci) Speci (1 − Sensi) * (1 − Speci) The logarithm of the DOR is the outcome of summary ROC-curve regression analysis [Moses, 1993; Pepe, 2003]. In the Ambilikai trial and the Kenya study, verification was limited to VIA-positive cases. In this circumstance we can only compute the test positivity rate, the positive predictive value and the detection rate of disease. In this situation, sensitivity and specificity cannot be computed anymore, since it requires verification of at least a random sample of test negative subjects. Nevertheless, an approximated specificity can be computed. The approximated specificity is defined as the number of test- negatives divided by the total number of tested cases from which the number of true positive cases is subtracted. # negative cases Specificityapprox = . Using the conventional 2x2 contingency table, N - # TP where a = #TP, b = #FP, c = #FN and d = # TN, the approximated specificity =(c+d)/(c+b+d); whereas the real specificity = d/(b+d). This approximated specificity is an acceptable proxy for the real specificity when the prevalence of disease is low 26 (<10%a), the sensitivity is high and/or the specificity is high. In these situations the value of "c" becomes so small that it can be ignored. The demonstration projects conducted in Thailand and Ghana, were not set up to study test accuracy. Nevertheless, information of the VIA-test positivity rates can be derived from both studies. 4.3. Statistical analyses Pooling methods The common variables and data values in all data files were recoded into identical variables and files were merged. We used random-effect meta-analytical methods to pool test accuracy measures from the different study sites for each category of CIN [Sharp, 1997; Sutton, 2000]. Such methods allow for inter-study heterogeneity, which is very common in diagnostic research. Two types of meta-analyses are performed: (1) pooling to the accuracy of separate tests and (2) pooling of the relative sensitivity and specificity of one test compared to another test. Regression models We applied a logistic regression model on the pooled individual patient files to explore the multi-variate impact of study characteristics on the positive predictive value, the sensitivity and the specificity [Hosmer, 1989]. From the estimated constant in the logit-transformed regression equation we could compute the studied accuracy parameter. For instance, the PPV of the basic or reference case is computed as: e b0 PPV0 = , where b0= is the constant term of the logistic regression equation. 1 + e b0 Most logistic software packages provide odds ratios (OR) for the respective categories. We transformed these odds ratios into relative risks (RR) using the following formula: ORi e ( b0 + bi ) RRi = .where ORi = and bi is the coefficient of the (1 − PPV0 + PPV0 .ORi ) e b0 logistic equation term corresponding with a certain category i [Zhang, 1998; McNut, 2003]. This computed RRi has a straightforward interpretation: the PPVi is RRi times that of PPV0. The summary Receiver Operating Characteristic (ROC) Curve allows simultaneous consideration of sensitivity and specificity of a diagnostic test assessed in a number of independent studies [Moses, 1993; Irwig, 1994; Irwig, 1995]. A fitted ROC curve is obtained by applying the following steps: a When the prevalence is low, the term c is automatically low as well since it represents only a part of the subjects with disease. When sensitivity is high, the FNR (and so also c) must be low. When specificity is high the weight of term d increases. 27 1. Fit a linear regression for D as dependent variable and S as explanatory variable, where D and S are respectively the difference and the sum of the logits of the sensitivity and the FPR (=1-Specificity). D = a + b* S ⎛ Sensitivity ⎞ ⎛ (1 − Specificity ) ⎞ D = Ln⎜ ⎜ (1 − Sensitivity ) ⎟ − Ln⎜ Specificity ⎟ ⎟ ⎜ ⎟ ⎝ ⎠ ⎝ ⎠ ⎛ Sensitivity ⎞ ⎛ (1 − Specificity ) ⎞ S = Ln⎜ ⎜ (1 − Sensitivity ) ⎟ + Ln⎜ Specificity ⎟ ⎟ ⎜ ⎟ ⎝ ⎠ ⎝ ⎠ The parameter D is equivalent to the log of the diagnostic odds-ratio (odds of a positive result among women with, for instance, CIN2+ to the odds of a positive test among women without CIN2+). S is a proxy for the test threshold; b1 is the slope and b0 the intercept of the linear regression equation. If b1 does not differ significantly from zero, then D=b0. In this case, a single constant odds ratio defines a single symmetric SROC. When the slope b1 is not different form zero, the odds ratio depends on the threshold yielding an asymmetric summary ROC curve [Glasziou, 2001]. Covariate variables can be included in the linear model allowing explanation of the variation of sensitivity and specificity by study characteristics. 2. Back transform the results into the ROC space where sensitivity is plotted ^ against FPR (=1-specificity). The relation between the fitted sensitivity ( Se ) and the false positive rate can be obtained as follows, from the estimated linear regression parameters b0 and b1: ⎡ (b + logit(FPR)(1 + b ) ⎤ exp ⎢ 1 0 ⎥ ⎢ (1 − b ) ⎥ Se = ˆ ⎣ 0 ⎦ ⎡ (b + logit(FPR)(1 + b ) ⎤ 1 + exp ⎢ 1 0 ⎥ ⎢ (1 − b ) ⎥ ⎣ 0 ⎦ The regression equation can be extended by inclusion of covariates: D= b0 + b1S + BaX+ BpY + BlZ, where b0 is the intercept, b1 is the slope of the relation between D and S, S is the sum of logit of the TPR and the FPR; Ba, Bp and Bl are the column vectors containing the coefficients corresponding to the age, period and location categories. Statistical software Statistical analyses were performed with software package STATA, version 7 and 8 (Stata Corp., College Station, Texas, US). 28 5. Results of the separate studies 5.1. Cross-sectional multi-center study on the test accuracy of VIA, VIAM, VILI, cytology and the HC2 assay 5.1.1. Test accuracy of VIA and VILI for CIN2+ The forest plots in Figure 3 display the variation of the sensitivity and specificity of VIA or VILI, considering CIN2 or worse disease as outcome. The sensitivity of VIA varied between 61.5% (CI= 53.5% to 69.0%), in Calcutta 1, and 91.1% (CI: 85.7% to 94.9%) in Conakry. The pooled sensitivity was 79.2% (CI= 73.3% to 85.0%). The minimal specificity of VIA (74.2%; CI= 72.2 to 76.1) was observed in Ouagadougou, and the highest specificity (93.8%; CI= 93.2% to 94.3%) was found again in Conakry. The pooled specificity of VIA was 84.7% (CI= 80.7% to 88.8%). In general, the sensitivity of VILI was higher than VIA with the exception of Jaipur (87.5%; CI= 76.8% to 83.2%) and Trivandrum-2 (80.2%; CI= 70.9% to 88.3%) where sensitivities were equal to those of VIA. The specificities of VILI varied over a similar range as VIA, between 73.0 % and 91.6%. The overall pooled sensitivity for VILI (91.2%; CI= 87.8% to 94.6%) was statistically significantly higher than for VIA. On the other hand, the pooled specificity of VILI 84.5% (CI= 81.3% to 87.8%) was not significantly different from that of VIA. Table 3 shows, besides sensitivity and specificity, also the predictive values, the prevalence of CIN2+ and the detection rates of CIN2+ in all settings where VIA and VILI were evaluated. The pooled prevalence of CIN2+ was 2.3% (CI= 1.6% to 3.0%) with a lowest value of 0.8%, noted in Niamey, and a highest value of 3.3% in Trivandrum 1. This yielded pooled positive predictive values (PPV) with wide overlapping confidence intervals, for VIA (11.6%; CI: 8.1% to 15.1%) and VILI (12.9%; CI: 8.0% to 17.9%). The negative predictive values (NPV) of both inspection methods were less heterogeneous and in general higher than 99%. Moreover the pooled negative predictive value of VILI (99.8%; CI: 99.7% to 99.9%) was significantly higher than that of VIA (99.4%; CI: 99.2-99.6%). The test positivity rates varied widely, but were highly correlated (Spearman’s correlation coefficient=0.93), ranging from 6.0% and 9.0%, observed in Niamey, to 24.2% and 21.3%, observed in Trivandrum 1, for respectively VIA and VILI. Summary of test accuracy of all screening tests for all categories of CIN The sensitivity and specificity for all tests and outcomes are summarized in Table 4. The sensitivity rose substantially with increasing severity of the outcome (>22% difference in sensitivity for CIN1+ and cancer), whereas the specificity decreased (≤3 % difference in specificity for CIN1+ and cancer). All accuracy measures showed statistically significant inter-study heterogeneity (p<0.01 for Cochrane’s Q test) with the exception of the sensitivity of the HC2 test for outcomes of CIN2+, CIN3+ and cancer, which were statistically homogeneous (p>0.2 for Cochrane’s Q test). 29 Table 3. Number of true and false positive and negative results, test accuracy of VIA and VILI to detect CIN2 or more severe cervical neoplasia; test positivity rate, prevalence of CIN2+ and detection rate of CIN2+; values from 11 centers in Africa and India and meta-analytically pooled values. VIA Study location TP FN FP TN Se Sp PPV NPV T+ rate Prev Det rate Bamako 130 34 494 4894 0.793 0.908 0.208 0.993 0.112 0.030 0.023 Brazzaville 313 76 1532 5014 0.805 0.766 0.170 0.985 0.266 0.056 0.045 Conakry 153 15 527 7932 0.911 0.938 0.225 0.998 0.079 0.019 0.018 Niamey 13 7 139 2375 0.650 0.945 0.086 0.997 0.060 0.008 0.005 Ouagadougou 45 5 516 1485 0.900 0.742 0.080 0.997 0.274 0.024 0.022 Calcutta 1 99 62 1023 4710 0.615 0.822 0.088 0.987 0.190 0.027 0.017 Calcutta 2 52 19 858 7151 0.732 0.893 0.057 0.997 0.113 0.009 0.006 Jaipur 56 8 1426 4296 0.875 0.751 0.038 0.998 0.256 0.011 0.010 Mumbai 50 31 467 3456 0.617 0.881 0.097 0.991 0.129 0.020 0.012 Trivandrum 1 132 17 945 3363 0.886 0.781 0.123 0.995 0.242 0.033 0.030 Trivandrum 2 65 16 514 4164 0.802 0.890 0.112 0.996 0.122 0.017 0.014 TOTAL/Pooled 1108 290 8441 48840 0.792 0.847 0.116 0.994 0.167 0.023 0.018 VILI Study location TP FN FP TN Se Sp PPV NPV T+ rate Det rate Bamako 159 5 568 4820 0.970 0.895 0.219 0.999 0.131 0.029 Brazzaville 371 17 723 5823 0.956 0.890 0.339 0.997 0.158 0.054 Conakry 163 5 811 7648 0.970 0.904 0.167 0.999 0.113 0.019 Niamey 18 2 210 2304 0.900 0.916 0.079 0.999 0.090 0.007 Ouagadougou 49 1 539 1462 0.980 0.731 0.083 0.999 0.287 0.024 Calcutta 2 57 13 1081 6925 0.814 0.865 0.050 0.998 0.141 0.007 Jaipur 56 8 1543 4179 0.875 0.730 0.035 0.998 0.276 0.010 Mumbai 60 21 626 3296 0.741 0.840 0.087 0.994 0.171 0.015 Trivandrum 1 136 13 813 3495 0.913 0.811 0.143 0.996 0.213 0.031 Trivandrum 2 65 16 615 4063 0.802 0.869 0.096 0.996 0.143 0.014 TOTAL/Pooled 1134 101 7529 44015 0.912 0.845 0.129 0.998 0.172 0.021 Table 4. Sensitivity and specificity of 5 screening tests for CIN1 or more severe disease (CIN1+), CIN2+, CIN3+ and cancer; minimum, maximum and meta-analytically pooled measures. Test Outcome Test cutoff Sensitivity Specificity Min Max Pooled (95% CI) Min Max Pooled (95% CI) VIA CIN 1+ AW lesions 0.425 0.900 0.618 (0.523 to 0.713) 0.752 0.951 0.865 (0.828 to 0.901) CIN2+ or growth 0.650 0.911 0.792 (0.733 to 0.850) 0.742 0.945 0.847 (0.807 to 0.888) CIN3+ 0.583 0.946 0.829 (0.771 to 0.887) 0.738 0.943 0.842 (0.800 to 0.883) Cancer 0.667 1.000 0.887 (0.831 to 0.943) 0.731 0.941 0.836 (0.793 to 0.880) VILI CIN 1+ Non iodine 0.503 0.941 0.737 (0.630 to 0.845) 0.741 0.928 0.866 (0.834 to 0.898) CIN2+ uptake 0.741 0.980 0.912 (0.878 to 0.946) 0.730 0.916 0.845 (0.813 to 0.878) CIN3+ yellow areas 0.729 1.000 0.938 (0.906 to 0.971) 0.726 0.914 0.838 (0.805 to 0.871) Cancer or growth 0.667 1.000 0.957 (0.918 to 0.997) 0.719 0.911 0.832 (0.798 to 0.865) VIAM CIN1+ AW lesions 0.425 0.684 0.585 (0.432 to 0.739) 0.864 0.901 0.881 (0.858 to 0.904) CIN2+ or growth 0.646 0.732 0.670 (0.618 to 0.722) 0.833 0.893 0.862 (0.824 to 0.900) CIN3+ 0.657 0.744 0.682 (0.618 to 0.747) 0.828 0.891 0.859 (0.820 to 0.898) Cancer 0.763 1.000 0.826 (0.677 to 0.976) 0.824 0.889 0.855 (0.815 to 0.896) 30 Test Outcome Test cutoff Sensitivity Specificity Min Max Pooled (95% CI) Min Max Pooled (95% CI) Pap CIN1+ ASCUS+ 0.230 0.655 0.343 (0.153 to 0. 532) 0.866 0.987 0.946 (0.915 to 0. 977) smear CIN2+ 0.333 0.819 0.570 (0.376 to 0. 763) 0.865 0.985 0.928 (0.887 to 0. 968) CIN3+ 0.356 0.964 0.630 (0.379 to 0. 882) 0.863 0.982 0.923 (0.881 to 0. 966) Cancer 0.400 1.000 0.725 (0.549 to 0. 900) 0.857 0.977 0.918 (0.875 to 0. 962) CIN1+ LSIL+ 0.172 0.633 0.306 (0.112 to 0. 499) 0.929 0.993 0.967 (0.948 to 0. 985) CIN2+ 0.238 0.779 0.512 (0.300 to 0. 724) 0.886 0.991 0.949 (0.921 to 0. 977) CIN3+ 0.267 0.893 0.561 (0.327 to 0. 796) 0.873 0.988 0.945 (0.916 to 0. 975) Cancer 0.200 1.000 0.651 (0.432 to 0. 871) 0.865 0.983 0.941 (0.910 to 0. 971) Pap CIN2+ HSIL+ 0.175 0.617 0.426 (0.265 to 0. 586) 0.977 0.997 0.993 (0.988 to 0. 997) Smear CIN3+ 0.222 0.768 0.516 (0.320 to 0. 711) 0.975 0.997 0.990 (0.984 to 0. 995) Cancer 0.200 1.000 0.651 (0.432 to 0. 871) 0.973 0.995 0.985 (0.978 to 0. 993) HC2 CIN1+ RLU>1 0.215 0.337 0.266 (0.215 to 0.316) 0.922 0.951 0.940 (0.929 to 0.951) CIN2+ 0.484 0.677 0.619 (0.562 to 0.677) 0.916 0.946 0.936 (0.924 to 0.948) CIN3+ 0.623 0.735 0.684 (0.615 to 0.754) 0.914 0.944 0.934 (0.922 to 0.946) Cancer 0.615 0.857 0.721 (0.603 to 0.838) 0.911 0.940 0.930 (0.918 to 0.942) VIA, Outcome=CIN2+ VIA, Outcome=CIN2+ Bamako Bamako Brazzavile Brazzavile Conakry Conakry Niamey Niamey Ouagadougou Ouagadougou Ca1cutta 1 Ca1cutta 1 Calcutta 2 Calcutta 2 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity VILI, Outcome=CIN2+ VILI, Outcome=CIN2+ Bamako Bamako Brazzavile Brazzavile Conakry Conakry Niamey Niamey Ouagadougou Ouagadougou Calcutta 2 Calcutta 2 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity Figure 3. Forest plot of the sensitivity and specificity of visual inspection of the cervix after application of acetic acid (VIA) or Lugol’s iodine (VILI) to detect CIN2 or more serious cervical disease. The forest plot at the bottom represents the 95% confidence interval of the pooled measure computed from a random effect model. 31 5.1.2. Pooled relative accuracy of the screening tests The forest plots in Figure 4 display the relative sensitivity and specificity of VILI compared to VIA with respect to the prediction of CIN2+ or CIN3+. The relative accuracy of the other screening tests is documented in Table 5. Overall, the sensitivity of VILI for CIN2+ and CIN3+ was respectively 10.5% (CI= 4.8% to 16.5%) and 7.4% (CI= 4.3% to 10.6%) higher than the sensitivity of VIA. The relative sensitivity of VILI was considerably higher in the African studies. Specificities of both tests were not statistically significantly different. The accuracy of VIAM was similar to that of VIA. The Pap smear had a significantly lower sensitivity for CIN2+, even at the lowest cytological cutoff of ASCUS+ (relative sensitivity= 0.742; CI= 0.576 to 0.958), but also showed a significantly higher specificity, and this difference rose with the test threshold. The sensitivity of HC2 was lower than that of VIA but this difference did not reach the level of statistical significance. On the other hand, the specificity of HC2 was 7% to 8% higher than that of VIA, and this difference was significant. For all histological outcomes and cytological cutoffs, the Pap smear was significantly less sensitive but more specific than VILI. The HC2 assay also had lower sensitivity than VILI, but this finding was only significant for CIN2+. On the other hand, the specificity of the HC2 test was significantly higher than VILI. The Pap smear showed a lower sensitivity and a higher specificity than the HC2 test. The difference in sensitivity was never significant but was significant for specificity when LSIL+ and HSIL+ were considered as cutoff. Outcome CIN2+ Outcome CIN3+ Mali Mali Congo Congo Guinea Guinea Niger Niger Burkina-Fasso Burkina-Fasso Kolkata 2 Kolkata 2 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .5 .67 1 1.5 2 .5 .67 1 1.5 2 Ratio of sensitivity of VILI/VIA Ratio of sensitivity of VILI/VIA Outcome CIN2+ Outcome CIN3+ Mali Mali Congo Congo Guinea Guinea Niger Niger Burkina-Fasso Burkina-Fasso Kolkata 2 Kolkata 2 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .5 .67 1 1.5 2 .5 .67 1 1.5 2 Ratio of specificity of VILI/VIA Ratio of specificity of VILI/VIA Figure 4. Forest plot of the relative sensitivity and specificity of VILI compared to VIA for the prediction of cervical intraepithelial neoplasia, grade 2 or more severe neoplasia (CIN2+) and for CIN3+. 32 Table 5. Relative sensitivity and specificity of VIA, VILI, VIAM, cytology (at cutoffs ASCUS, LSIL and HSIL) and the HC2 assay versus another screening test to detect CIN2 or CIN3 or more serious cervical disease. Test combinations Outcome Relative sensitivity Relative specificity Pooled (95%) Pooled (95%) VILI/VIA CIN2+ 1.105 (1.048 to 1.165)* 0.996 (0.962 to 1.031) CIN3+ 1.074 (1.043 to 1.106)* 0.994 (0.962 to 1.027) VIAM/VIA CIN2+ 1.036 (0.926 to 1.159) 0.997 (0.978 to 1.016) CIN3+ 1.028 ( 0.897 to 1.180) 0.997 (0.978 to 1.016) Pap smear (ASCUS+)/VIA CIN2+ 0.742 (0.576 to 0.958)* 1.125 (1.072 to 1.181)* CIN3+ 0.795 (0.613 to 1.032)* 1.127 (1.073 to 1.183)* Pap smear (LSIL+)/VIA CIN2+ 0.659 (0.491 to 0.884)* 1.152 (1.099 to 1.208)* CIN3+ 0.708 (0.535 to 0.937)* 1.154 (1.099 to 1.211)* Pap smear (HSIL+)/VIA CIN2+ 0.552 (0.400 to 0.763)* 1.205 (1.132 to 1.284)* CIN3+ 0.660 (0.499 to 0.872)* 1.209 (1.133 to 1.291)* HC2/VIA CIN2+ 0.883 (0.775 to 1.007) 1.074 (1.051 to 1.097)* CIN3+ 0.956 (0.781 to 1.169) 1.075 (1.051 to 1.099)* Pap smear (ASCUS+)/VILI CIN2+ 0.735 (0.570 to 0.948)* 1.162 (1.092 to 1.236)* CIN3+ 0.796 (0.618 to 1.024) 1.164 (1.094 to 1.239)* Pap smear (LSIL+)/VILI CIN2+ 0.668 (0.500 to 0.892)* 1.176 (1.096 to 1.261)* CIN3+ 0.721 (0.550 to 0.945)* 1.178 (1.097 to 1.265)* Pap smear (HSIL+)/VILI CIN2+ 0.552 (0.392 to 0.777)* 1.228 (1.144 to 1.318)* CIN3+ 0.676 (0.518 to 0.882)* 1.233 (1.145 to 1.327)* HC2/VILI CIN2+ 0.834 (0.740 to 0.939)* 1.097 (1.085 to 1.110)* CIN3+ 0.857 (0.714 to 1.028) 1.098 (1.084 to 1.112)* Pap smear (ASCUS+)/HC2 CIN2+ 0.957 (0.825 to 1.109) 1.008 (0.956 to 1.063) CIN3+ 0.915 (0.742 to 1.130) 1.008 (0.956 to 1.064) Pap smear (LSIL+)/HC2 CIN2+ 0.870 (0.723 to 1.047) 1.037 (1.013 to 1.061)* CIN3+ 0.816 (0.609 to 1.093) 1.037 (1.014 to 1.060)* Pap smear (HSIL+)/HC2 CIN2+ 0.786 (0.562 to 1.099) 1.061 (1.052 to 1.070)* CIN3+ 0.974 (0.756 to 1.256) 1.061 (1.052 to 1.071)* 5.2. Cross-sectional study on the test accuracy of VIA, cytology and the HC2 assay in Zimbabwe As with the previous multi-center study, the Zimbabwe study also allows computation of test accuracy without verification bias for 3 screening tests: VIA, HPV DNA detection using the Hybrid Capture II assay and the conventional Pap smear. Three different outcomes were considered: histologically confirmed CIN1+, CIN2+ and cervical cancer. No distinction between CIN2 and CIN3 was made. The absolute values and accuracy parameters derived from them are shown in Table 6. 33 Table 6. Number women with true- and false positives and negatives for 3 screening tests considering histologically confirmed CIN1+, CIN2+ or cancer as outcome (part A); derived accuracy parameters (part B); from a cross-sectional cervical cancer screening study conducted by the University of Harare and JHPIEGO, where all tested women were submitted to gold standard verification [University Zimbabwe, 1999; Womack, 2000a; Womack, 2000b, Blumenthal, 2001]. Test Outcome test cutoff tp fn fp tn VIA CIN1+ VIA+ 347 217 502 1064 CIN2+ VIA+ 158 48 691 1233 Cancer VIA+ 3 0 846 1281 Cytology CIN1+ ASCUS+ 257 295 324 1216 LSIL+ 158 394 109 1431 CIN2+ ASCUS+ 130 71 451 1440 LSIL+ 89 112 178 1713 HSIL+ 33 168 40 1851 Cancer ASCUS+ 3 0 578 1511 LSIL+ 3 0 264 1825 HSIL+ 3 0 70 2019 HC2 CIN1+ HC2+ 361 205 556 1023 CIN2+ HC2+ 166 41 751 1187 Cancer HC2+ 2 1 915 1227 Test Outcome test cutoff Se Sp PPV NPV T+ rate Prev Det rate VIA CIN1+ VIA+ 61.5% 67.9% 40.9% 83.1% 39.9% 26.5% 16.3% CIN2+ VIA+ 76.7% 64.1% 18.6% 96.3% 39.9% 9.7% 7.4% Cancer VIA+ 100.0% 60.2% 0.4% 100.0% 39.9% 0.1% 0.1% Cytology CIN1+ ASCUS+ 46.6% 79.0% 44.2% 80.5% 27.8% 26.4% 12.3% LSIL+ 28.6% 92.9% 59.2% 78.4% 12.8% 26.4% 7.6% CIN2+ ASCUS+ 64.7% 76.2% 22.4% 95.3% 27.8% 9.6% 6.2% LSIL+ 44.3% 90.6% 33.3% 93.9% 12.8% 9.6% 4.3% HSIL+ 16.4% 97.9% 45.2% 91.7% 3.5% 9.6% 1.6% Cancer ASCUS+ 100.0% 72.3% 0.5% 100.0% 27.8% 0.1% 0.1% LSIL+ 100.0% 87.4% 1.1% 100.0% 12.8% 0.1% 0.1% HSIL+ 100.0% 96.6% 4.1% 100.0% 3.5% 0.1% 0.1% HC2 CIN1+ HC2+ 63.8% 64.8% 39.4% 83.3% 42.8% 26.4% 16.8% CIN2+ HC2+ 80.2% 61.2% 18.1% 96.7% 42.8% 9.7% 7.7% Cancer HC2+ 66.7% 57.3% 0.2% 99.9% 42.8% 0.1% 0.1% The ratios of the sensitivities and specificities of one test relative to another test for all considered outcomes are presented in Table 7. The prevalence of CIN was particularly high: more than one quarter had histologically confirmed CIN1 or worse and one in ten showed high-grade CIN. The sensitivity of VIA and HC2 for CIN1+ and CIN2+ was always higher than that of the Pap smear, even at the lowest cytological cutoff. On the other hand, the specificity of cytology always was higher than the other two tests. The sensitivity of HC2 and VIA was similar, but the specificity of VIA was marginally significantly higher than the specificity ofHC2. 34 Table 7. Relative sensitivity and specificity of the Pap smear, VIA and HPV DNA detection to detect different degrees of cervical dysplasia and cancer, derived from the cross-sectional study conducted in Zimbabwe. A Cytology,ASCUS+ VIA+ Ratio (VIA/Cyto) Outcome Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib CIN1+ 0.466 0.424 0.507 0.790 0.769 0.809 0.615 0.574 0.654 0.679 0.656 0.702 1.32 1.18 1.48 0.86 0.82 0.90 CIN2+ 0.647 0.578 0.710 0.762 0.742 0.780 0.767 0.705 0.820 0.641 0.619 0.662 1.19 1.04 1.35 0.84 0.81 0.88 Ca 1.000 0.438 1.000 0.723 0.704 0.742 1.000 0.438 1.000 0.602 0.581 0.623 1.00 1.00 1.00 0.83 0.80 0.87 B Cytology,LSIL+ VIA+ Ratio (VIA/Cyto) Outcome Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib CIN1+ 0.286 0.250 0.325 0.929 0.915 0.941 0.615 0.574 0.654 0.679 0.656 0.702 2.15 1.86 2.49 0.73 0.70 0.76 CIN2+ 0.443 0.376 0.512 0.906 0.892 0.918 0.767 0.705 0.820 0.641 0.619 0.662 1.73 1.46 2.06 0.71 0.68 0.73 Ca 1.000 0.438 1.000 0.874 0.859 0.887 1.000 0.438 1.000 0.602 0.581 0.623 1.00 1.00 1.00 0.69 0.66 0.72 C Cytology,HSIL+ VIA+ Ratio (VIA/Cyto) Outcome Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib CIN2+ 0.164 0.119 0.222 0.979 0.971 0.984 0.767 0.705 0.820 0.641 0.619 0.662 4.67 3.39 6.44 0.65 0.63 0.68 Ca 1.000 0.438 1.000 0.966 0.958 0.973 1.000 0.438 1.000 0.602 0.581 0.623 1.00 1.00 1.00 0.62 0.60 0.65 D Cytology,ASCUS+ HC2+ Ratio (HC2/Cyto) Outcome Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib Ucib CIN1+ 0.466 0.424 0.507 0.790 0.769 0.809 0.638 0.597 0.676 0.648 0.624 0.671 1.37 1.23 1.53 0.82 0.78 0.86 CIN2+ 0.647 0.578 0.710 0.762 0.742 0.780 0.802 0.742 0.851 0.612 0.591 0.634 1.24 1.10 1.40 0.80 0.77 0.84 Ca 1.000 0.438 1.000 0.723 0.704 0.742 0.667 0.208 0.939 0.573 0.552 0.594 0.67 0.30 1.48 0.79 0.76 0.83 E Cytology,LSIL+ HC2+ Ratio (HC2/Cyto) Outcome Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib Ucib CIN1+ 0.286 0.250 0.325 0.929 0.915 0.941 0.638 0.597 0.676 0.648 0.624 0.671 2.23 1.93 2.58 0.70 0.67 0.72 CIN2+ 0.443 0.376 0.512 0.906 0.892 0.918 0.802 0.742 0.851 0.612 0.591 0.634 1.81 1.53 2.15 0.68 0.65 0.70 Ca 1.000 0.438 1.000 0.874 0.859 0.887 0.667 0.208 0.939 0.573 0.552 0.594 0.67 0.30 1.48 0.66 0.63 0.68 F Cytology,HSIL+ HC2+ Ratio (HC2/Cyto) Outcome Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib Ucib CIN2+ 0.164 0.119 0.222 0.979 0.971 0.984 0.638 0.597 0.676 0.648 0.624 0.671 3.88 2.83 5.34 0.66 0.64 0.69 Ca 1.000 0.438 1.000 0.966 0.958 0.973 0.802 0.742 0.851 0.612 0.591 0.634 0.80 0.75 0.86 0.63 0.61 0.66 E HC2 VIA Ratio (VIA/HC2) Outcome Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib Ucib CIN1+ 0.638 0.597 0.676 0.648 0.624 0.671 0.615 0.574 0.654 0.679 0.656 0.702 0.96 0.88 1.06 1.05 1.00 1.10 CIN2+ 0.802 0.742 0.851 0.612 0.591 0.634 0.767 0.705 0.820 0.641 0.619 0.662 0.96 0.86 1.06 1.05 1.00 1.10 Ca 0.667 0.208 0.939 0.573 0.552 0.594 1.000 0.438 1.000 0.602 0.581 0.623 1.50 0.67 3.34 1.05 1.00 1.11 5.3. Randomized trial of VIA screening in rural South India In a series of tables, we summarize the main short-term results of the randomized trial conducted in the Ambilikai are in the Dindigul district, which were reported in the International Journal of Cancer [Sankaranarayanan, 2004b] (see Table 8 to Table 11). 35 More than 30,000 women or 63.4% of the total eligible population participated in arm A. In the Indian trial only a limited number of test accuracy outcomes are available: the test positivity rate for VIA was 9.6% (CI: 9.3% to 9.9%); the PPV of a positive VIA test result for an outcome of CIN2 or CIN3 was 9.5% (CI: 8.3% to 10.8%) and the detection rate of CIN2+ was 0.73% (CI: 0.63% to 0.83%). Among the 97 invasive cervical cancers found in group A, 69 (71.1%; CI: 61.0% to 79.9%) were found at screening. The observed incidence was 78.1 (CI: 63.4 to 95.3) per 100,000 women-years in group A, whereas only 37.7 (CI: 26.1 to 52.7) per 100,000 women-years in group B. But in group A, a much more favorable stage distribution was observed (see Table 10). Table 8. Attendance at screening and test performance indicators in group A. [from Sankaranarayanan, 2004b]. Process indicators number % Definition of the Indicator Eligible population 48,225 Screened with VIA 30,577 63.4% Participation rate, among eligible population Non-responders 17,648 VIA+ 2,939 9.6% Test+ rate among women who received VIA Number colposcoped 2,939 100.0% Compliance with colposcopy Number biopsied 2,777 94.5% % with colposcopy who were biopsied Number of VIA+ with CIN+ CIN 1 1,778 5.8% Detection rate, among screened women CIN2/3 222 0.7% Cancer 69 0.2% Total with lesions 2,069 6.8% PPV of VIA for CIN 1+ 2,069 70.4% PPV, among VIA+ women CIN2+ 291 9.9% PPV, among VIA+ women Cancer 69 2.3% PPV, among VIA+ women Table 9. Occasion of cancer detection in group A [from Sankaranarayanan, 2004b]. Cancers detected in the FU period, by occasion Screen detected ca 69 71.1% program sensitivity Interval ca (VIA-, detected within 2yr after screen) 7 Ca among non responders to screening 13 Ca after randomization but before screening 8 TOTAL 97 Table 10. Comparison of short term longitudinal outcomes, number of included women, number of women-years in the follow-up period, number of cervical cancers per stage and cancer incidence rate, in the two arms A (VIA-screened population) and B (control population) [from Sankaranarayanan, 2004b]. Comparison between arms A B Number of women 30 577 30 167 Number of women-years in FU period 124 144 90 172 Number of cancers in FU period, per stage (%) I 34 (35.1%) 0 (0.0%) 36 II 18 (18.6%) 6 (17.6%) III 45 (46.4%) 26 (76.5%) IV 0 (0.0%) 2 (5.9%) Total 97 (100.0%) 34 (100.0%) Cancer incidence rate over FU period (per 100 000 women-years) 78.1 37.7 More than 70% of CIN1 and 80% of CIN2/3 cases were treated (see Table 11). Most of CIN1 cases were treated with cryotherapy the same day (93% of treated cases), whereas almost 4 in 10 cases with high-grade CIN were treated with LEEP or cold knife conization. Fifty-two of the 69 women with screen detected cancer received appropriate cancer treatment. Fifteen percent of women treated for CIN came back for a follow-up visit after one year. Table 11. Treatment compliance in group A [from Sankaranarayanan, 2004b]. To be T with % of Cryo % of LEEP/ % of FU after % of CIN treated Treated % T cryo treated same day treated conization treated 1 yr treated CIN1 1778 1263 71.0% 1179 93.3% 84 6.7% CIN2/3 222 178 80.2% 109 61.2% 69 38.8% Total 2000 1441 72.1% 1288 89.4% 983 68.2% 153 10.6% 222 15.4% 5.4. Western Kenya Cervical Cancer Prevention Project (WKCCPP) Since the results of the WKCCPP study are not yet published, we produced a report, from which we extract the main tables below. About 19% of the examined women had a positive result for VIA, 2% showed an indeterminate result and in 0.7% cancer was suspected. In total 489 among 2295 tested women, or 21.3%, had an abnormal VIA result. Table 12. Distribution of VIA test results in the WKCCPP study. VIA results Number % Negative 1806 78.7% Indeterminate 45 2.0% Positive 427 18.6% Suspect for cancer 17 0.7% Total 2295 100.0% 5.4.1. Positive predictive value and approximated specificity of VIA Three hundred and twelve women among 489 women with abnormal VIA result (63.8%) presented at the district hospital for colposcopic exploration. For 168 women biopsy results are available. We constructed a pseudo-gold standard variable based on the biopsy results, accepting expert colposcopya if no biopsy was taken and finally accepting the nurse colposcopy result when no biopsy or expert colposcopy a Expert colposcopy: Dr. E. Kauffmann, expert colposcopist, did colposcopy for 133 women. 37 results were available. This yielded a pseudo-gold standard diagnosis for 327 VIAindeter+ cases. The positive predictive value of respectively VIA+ or VIAindeter+ can be computed from the verified test positives. Assuming the PPV among the not verified cases is equal to the PPV among verified cases, allows us to compute the extrapolated number of tp' and fp' for all VIA+ (#tp'= PPV*T+rate). The detection rate of disease is computed as the number of true-positives over the number of screened subjects. This parameter reflects sensitivity (but is also dependent on prevalence of the disease and the verification). Finally we computed the approximated specificity, which is an acceptable proxy for the real specificity when the prevalence of disease is low (<10%). All these parameters are shown in Table 13. The PPV for CIN2+ was respectively 11.0% (CI: 7.6% to 15.1%) (when manifest VIA+ is considered as cutoff) or 10.4% (CI: 7.3% to 14.2%) (when indeterminate cases are considered also as VIA+). The approximate specificity to exclude CIN2+ was respectively 82.4% (CI: 80.8% to 84.4%) and 80.5% (CI: 78.8% to 82.1%). Table 13. Number of true- and false positives, number of verified test-positives, PPV, number of expected true- & false-positives assuming PPV is equal for non verified cases, detection rate of disease and approximate specificity of VIA for different degrees of CIN (WKCCPP study). TP/ TP/ (TP+FP) (T+v+T-) T-/(N-tp') test outcome Test cutoff threshold TP FP T+verif PPV T+ tp' fp' T- N Det rate App Spec VIA Indeter- CIN1+ 124 203 327 37.30% 489 182 307 1806 2295 5.8% 85.5% minate+ CIN2+ 34 293 327 10.00% 489 49 440 1806 2295 1.6% 80.4% CIN3+ 17 310 327 5.30% 489 26 463 1806 2295 0.8% 79.6% Ca 6 321 327 1.90% 489 9 480 1806 2295 0.3% 79.0% VIA+ CIN1+ 109 183 292 36.80% 444 163 281 1851 2295 5.1% 86.8% CIN2+ 32 260 292 10.50% 444 47 397 1851 2295 1.5% 82.3% CIN3+ 15 277 292 5.30% 444 24 420 1851 2295 0.7% 81.5% Ca 6 286 292 2.10% 444 9 435 1851 2295 0.3% 81.0% 5.4.2. Performance of triage methods Respectively 271, 280, and 76 women with an abnormal VIA screen test result were submitted to triage testing with VIAM, cytology and VILI and have a pseudo-gold standard result. The number of TP, FN, FP, TN and derived accuracy parameters are shown the following tables (see Table 14). 38 Table 14. Number of true- and false-positive cases and derived accuracy for different degrees of CIN of three triage methods (Pap smear, VIAM and VILI) applied to women referred for an abnormal VIA screen test. Test Test cutoffOutcome threshold TP FN FP TN Pap ASCUS+ CIN1+ 60 43 86 87 CIN2+ 18 7 128 123 CIN3+ 9 2 137 128 Ca 1 0 145 130 AGUS/LSIL+ CIN1+ 52 51 57 116 CIN2+ 17 8 92 159 CIN3+ 9 2 100 165 Ca 1 0 108 167 AGUS/HSIL+ CIN2+ 8 17 10 241 CIN3+ 6 5 12 253 Ca 1 0 17 258 Outcome Test Test cutoffthreshold Sens Spec T+ PPV NPV Prev Pap ASCUS+ CIN1+ 58.3% 50.3% 52.9% 41.1% 66.9% 37.3% CIN2+ 72.0% 49.0% 52.9% 12.3% 94.6% 9.1% CIN3+ 81.8% 48.3% 52.9% 6.2% 98.5% 4.0% Ca 100.0% 47.3% 52.9% 0.7% 100.0% 0.4% LSIL CIN1+ 50.5% 67.1% 39.5% 47.7% 69.5% 37.3% CIN2+ 68.0% 63.3% 39.5% 15.6% 95.2% 9.1% CIN3+ 81.8% 62.3% 39.5% 8.3% 98.8% 4.0% Ca 100.0% 60.7% 39.5% 0.9% 100.0% 0.4% AGUS/HSIL+ CIN2+ 32.0% 96.0% 6.5% 44.4% 93.4% 9.1% CIN3+ 54.5% 95.5% 6.5% 33.3% 98.1% 4.0% Ca 100.0% 93.8% 6.5% 5.6% 100.0% 0.4% Test Test cutoff Outcome threshold TP FN FP TN Tot VIAM VIAM indet+ CIN1+ 86 14 91 80 271 CIN2+ 26 2 151 92 271 CIN3+ 14 1 163 93 271 Ca 5 0 172 94 271 VIAM+ CIN1+ 86 14 90 81 271 CIN2+ 26 2 150 93 271 CIN3+ 14 1 162 94 271 Ca 5 0 171 95 271 Outcome Test Test cutoff threshold Sens Spec T+ PPV NPV Prev VIAM VIAM indet+ CIN1+ 86.0% 46.8% 65.3% 48.6% 85.1% 36.9% CIN2+ 92.9% 37.9% 65.3% 14.7% 97.9% 10.3% CIN3+ 93.3% 36.3% 65.3% 7.9% 98.9% 5.5% Ca 100.0% 35.3% 65.3% 2.8% 100.0% 1.8% VIAM+ CIN1+ 86.0% 47.4% 64.9% 48.9% 85.3% 36.9% CIN2+ 92.9% 38.3% 64.9% 14.8% 97.9% 10.3% CIN3+ 93.3% 36.7% 64.9% 8.0% 98.9% 5.5% Ca 100.0% 35.7% 64.9% 2.8% 100.0% 1.8% 39 Outcome Test Test cutoff threshold TP FN FP TN VILI ASCUS+ CIN1+ 19 5 4 48 CIN2+ 6 0 17 53 CIN3+ 2 0 21 53 Ca 1 0 22 53 Outcome Test Test cutoff threshold Sens Spec T+ PPV NPV Prev VILI VILI+ CIN1+ 79.2% 92.3% 30.3% 82.6% 90.6% 31.6% CIN2+ 100.0% 75.7% 30.3% 26.1% 100.0% 7.9% CIN3+ 100.0% 71.6% 30.3% 8.7% 100.0% 2.6% Ca 100.0% 70.7% 30.3% 4.3% 100.0% 1.3% The relative performance of the triage tests is evaluated by computing the relative sensitivity and specificity (see Table 15). Table 15. Ratio of sensitivities and specificities of the 3 applied triage tests (cytology, VIAM and VILI). A Cytology,ASCUS+ VIAM, Indeterminate+ Ratio (VIAM/Cyto) Outcome Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib CIN1+ 0.583 0.486 0.673 0.503 0.429 0.577 0.860 0.779 0.915 0.468 0.395 0.543 1.48 1.23 1.77 0.93 0.75 1.16 CIN2+ 0.720 0.524 0.857 0.490 0.429 0.552 0.929 0.774 0.980 0.379 0.320 0.441 1.29 0.99 1.68 0.77 0.63 0.95 CIN3+ 0.818 0.523 0.949 0.483 0.424 0.543 0.933 0.702 0.988 0.363 0.307 0.424 1.14 0.84 1.55 0.75 0.61 0.92 Ca 1.000 0.207 1.000 0.473 0.415 0.532 1.000 0.566 1.000 0.353 0.298 0.413 1.00 1.00 1.00 0.75 0.61 0.92 B Cytology,LSIL+/AGUS VIAM, Indeterminate+ Ratio (VIAM/Cyto) Outcome Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib CIN1+ 0.505 0.410 0.599 0.671 0.597 0.736 0.860 0.779 0.915 0.468 0.395 0.543 1.70 1.38 2.10 0.70 0.58 0.84 CIN2+ 0.680 0.484 0.828 0.633 0.572 0.691 0.929 0.774 0.980 0.379 0.320 0.441 1.37 1.02 1.82 0.60 0.50 0.72 CIN3+ 0.818 0.523 0.949 0.623 0.563 0.679 0.933 0.702 0.988 0.363 0.307 0.424 1.14 0.84 1.55 0.58 0.48 0.70 Ca 1.000 0.207 1.000 0.607 0.548 0.663 1.000 0.566 1.000 0.353 0.298 0.413 1.00 1.00 1.00 0.58 0.48 0.70 C Cytology,HSIL+/AGUS VIAM, Indeterminate+ Ratio (VIAM/Cyto) Outcome Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib CIN2+ 0.320 0.172 0.516 0.960 0.928 0.978 0.929 0.774 0.980 0.379 0.320 0.441 2.90 1.62 5.19 0.39 0.33 0.46 CIN3+ 0.545 0.280 0.787 0.955 0.923 0.974 0.933 0.702 0.988 0.363 0.307 0.424 1.71 0.98 2.98 0.38 0.32 0.45 Ca 1.000 0.207 1.000 0.938 0.903 0.961 1.000 0.566 1.000 0.353 0.298 0.413 1.00 1.00 1.00 0.38 0.32 0.44 D Cytology,ASCUS+ VILI Ratio (VILI/Cyto) Outcome Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib CIN1+ 0.583 0.486 0.673 0.503 0.429 0.577 0.792 0.595 0.908 0.923 0.818 0.970 1.36 1.05 1.77 1.84 1.55 2.17 CIN2+ 0.720 0.524 0.857 0.490 0.429 0.552 1.000 0.610 1.000 0.757 0.645 0.842 1.39 1.09 1.77 1.55 1.29 1.86 CIN3+ 0.818 0.523 0.949 0.483 0.424 0.543 1.000 0.342 1.000 0.716 0.605 0.806 1.22 0.93 1.61 1.48 1.23 1.79 Ca 1.000 0.207 1.000 0.473 0.415 0.532 1.000 0.207 1.000 0.707 0.596 0.798 1.00 1.00 1.00 1.49 1.23 1.81 E Cytology,LSIL+/AGUS VILI Ratio (VILI/Cyto) Outcome Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib CIN1+ 0.505 0.410 0.599 0.671 0.597 0.736 0.792 0.595 0.908 0.923 0.818 0.970 1.57 1.18 2.08 1.38 1.21 1.57 CIN2+ 0.680 0.484 0.828 0.633 0.572 0.691 1.000 0.610 1.000 0.757 0.645 0.842 1.47 1.12 1.92 1.20 1.02 1.41 CIN3+ 0.818 0.523 0.949 0.623 0.563 0.679 1.000 0.342 1.000 0.716 0.605 0.806 1.22 0.93 1.61 1.15 0.97 1.37 Ca 1.000 0.207 1.000 0.607 0.548 0.663 1.000 0.207 1.000 0.707 0.596 0.798 1.00 1.00 1.00 1.16 0.98 1.38 F Cytology,HSIL+/AGUS VILI Ratio (VILI/Cyto) Outcome Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib CIN2+ 0.320 0.172 0.516 0.960 0.928 0.978 1.000 0.610 1.000 0.757 0.645 0.842 3.13 1.76 5.53 0.79 0.69 0.90 CIN3+ 0.545 0.280 0.787 0.955 0.923 0.974 1.000 0.342 1.000 0.716 0.605 0.806 1.83 1.07 3.14 0.75 0.65 0.87 Ca 1.000 0.207 1.000 0.938 0.903 0.961 1.000 0.207 1.000 0.707 0.596 0.798 1.00 1.00 1.00 0.75 0.65 0.87 40 E VIAM VILI Ratio (VILI/VIA) Outcome Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib Se lcib ucib Sp lcib ucib CIN1+ 0.860 0.779 0.915 0.468 0.395 0.543 0.792 0.595 0.908 0.923 0.818 0.970 0.92 0.74 1.15 1.97 1.65 2.36 CIN2+ 0.929 0.774 0.980 0.379 0.320 0.441 1.000 0.610 1.000 0.757 0.645 0.842 1.08 0.97 1.19 2.00 1.62 2.46 CIN3+ 0.933 0.702 0.988 0.363 0.307 0.424 1.000 0.342 1.000 0.716 0.605 0.806 1.07 0.94 1.23 1.97 1.59 2.45 Ca 1.000 0.566 1.000 0.353 0.298 0.413 1.000 0.207 1.000 0.707 0.596 0.798 1.00 1.00 1.00 2.00 1.61 2.49 VIAM is more sensitive, but less specific than cytology at all cytological cutoffs for detection of CIN1+, CIN2+ and CIN3+. The sensitivity of detecting cancer is the same. Differences in sensitivity were significant (using independent comparisonsa) when: • the outcome is CIN1+ for all cytological cutoffs; • the outcome is CIN2+ for the cytological cutoffs LSIL/AGUS+ and HSIL/AGUS+. Differences in specificity are significant (using independent comparisons) when: • the outcome is CIN1+ for the cytological cutoffs LSIL/AGUS+ and HSIL/AGUS+ • the outcome is CIN2+ for all cytological cutoffs. The difference in sensitivity was marginally not significant for the outcome CIN2+ when the cytological cutoff was ASCUS+ (relative sensitivity of 1.29 (CI: 0.99-1.68). Although VILI triage was applied to a limited number of screen positives, some statistically significant results were obtained. VILI is more sensitive than cytology even at the lowest cytological cutoff and this difference is significant for the outcome CIN1+ and CIN2+. The sensitivity of VILI triage is similar to that of VIAM but the specificity of VILI was higher. This difference in specificity was statistically significant for all outcomes. To conclude: VILI looks the most accurate triage method, but this finding is based on a limited number of observations. Moreover, VILI does not require technical equipment and allows a diagnosis on the spot. 5.4.3. Combination of VIA screening with subsequent triage The advantage of adding a triage method is an increase in PPV and a decrease in the proportion of treated women who did not require treatment. The PPV of the combination screen test followed by triage can be computed as following: 1-(1-PPVscreen test)*(1-Spec triage test). The PPV of VIA for CIN2+ and CIN3+ was respectively 10.0% and 5.3%. The specificity of VILI triage of VIA+ cases considering the same outcomes was respectively 75.7% and 71.6%. This means that the PPV of the combination becomes, for CIN2+: 1-(1-0.10)*(1-0.757) =78.1% and for CIN3+: 1-(1-0.053)*(1-0.716) =73.1%. This means that "see and treat" VIA+ cases includes treatment of women where 90% (1-PPVVIA) do not have CIN2+. With VILI triage (see-triage-treat strategy), only 22% of treated women will not have a CIN2+ lesion. a The statistical significance of differences did not change using paired comparisons (McNemar test). 41 A see and treat strategy based on VIA-positivity, applied to the WKCCPP situation would require treatment of 18.6% of the eligible screening population (or even 20.6% if indeterminate cases are included). Adding VILI triage should decrease this proportion. This proportion to be treated when combining screening and triage can be computed as following: T+ ratescreen test * T+ ratetriage test. For the combination of VIA screening followed by VILI triage, the proportion of eligible screened population that is treated would be: 18.6%*33.3%=5.6%. In absolute figures: in a population of 10,000 eligible screened women: 1860 women should require treatment in a see and treat strategy based on VIA screening alone; whereas only 560 women should need to be treated in a see-triage-treat strategy. Triage increases the specificity and the PPV of the screening strategy, but includes a risk of drop out when triage is organized at another occasion at a distant place. Another risk is to miss cases when the triage method has a sensitivity lower than 100%. In the WKCCPP project 36.2% of VIA+ women did not present at the district hospital for diagnostic follow-up. The prevalence of disease among VIA+ cases was: 7.9% of CIN2+, 2.6% of CIN3+, and 1.3% of cancer. This dropout rate means that in a population of 10,000 women eligible for screening, 147 of CIN2+ would not receive treatment, as well as 49 with CIN3+ and 24 with cancer (10000*test positivity rate*drop out rate*prevalence of disease among VIA-positives). The see-triage-treatment procedure can reduce the number of women to be treated but is only acceptable when a high follow-up rate can be assured or triage is done at the time of screening. It also reduces the number of true cases of disease picked up due to sensitivity problems resulting in decreased effectiveness. 5.4.4. Experience of women in the 1-3 month period after cryotherapy The quality of the information concerning safety and acceptability of cryotherapy is limited given the low attendance of treated women at the foreseen follow-up visits. Further the post -treatment follow-up procedures and the definition of follow-up variables changed over time. In a first phase a contact was foreseen at 1 and 3 months; in the second period there was only one visit between 1 to 3 months after treatment. Among the 116 women having received cryotherapy, 31 had a follow-up in the first pilot phase or in the expansion phase. Below, we present a few mono-variate frequency distributions for the main follow-up questions: 42 Lower back or abdominal pain | Freq. Percent Cum. ------------+----------------------------------- No | 9 29.03 29.03 Yes | 19 61.29 90.32 missing | 3 9.68 100.00 ------------+----------------------------------- Total | 31 100.00 Severe back pain or abdominal pain after cryotherapy | Freq. Percent Cum. ------------+----------------------------------- mild | 4 12.90 12.90 moderate | 3 9.68 22.58 severe | 2 6.45 29.03 missing | 22 70.97 100.00 ------------+----------------------------------- Total | 31 100.00 Vaginal bleeding 1-3M post-cryotherapy | Freq. Percent Cum. ------------+----------------------------------- No | 23 74.19 74.19 Yes | 8 25.81 100.00 ------------+----------------------------------- Total | 31 100.00 Severity of vaginal bleeding 1-3 months post-cryo | Freq. Percent Cum. ------------+----------------------------------- mild | 5 16.13 16.13 moderate | 3 9.68 25.81 missing | 23 74.19 100.00 ------------+----------------------------------- Total | 31 100.00 Cervical abnormality noted (in expansion phase only, 1-3 months post-cryotherapy). | Freq. Percent Cum. ------------+----------------------------------- No | 9 100.00 100.00 ------------+----------------------------------- Total | 9 100.00 Colposcopy results at 3 months follow-up (exam-pilot phase only). | Freq. Percent Cum. ------------+----------------------------------- negative | 13 81.25 81.25 CIN 1 | 2 12.50 93.75 CIN 2 | 1 6.25 100.00 ------------+----------------------------------- Total | 16 100.00 Biopsy results at 3 month follow-up (exam-pilot phase only) | Freq. Percent Cum. ------------+----------------------------------- CIN 1 | 1 5.88 5.88 missing | 16 94.12 100.00 ------------+----------------------------------- Total | 17 100.00 43 Pap smear results at 3 month follow-up exam-pilot phase only | Freq. Percent Cum. ------------+----------------------------------- negative | 3 60.00 60.00 ASCUS | 1 20.00 80.00 LSIL | 1 20.00 100.00 ------------+----------------------------------- Total | 5 100.00 Final diagnosis at 3 month follow-up exam-pilot phase only | Freq. Percent Cum. ------------+----------------------------------- negative | 1 5.88 5.88 CIN 1+ | 1 5.88 11.76 missing | 15 88.24 100.00 ------------+----------------------------------- Total | 17 100.00 Sex since cryotherapy? | Freq. Percent Cum. ------------+----------------------------------- No | 6 19.35 19.35 Yes | 21 67.74 87.10 missing | 4 12.90 100.00 ------------+----------------------------------- Total | 31 100.00 Bleeding with sex post-cryotherapy? | Freq. Percent Cum. ------------+----------------------------------- No | 8 25.81 25.81 Yes | 1 3.23 29.03 missing | 22 70.97 100.00 ------------+----------------------------------- Total | 31 100.00 Pain with sex post-cryotherapy? | Freq. Percent Cum. ------------+----------------------------------- No | 6 19.35 19.35 Yes | 4 12.90 32.26 missing | 21 67.74 100.00 ------------+----------------------------------- Total | 31 100.00 Other problems with sex post-cryotherapy? | Freq. Percent Cum. ------------+----------------------------------- No | 8 25.81 25.81 Yes | 1 3.23 29.03 missing | 22 70.97 100.00 ------------+----------------------------------- Total | 31 100.00 Other problems with sex post-cryotherapy | Freq. Percent Cum. -----------------+----------------------------------- post-coital pain | 1 100.00 100.00 -----------------+----------------------------------- Total | 1 100.00 44 Days between treatment and sex | Freq. Percent Cum. ------------+----------------------------------- 7 | 1 3.23 3.23 10 | 2 6.45 9.68 14 | 1 3.23 12.90 21 | 1 3.23 16.13 25 | 1 3.23 19.35 28 | 3 9.68 29.03 30 | 3 9.68 38.71 35 | 1 3.23 41.94 42 | 1 3.23 45.16 45 | 1 3.23 48.39 60 | 1 3.23 51.61 99 | 1 3.23 54.84 missing | 14 45.16 100.00 ------------+----------------------------------- Total | 31 100.00 Did client wait 1 month post-cryotherapy before having sex? | Freq. Percent Cum. ------------+----------------------------------- Yes | 9 29.03 29.03 No | 14 45.16 74.19 missing | 8 25.81 100.00 ------------+----------------------------------- Total | 31 100.00 Were condoms used during sex? Expansion phase only. | Freq. Percent Cum. ------------+----------------------------------- No | 4 44.44 44.44 missing | 5 55.56 100.00 ------------+----------------------------------- Total | 9 100.00 Vaginal discharge post-cryotherapy? | Freq. Percent Cum. ------------+----------------------------------- No | 2 6.45 6.45 Yes | 29 93.55 100.00 ------------+----------------------------------- Total | 31 100.00 Severity of the post-cryo vaginal discharge. | Freq. Percent Cum. ------------+----------------------------------- mild | 8 25.81 25.81 moderate | 20 64.52 90.32 severe | 1 3.23 93.55 missing | 2 6.45 100.00 ------------+----------------------------------- Total | 31 100.00 Has the discharge stopped? | Freq. Percent Cum. ------------+----------------------------------- Yes | 28 90.32 90.32 missing | 3 9.68 100.00 ------------+----------------------------------- Total | 31 100.00 45 Number of days until discharge stopped | Freq. Percent Cum. ------------+----------------------------------- 0 | 1 3.23 3.23 2 | 1 3.23 6.45 3 | 1 3.23 9.68 4 | 1 3.23 12.90 5 | 1 3.23 16.13 10 | 2 6.45 22.58 14 | 10 32.26 54.84 18 | 1 3.23 58.06 21 | 6 19.35 77.42 30 | 1 3.23 80.65 999 | 6 19.35 100.00 ------------+----------------------------------- Total | 31 100.00 Type of post-cryotherapy vaginal discharge- 1month pilot phase & 1-3 month expansion phase only. | Freq. Percent Cum. --------------+----------------------------------- watery | 23 74.19 74.19 foul-smelling | 3 9.68 83.87 missing | 5 16.13 100.00 --------------+----------------------------------- Total | 31 100.00 Other treatment sought for post-cryotherapy problems thought to be associated with treatment? | Freq. Percent Cum. ------------+----------------------------------- No | 19 61.29 61.29 Yes | 8 25.81 87.10 missing | 4 12.90 100.00 ------------+----------------------------------- Total | 31 100.00 5.4.5. Follow-up 1 year after cryotherapy Of the 116 treated women, 34 presented at the final visit 1 year after treatment at the moment that the data file was prepared. We will focus at the VILI, colposcopy and biopsy results. Based on biopsy and colposcopy we constructed a final gold standard outcome as defined before at the initial diagnostic assessment (using the biopsy result if available and considering a negative colposcopy result as a truly negative result). For 28 women a VILI result was available and 26 colposcopy and/or biopsy results were provided. VILI at 1Yr | Freq. Percent Cum. ------------------------+----------------------------------- negative | 20 71.43 71.43 positive | 7 25.00 96.43 suspect for ca | 1 3.57 100.00 ------------------------+----------------------------------- Total | 28 100.00 46 | gold standard outcome VILI at 1Yr | negative CIN 1 CIN 2 CIN3 | Total -----------------+--------------------------------------------+---------- negative | 17 1 0 0 | 18 positive | 1 3 2 1 | 7 suspect for ca | 0 0 0 1 | 1 -----------------+--------------------------------------------+---------- Total | 18 4 2 2 | 26 There were still 4/26 cases with CIN2+ (2 CIN2 and 2 CIN3). Three of them were retreated. They were all detected by VILI. 18/26 did not show any sign of CIN and 4 cases only showed low grade CIN. The high frequency of missing values in the follow-up files and the low attendance at triage and post-treatment follow-up seriously compromises the quality of information on the safety, acceptability and effectiveness of cryotherapy in this study. Safety and acceptability of cryotherapy will be addressed in extenso in the next two studies, conducted in Thailand and Ghana. 47 5.5. Safety, acceptability, and feasibility of a single-visit approach to cervical-cancer prevention in rural Thailand Among the 5999 women tested by VIA in the Thai demonstration project, 5146 were negative, 51 had an indeterminate result, 798 had a positive result and 4 showed clinical signs compatible with suspicion of cervical cancer (see flowchart in Figure 5). Figure 5. Flow of participants through the different steps of the Thai demonstration project. 5999 women tested 51 4 5146 798 indeterminate suspect for negative VIA test positive VIA test VIA test cancer (85.8%) (13.3%) (0.9%) (0.067%) 33 60 4 no cryo-therapy no cryo-therapy no cryo-therapy 18 738 received cryo received cryo (35.3%) (92.5%) 8 10 607 131 immediate cryo postponed cryo immediate cryo postponed cryo (44.4%) (55.6%) (82.3%) (17,7%) 629 folow-up at 3-4M (83.2% of 756) 707 folow-up at 1Y (93.5% of 756) The VIA test positivity rate did not differ significantly between fixed services (district hospitals) or mobile services set up at the health centers (p for χ2=0.13), but differed significantly by district (p for χ2=0.00). Nevertheless, the variation by district was not so large (between 9.4% in Pathumrat and 14.3% in Kasetwisai). 48 Table 16. Frequency distribution of the different VIA categories for all districts combined and for the districts Katsetwisai, Pathumrat, Chaturapakpiman and Phanomphrai. All districts 95% CI VIA N % lower upper Negative 5146 85.8% 84.9% 86.7% Indeterminate 51 0.9% 0.6% 1.1% Positive 798 13.3% 12.4% 14.2% Suspect Ca 4 0.1% 0.0% 0.1% Total 5999 100% Kasetwisai 95% CI Pathumrat 95% CI VIA N % lower upper N % lower Upper Negative 1051 85.7% 83.7% 87.6% 1130 85.6% 83.7% 87.5% Indeterminate 1 0.1% 0.0% 0.2% 43 3.3% 0.0% 4.2% Positive 175 14.3% 12.3% 16.2% 147 11.1% 9.4% 12.8% Suspect Ca 0 0.0% 0.0% 0.0% 0 0.0% 0.0% 0.0% Total 1227 100.0% 1320 100.0% Chaturapakpiman 95% CI Phanomphrai 95% CI VIA N % lower upper N % lower Upper Negative 1675 85.8% 84.2% 87.3% 1290 85.8% 84.0% 87.6% Indeterminate 6 0.9% 0.4% 1.3% 1 0.9% 0.4% 1.3% Positive 307 13.3% 11.8% 14.8% 169 13.3% 11.6% 15.0% Suspect Ca 2 0.0% 0.0% 0.1% 2 0.0% 0.0% 0.1% Total 1990 100% 1462 100% The VIA positivity rate did not differ significantly by age group. The rate differed by VIA provider, but when controlling for district it differed by provider only in the district of Chaturapakpiman. The inter-provider range was 9.4%-17.8%. In total 756 women received cryotherapy (18/51 = 35.3% of women with an intermediate result; 738/798=92.5% of women being VIA+). Of 756 treated women, 615 (81.3%) received cryotherapy the same day. Of 756 treated women, 629 (83.2%) presented for the planned follow-up at 3 months after treatment and 707 (93.5%) of treated women presented for the final follow-up visit at one year. The main experiences of tested and treated women are presented in Table 17 below. The data in the tables clearly show that the level of acceptance among tested and treated women is high, the number of secondary effects is low, and if present they are mild and of short duration. The rate of residual disease 1 year after treatment is low as well. 49 Table 17. Selection of answers to questionnaires, observations of health providers and experiences of women after testing, cryotherapy and post-therapy follow-up Question (Thailand). Definition of denominator Denominator Percent Experience about testing Satisfaction about being tested Tested woman 5742 98.5% Experience of being tested was better than expected Tested woman 5729 89.6% Declared having received enough information Tested woman 5709 99.8% Compliance with treatment decision Management decision to treat immediately Women being VIA+ 780 77.7% Management decision to treat immediately Women being VIA indeterminate 22 36.4% Management decision to treat immediately Women being VIA+ or indeterminate 802 76.6% Received cryotherapy Women VIA+,indet,ca 802 92.0% Received cryotherapy Women being VIA+ 798 92.5% Received cryotherapy Women being VIA indeterminate 51 35.3% Received cryotherapy Women being VIA+ or indeterminate 853 88.6% Received cryotherapy on same day Women being VIA+ and treated 738 82.2% Received cryotherapy on same day Women being VIA indeterminate and treated 18 44.4% Received cryotherapy on same day Women being VIA+ or indeterminate and treated 756 81.3% Reason to postpone: menstruation Management decision to postpone treatment 125 80.8% Reason to postpone: lacking material Management decision to postpone treatment 125 8.0% Treatment of postponed women Management decision to postpone treatment 126 91.3% Treatment of referred women Management decision to refer 65 16.9% Referred women All 5999 1.2% Reason of reference: suspicion of cancer Referred women 74 5.4% Reason of reference: lesion occupying >75% of TZ Referred women 74 31.1% Reason of reference: lesion extending to vagina Referred women 74 28.4% Reason of reference: large fibroma Referred women 74 8.1% Reason of reference: adnexal mass Referred women 74 2.7% Reason of reference: other Referred women 74 23.0% Reason of reference: bleeding Referred women 74 0.0% 50 Question (Thailand). Definition of denominator Denominator Percent Experience at the moment of treatment Pain reported by provider Treated women 756 39.6% Pain was mild Treated women reporting pain 290 79.0% Pain was moderate Treated women reporting pain 290 18.3% Pain was severe Treated women reporting pain 290 2.8% An analgesic drug was given Treated women reporting pain 746 6.2% Pain reported by women immediately after cryo Treated women 753 40.8% Bleeding after cryotherapy Treated women 748 1.7% Bleeding was mild Treated women reporting bleeding 11 100.0% Bleeding could be treated by pressure Treated women reporting bleeding 10 100.0% Satisfaction about being treated Treated women 752 99.1% Experience of treatment was better than expected Treated women 752 93.0% Declared having received enough information about treatment Treated women 744 99.5% Woman will recommend VIA testing to others Women with negative VIA result 5122 99.8% Woman will recommend VIA testing to others Women with positive VIA result 751 97.5% Problem visit of a health structure Women contacted health service after cryotherapy Treated women 756 4.4% The visit: related to the cryotherapy Women having contacted a health service 33 21.2% Reason of problem visit: bleeding related to menses Women having contacted a health service 33 9.1% Reason of problem visit: bleeding unrelated to menses Women having contacted a health service 33 12.1% Reason of problem visit: pain/cramping unrelated to menses Women having contacted a health service 33 27.3% Reason of problem visit was: abnormal vaginal discharge Women having contacted a health service 34 44.1% Reason of problem visit: chills/hot sweats/fever Women having contacted a health service 35 2.9% Reason of problem visit: another problem Women having contacted a health service 36 2.8% Planned follow-up visit at 3 months Attended at the visit Treated women 756 83.2% Had bleeding not associated with menses Treated women presenting at 3M follow-up visit 627 2.9% 51 Question (Thailand). Definition of denominator Denominator Percent Observed presence of blood clots Treated women presenting at 3M follow-up visit 627 0.5% Had pain or cramping not associated with menses Treated women presenting at 3M follow-up visit 626 13.6% Observed changes in menstruation Treated women presenting at 3M follow-up visit 552 10.0% Less menstrual blood loss since cryotherapy Treated women presenting at 3M follow-up visit 552 5.6% More menstrual blood loss since cryotherapy Treated women presenting at 3M follow-up visit 552 4.3% Changed duration of menstruation Treated women presenting at 3M follow-up visit 552 7.4% Shorter menses Treated women presenting at 3M follow-up visit 552 4.3% Longer menses Treated women presenting at 3M follow-up visit 552 3.1% Change in menstrual cramping Treated women presenting at 3M follow-up visit 552 11.4% More cramping Treated women presenting at 3M follow-up visit 552 10.0% Less cramping Treated women presenting at 3M follow-up visit 552 1.4% Abnormal vaginal discharge Treated women presenting at 3M follow-up visit 628 0.5% Woman will recommend VIA testing to others Treated women presenting at 3M follow-up visit 627 97.9% Compliance with recommendation concerning sexual hygiene Woman had sex after treatment Treated women presenting at 3M follow-up visit 628 72.3% Woman had sex only after 4W post-treatment Treated women presenting at 3M follow-up visit 459 69.1% Woman had sex within 4W post-treatment Treated women presenting at 3M follow-up visit 459 30.9% Condom use if sex within 4W post-treatment Having had sex <4W post-treatment 142 89.4% Condom use all the time if sex within 4W post-treatment Having had sex <4W post-treatment 142 79.6% No condom use if sex within 4W post-treatment Having had sex <4W post-treatment 142 10.6% Declared that abstinence from sex should not be a problem Treated women 756 92.2% Had problem convincing partner to abstain Treated women presenting at 3M follow-up visit 628 7.2% Had problem convincing partner to use condoms Treated women presenting at 3M follow-up visit 454 2.9% Husband was satisfied about decision for treatment Treated women presenting at 3M follow-up visit 628 94.9% 52 Question (Thailand). Definition of denominator Denominator Percent Planned follow-up at 12 months after treatment Attended at follow-up Treated women 756 93.5% Suspect for cancer at VIA examination Treated women, attending at 12 month visit 704 0.1% VIA positive Treated women, attending at 12 month visit 704 5.7% Indeterminate VIA result Treated women, attending at 12 month visit 704 0.3% 53 5.6. Safety, acceptability, and feasibility of a single-visit approach to cervical cancer prevention in Ghana Of the 3665 women who participated in the Ghanaian demonstration project,3177 had a negative VIA test result, 484 were positive and 4 were suspect for cancer (see flowchart in Figure 6). Figure 6. Flow of participants through the different steps of the Ghanaian demonstration project. 3665 women tested 4 3177 484 suspect for negative VIA test positive VIA test cancer (86.7%) (13.2%) (0.11%) 47 2 no cryo-therapy no cryo-therapy 437 2 received cryo received cryo (90.3%) (50%%) 308 129 0 2 immediate cryo postponed cryo received cryo postponed cryo (70.5%) (29.5%) (0%) (100%) 339 folow-up at 3-4M (77.2% of 439) 241 folow-up at 1Y (54.9% of 439) 54 The VIA test positivity rate varied significantly by age group: it was highest in the youngest age group (16.0% in the 25-29 age group) and decreased until 10.6% in the age group 35-39 (see Table 18). After the age of 40 the positivity rate rose about 2%. Table 18. Frequency distribution of the different VIA categories by 5-year age group. Age group Negative VIA+ Suspect Ca Total 25-29 852 163 1 1016 83.9% 16.0% 0.1% 100.0% 30-34 820 132 1 953 86.0% 13.9% 0.1% 100.0% 35-39 830 98 0 928 89.4% 10.6% 0.0% 100.0% 40-49 563 71 1 635 88.7% 11.2% 0.2% 100.0% 45-49 107 16 1 124 86.3% 12.9% 0.8% 100.0% Total 3175 481 4 3660 Pearson chi2(8) = 22.2016 Pr = 0.005 The test positivity rate also varied significantly by provider (range: 10.1%-17%). In total, 437 of the 484 (90.3%) VIA-positive women received cryotherapy, among whom 308 were treated the same day (70.5%). Also 2 cases, suspect for cancer at VIA but confirmed free of cancer at the reference hospital, received cryotherapy. Of 439 treated women, 339 (77.2%) presented for the planned follow-up at 3 months after treatment and 241 (54.9%) of treated women were present at the final follow-up visit at one year. The main experiences of tested and treated women are presented in Table 17 below. No major complications were noted. The reported side effects were rare or mild. Twenty-three percent mentioned pain after treatment, which was mild in 86% of the cases reporting pain. In two percent of women mild blood loss was observed, which was easily treated with simple pressure. The reported side effects were rare. Only 1% of treated women contacted the health service after treatment and more than 98% of treated women declared themselves to be satisfied with the received procedure. As in the Thai study, we can conclude that cryotherapy offered to VIA-positive women is an acceptable and safe strategy. In the Amasaman Health Center, 3206 women participated and were tested with VIA. The test positivity rate was 5.9%. The Amasaman data will be analyzed when a completely documented data file will be obtained. 55 Table 19. Selection of answers to questionnaires, observations of health providers and experiences of women after testing, cryotherapy and post-therapy follow-up Question (Ghana, Accra) Definition of denominator Denominator Percent Experience about testing Satisfaction about being tested Tested woman 3621 79.8% Experience of being tested was better than expected Tested woman 3598 83.6% Declared having received enough information Tested woman 3582 99.7% Compliance with treatment decision Management decision to treat immediately Women being VIA+ 480 74.4% Management decision to treat immediately Women being VIA+ or indeterminate 480 74.4% Received cryotherapy Women VIA+,indet,ca 488 90.0% Received cryotherapy Women being VIA+ 484 90.3% Received cryotherapy Women being VIA+ or indeterminate 488 90.0% Received cryotherapy on same day Women being VIA+ and treated 437 70.5% Received cryotherapy on same day Women being VIA+ or indeterminate and treated 437 70.5% Reason to postpone: menstruation Management decision to postpone treatment 141 1.4% Reason to postpone: lacking material Management decision to postpone treatment 141 5.7% Treatment of postponed women Management decision to postpone treatment 113 69.9% Treatment of referred women Management decision to refer 13 61.5% Reason of reference: suspicion of cancer Referred women 25 16.0% Reason of reference: lesion occupying >75% of TZ Referred women 25 0.0% Reason of reference: lesion extending to vagina Referred women 25 0.0% Reason of reference is fibroma (size ~>20W pregnant uterus) Referred women 25 28.0% Reason of reference: adnexal mass Referred women 25 0.0% Reason of reference: other Referred women 25 48.0% Reason of reference: bleeding Referred women 25 20.0% Experience at the moment of treatment Pain reported by provider Treated women 412 22.6% Pain was mild Treated women reporting pain 90 85.6% Pain was moderate Treated women reporting pain 90 12.2% Pain was severe Treated women reporting pain 90 2.2% 56 Question (Ghana, Accra) Definition of denominator Denominator Percent An analgesic drug was given Treated women reporting pain 422 1.7% Pain reported by women immediately after cryotherapy Treated women 413 34.9% Bleeding after cryotherapy Treated women 400 2.0% Bleeding was mild Treated women reporting bleeding 5 100.0% Bleeding could be treated by pressure Treated women reporting bleeding 8 12.5% Satisfaction about being treated Treated women 421 98.6% Experience of treatment was better than expected Treated women 428 82.5% Declared having received enough information about treatment Treated women 412 98.8% Woman will recommend VIA testing to others Women with negative VIA result 3115 99.2% Woman will recommend VIA testing to others Women with positive VIA result 417 99.8% Problem visit of a health structurea Women contacted health service after cryotherapy Treated women 439 1.1% The visit: related to the cryotherapy Women having contacted a health service - - Reason of problem visit: bleeding related to menses Women having contacted a health service - - Reason of problem visit: bleeding unrelated to menses Women having contacted a health service - - Reason of problem visit: pain/cramping unrelated to menses Women having contacted a health service - - Reason of problem visit was: abnorm vaginal discharge Women having contacted a health service - - Reason of problem visit: chills/hot sweats/fever Women having contacted a health service - - Reason of problem visit: another problem Women having contacted a health service - - Planned follow-up visit at 3 months Attended at the visit Treated women 439 77.2% Had bleeding not associated with menses Treated women presenting at 3M follow-up visit 324 6.2% Observed presence of blood clots Treated women presenting at 3M follow-up visit 337 2.4% Had pain or cramping not associated with menses Treated women presenting at 3M follow-up visit 316 10.1% Observed changes in menstruation Treated women presenting at 3M follow-up visit 337 20.5% Less menstrual blood loss since cryotherapy Treated women presenting at 3M follow-up visit 337 5.9% More menstrual blood loss since cryotherapy Treated women presenting at 3M follow-up visit 337 14.5% a The problem visit variables in the Accra file could not be analysed because of lack of codebook and data labels. 57 Question (Ghana, Accra) Definition of denominator Denominator Percent Changed duration of menstruation Treated women presenting at 3M follow-up visit 337 16.0% Shorter menses Treated women presenting at 3M follow-up visit 337 5.0% Longer menses Treated women presenting at 3M follow-up visit 337 11.0% Change in menstrual cramping Treated women presenting at 3M follow-up visit 337 17.5% More cramping Treated women presenting at 3M follow-up visit 337 8.0% Less cramping Treated women presenting at 3M follow-up visit 337 9.5% Abnormal vaginal discharge Treated women presenting at 3M follow-up visit 310 5.5% Woman will recommend VIA testing to others Treated women presenting at 3M follow-up visit 335 98.5% Compliance with recommendation concerning sexual hygiene Woman had sex after treatment Treated women presenting at 3M follow-up visit 337 68.0% Woman had sex only after 4W post-treatment Treated women presenting at 3M follow-up visit 229 85.6% Woman had sex within 4W post-treatment Treated women presenting at 3M follow-up visit 229 14.4% Condom use if sex within 4W post-treatment Having had sex <4W post-treatment 32 96.9% Condom use all the time if sex within 4W post-treatment Having had sex <4W post-treatment 30 86.7% No condom use if sex within 4W post-treatment Having had sex <4W post-treatment 32 3.1% Declared that abstinence from sex should not be a problem Treated women 422 85.1% Had problem convincing partner to abstinate Treated women presenting at 3M follow-up visit 337 7.7% Had problem convincing partner to use condoms Treated women presenting at 3M follow-up visit 337 8.3% Husband was satisfied about decision for treatment Treated women presenting at 3M follow-up visit 337 78.9% Planned follow-up at 12 months after treatment Attended at follow-up Treated women 439 54.9% Suspect for cancer at VIA examination Treated women, attending at 12 month visit 232 0.0% VIA positive Treated women, attending at 12 month visit 232 2.6% Indeterminate VIA result Treated women, attending at 12 month visit 232 0.0% VIA technically impossible Treated women, attending at 12 month visit 232 2.2% 58 6. Pooled analyses In the next subchapter we will pool the available test characteristics of the five screening methods applied in the respective ACCP studies. Test positivity rates of VIA are documented in all studies. The positive predictive value of VIA can be evaluated from four studies, where screen positive women are submitted to gold standard verification: the cross sectional multi-center study, the Zimbabwe cross-sectional study, the Ambilikai RCT and the Kenyan WKCCPP study. In these four studies also the detection rate of disease and the approximated specificity can be derived. Finally, the sensitivity, specificity and negative predictive value can only be assessed in the multi-center cross-sectional study and in the Zimbabwean study, where all screened women are verified with a gold-standard. In the next five subchapters we describe subsequently the test characteristics of VIA, VILI, VIAM, high-risk HPV type detection by Hybrid Capture II and the conventional Pap smear, considered at 3 cytological cutoffs (ASCUS+, LSIL+ and HSIL+) for each outcome (CIN1+, CIN2+, CIN3+ and cancer). We present for each parameter of a given test the forest plot and the meta-analytically pooled measure. Subsequently, we pool the ratio of the sensitivity and specificity from the multi-center and Zimbabwe cross-sectional studies. 6.1. Meta-analysis of test characteristics of applied screening methods 6.1.1. Test characteristics of VIA The accuracy of VIA is assessed for the outcomes CIN1+, CIN2+, CIN3+ and cancer (see Table 20). Table 20. Test characteristics of VIA assessed in 14 locations. A: Outcome=CIN1+ Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Bamako Mali 253 278 371 4,650 624 5,552 0.476 0.926 0.405 0.944 0.112 0.096 0.046 Brazzaville Congo 452 179 1,393 4,911 1,845 6,935 0.716 0.779 0.245 0.965 0.266 0.091 0.065 Conakry Guinea 205 252 475 7,695 680 8,627 0.449 0.942 0.301 0.968 0.079 0.053 0.024 Jaipur India 207 23 1,275 4,281 1,482 5,786 0.900 0.771 0.140 0.995 0.256 0.040 0.036 Kolkata 1 India 316 194 806 4,578 1,122 5,894 0.620 0.850 0.282 0.959 0.190 0.087 0.054 Kolkata 2 India 135 183 775 6,987 910 8,080 0.425 0.900 0.148 0.974 0.113 0.039 0.017 Mumbai India 129 77 388 3,410 517 4,004 0.626 0.898 0.250 0.978 0.129 0.051 0.032 Niamey India 31 20 121 2,362 152 2,534 0.608 0.951 0.204 0.992 0.060 0.020 0.012 Ouagadougou B-Fasso 73 11 488 1,479 561 2,051 0.869 0.752 0.130 0.993 0.274 0.041 0.036 Trivandrum 1 India 397 316 680 3,064 1,077 4,457 0.557 0.818 0.369 0.907 0.242 0.160 0.089 Trivandrum 2 India 231 187 348 3,993 579 4,759 0.553 0.920 0.399 0.955 0.122 0.088 0.049 Harare Zimbabwe 347 217 502 1,064 849 2,130 0.615 0.679 0.409 0.831 0.399 0.265 0.163 Total 1a 2,776 1,937 7,622 48,474 10,398 60,809 0.589 0.864 0.267 0.962 0.171 0.078 0.046 Ambilikai India 1,871 1,110 2,981 31,282 0.628 0.095 0.060 Busia Kenya 163 281 444 2,295 0.367 0.193 0.071 Total 2b 4,810 9,013 13,823 94,386 0.348 0.146 0.051 a Total 1: total of all studies with complete verification using colposcopy and biopsy from colposcopically suspect areas. b Total 2: total of all studies with complete and incomplete verification; where incomplete verification involved only test positive cases. Total 2 includes Total 1 & Ambilikai RCT and the Busia study. 59 B: Outcome=CIN2+ Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Bamako Mali 130 34 494 4,894 624 5,552 0.793 0.908 0.208 0.993 0.112 0.030 0.023 Brazzaville Congo 313 76 1,532 5,014 1,845 6,935 0.805 0.766 0.170 0.985 0.266 0.056 0.045 Conakry Guinea 153 15 527 7,932 680 8,627 0.911 0.938 0.225 0.998 0.079 0.019 0.018 Jaipur India 56 8 1,426 4,296 1,482 5,786 0.875 0.751 0.038 0.998 0.256 0.011 0.010 Kolkata 1 India 99 62 1,023 4,710 1,122 5,894 0.615 0.822 0.088 0.987 0.190 0.027 0.017 Kolkata 2 India 52 19 858 7,151 910 8,080 0.732 0.893 0.057 0.997 0.113 0.009 0.006 Mumbai India 50 31 467 3,456 517 4,004 0.617 0.881 0.097 0.991 0.129 0.020 0.012 Niamey India 13 7 139 2,375 152 2,534 0.650 0.945 0.086 0.997 0.060 0.008 0.005 Ouagadougou B-Fasso 45 5 516 1,485 561 2,051 0.900 0.742 0.080 0.997 0.274 0.024 0.022 Trivandrum 1 India 132 17 945 3,363 1,077 4,457 0.886 0.781 0.123 0.995 0.242 0.033 0.030 Trivandrum 2 India 65 16 514 4,164 579 4,759 0.802 0.890 0.112 0.996 0.122 0.017 0.014 Harare Zimbabwe 158 48 691 1,233 849 2,130 0.767 0.641 0.186 0.963 0.399 0.097 0.074 Total 1 1,266 338 9,132 50,073 10,398 60,809 0.789 0.846 0.122 0.993 0.171 0.026 0.021 Ambilikai India 272 2,709 2,981 31,282 0.091 0.095 0.009 Busia Kenya 47 397 444 2,295 0.106 0.193 0.020 Total 2 1,585 12,238 13,823 94,386 0.115 0.146 0.017 C: Outcome=CIN3+ Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Bamako Mali 75 8 549 4,920 624 5,552 0.904 0.900 0.120 0.998 0.112 0.015 0.014 Brazzaville Congo 176 24 1,669 5,066 1,845 6,935 0.880 0.752 0.095 0.995 0.266 0.029 0.025 Conakry Guinea 113 10 567 7,937 680 8,627 0.919 0.933 0.166 0.999 0.079 0.014 0.013 Jaipur India 40 6 1,442 4,298 1,482 5,786 0.870 0.749 0.027 0.999 0.256 0.008 0.007 Kolkata 1 India 64 35 1,058 4,737 1,122 5,894 0.646 0.817 0.057 0.993 0.190 0.017 0.011 Kolkata 2 India 29 10 881 7,160 910 8,080 0.744 0.890 0.032 0.999 0.113 0.005 0.004 Mumbai India 37 22 480 3,465 517 4,004 0.627 0.878 0.072 0.994 0.129 0.015 0.009 Niamey India 7 5 145 2,377 152 2,534 0.583 0.943 0.046 0.998 0.060 0.005 0.003 Ouagadougou B-Fasso 33 4 528 1,486 561 2,051 0.892 0.738 0.059 0.997 0.274 0.018 0.016 Trivandrum 1 India 53 3 1,024 3,377 1,077 4,457 0.946 0.767 0.049 0.999 0.242 0.013 0.012 Trivandrum 2 India 51 10 528 4,170 579 4,759 0.836 0.888 0.088 0.998 0.122 0.013 0.011 Total 1 678 137 8,871 48,993 9,549 58,679 0.832 0.847 0.071 0.997 0.163 0.014 0.012 Ambilikai India 131 2,850 2,981 31,282 0.044 0.095 0.004 Busia Kenya 24 420 444 2,295 0.054 0.193 0.010 Total 2 833 12,141 12,974 92,256 0.064 0.141 0.009 D: Outcome=Cancer Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Ouagadougou B-Fasso 14 3 547 1,487 561 2,051 0.824 0.731 0.025 0.998 0.274 0.008 0.007 Kolkata 1 India 30 8 1,092 4,764 1,122 5,894 0.789 0.814 0.027 0.998 0.190 0.006 0.005 Kolkata 2 India 8 0 902 7,170 910 8,080 1.000 0.888 0.009 1.000 0.113 0.001 0.001 Brazzaville Congo 22 4 1,823 5,086 1,845 6,935 0.846 0.736 0.012 0.999 0.266 0.004 0.003 Conakry Guinea 84 5 596 7,942 680 8,627 0.944 0.930 0.124 0.999 0.079 0.010 0.010 Jaipur India 4 2 1,478 4,302 1,482 5,786 0.667 0.744 0.003 1.000 0.256 0.001 0.001 Bamako Mali 52 2 572 4,926 624 5,552 0.963 0.896 0.083 1.000 0.112 0.010 0.009 Mumbai India 20 7 497 3,480 517 4,004 0.741 0.875 0.039 0.998 0.129 0.007 0.005 Niamey India 3 2 149 2,380 152 2,534 0.600 0.941 0.020 0.999 0.060 0.002 0.001 Trivandrum 1 India 6 0 1,071 3,380 1,077 4,457 1.000 0.759 0.006 1.000 0.242 0.001 0.001 Trivandrum 2 India 13 2 566 4,178 579 4,759 0.867 0.881 0.022 1.000 0.122 0.003 0.003 Harare Zimbabwe 3 0 846 1,281 849 2,130 1.000 0.602 0.004 1.000 0.399 0.001 0.001 Total 1 259 35 10,139 50,376 10,398 60,809 0.881 0.832 0.025 0.999 0.171 0.005 0.004 Ambilikai India 66 2,915 2,981 31,282 0.022 0.095 0.002 Busia Kenya 9 435 444 2,295 0.020 0.193 0.004 Total 2 334 13,489 13,823 94,386 0.024 0.146 0.004 60 E: Test positivity rate Setting Country T+ N T+ rate Bamako Mali 624 5,552 0.112 Brazzaville Congo 1,845 6,935 0.266 Conakry Guinea 680 8,627 0.079 Jaipur India 1,482 5,786 0.256 Kolkata 1 India 1,122 5,894 0.190 Kolkata 2 India 910 8,080 0.113 Mumbai India 517 4,004 0.129 Niamey India 152 2,534 0.060 Ouagadougou B-Fasso 561 2,051 0.274 Trivandrum 1 India 1,077 4,457 0.242 Trivandrum 2 India 579 4,759 0.122 Harare Zimbabwe 849 2,130 0.399 Ambilikai India 2,981 31,282 0.095 Busia Kenya 444 2,295 0.193 Chaturapakpiman Thailand 309 1,990 0.155 Kasetwisai Thailand 175 1,227 0.143 Pathumrat Thailand 147 1,320 0.111 Phanomphrai Thailand 171 1,462 0.117 Accra Ghana 488 3,665 0.133 Amasaman Ghana 194 3,206 0.061 Total 15,307 107,256 0.143 VIA; all studies Bamako Brazzaville Conakry Jaipur Kolkata 1 Kolkata 2 Mumbai Niamey Ouagadougou Trivandrum 1 Trivandrum 2 Harare Ambilikai Busia Chaturapakpiman Kasetwisai Pathumrat Phanomphrai Accra Amasaman Combined 0 .1 .2 .3 .4 .5 Test positivity rate Meta-analysis of the test-positivity rate of VIA (including all studies) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.121 0.119 0.122 123.289 0.000 20 Random | 0.162 0.134 0.191 11.164 0.000 Test for heterogeneity: Q= 3708.324 on 19 degrees of freedom (p= 0.000) 61 VIA, Outcome=CIN1+ VIA, Outcome=CIN1+ Bamako Bamako Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Niamey Niamey Ouagadougou Ouagadougou Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity VIA, Outcome=CIN1+ VIA, Outcome=CIN1+ Bamako Bamako Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Kolkata 1 Kolkata 2 Kolkata 1 Mumbai Kolkata 2 Niamey Mumbai Ouagadougou Niamey Trivandrum 1 Ouagadougou Trivandrum 2 Trivandrum 1 Harare Ambilikai Trivandrum 2 Busia Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of VIA+ (outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.621 0.608 0.634 91.859 0.000 12 Random | 0.618 0.532 0.703 14.170 0.000 Test for heterogeneity: Q= 441.564 on 11 degrees of freedom (p= 0.000) Meta-analysis of specificity of VIA+ (outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.894 0.891 0.896 701.545 0.000 12 Random | 0.849 0.812 0.887 44.271 0.000 Test for heterogeneity: Q= 2368.152 on 11 degrees of freedom (p= 0.000) Meta-analysis of the PPV of VIA+ (outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.310 0.303 0.317 84.731 0.000 14 Random | 0.306 0.211 0.400 6.346 0.000 Test for heterogeneity: Q= 2162.335 on 13 degrees of freedom (p= 0.000) Meta-analysis of the NPV of VIA+ (outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.979 0.978 0.980 1554.227 0.000 12 Random | 0.957 0.945 0.969 152.887 0.000 Test for heterogeneity: Q= 975.985 on 11 degrees of freedom (p= 0.000) 62 VIA, Outcome=CIN1+ VIA, Outcome=CIN1+ Ouagadougou Ouagadougou Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Bamako Bamako Mumbai Mumbai Niamey Niamey Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Ambilikai Harare Busia Combined Combined 0 .1 .2 .3 .4 .5 0 .1 .2 .3 .4 .5 Prevalence of disease Detection rate of disease VIA, Outcome=CIN1+ Ouagadougou Kolkata 1 Kolkata 2 Brazzaville Conakry Jaipur Bamako Mumbai Niamey Trivandrum 1 Trivandrum 2 Harare Ambilikai Busia Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN1+) in VIA studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.059 0.057 0.061 62.319 0.000 12 Random | 0.085 0.063 0.107 7.721 0.000 Test for heterogeneity: Q= 1440.016 on 11 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN1+) using VIA | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.039 0.038 0.040 62.587 0.000 14 Random | 0.053 0.040 0.066 8.277 0.000 Test for heterogeneity: Q= 1260.909 on 13 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of VIA (outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.932 0.931 0.934 1135.602 0.000 14 Random | 0.865 0.832 0.897 51.610 0.000 Test for heterogeneity: Q= 3926.540 on 13 degrees of freedom (p= 0.000) 63 VIA, Outcome=CIN2+ VIA, Outcome=CIN2+ Bamako Bamako Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Niamey Niamey Ouagadougou Ouagadougou Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity VIA, Outcome=CIN2+ VIA, Outcome=CIN2+ Bamako Bamako Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Kolkata 1 Kolkata 2 Kolkata 1 Mumbai Kolkata 2 Niamey Mumbai Ouagadougou Niamey Trivandrum 1 Ouagadougou Trivandrum 2 Trivandrum 1 Harare Ambilikai Trivandrum 2 Busia Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .8 .85 .9 .95 1 PPV NPV Meta-analysis of the sensitivity of VIA+ (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.817 0.799 0.836 86.848 0.000 12 Random | 0.790 0.736 0.843 28.927 0.000 Test for heterogeneity: Q= 81.387 on 11 degrees of freedom (p= 0.000) Meta-analysis of specificity of VIA+ (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.877 0.875 0.880 666.421 0.000 12 Random | 0.830 0.788 0.872 38.513 0.000 Test for heterogeneity: Q= 2794.893 on 11 degrees of freedom (p= 0.000) Meta-analysis of the PPV of VIA+ (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.092 0.087 0.097 38.023 0.000 14 Random | 0.118 0.091 0.146 8.347 0.000 Test for heterogeneity: Q= 403.079 on 13 degrees of freedom (p= 0.000) Meta-analysis of the NPV of VIA+ (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.996 0.996 0.997 3748.138 0.000 12 Random | 0.993 0.991 0.996 889.063 0.000 Test for heterogeneity: Q= 160.584 on 11 degrees of freedom (p= 0.000) 64 VIA, Outcome=CIN2+ VIA, Outcome=CIN2+ Ouagadougou Ouagadougou Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Bamako Bamako Mumbai Mumbai Niamey Niamey Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Ambilikai Harare Busia Combined Combined 0 .05 .1 .15 .2 0 .05 .1 .15 .2 Prevalence of disease Detection rate of disease VIA, Outcome=CIN2+ Ouagadougou Kolkata 1 Kolkata 2 Brazzaville Conakry Jaipur Bamako Mumbai Niamey Trivandrum 1 Trivandrum 2 Harare Ambilikai Busia Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN2+) in VIA studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.059 0.057 0.061 62.319 0.000 12 Random | 0.085 0.063 0.107 7.721 0.000 Test for heterogeneity: Q= 1440.016 on 11 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN2+) using VIA | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.012 0.011 0.012 33.537 0.000 14 Random | 0.021 0.016 0.025 8.351 0.000 Test for heterogeneity: Q= 522.856 on 13 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of VIA (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.891 0.889 0.893 884.553 0.000 14 Random | 0.837 0.804 0.870 49.710 0.000 Test for heterogeneity: Q= 3114.155 on 13 degrees of freedom (p= 0.000) 65 VIA, Outcome=CIN3+ VIA, Outcome=CIN3+ Bamako Bamako Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Niamey Niamey Ouagadougou Ouagadougou Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 .6 .7 .8 .9 1 Sensitivity Specificity VIA, Outcome=CIN3+ VIA, Outcome=CIN3+ Bamako Bamako Brazzaville Brazzaville Conakry Conakry Jaipur Kolkata 1 Jaipur Kolkata 2 Kolkata 1 Mumbai Kolkata 2 Niamey Mumbai Ouagadougou Niamey Trivandrum 1 Ouagadougou Trivandrum 2 Trivandrum 1 Ambilikai Busia Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .8 .85 .9 .95 1 PPV NPV Meta-analysis of the sensitivity of VIA+ (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.871 0.849 0.894 76.818 0.000 11 Random | 0.829 0.771 0.887 28.126 0.000 Test for heterogeneity: Q= 56.160 on 10 degrees of freedom (p= 0.000) Meta-analysis of the specificity of VIA+ (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.875 0.872 0.878 649.327 0.000 11 Random | 0.842 0.800 0.883 39.554 0.000 Test for heterogeneity: Q= 2382.681 on 10 degrees of freedom (p= 0.000) Meta-analysis of the PPV of VIA+ (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.051 0.048 0.055 26.703 0.000 13 Random | 0.068 0.052 0.085 8.228 0.000 Test for heterogeneity: Q= 195.937 on 12 degrees of freedom (p= 0.000) Meta-analysis of the NPV of VIA+ (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.998 0.998 0.999 5201.754 0.000 11 Random | 0.997 0.997 0.998 2178.608 0.000 Test for heterogeneity: Q= 48.227 on 10 degrees of freedom (p= 0.000) 66 VIA, Outcome=CIN3+ VIA, Outcome=CIN3+ Ouagadougou Ouagadougou Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Bamako Bamako Mumbai Mumbai Niamey Niamey Trivandrum 1 Trivandrum 1 Trivandrum 2 Ambilikai Trivandrum 2 Busia Combined Combined 0 .05 .1 .15 .2 0 .05 .1 .15 .2 Prevalence of disease Detection rate of disease VIA, Outcome=CIN3+ Ouagadougou Kolkata 1 Kolkata 2 Brazzaville Conakry Jaipur Bamako Mumbai Niamey Trivandrum 1 Trivandrum 2 Ambilikai Busia Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN3+) in VIA studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.010 0.010 0.011 24.979 0.000 11 Random | 0.014 0.010 0.017 6.819 0.000 Test for heterogeneity: Q= 205.416 on 10 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN3+) using VIA | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.006 0.006 0.007 23.963 0.000 13 Random | 0.010 0.008 0.013 7.388 0.000 Test for heterogeneity: Q= 267.316 on 12 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of VIA (CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.887 0.885 0.890 861.713 0.000 13 Random | 0.845 0.813 0.878 50.692 0.000 Test for heterogeneity: Q= 2705.057 on 12 degrees of freedom (p= 0.000) 67 VIA, Outcome=Cancer VIA, Outcome=Cancer Bamako Bamako Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Niamey Niamey Ouagadougou Ouagadougou Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 .6 .7 .8 .9 1 Sensitivity Specificity VIA, Outcome=Cancer VIA, Outcome=Cancer Bamako Bamako Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Kolkata 1 Kolkata 2 Kolkata 1 Mumbai Kolkata 2 Niamey Mumbai Ouagadougou Niamey Trivandrum 1 Ouagadougou Trivandrum 2 Trivandrum 1 Harare Ambilikai Trivandrum 2 Busia Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .8 .85 .9 .95 1 PPV NPV Meta-analysis of the sensitivity of VIA+ (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.925 0.895 0.955 60.452 0.000 12 Random | 0.891 0.837 0.945 32.455 0.000 Test for heterogeneity: Q= 19.855 on 11 degrees of freedom (p= 0.047) Meta-analysis of the specificity of VIA+ (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.866 0.863 0.868 640.193 0.000 12 Random | 0.817 0.771 0.863 34.889 0.000 Test for heterogeneity: Q= 3143.746 on 11 degrees of freedom (p= 0.000) Meta-analysis of the PPV of VIA+ (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.009 0.008 0.011 11.744 0.000 14 Random | 0.025 0.017 0.032 6.429 0.000 Test for heterogeneity: Q= 220.219 on 13 degrees of freedom (p= 0.000) Meta-analysis of the NPV of VIA+ (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.000 1.000 1.00011411.930 0.000 12 Random | 0.999 0.999 1.000 5265.901 0.000 Test for heterogeneity: Q= 35.123 on 11 degrees of freedom (p= 0.000) 68 VIA, Outcome=Cancer VIA, Outcome=Cancer Ouagadougou Ouagadougou Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Bamako Bamako Mumbai Mumbai Niamey Niamey Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Ambilikai Harare Busia Combined Combined 0 .05 .1 .15 .2 0 .05 .1 .15 .2 Prevalence of disease Detection rate of disease VIA, Outcome=Cancer Ouagadougou Kolkata 1 Kolkata 2 Brazzaville Conakry Jaipur Bamako Mumbai Niamey Trivandrum 1 Trivandrum 2 Harare Ambilikai Busia Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (cancer) in VIA studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.002 0.002 0.003 11.886 0.000 12 Random | 0.004 0.003 0.006 5.439 0.000 Test for heterogeneity: Q= 154.944 on 11 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (cancer) using VIA | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.002 0.002 0.002 13.716 0.000 14 Random | 0.003 0.002 0.004 6.318 0.000 Test for heterogeneity: Q= 141.677 on 13 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of VIA (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.881 0.879 0.883 850.124 0.000 14 Random | 0.823 0.787 0.859 44.269 0.000 Test for heterogeneity: Q= 3590.107 on 13 degrees of freedom (p= 0.000) 69 6.1.2. Test characteristics of VILI The accuracy of VILI is assessed for the outcomes CIN1+, CIN2+, CIN3+ and cancer (see Table 21). Table 21. Test characteristics of VILI assessed in 10 locations (included in the multi-center cross-sectional study [Tables A-D]) and the test positivity rate of VILI observed at 11 locations (multi-center and Kenyan study [Table E]). A: Outcome=CIN1+ Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Bamako Mali 336 195 391 4,630 727 5,552 0.633 0.922 0.462 0.960 0.131 0.096 0.061 Brazzaville Congo 556 74 538 5,766 1,094 6,934 0.883 0.915 0.508 0.987 0.158 0.091 0.080 Conakry Guinea 301 156 673 7,497 974 8,627 0.659 0.918 0.309 0.980 0.113 0.053 0.035 Jaipur India 212 18 1,387 4,169 1,599 5,786 0.922 0.750 0.133 0.996 0.276 0.040 0.037 Kolkata 2 India 159 157 979 6,781 1,138 8,076 0.503 0.874 0.140 0.977 0.141 0.039 0.020 Mumbai India 159 47 527 3,270 686 4,003 0.772 0.861 0.232 0.986 0.171 0.051 0.040 Niamey Niger 48 3 180 2,303 228 2,534 0.941 0.928 0.211 0.999 0.090 0.020 0.019 Ouagadougou B-Fasso 78 6 510 1,457 588 2,051 0.929 0.741 0.133 0.996 0.287 0.041 0.038 Trivandrum 1 India 365 348 584 3,160 949 4,457 0.512 0.844 0.385 0.901 0.213 0.160 0.082 Trivandrum 2 India 261 157 419 3,922 680 4,759 0.624 0.903 0.384 0.962 0.143 0.088 0.055 Overall 2,475 1,161 6,188 42,955 8,663 52,779 0.681 0.874 0.286 0.974 0.164 0.069 0.047 B: Outcome=CIN2+ Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Bamako Mali 159 5 568 4,820 727 5,552 0.970 0.895 0.219 0.999 0.131 0.030 0.029 Brazzaville Congo 371 17 723 5,823 1,094 6,934 0.956 0.890 0.339 0.997 0.158 0.056 0.054 Conakry Guinea 163 5 811 7,648 974 8,627 0.970 0.904 0.167 0.999 0.113 0.019 0.019 Jaipur India 56 8 1,543 4,179 1,599 5,786 0.875 0.730 0.035 0.998 0.276 0.011 0.010 Kolkata 2 India 57 13 1,081 6,925 1,138 8,076 0.814 0.865 0.050 0.998 0.141 0.009 0.007 Mumbai India 60 21 626 3,296 686 4,003 0.741 0.840 0.087 0.994 0.171 0.020 0.015 Niamey Niger 18 2 210 2,304 228 2,534 0.900 0.916 0.079 0.999 0.090 0.008 0.007 Ouagadougou B-Fasso 49 1 539 1,462 588 2,051 0.980 0.731 0.083 0.999 0.287 0.024 0.024 Trivandrum 1 India 136 13 813 3,495 949 4,457 0.913 0.811 0.143 0.996 0.213 0.033 0.031 Trivandrum 2 India 65 16 615 4,063 680 4,759 0.802 0.869 0.096 0.996 0.143 0.017 0.014 Overall 1,134 101 7,529 44,015 8,663 52,779 0.918 0.854 0.131 0.998 0.164 0.023 0.021 C: Outcome=CIN3+ Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Bamako Mali 81 2 646 4,823 727 5,552 0.976 0.882 0.111 1.000 0.131 0.015 0.015 Brazzaville Congo 192 7 902 5,833 1,094 6,934 0.965 0.866 0.176 0.999 0.158 0.029 0.028 Conakry Guinea 121 2 853 7,651 974 8,627 0.984 0.900 0.124 1.000 0.113 0.014 0.014 Jaipur India 38 8 1,561 4,179 1,599 5,786 0.826 0.728 0.024 0.998 0.276 0.008 0.007 Kolkata 2 India 33 6 1,105 6,932 1,138 8,076 0.846 0.863 0.029 0.999 0.141 0.005 0.004 Mumbai India 43 16 643 3,301 686 4,003 0.729 0.837 0.063 0.995 0.171 0.015 0.011 Niamey Niger 11 1 217 2,305 228 2,534 0.917 0.914 0.048 1.000 0.090 0.005 0.004 Ouagadougou B-Fasso 37 0 551 1,463 588 2,051 1.000 0.726 0.063 1.000 0.287 0.018 0.018 Trivandrum 1 India 55 1 894 3,507 949 4,457 0.982 0.797 0.058 1.000 0.213 0.013 0.012 Trivandrum 2 India 52 9 628 4,070 680 4,759 0.852 0.866 0.076 0.998 0.143 0.013 0.011 Setting 663 52 8,000 44,064 8,663 52,779 0.927 0.846 0.077 0.999 0.164 0.014 0.013 70 D: Outcome=Cancer Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Bamako Mali 52 2 675 4,823 727 5,552 0.963 0.877 0.072 1.000 0.131 0.010 0.009 Brazzaville Congo 25 1 1,069 5,839 1,094 6,934 0.962 0.845 0.023 1.000 0.158 0.004 0.004 Conakry Guinea 88 1 886 7,652 974 8,627 0.989 0.896 0.090 1.000 0.113 0.010 0.010 Jaipur India 4 2 1,595 4,185 1,599 5,786 0.667 0.724 0.003 1.000 0.276 0.001 0.001 Kolkata 2 India 8 0 1,130 6,938 1,138 8,076 1.000 0.860 0.007 1.000 0.141 0.001 0.001 Mumbai India 20 7 666 3,310 686 4,003 0.741 0.832 0.029 0.998 0.171 0.007 0.005 Niamey Niger 4 1 224 2,305 228 2,534 0.800 0.911 0.018 1.000 0.090 0.002 0.002 Ouagadougou B-Fasso 17 0 571 1,463 588 2,051 1.000 0.719 0.029 1.000 0.287 0.008 0.008 Trivandrum 1 India 6 0 943 3,508 949 4,457 1.000 0.788 0.006 1.000 0.213 0.001 0.001 Trivandrum 2 India 13 2 667 4,077 680 4,759 0.867 0.859 0.019 1.000 0.143 0.003 0.003 Overall 237 16 8,426 44,100 8,663 52,779 0.937 0.840 0.027 1.000 0.164 0.005 0.004 E. PS: the test positivity rate of VILI is also documented in a small subpopulation of the WKCCPP (Busia): 19/137=13.9% VILI Ouagadougou Kolkata 2 Brazzaville Conakry Jaipur Bamako Mumbai Niamey Trivandrum 1 Trivandrum 2 Busia Combined 0 .1 .2 .3 .4 .5 Test positivity rate Meta-analysis of the test-positivity rate of VILI | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.152 0.149 0.155 98.351 0.000 11 Random | 0.170 0.138 0.201 10.525 0.000 Test for heterogeneity: Q= 1021.571 on 10 degrees of freedom (p= 0.000) 71 VILI, Outcome=CIN1+ VILI, Outcome=CIN1+ Ouagadougou Ouagadougou Kolkata 2 Kolkata 2 Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Bamako Bamako Mumbai Mumbai Niamey Niamey Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 .6 .7 .8 .9 1 Sensitivity Specificity VILI, Outcome=CIN1+ VILI, Outcome=CIN1+ Ouagadougou Ouagadougou Kolkata 2 Kolkata 2 Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Bamako Bamako Mumbai Mumbai Niamey Niamey Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .8 .85 .9 .95 1 PPV NPV Meta-analysis of the sensitivity of VILI (CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.755 0.742 0.768 114.091 0.000 10 Random | 0.737 0.630 0.845 13.425 0.000 Test for heterogeneity: Q= 590.110 on 9 degrees of freedom (p= 0.000) Meta-analysis of the specificity of VILI (CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.890 0.888 0.893 639.643 0.000 10 Random | 0.866 0.834 0.898 53.400 0.000 Test for heterogeneity: Q= 1176.071 on 9 degrees of freedom (p= 0.000) Meta-analysis of the PPV of VILI (CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.241 0.233 0.250 55.266 0.000 10 Random | 0.289 0.199 0.379 6.308 0.000 Test for heterogeneity: Q= 940.246 on 9 degrees of freedom (p= 0.000) Meta-analysis of the NPV of VILI (CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.991 0.990 0.991 2167.038 0.000 10 Random | 0.975 0.966 0.984 211.414 0.000 Test for heterogeneity: Q= 792.965 on 9 degrees of freedom (p= 0.000) 72 VILI, Outcome=CIN1+ VILI, Outcome=CIN1+ Ouagadougou Ouagadougou Kolkata 2 Kolkata 2 Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Bamako Bamako Mumbai Mumbai Niamey Niamey Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .05 .1 .15 .2 0 .05 .1 .15 .2 Prevalence of disease Detection rate of disease VILI, Outcome=CIN1+ Ouagadougou Kolkata 2 Brazzaville Conakry Jaipur Bamako Mumbai Niamey Trivandrum 1 Trivandrum 2 Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN1+) in VILI studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.054 0.053 0.056 55.624 0.000 10 Random | 0.068 0.048 0.087 6.776 0.000 Test for heterogeneity: Q= 900.857 on 9 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN1+) using VILI | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.037 0.036 0.039 45.563 0.000 10 Random | 0.046 0.033 0.059 6.993 0.000 Test for heterogeneity: Q= 549.588 on 9 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of VILI (CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.893 0.890 0.896 655.823 0.000 10 Random | 0.869 0.838 0.900 54.706 0.000 Test for heterogeneity: Q= 1179.316 on 9 degrees of freedom (p= 0.000) 73 VILI, Outcome=CIN2+ VILI, Outcome=CIN2+ Burkina-Fasso Burkina-Fasso Kolkata 2 Kolkata 2 Congo Congo Guinea Guinea Jaipur Jaipur Mali Mali Mumbai Mumbai Niger Niger Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 .6 .7 .8 .9 1 Sensitivity Specificity VILI, Outcome=CIN2+ VILI, Outcome=CIN2+ Burkina-Fasso Burkina-Fasso Kolkata 2 Kolkata 2 Congo Congo Guinea Guinea Jaipur Jaipur Mali Mali Mumbai Mumbai Niger Niger Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .8 .85 .9 .95 1 PPV NPV Meta-analysis of the sensitivity of VILI (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.950 0.938 0.962 156.881 0.000 10 Random | 0.912 0.878 0.946 52.760 0.000 Test for heterogeneity: Q= 51.760 on 9 degrees of freedom (p= 0.000) Meta-analysis of the specificity of VILI (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.868 0.865 0.871 589.058 0.000 10 Random | 0.845 0.812 0.879 50.177 0.000 Test for heterogeneity: Q= 1135.139 on 9 degrees of freedom (p= 0.000) Meta-analysis of the PPV of VILI (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.083 0.077 0.088 28.873 0.000 10 Random | 0.129 0.080 0.179 5.111 0.000 Test for heterogeneity: Q= 612.294 on 9 degrees of freedom (p= 0.000) Meta-analysis of the NPV of VILI (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.999 0.998 0.999 5615.658 0.000 10 Random | 0.998 0.997 0.999 2408.406 0.000 Test for heterogeneity: Q= 39.520 on 9 degrees of freedom (p= 0.000) 74 VILI, Outcome=CIN2+ VILI, Outcome=CIN2+ Burkina-Fasso Burkina-Fasso Kolkata 2 Kolkata 2 Congo Congo Guinea Guinea Jaipur Jaipur Mali Mali Mumbai Mumbai Niger Niger Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .05 .1 .15 .2 0 .05 .1 .15 .2 Prevalence of disease Detection rate of disease VILI, Outcome=CIN2+ Burkina-Fasso Kolkata 2 Congo Guinea Jaipur Mali Mumbai Niger Trivandrum 1 Trivandrum 2 Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN2+) in VILI studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.016 0.015 0.018 29.877 0.000 10 Random | 0.023 0.015 0.030 6.043 0.000 Test for heterogeneity: Q= 380.122 on 9 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN2+) using VILI | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.014 0.013 0.015 27.998 0.000 10 Random | 0.021 0.014 0.028 5.736 0.000 Test for heterogeneity: Q= 411.348 on 9 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of VILI (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.868 0.866 0.871 590.181 0.000 10 Random | 0.846 0.813 0.879 50.412 0.000 Test for heterogeneity: Q= 1129.326 on 9 degrees of freedom (p= 0.000) 75 VILI, Outcome=CIN3+ VILI, Outcome=CIN3+ Burkina-Fasso Burkina-Fasso Kolkata 2 Kolkata 2 Congo Congo Guinea Guinea Jaipur Jaipur Mali Mali Mumbai Mumbai Niger Niger Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 .6 .7 .8 .9 1 Sensitivity Specificity VILI, Outcome=CIN3+ VILI, Outcome=CIN3+ Burkina-Fasso Burkina-Fasso Kolkata 2 Kolkata 2 Congo Congo Guinea Guinea Jaipur Jaipur Mali Mali Mumbai Mumbai Niger Niger Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .8 .85 .9 .95 1 PPV NPV Meta-analysis of the sensitivity of VILI (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.969 0.956 0.981 148.626 0.000 10 Random | 0.938 0.906 0.971 56.176 0.000 Test for heterogeneity: Q= 39.512 on 9 degrees of freedom (p= 0.000) Meta-analysis of the specificity of VILI (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.860 0.857 0.863 570.570 0.000 10 Random | 0.838 0.805 0.871 49.823 0.000 Test for heterogeneity: Q= 1084.136 on 9 degrees of freedom (p= 0.000) Meta-analysis of the PPV of VILI (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.050 0.046 0.055 21.804 0.000 10 Random | 0.077 0.050 0.103 5.598 0.000 Test for heterogeneity: Q= 272.054 on 9 degrees of freedom (p= 0.000) Meta-analysis of the NPV of VILI (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.000 0.999 1.000 8864.979 0.000 10 Random | 0.999 0.999 1.000 3719.816 0.000 Test for heterogeneity: Q= 38.767 on 9 degrees of freedom (p= 0.000) 76 VILI, Outcome=CIN3+ VILI, Outcome=CIN3+ Burkina-Fasso Burkina-Fasso Kolkata 2 Kolkata 2 Congo Congo Guinea Guinea Jaipur Jaipur Mali Mali Mumbai Mumbai Niger Niger Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .05 .1 .15 .2 0 .05 .1 .15 .2 Prevalence of disease Detection rate of disease VILI, Outcome=CIN3+ Burkina-Fasso Kolkata 2 Congo Guinea Jaipur Mali Mumbai Niger Trivandrum 1 Trivandrum 2 Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN3+) in VILI studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.010 0.009 0.011 23.503 0.000 10 Random | 0.013 0.009 0.017 6.415 0.000 Test for heterogeneity: Q= 188.156 on 9 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN3+) using VILI | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.009 0.008 0.010 21.875 0.000 10 Random | 0.012 0.008 0.016 5.948 0.000 Test for heterogeneity: Q= 206.184 on 9 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of VILI (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.860 0.857 0.863 571.052 0.000 10 Random | 0.838 0.806 0.871 50.002 0.000 Test for heterogeneity: Q= 1078.657 on 9 degrees of freedom (p= 0.000) 77 VILI, Outcome=Cancer VILI, Outcome=Cancer Burkina-Fasso Burkina-Fasso Kolkata 2 Kolkata 2 Congo Congo Guinea Guinea Jaipur Jaipur Mali Mali Mumbai Mumbai Niger Niger Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 .6 .7 .8 .9 1 Sensitivity Specificity VILI, Outcome=Cancer VILI, Outcome=Cancer Burkina-Fasso Burkina-Fasso Kolkata 2 Kolkata 2 Congo Congo Guinea Guinea Jaipur Jaipur Mali Mali Mumbai Mumbai Niger Niger Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .8 .85 .9 .95 1 PPV NPV Meta-analysis of the sensitivity of VILI (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.978 0.960 0.997 103.304 0.000 10 Random | 0.957 0.918 0.997 47.504 0.000 Test for heterogeneity: Q= 14.870 on 9 degrees of freedom (p= 0.095) Meta-analysis of the specificity of VILI (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.853 0.850 0.856 557.448 0.000 10 Random | 0.831 0.798 0.865 48.341 0.000 Test for heterogeneity: Q= 1100.040 on 9 degrees of freedom (p= 0.000) Meta-analysis of the PPV of VILI (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.008 0.006 0.010 8.677 0.000 10 Random | 0.027 0.017 0.037 5.163 0.000 Test for heterogeneity: Q= 177.929 on 9 degrees of freedom (p= 0.000) Meta-analysis of the NPV of VILI (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.000 1.000 1.00014107.972 0.000 10 Random | 1.000 1.000 1.00014107.972 0.000 Test for heterogeneity: Q= 6.250 on 9 degrees of freedom (p= 0.715) 78 VILI, Outcome=Cancer VILI, Outcome=Cancer Burkina-Fasso Burkina-Fasso Kolkata 2 Kolkata 2 Congo Congo Guinea Guinea Jaipur Jaipur Mali Mali Mumbai Mumbai Niger Niger Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .05 .1 .15 .2 0 .05 .1 .15 .2 Prevalence of disease Detection rate of disease VILI, Outcome=Cancer Burkina-Fasso Kolkata 2 Congo Guinea Jaipur Mali Mumbai Niger Trivandrum 1 Trivandrum 2 Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (cancer) in VILI studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.002 0.002 0.003 10.616 0.000 10 Random | 0.004 0.003 0.006 4.948 0.000 Test for heterogeneity: Q= 140.216 on 9 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (cancer) using VILI | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.002 0.002 0.002 10.647 0.000 10 Random | 0.004 0.002 0.006 5.074 0.000 Test for heterogeneity: Q= 129.537 on 9 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of VILI (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.853 0.850 0.856 557.768 0.000 10 Random | 0.832 0.798 0.865 48.340 0.000 Test for heterogeneity: Q= 1101.527 on 9 degrees of freedom (p= 0.000) 79 6.1.3. Test characteristics of VIAM The accuracy of VIAM is assessed for the outcomes CIN1+, CIN2+, CIN3+ and cancer (see Table 22). Table 22. Test characteristics of VIAM, assessed at 3 locations included in the multi-center cross-sectional study [tables A-D] A: Outcome=CIN1+ Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Kolkata 1 India 330 180 732 4,651 1,062 1,242 0.647 0.864 0.311 0.963 0.180 0.411 0.266 Kolkata 2 India 135 183 772 6,990 907 1,090 0.425 0.901 0.149 0.974 0.112 0.292 0.124 Mumbai India 141 65 463 3,334 604 669 0.684 0.878 0.233 0.981 0.151 0.308 0.211 Overall 606 428 1,967 14,975 2,573 3,001 0.586 0.884 0.236 0.972 0.143 0.345 0.202 B: Outcome=CIN2+ Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Kolkata 1 India 104 57 958 4,774 1,062 1,119 0.646 0.833 0.098 0.988 0.180 0.144 0.093 Kolkata 2 India 52 19 855 7,154 907 926 0.732 0.893 0.057 0.997 0.112 0.077 0.056 Mumbai India 53 28 551 3,371 604 632 0.654 0.860 0.088 0.992 0.151 0.128 0.084 Overall 209 104 2,364 15,299 2,573 2,677 0.668 0.866 0.081 0.993 0.143 0.117 0.078 C: Outcome=CIN3+ Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Kolkata 1 India 65 34 997 4,797 1,062 1,096 0.657 0.828 0.061 0.993 0.180 0.090 0.059 Kolkata 2 India 29 10 878 7,163 907 917 0.744 0.891 0.032 0.999 0.112 0.043 0.032 Mumbai India 40 19 564 3,380 604 623 0.678 0.857 0.066 0.994 0.151 0.095 0.064 Overall 134 63 2,439 15,340 2,573 2,636 0.680 0.863 0.052 0.996 0.143 0.075 0.051 C: Outcome=Cancer Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Kolkata 1 India 29 9 1,033 4,822 1,062 1,071 0.763 0.824 0.027 0.998 0.180 0.035 0.027 Kolkata 2 India 8 0 899 7,173 907 907 1.000 0.889 0.009 1.000 0.112 0.009 0.009 Mumbai India 20 7 584 3,392 604 611 0.741 0.853 0.033 0.998 0.151 0.044 0.033 Overall 57 16 2,516 15,387 2,573 2,589 0.781 0.859 0.022 0.999 0.143 0.028 0.022 80 VIAM Kolkata 1 Kolkata 2 Mumbai Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Test positivity rate Meta-analysis of the test-positivity rate of VIAM | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.138 0.133 0.143 53.772 0.000 3 Random | 0.148 0.105 0.190 6.790 0.000 Test for heterogeneity: Q= 130.451 on 2 degrees of freedom (p= 0.000) 81 VIAM, Outcome=CIN1+ VIAM, Outcome=CIN1+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity VIAM, Outcome=CIN1+ VIAM, Outcome=CIN1+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of VIAM (outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.590 0.561 0.620 39.561 0.000 3 Random | 0.585 0.432 0.738 7.491 0.000 Test for heterogeneity: Q= 51.120 on 2 degrees of freedom (p= 0.000) Meta-analysis of the specificity of VIAM (outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.886 0.881 0.891 363.108 0.000 3 Random | 0.881 0.858 0.905 73.886 0.000 Test for heterogeneity: Q= 43.943 on 2 degrees of freedom (p= 0.000) Meta-analysis of the PPV of VIAM (outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.219 0.203 0.235 27.278 0.000 3 Random | 0.231 0.130 0.331 4.500 0.000 Test for heterogeneity: Q= 77.710 on 2 degrees of freedom (p= 0.000) Meta-analysis of the NPV of VIAM (outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.974 0.971 0.976 756.946 0.000 3 Random | 0.973 0.964 0.982 207.635 0.000 Test for heterogeneity: Q= 25.100 on 2 degrees of freedom (p= 0.000) 82 VIAM, Outcome=CIN1+ VIAL, Outcome=CIN1+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined 0 .1 .2 .3 .4 .5 0 .1 .2 .3 .4 .5 Prevalence of disease Detection rate of disease VIAM, Outcome=CIN1+ Kolkata 1 Kolkata 2 Mumbai Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN1+) in VIAM studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.051 0.048 0.055 31.229 0.000 3 Random | 0.059 0.031 0.087 4.177 0.000 Test for heterogeneity: Q= 127.346 on 2 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN1+) using VIAM | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.026 0.024 0.028 22.053 0.000 3 Random | 0.036 0.012 0.059 3.000 0.003 Test for heterogeneity: Q= 149.708 on 2 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of VIAM (outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.889 0.884 0.893 372.620 0.000 3 Random | 0.884 0.862 0.906 78.697 0.000 Test for heterogeneity: Q= 40.812 on 2 degrees of freedom (p= 0.000) 83 VIAM, Outcome=CIN2+ VIAM, Outcome=CIN2+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity VIAM, Outcome=CIN2+ VIAM, Outcome=CIN2+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 1.1 PPV NPV Meta-analysis of the sensitivity of VIAM (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.670 0.618 0.722 25.282 0.000 3 Random | 0.670 0.618 0.722 25.282 0.000 Test for heterogeneity: Q= 1.890 on 2 degrees of freedom (p= 0.389) Meta-analysis of the specificity of VIAM (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.871 0.866 0.875 345.641 0.000 3 Random | 0.862 0.825 0.899 45.209 0.000 Test for heterogeneity: Q= 104.007 on 2 degrees of freedom (p= 0.000) Meta-analysis of the PPV of VIAM (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.077 0.067 0.087 14.680 0.000 3 Random | 0.080 0.054 0.107 5.867 0.000 Test for heterogeneity: Q= 12.928 on 2 degrees of freedom (p= 0.002) Meta-analysis of the NPV of VIAM (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.995 0.994 0.996 1874.234 0.000 3 Random | 0.992 0.987 0.998 341.151 0.000 Test for heterogeneity: Q= 34.444 on 2 degrees of freedom (p= 0.000) 84 VIAM, Outcome=CIN2+ VIAL, Outcome=CIN2+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 Prevalence of disease Detection rate of disease VIAM, Outcome=CIN2+ Kolkata 1 Kolkata 2 Mumbai Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN2+) in VIAM studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.014 0.012 0.015 15.807 0.000 3 Random | 0.019 0.007 0.030 3.138 0.002 Test for heterogeneity: Q= 67.719 on 2 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN2+) using VIAM | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.009 0.008 0.011 12.795 0.000 3 Random | 0.012 0.005 0.020 3.139 0.002 Test for heterogeneity: Q= 43.681 on 2 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of VIAM (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.871 0.866 0.876 347.574 0.000 3 Random | 0.863 0.826 0.900 45.952 0.000 Test for heterogeneity: Q= 102.125 on 2 degrees of freedom (p= 0.000) 85 VIAM, Outcome=CIN3+ VIAM, Outcome=CIN3+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity VIAM, Outcome=CIN3+ VIAM, Outcome=CIN3+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 1.1 PPV NPV Meta-analysis of the sensitivity of VIAM (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.683 0.618 0.748 20.637 0.000 3 Random | 0.683 0.618 0.748 20.637 0.000 Test for heterogeneity: Q= 1.065 on 2 degrees of freedom (p= 0.587) Meta-analysis of the specificity of VIAM (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.867 0.863 0.872 342.104 0.000 3 Random | 0.859 0.820 0.898 43.007 0.000 Test for heterogeneity: Q= 112.657 on 2 degrees of freedom (p= 0.000) Meta-analysis of the PPV of VIAM (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.047 0.039 0.055 11.295 0.000 3 Random | 0.052 0.030 0.075 4.533 0.000 Test for heterogeneity: Q= 13.742 on 2 degrees of freedom (p= 0.001) Meta-analysis of the NPV of VIAM (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.998 0.997 0.999 2535.322 0.000 3 Random | 0.995 0.991 1.000 449.797 0.000 Test for heterogeneity: Q= 32.132 on 2 degrees of freedom (p= 0.000) 86 VIAM, Outcome=CIN3+ VIAL, Outcome=CIN3+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined 0 .1 .2 .3 .4 0 .1 .2 .3 .4 Prevalence of disease Detection rate of disease VIAM, Outcome=CIN3+ Kolkata 1 Kolkata 2 Mumbai Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN3+) in VIAM studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.008 0.007 0.009 12.233 0.000 3 Random | 0.012 0.004 0.021 2.790 0.005 Test for heterogeneity: Q= 57.541 on 2 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN3+) using VIAM | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.006 0.005 0.007 10.629 0.000 3 Random | 0.008 0.003 0.013 3.185 0.001 Test for heterogeneity: Q= 29.005 on 2 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of VIAM (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.868 0.863 0.873 343.202 0.000 3 Random | 0.859 0.821 0.898 43.752 0.000 Test for heterogeneity: Q= 109.808 on 2 degrees of freedom (p= 0.000) 87 VIAM, Outcome=Cancer VIAM, Outcome=Cancer Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity VIAM, Outcome=Cancer VIAM, Outcome=Cancer Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 1.1 PPV NPV Meta-analysis of the sensitivity of VIAM (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.814 0.723 0.905 17.521 0.000 3 Random | 0.826 0.677 0.976 10.831 0.000 Test for heterogeneity: Q= 5.191 on 2 degrees of freedom (p= 0.075) Meta-analysis of the specificity of VIAM (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.865 0.860 0.870 338.802 0.000 3 Random | 0.855 0.815 0.896 41.389 0.000 Test for heterogeneity: Q= 119.248 on 2 degrees of freedom (p= 0.000) Meta-analysis of the PPV of VIAM (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.016 0.011 0.021 6.537 0.000 3 Random | 0.022 0.007 0.038 2.809 0.005 Test for heterogeneity: Q= 15.366 on 2 degrees of freedom (p= 0.000) Meta-analysis of the NPV of VIAM (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.000 1.000 1.000 7897.081 0.000 3 Random | 0.999 0.997 1.000 1216.134 0.000 Test for heterogeneity: Q= 15.848 on 2 degrees of freedom (p= 0.000) 88 VIAM, Outcome=Cancer VIAL, Outcome=Cancer Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined 0 .1 .2 0 .1 .2 Prevalence of disease Detection rate of disease VIAM, Outcome=Cancer Kolkata 1 Kolkata 2 Mumbai Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (cancer) in VIAM studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.002 0.001 0.002 5.742 0.000 3 Random | 0.005 0.000 0.009 2.088 0.037 Test for heterogeneity: Q= 37.591 on 2 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (cancer) using VIAM | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.002 0.001 0.002 5.709 0.000 3 Random | 0.004 0.000 0.007 2.225 0.026 Test for heterogeneity: Q= 25.795 on 2 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of VIAM (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.864 0.859 0.869 338.959 0.000 3 Random | 0.855 0.815 0.896 41.611 0.000 Test for heterogeneity: Q= 118.211 on 2 degrees of freedom (p= 0.000) 89 6.1.4. Test characteristics of HPV DNA detection by HC2 The accuracy of HPV is assessed for the outcomes CIN1+, CIN2+, CIN3+ and cancer (see Table 23). Table 23. Test characteristics of the Hybrid Capture II assay assessed at 5 study locations included in the multi-center and Zimbabwe cross-sectional studies [Tables A-D]. A: Outcome=CIN1+ Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Kolkata 1 India 62 151 264 3,126 326 3,603 0.291 0.922 0.190 0.954 0.090 0.059 0.017 Kolkata 2 India 66 207 337 6,066 403 6,676 0.242 0.947 0.164 0.967 0.060 0.041 0.010 Mumbai India 57 112 211 3,163 268 3,543 0.337 0.937 0.213 0.966 0.076 0.048 0.016 Trivandrum 2 India 90 328 213 4,127 303 4,758 0.215 0.951 0.297 0.926 0.064 0.088 0.019 Harare Zimbabwe 361 205 556 1,023 917 2,145 0.638 0.648 0.394 0.833 0.428 0.264 0.168 Overall 275 798 1,025 16,482 1,300 18,580 0.256 0.941 0.212 0.954 0.070 0.058 0.015 B: Outcome=CIN2+ Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Kolkata 1 India 30 32 296 3,245 326 3,603 0.484 0.916 0.092 0.990 0.090 0.017 0.008 Kolkata 2 India 42 20 361 6,253 403 6,676 0.677 0.945 0.104 0.997 0.060 0.009 0.006 Mumbai India 43 23 225 3,252 268 3,543 0.652 0.935 0.160 0.993 0.076 0.019 0.012 Trivandrum 2 India 52 29 251 4,426 303 4,758 0.642 0.946 0.172 0.993 0.064 0.017 0.011 Harare Zimbabwe 166 41 751 1,187 917 2,145 0.802 0.612 0.181 0.967 0.428 0.097 0.077 Overall 167 104 1,133 17,176 1,300 18,580 0.616 0.938 0.128 0.994 0.070 0.015 0.009 C: Outcome=CIN3+ Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Kolkata 1 India 19 10 307 3,267 326 3,603 0.655 0.914 0.058 0.997 0.090 0.008 0.005 Setting India 25 9 378 6,264 403 6,676 0.735 0.943 0.062 0.999 0.060 0.005 0.004 Mumbai India 35 13 233 3,262 268 3,543 0.729 0.933 0.131 0.996 0.076 0.014 0.010 Trivandrum 2 India 38 23 265 4,432 303 4,758 0.623 0.944 0.125 0.995 0.064 0.013 0.008 Overall 117 55 1,183 17,225 1,300 18,580 0.680 0.936 0.090 0.997 0.070 0.009 0.006 D: Outcome=Cancer Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Kolkata 1 India 8 5 318 3,272 326 3,603 0.615 0.911 0.025 0.998 0.090 0.004 0.002 Kolkata 2 India 6 1 397 6,272 403 6,676 0.857 0.940 0.015 1.000 0.060 0.001 0.001 Mumbai India 14 5 254 3,270 268 3,543 0.737 0.928 0.052 0.998 0.076 0.005 0.004 Trivandrum 2 India 10 5 293 4,450 303 4,758 0.667 0.938 0.033 0.999 0.064 0.003 0.002 Harare Zimbabwe 2 1 915 1,227 917 2,145 0.667 0.573 0.002 0.999 0.428 0.001 0.001 Overall 38 16 1,262 17,264 1,300 18,580 0.704 0.932 0.029 0.999 0.070 0.003 0.002 90 HPV Kolkata 1 Kolkata 2 Mumbai Trivandrum 2 Harare Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Test positivity rate Meta-analysis of the test-positivity rate of Hybrid Capture II | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.079 0.075 0.082 43.240 0.000 5 Random | 0.143 0.079 0.206 4.415 0.000 Test for heterogeneity: Q= 1133.683 on 4 degrees of freedom (p= 0.000) 91 HPV, Outcome=CIN1+ HPV, Outcome=CIN1+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity HPV, Outcome=CIN1+ HPV, Outcome=CIN1+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of HC2 (outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.367 0.345 0.389 33.246 0.000 5 Random | 0.345 0.164 0.526 3.731 0.000 Test for heterogeneity: Q= 266.591 on 4 degrees of freedom (p= 0.000) Meta-analysis of the specificity of HC2 (outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.937 0.933 0.940 540.691 0.000 5 Random | 0.883 0.839 0.928 38.737 0.000 Test for heterogeneity: Q= 620.852 on 4 degrees of freedom (p= 0.000) Meta-analysis of the PPV of HC2 (outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.265 0.247 0.283 28.980 0.000 5 Random | 0.252 0.155 0.348 5.112 0.000 Test for heterogeneity: Q= 111.479 on 4 degrees of freedom (p= 0.000) Meta-analysis of the NPV of HC2 (outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.956 0.953 0.959 637.271 0.000 5 Random | 0.932 0.908 0.955 76.938 0.000 Test for heterogeneity: Q= 224.420 on 4 degrees of freedom (p= 0.000) 92 HPV, Outcome=CIN1+ HPV, Outcome=CIN1+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 0 .1 .2 .3 .4 .5 Prevalence of disease Detection rate of disease HPV, Outcome=CIN1+ Kolkata 1 Kolkata 2 Mumbai Trivandrum 2 Harare Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN1+) in HC2 studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.059 0.056 0.062 36.564 0.000 5 Random | 0.099 0.058 0.139 4.766 0.000 Test for heterogeneity: Q= 578.683 on 4 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN1+) using HC2 | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.015 0.014 0.017 18.180 0.000 5 Random | 0.043 0.025 0.061 4.679 0.000 Test for heterogeneity: Q= 382.955 on 4 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of HC2 (outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.940 0.936 0.943 567.542 0.000 5 Random | 0.893 0.852 0.934 42.804 0.000 Test for heterogeneity: Q= 570.773 on 4 degrees of freedom (p= 0.000) 93 HPV, Outcome=CIN2+ HPV, Outcome=CIN2+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity HPV, Outcome=CIN2+ HPV, Outcome=CIN2+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of HC2 (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.715 0.676 0.755 35.616 0.000 5 Random | 0.658 0.545 0.770 11.451 0.000 Test for heterogeneity: Q= 26.562 on 4 degrees of freedom (p= 0.000) Meta-analysis of the specificity of HC2 (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.932 0.928 0.935 536.921 0.000 5 Random | 0.873 0.819 0.926 32.044 0.000 Test for heterogeneity: Q= 887.408 on 4 degrees of freedom (p= 0.000) Meta-analysis of the PPV of HC2 (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.141 0.127 0.155 19.203 0.000 5 Random | 0.141 0.102 0.180 7.060 0.000 Test for heterogeneity: Q= 27.858 on 4 degrees of freedom (p= 0.000) Meta-analysis of the NPV (of HC2 (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.995 0.994 0.996 1863.343 0.000 5 Random | 0.991 0.986 0.995 446.571 0.000 Test for heterogeneity: Q= 47.978 on 4 degrees of freedom (p= 0.000) 94 HPV, Outcome=CIN2+ HPV, Outcome=CIN2+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 0 .1 .2 .3 .4 .5 Prevalence of disease Detection rate of disease HPV, Outcome=CIN2+ Kolkata 1 Kolkata 2 Mumbai Trivandrum 2 Harare Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN2+) in HC2 studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.015 0.013 0.016 17.683 0.000 5 Random | 0.030 0.017 0.042 4.622 0.000 Test for heterogeneity: Q= 191.753 on 4 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN2+) using HC2 | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.009 0.008 0.011 13.987 0.000 5 Random | 0.020 0.011 0.029 4.467 0.000 Test for heterogeneity: Q= 152.905 on 4 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of HC2 (outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.932 0.929 0.935 540.065 0.000 5 Random | 0.875 0.822 0.927 32.626 0.000 Test for heterogeneity: Q= 870.395 on 4 degrees of freedom (p= 0.000) 95 HPV, Outcome=CIN3+ HPV, Outcome=CIN3+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity HPV, Outcome=CIN3+ HPV, Outcome=CIN3+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of HC2 (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.684 0.615 0.754 19.430 0.000 4 Random | 0.684 0.615 0.754 19.430 0.000 Test for heterogeneity: Q= 2.030 on 3 degrees of freedom (p= 0.566) Meta-analysis of the specificity of HC2 (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.937 0.934 0.941 524.931 0.000 4 Random | 0.934 0.922 0.946 153.049 0.000 Test for heterogeneity: Q= 33.008 on 3 degrees of freedom (p= 0.000) Meta-analysis of the PPV of HC2 (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.080 0.065 0.094 10.666 0.000 4 Random | 0.091 0.055 0.127 4.952 0.000 Test for heterogeneity: Q= 16.773 on 3 degrees of freedom (p= 0.001) Meta-analysis of the NPV of HC2 (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.998 0.997 0.998 2667.737 0.000 4 Random | 0.997 0.995 0.999 1095.700 0.000 Test for heterogeneity: Q= 13.194 on 3 degrees of freedom (p= 0.004) 96 HPV, Outcome=CIN3+ HPV, Outcome=CIN3+ Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 0 .1 .2 .3 .4 .5 Prevalence of disease Detection rate of disease HPV, Outcome=CIN3+ Kolkata 1 Kolkata 2 Mumbai Trivandrum 2 Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN3+) in HC2 studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.008 0.007 0.009 12.081 0.000 4 Random | 0.010 0.005 0.014 4.479 0.000 Test for heterogeneity: Q= 27.917 on 3 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN3+) using HC2 | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.005 0.004 0.006 10.083 0.000 4 Random | 0.006 0.004 0.009 4.754 0.000 Test for heterogeneity: Q= 16.193 on 3 degrees of freedom (p= 0.001) Meta-analysis of the approximate specificity of HC2 (outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.937 0.934 0.941 526.551 0.000 4 Random | 0.934 0.922 0.946 153.593 0.000 Test for heterogeneity: Q= 32.993 on 3 degrees of freedom (p= 0.000) 97 HPV, Outcome=Cancer HPV, Outcome=Cancer Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity HPV, Outcome=Cancer HPV, Outcome=Cancer Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of HC2 (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.718 0.603 0.833 12.250 0.000 5 Random | 0.718 0.603 0.833 12.250 0.000 Test for heterogeneity: Q= 1.933 on 4 degrees of freedom (p= 0.748) Meta-analysis of the specificity of HC2 (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.923 0.919 0.927 511.311 0.000 5 Random | 0.859 0.797 0.922 26.880 0.000 Test for heterogeneity: Q= 1135.945 on 4 degrees of freedom (p= 0.000) Meta-analysis of the PPV of HC2 (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.005 0.002 0.008 3.269 0.001 5 Random | 0.022 0.007 0.038 2.764 0.006 Test for heterogeneity: Q= 30.725 on 4 degrees of freedom (p= 0.000) Meta-analysis of the NPV of HC2 (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.000 0.999 1.000 7007.301 0.000 5 Random | 0.999 0.998 1.000 2774.918 0.000 Test for heterogeneity: Q= 10.038 on 4 degrees of freedom (p= 0.040) 98 HPV, Outcome=Cancer HPV, Outcome=Cancer Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .05 .1 .15 .2 0 .05 .1 .15 .2 Prevalence of disease Detection rate of disease HPV, Outcome=Cancer Kolkata 1 Kolkata 2 Mumbai Trivandrum 2 Harare Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (cancer) in HC2 studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.002 0.001 0.002 6.241 0.000 5 Random | 0.003 0.001 0.004 3.599 0.000 Test for heterogeneity: Q= 18.263 on 4 degrees of freedom (p= 0.001) Meta-analysis of the detection rate of disease (cancer) using HC2 | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.001 0.001 0.002 5.447 0.000 5 Random | 0.002 0.001 0.003 3.740 0.000 Test for heterogeneity: Q= 10.431 on 4 degrees of freedom (p= 0.034) Meta-analysis of the approximate specificity of HC2 (outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.923 0.920 0.927 511.736 0.000 5 Random | 0.859 0.797 0.922 26.911 0.000 Test for heterogeneity: Q= 1135.545 on 4 degrees of freedom (p= 0.000) 99 6.1.5. Test characteristics of cervical cytology The accuracy of cytology is assessed at 3 different cut-offs ASCUS, LSIL and HSIL for the outcome of respectively CIN1+, CIN2+, CIN3+ and cancer (see Table 24). Table 24. Test characteristics of the Pap smear assessed at six locations, included in the multi-center and Zimbabwe cross-sectional studies [Tables A-D]. A: Outcome=CIN1+, threshold>=ASCUS Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Jaipur India 64 162 343 5,187 407 5,756 0.283 0.938 0.157 0.970 0.071 0.039 0.011 Kolkata 1 India 112 375 662 4,263 774 5,412 0.230 0.866 0.145 0.919 0.143 0.090 0.021 Mumbai India 61 136 49 3,709 110 3,955 0.310 0.987 0.555 0.965 0.028 0.050 0.015 Trivandrum 1 India 467 246 175 3,569 642 4,457 0.655 0.953 0.727 0.936 0.144 0.160 0.105 Trivandrum 2 India 98 320 66 4,275 164 4,759 0.234 0.985 0.598 0.930 0.034 0.088 0.021 Harare Zimbabwe 257 295 324 1,216 581 2,092 0.466 0.790 0.442 0.805 0.278 0.264 0.123 Overall 802 1,239 1,295 21,003 2,097 24,339 0.393 0.942 0.382 0.944 0.086 0.084 0.033 B: Outcome=CIN1+, threshold>=LSIL Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Jaipur India 56 170 197 5,333 253 5,756 0.248 0.964 0.221 0.969 0.044 0.039 0.010 Kolkata 1 India 84 403 349 4,576 433 5,412 0.172 0.929 0.194 0.919 0.080 0.090 0.016 Mumbai India 54 143 28 3,730 82 3,955 0.274 0.993 0.659 0.963 0.021 0.050 0.014 Trivandrum 1 India 451 262 158 3,586 609 4,457 0.633 0.958 0.741 0.932 0.137 0.160 0.101 Trivandrum 2 India 84 334 47 4,294 131 4,759 0.201 0.989 0.641 0.928 0.028 0.088 0.018 Harare Zimbabwe 158 394 109 1,431 267 2,092 0.286 0.929 0.592 0.784 0.128 0.264 0.076 Overall 729 1,312 779 21,519 1,508 24,339 0.357 0.965 0.483 0.943 0.062 0.084 0.030 C: Outcome=CIN2+, threshold>= ASCUS Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Jaipur India 21 42 386 5,307 407 5,756 0.333 0.932 0.052 0.992 0.071 0.011 0.004 Kolkata 1 India 62 93 712 4,545 774 5,412 0.400 0.865 0.080 0.980 0.143 0.029 0.011 Mumbai India 53 26 57 3,819 110 3,955 0.671 0.985 0.482 0.993 0.028 0.020 0.013 Trivandrum 1 India 122 27 520 3,788 642 4,457 0.819 0.879 0.190 0.993 0.144 0.033 0.027 Trivandrum 2 India 50 31 114 4,564 164 4,759 0.617 0.976 0.305 0.993 0.034 0.017 0.011 Harare Zimbabwe 130 71 451 1,440 581 2,092 0.647 0.762 0.224 0.953 0.278 0.096 0.062 Overall 308 219 1,789 22,023 2,097 24,339 0.584 0.925 0.147 0.990 0.086 0.022 0.013 D: Outcome=CIN2+, threshold>=LSIL Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Jaipur India 15 48 238 5,455 253 5,756 0.238 0.958 0.059 0.991 0.044 0.011 0.003 Kolkata 1 India 51 104 382 4,875 433 5,412 0.329 0.927 0.118 0.979 0.080 0.029 0.009 Mumbai India 47 32 35 3,841 82 3,955 0.595 0.991 0.573 0.992 0.021 0.020 0.012 Trivandrum 1 India 116 33 493 3,815 609 4,457 0.779 0.886 0.190 0.991 0.137 0.033 0.026 Trivandrum 2 India 50 31 81 4,597 131 4,759 0.617 0.983 0.382 0.993 0.028 0.017 0.011 Harare Zimbabwe 89 112 178 1,713 267 2,092 0.443 0.906 0.333 0.939 0.128 0.096 0.043 Overall 279 248 1,229 22,583 1,508 24,339 0.529 0.948 0.185 0.989 0.062 0.022 0.011 E: Outcome=CIN2+, threshold>=HSIL Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Jaipur India 11 52 17 5,676 28 5,756 0.175 0.997 0.393 0.991 0.005 0.011 0.002 Kolkata 1 India 44 111 121 5,136 165 5,412 0.284 0.977 0.267 0.979 0.030 0.029 0.008 Mumbai India 43 36 17 3,859 60 3,955 0.544 0.996 0.717 0.991 0.015 0.020 0.011 Trivandrum 1 India 77 72 20 4,288 97 4,457 0.517 0.995 0.794 0.983 0.022 0.033 0.017 Trivandrum 2 India 50 31 18 4,660 68 4,759 0.617 0.996 0.735 0.993 0.014 0.017 0.011 Harare Zimbabwe 33 168 40 1,851 73 2,092 0.164 0.979 0.452 0.917 0.035 0.096 0.016 Overall 225 302 193 23,619 418 24,339 0.427 0.992 0.538 0.987 0.017 0.022 0.009 100 F: Outcome=CIN3+, threshold>= ASCUS Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Jaipur India 16 29 391 5,320 407 5,756 0.356 0.932 0.039 0.995 0.071 0.008 0.003 Kolkata 1 India 44 52 730 4,586 774 5,412 0.458 0.863 0.057 0.989 0.143 0.018 0.008 Mumbai India 41 16 69 3,829 110 3,955 0.719 0.982 0.373 0.996 0.028 0.014 0.010 Trivandrum 1 India 54 2 588 3,813 642 4,457 0.964 0.866 0.084 0.999 0.144 0.013 0.012 Trivandrum 2 India 39 22 125 4,573 164 4,759 0.639 0.973 0.238 0.995 0.034 0.013 0.008 Overall 194 121 1,903 22,121 2,097 24,339 0.616 0.921 0.093 0.995 0.086 0.013 0.008 G: Outcome=CIN3+, threshold>=LSIL Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Jaipur India 12 33 241 5,470 253 5,756 0.267 0.958 0.047 0.994 0.044 0.008 0.002 Kolkata 1 India 37 59 396 4,920 433 5,412 0.385 0.926 0.085 0.988 0.080 0.018 0.007 Mumbai India 35 22 47 3,851 82 3,955 0.614 0.988 0.427 0.994 0.021 0.014 0.009 Trivandrum 1 India 50 6 559 3,842 609 4,457 0.893 0.873 0.082 0.998 0.137 0.013 0.011 Trivandrum 2 India 39 22 92 4,606 131 4,759 0.639 0.980 0.298 0.995 0.028 0.013 0.008 Overall 173 142 1,335 22,689 1,508 24,339 0.549 0.944 0.115 0.994 0.062 0.013 0.007 H: Outcome=CIN3+, threshold>=HSIL Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Jaipur India 10 35 18 5,693 28 5,756 0.222 0.997 0.357 0.994 0.005 0.008 0.002 Kolkata 1 India 34 62 131 5,185 165 5,412 0.354 0.975 0.206 0.988 0.030 0.018 0.006 Mumbai India 34 23 26 3,872 60 3,955 0.596 0.993 0.567 0.994 0.015 0.014 0.009 Trivandrum 1 India 43 13 54 4,347 97 4,457 0.768 0.988 0.443 0.997 0.022 0.013 0.010 Trivandrum 2 India 39 22 29 4,669 68 4,759 0.639 0.994 0.574 0.995 0.014 0.013 0.008 Overall 160 155 258 23,766 418 24,339 0.508 0.989 0.383 0.994 0.017 0.013 0.007 I: Outcome=Cancer, threshold>= ASCUS Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Jaipur India 2 3 405 5,346 407 5,756 0.400 0.930 0.005 0.999 0.071 0.001 0.000 Kolkata 1 India 20 17 754 4,621 774 5,412 0.541 0.860 0.026 0.996 0.143 0.007 0.004 Setting India 19 6 91 3,839 110 3,955 0.760 0.977 0.173 0.998 0.028 0.006 0.005 Trivandrum 1 India 6 0 636 3,815 642 4,457 1.000 0.857 0.009 1.000 0.144 0.001 0.001 Trivandrum 2 India 12 3 152 4,592 164 4,759 0.800 0.968 0.073 0.999 0.034 0.003 0.003 Harare Zimbabwe 3 0 578 1,511 581 2,092 1.000 0.723 0.005 1.000 0.278 0.001 0.001 Overall 59 29 2,038 22,213 2,097 24,339 0.670 0.916 0.028 0.999 0.086 0.004 0.002 J: Outcome=Cancer, threshold>=LSIL Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Jaipur India 1 4 252 5,499 253 5,756 0.200 0.956 0.004 0.999 0.044 0.001 0.000 Kolkata 1 India 18 19 415 4,960 433 5,412 0.486 0.923 0.042 0.996 0.080 0.007 0.003 Mumbai India 17 8 65 3,865 82 3,955 0.680 0.983 0.207 0.998 0.021 0.006 0.004 Trivandrum 1 India 6 0 603 3,848 609 4,457 1.000 0.865 0.010 1.000 0.137 0.001 0.001 Trivandrum 2 India 12 3 119 4,625 131 4,759 0.800 0.975 0.092 0.999 0.028 0.003 0.003 Harare Zimbabwe 3 0 264 1,825 267 2,092 1.000 0.874 0.011 1.000 0.128 0.001 0.001 Overall 54 34 1,454 22,797 1,508 24,339 0.614 0.940 0.036 0.999 0.062 0.004 0.002 K: Outcome=Cancer, threshold>=HSIL Setting Country TP FN FP TN T+ N Se Sp PPV NPV T+ rate Prev Det rate Jaipur India 1 4 27 5,724 28 5,756 0.200 0.995 0.036 0.999 0.005 0.001 0.000 Kolkata 1 India 18 19 147 5,228 165 5,412 0.486 0.973 0.109 0.996 0.030 0.007 0.003 Mumbai India 17 8 43 3,887 60 3,955 0.680 0.989 0.283 0.998 0.015 0.006 0.004 Trivandrum 1 India 6 0 91 4,360 97 4,457 1.000 0.980 0.062 1.000 0.022 0.001 0.001 Trivandrum 2 India 12 3 56 4,688 68 4,759 0.800 0.988 0.176 0.999 0.014 0.003 0.003 Harare Zimbabwe 3 0 70 2,019 73 2,092 1.000 0.966 0.041 1.000 0.035 0.001 0.001 Overall 54 34 364 23,887 418 24,339 0.614 0.985 0.129 0.999 0.017 0.004 0.002 101 ASCUS Kolkata 1 Jaipur Mumbai Trivandrum 1 Trivandrum 2 Harare Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Test positivity rate Meta-analysis of the test-positivity rate at cutoff ASCUS+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.063 0.060 0.066 43.281 0.000 6 Random | 0.116 0.067 0.164 4.657 0.000 Test for heterogeneity: Q= 1306.894 on 5 degrees of freedom (p= 0.000) 102 ASCUS, Outcome=CIN1+ ASCUS, Outcome=CIN1+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity ASCUS, Outcome=CIN1+ ASCUS, Outcome=CIN1+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of cyto (ASCUS+)(outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.391 0.373 0.408 43.657 0.000 6 Random | 0.363 0.208 0.519 4.583 0.000 Test for heterogeneity: Q= 379.623 on 5 degrees of freedom (p= 0.000) Meta-analysis of the specificity of cyto (ASCUS+)(outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.968 0.966 0.971 872.974 0.000 6 Random | 0.921 0.887 0.955 53.379 0.000 Test for heterogeneity: Q= 1029.603 on 5 degrees of freedom (p= 0.000) Meta-analysis of the PPV of cyto (ASCUS+)(outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.339 0.324 0.354 42.941 0.000 6 Random | 0.436 0.214 0.658 3.852 0.000 Test for heterogeneity: Q= 917.451 on 5 degrees of freedom (p= 0.000) Meta-analysis of the NPV of cyto (ASCUS+)(outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.949 0.946 0.952 671.877 0.000 6 Random | 0.922 0.896 0.948 69.730 0.000 Test for heterogeneity: Q= 397.482 on 5 degrees of freedom (p= 0.000) 103 ASCUS, Outcome=CIN1+ ASCUS, Outcome=CIN1+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 0 .1 .2 .3 .4 Prevalence of disease Prevalence of disease ASCUS, Outcome=CIN1+ Kolkata 1 Jaipur Mumbai Trivandrum 1 Trivandrum 2 Harare Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN1+) in cytology studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.073 0.070 0.076 46.268 0.000 6 Random | 0.114 0.071 0.158 5.194 0.000 Test for heterogeneity: Q= 893.924 on 5 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN1+) using cytology (ASCUS+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.020 0.019 0.022 23.709 0.000 6 Random | 0.048 0.028 0.068 4.805 0.000 Test for heterogeneity: Q= 593.645 on 5 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of cyto (ASCUS+)(outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.970 0.968 0.972 922.228 0.000 6 Random | 0.929 0.897 0.960 57.768 0.000 Test for heterogeneity: Q= 997.015 on 5 degrees of freedom (p= 0.000) 104 ASCUS, Outcome=CIN2+ ASCUS, Outcome=CIN2+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity ASCUS, Outcome=CIN2+ ASCUS, Outcome=CIN2+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of cyto (ASCUS+)(outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.625 0.592 0.658 37.042 0.000 6 Random | 0.584 0.435 0.732 7.686 0.000 Test for heterogeneity: Q= 95.717 on 5 degrees of freedom (p= 0.000) Meta-analysis of the specificity of cyto (ASCUS+)(outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.956 0.954 0.959 771.272 0.000 6 Random | 0.900 0.857 0.944 40.496 0.000 Test for heterogeneity: Q= 1364.330 on 5 degrees of freedom (p= 0.000) Meta-analysis of the PPV of cyto (ASCUS+)(outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.118 0.107 0.130 19.639 0.000 6 Random | 0.212 0.131 0.293 5.137 0.000 Test for heterogeneity: Q= 196.150 on 5 degrees of freedom (p= 0.000) Meta-analysis of the NPV of cyto (ASCUS+)(outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.991 0.990 0.992 1646.090 0.000 6 Random | 0.986 0.981 0.992 370.110 0.000 Test for heterogeneity: Q= 86.572 on 5 degrees of freedom (p= 0.000) 105 ASCUS, Outcome=CIN2+ ASCUS, Outcome=CIN2+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 0 .1 .2 .3 .4 Prevalence of disease Detection rate of disease ASCUS, Outcome=CIN2+ Kolkata 1 Jaipur Mumbai Trivandrum 1 Trivandrum 2 Harare Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN2+) in cytology studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.020 0.018 0.021 23.193 0.000 6 Random | 0.033 0.021 0.045 5.390 0.000 Test for heterogeneity: Q= 224.884 on 5 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN2+) using cytology (ASCUS+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.009 0.008 0.010 15.465 0.000 6 Random | 0.020 0.012 0.028 4.668 0.000 Test for heterogeneity: Q= 212.913 on 5 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of cyto (ASCUS+)(outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.957 0.954 0.959 777.413 0.000 6 Random | 0.903 0.860 0.945 41.160 0.000 Test for heterogeneity: Q= 1349.328 on 5 degrees of freedom (p= 0.000) 106 ASCUS, Outcome=CIN3+ ASCUS, Outcome=CIN3+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity ASCUS, Outcome=CIN3+ ASCUS, Outcome=CIN3+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of cyto (ASCUS+)(outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.794 0.756 0.831 41.664 0.000 5 Random | 0.630 0.379 0.882 4.918 0.000 Test for heterogeneity: Q= 136.376 on 4 degrees of freedom (p= 0.000) Meta-analysis of the specificity of cyto (ASCUS+)(outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.955 0.952 0.957 725.854 0.000 5 Random | 0.923 0.881 0.966 42.886 0.000 Test for heterogeneity: Q= 958.654 on 4 degrees of freedom (p= 0.000) Meta-analysis of the PPV of cyto (ASCUS+)(outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.067 0.056 0.077 12.499 0.000 5 Random | 0.136 0.081 0.190 4.870 0.000 Test for heterogeneity: Q= 82.562 on 4 degrees of freedom (p= 0.000) Meta-analysis of the NPV of cyto (ASCUS+)(outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.998 0.997 0.998 3246.354 0.000 5 Random | 0.995 0.992 0.998 585.494 0.000 Test for heterogeneity: Q= 74.665 on 4 degrees of freedom (p= 0.000) 107 ASCUS, Outcome=CIN3+ ASCUS, Outcome=CIN3+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 0 .1 .2 .3 .4 Prevalence of disease Detection rate of disease ASCUS, Outcome=CIN3+ Kolkata 1 Jaipur Mumbai Trivandrum 1 Trivandrum 2 Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN3+) in cytology studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.012 0.011 0.013 17.154 0.000 5 Random | 0.013 0.009 0.016 7.218 0.000 Test for heterogeneity: Q= 25.183 on 4 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN3+) using cytology (ASCUS+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.006 0.005 0.007 12.130 0.000 5 Random | 0.008 0.004 0.012 4.325 0.000 Test for heterogeneity: Q= 48.679 on 4 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of cyto (ASCUS+)(outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.955 0.952 0.957 729.283 0.000 5 Random | 0.924 0.882 0.966 43.139 0.000 Test for heterogeneity: Q= 957.002 on 4 degrees of freedom (p= 0.000) 108 ASCUS, Outcome=Cancer ASCUS, Outcome=Cancer Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity ASCUS, Outcome=Cancer ASCUS, Outcome=Cancer Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of cyto (ASCUS+)(outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.737 0.650 0.824 16.549 0.000 6 Random | 0.758 0.591 0.924 8.928 0.000 Test for heterogeneity: Q= 15.724 on 5 degrees of freedom (p= 0.008) Meta-analysis of the specificity of cyto (ASCUS+)(outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.942 0.939 0.945 669.879 0.000 6 Random | 0.886 0.838 0.935 35.729 0.000 Test for heterogeneity: Q= 1390.387 on 5 degrees of freedom (p= 0.000) Meta-analysis of the PPV of cyto (ASCUS+)(outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.009 0.006 0.013 5.016 0.000 6 Random | 0.020 0.007 0.033 3.082 0.002 Test for heterogeneity: Q= 42.368 on 5 degrees of freedom (p= 0.000) Meta-analysis of the NPV of cyto (ASCUS+)(outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.000 0.999 1.000 6270.464 0.000 6 Random | 0.999 0.998 1.000 2748.884 0.000 Test for heterogeneity: Q= 20.421 on 5 degrees of freedom (p= 0.001) 109 ASCUS, Outcome=Cancer ASCUS, Outcome=Cancer Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .05 .1 .15 .2 0 .05 .1 .15 .2 Prevalence of disease Detection rate of disease ASCUS, Outcome=Cancer Kolkata 1 Jaipur Mumbai Trivandrum 1 Trivandrum 2 Harare Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (cancer) in cytology studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.002 0.001 0.002 7.002 0.000 6 Random | 0.003 0.001 0.005 3.595 0.000 Test for heterogeneity: Q= 42.456 on 5 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (cancer) using cytology (ASCUS+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.001 0.001 0.001 5.243 0.000 6 Random | 0.002 0.001 0.003 3.284 0.001 Test for heterogeneity: Q= 34.724 on 5 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of cyto (ASCUS+)(outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.942 0.939 0.945 670.628 0.000 6 Random | 0.887 0.838 0.935 35.772 0.000 Test for heterogeneity: Q= 1390.189 on 5 degrees of freedom (p= 0.000) 110 LSIL+ Kolkata 1 Jaipur Mumbai Trivandrum 1 Trivandrum 2 Harare Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Test positivity rate Meta-analysis of the test positivity rate of cytology at the cutoff LSIL+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.045 0.042 0.047 35.708 0.000 6 Random | 0.073 0.042 0.103 4.680 0.000 Test for heterogeneity: Q= 702.927 on 5 degrees of freedom (p= 0.000) 111 LSIL+, Outcome=CIN1+ LSIL+, Outcome=CIN1+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity LSIL+, Outcome=CIN1+ LSIL+, Outcome=CIN1+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of cyto (LSIL+)(outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.314 0.298 0.331 37.180 0.000 6 Random | 0.303 0.148 0.457 3.831 0.000 Test for heterogeneity: Q= 421.821 on 5 degrees of freedom (p= 0.000) Meta-analysis of the specificity of cyto (LSIL+)(outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.980 0.979 0.982 1102.832 0.000 6 Random | 0.961 0.943 0.979 104.176 0.000 Test for heterogeneity: Q= 452.748 on 5 degrees of freedom (p= 0.000) Meta-analysis of the PPV of cyto (LSIL+)(outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.473 0.453 0.493 45.979 0.000 6 Random | 0.507 0.279 0.735 4.354 0.000 Test for heterogeneity: Q= 579.698 on 5 degrees of freedom (p= 0.000) Meta-analysis of the NPV of cyto (LSIL+)(outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.947 0.944 0.950 669.694 0.000 6 Random | 0.917 0.888 0.946 62.168 0.000 Test for heterogeneity: Q= 495.492 on 5 degrees of freedom (p= 0.000) 112 LSIL+, Outcome=CIN1+ LSIL+, Outcome=CIN1+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 0 .1 .2 .3 .4 Prevalence of disease Detection rate of disease LSIL+, Outcome=CIN1+ Kolkata 1 Jaipur Mumbai Trivandrum 1 Trivandrum 2 Harare Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN1+) in cytology studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.073 0.070 0.076 46.268 0.000 6 Random | 0.114 0.071 0.158 5.194 0.000 Test for heterogeneity: Q= 893.924 on 5 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN1+) using cytology (LSIL+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.017 0.016 0.019 21.659 0.000 6 Random | 0.038 0.022 0.054 4.579 0.000 Test for heterogeneity: Q= 485.705 on 5 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of cyto (LSIL+)(outcome CIN1+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.981 0.980 0.983 1168.192 0.000 6 Random | 0.965 0.948 0.982 113.109 0.000 Test for heterogeneity: Q= 438.502 on 5 degrees of freedom (p= 0.000) 113 LSIL+, Outcome=CIN2+ LSIL+, Outcome=CIN2+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity LSIL+, Outcome=CIN2+ LSIL+, Outcome=CIN2+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of cyto (LSIL+)(outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.516 0.482 0.550 30.083 0.000 6 Random | 0.501 0.332 0.669 5.818 0.000 Test for heterogeneity: Q= 120.847 on 5 degrees of freedom (p= 0.000) Meta-analysis of the specificity of cyto (LSIL+)(outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.973 0.971 0.975 969.112 0.000 6 Random | 0.942 0.915 0.969 68.644 0.000 Test for heterogeneity: Q= 784.259 on 5 degrees of freedom (p= 0.000) Meta-analysis of the PPV of cyto (LSIL+)(outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.157 0.141 0.173 19.077 0.000 6 Random | 0.266 0.163 0.370 5.047 0.000 Test for heterogeneity: Q= 177.576 on 5 degrees of freedom (p= 0.000) Meta-analysis of the NPV of cyto (LSIL+)(outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.990 0.989 0.991 1575.625 0.000 6 Random | 0.983 0.977 0.990 298.052 0.000 Test for heterogeneity: Q= 123.395 on 5 degrees of freedom (p= 0.000) 114 LSIL+, Outcome=CIN2+ LSIL+, Outcome=CIN2+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 0 .1 .2 .3 .4 Prevalence of disease Detection rate of disease LSIL+, Outcome=CIN2+ Kolkata 1 Jaipur Mumbai Trivandrum 1 Trivandrum 2 Harare Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN2+) in cytology studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.020 0.018 0.021 23.193 0.000 6 Random | 0.033 0.021 0.045 5.390 0.000 Test for heterogeneity: Q= 224.884 on 5 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN2+) using cytology (LSIL+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.007 0.006 0.008 13.723 0.000 6 Random | 0.016 0.009 0.024 4.375 0.000 Test for heterogeneity: Q= 188.353 on 5 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of cyto (LSIL+)(outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.973 0.971 0.975 978.653 0.000 6 Random | 0.943 0.917 0.970 69.847 0.000 Test for heterogeneity: Q= 776.718 on 5 degrees of freedom (p= 0.000) 115 LSIL+, Outcome=CIN3+ LSIL+, Outcome=CIN3+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity LSIL+, Outcome=CIN3+ LSIL+, Outcome=CIN3+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of cyto (LSIL+)(outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.612 0.565 0.660 25.445 0.000 5 Random | 0.561 0.327 0.796 4.693 0.000 Test for heterogeneity: Q= 94.624 on 4 degrees of freedom (p= 0.000) Meta-analysis of the specificity of cyto (LSIL+)(outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.969 0.967 0.971 883.082 0.000 5 Random | 0.945 0.916 0.975 62.518 0.000 Test for heterogeneity: Q= 678.728 on 4 degrees of freedom (p= 0.000) Meta-analysis of the PPV of cyto (LSIL+)(outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.086 0.072 0.100 12.192 0.000 5 Random | 0.164 0.096 0.232 4.738 0.000 Test for heterogeneity: Q= 75.473 on 4 degrees of freedom (p= 0.000) Meta-analysis of the NPV of cyto (LSIL+)(outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.996 0.995 0.997 2331.996 0.000 5 Random | 0.994 0.991 0.997 637.215 0.000 Test for heterogeneity: Q= 46.902 on 4 degrees of freedom (p= 0.000) 116 LSIL+, Outcome=CIN3+ LSIL+, Outcome=CIN3+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 0 .1 .2 .3 .4 Prevalence of disease Detection rate of disease LSIL+, Outcome=CIN3+ Kolkata 1 Jaipur Mumbai Trivandrum 1 Trivandrum 2 Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN3+) in cytology studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.012 0.011 0.013 17.154 0.000 5 Random | 0.013 0.009 0.016 7.218 0.000 Test for heterogeneity: Q= 25.183 on 4 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN3+) using cytology (LSIL+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.005 0.004 0.006 10.961 0.000 5 Random | 0.007 0.004 0.011 3.928 0.000 Test for heterogeneity: Q= 54.348 on 4 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of cyto (LSIL+)(outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.969 0.967 0.972 888.178 0.000 5 Random | 0.946 0.916 0.975 62.917 0.000 Test for heterogeneity: Q= 678.215 on 4 degrees of freedom (p= 0.000) 117 LSIL+, Outcome=Cancer LSIL+, Outcome=Cancer Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity LSIL+, Outcome=Cancer LSIL+, Outcome=Cancer Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of cyto (LSIL+)(outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.766 0.660 0.871 14.188 0.000 5 Random | 0.748 0.522 0.975 6.470 0.000 Test for heterogeneity: Q= 16.639 on 4 degrees of freedom (p= 0.002) Meta-analysis of the specificity of cyto (LSIL+)(outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.964 0.962 0.967 765.223 0.000 5 Random | 0.931 0.898 0.965 54.470 0.000 Test for heterogeneity: Q= 645.168 on 4 degrees of freedom (p= 0.000) Meta-analysis of the PPV of cyto (LSIL+)(outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.009 0.004 0.014 3.566 0.000 5 Random | 0.024 0.006 0.043 2.552 0.011 Test for heterogeneity: Q= 32.170 on 4 degrees of freedom (p= 0.000) Meta-analysis of the NPV of cyto (LSIL+)(outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.000 0.999 1.000 6088.234 0.000 5 Random | 0.999 0.999 1.000 3063.869 0.000 Test for heterogeneity: Q= 13.033 on 4 degrees of freedom (p= 0.011)) 118 LSIL+, Outcome=Cancer LSIL+, Outcome=Cancer Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .05 .1 .15 .2 0 .05 .1 .15 .2 Prevalence of disease Detection rat of disease LSIL+, Outcome=Cancer Jaipur Mumbai Trivandrum 1 Trivandrum 2 Harare Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (cancer) in cytology studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.001 0.001 0.002 5.359 0.000 5 Random | 0.002 0.001 0.003 3.113 0.002 Test for heterogeneity: Q= 18.984 on 4 degrees of freedom (p= 0.001) Meta-analysis of the detection rate of disease (cancer) using cytology (LSIL+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.000 0.000 0.001 2.217 0.027 5 Random | 0.002 0.000 0.003 2.288 0.022 Test for heterogeneity: Q= 31.676 on 4 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of cyto (LSIL+)(outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.964 0.962 0.967 766.239 0.000 5 Random | 0.931 0.897 0.965 54.143 0.000 Test for heterogeneity: Q= 654.200 on 4 degrees of freedom (p= 0.000) 119 HSIL+ Kolkata 1 Jaipur Mumbai Trivandrum 1 Trivandrum 2 Harare Combined 0 .05 .1 .15 .2 Test positivity rate Meta-analysis of the test-positivity rate of cytology at cut-off HSIL+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.012 0.011 0.013 17.961 0.000 6 Random | 0.020 0.011 0.029 4.453 0.000 Test for heterogeneity: Q= 175.116 on 5 degrees of freedom (p= 0.000) 120 HSIL, Outcome=CIN2+ HSIL, Outcome=CIN2+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity HSIL, Outcome=CIN2+ HSIL, Outcome=CIN2+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of cyto (HSIL+)(outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.313 0.281 0.344 19.585 0.000 6 Random | 0.381 0.224 0.537 4.766 0.000 Test for heterogeneity: Q= 115.023 on 5 degrees of freedom (p= 0.000) Meta-analysis of the specificity of cyto (HSIL+)(outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.995 0.994 0.996 2286.022 0.000 6 Random | 0.991 0.986 0.995 450.614 0.000 Test for heterogeneity: Q= 109.507 on 5 degrees of freedom (p= 0.000) Meta-analysis of the PPV of cyto (HSIL+)(outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.539 0.500 0.578 26.905 0.000 6 Random | 0.561 0.354 0.768 5.309 0.000 Test for heterogeneity: Q= 128.581 on 5 degrees of freedom (p= 0.000) Meta-analysis of the NPV of cyto (HSIL+)(outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.989 0.987 0.990 1507.454 0.000 6 Random | 0.978 0.970 0.987 228.522 0.000 Test for heterogeneity: Q= 189.425 on 5 degrees of freedom (p= 0.000) 121 HSIL, Outcome=CIN2+ HSIL, Outcome=CIN2+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .05 .1 .15 .2 0 .05 .1 .15 .2 Prevalence of disease Detection rate of disease HSIL, Outcome=CIN2+ Kolkata 1 Jaipur Mumbai Trivandrum 1 Trivandrum 2 Harare Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN2+) in cytology studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.020 0.018 0.021 23.193 0.000 6 Random | 0.033 0.021 0.045 5.390 0.000 Test for heterogeneity: Q= 224.884 on 5 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN2+) using cytology (HSIL+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.005 0.005 0.006 12.044 0.000 6 Random | 0.011 0.005 0.016 4.002 0.000 Test for heterogeneity: Q= 116.210 on 5 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of cyto (HSIL+)(outcome CIN2+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.995 0.994 0.996 2313.156 0.000 6 Random | 0.991 0.987 0.995 461.447 0.000 Test for heterogeneity: Q= 107.786 on 5 degrees of freedom (p= 0.000) 122 HSIL, Outcome=CIN3+ HSIL, Outcome=CIN3+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity HSIL, Outcome=CIN3+ HSIL, Outcome=CIN3+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of cyto (HSIL+)(outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.507 0.456 0.558 19.604 0.000 5 Random | 0.516 0.320 0.711 5.164 0.000 Test for heterogeneity: Q= 58.826 on 4 degrees of freedom (p= 0.000) Meta-analysis of the specificity of cyto (HSIL+)(outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.994 0.993 0.995 1925.546 0.000 5 Random | 0.990 0.984 0.995 343.901 0.000 Test for heterogeneity: Q= 105.361 on 4 degrees of freedom (p= 0.000) Meta-analysis of the PPV of cyto (HSIL+)(outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.354 0.310 0.397 15.996 0.000 5 Random | 0.427 0.260 0.594 5.009 0.000 Test for heterogeneity: Q= 49.658 on 4 degrees of freedom (p= 0.000) Meta-analysis of the NPV of cyto (HSIL+)(outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.995 0.994 0.996 2116.522 0.000 5 Random | 0.994 0.991 0.996 776.478 0.000 Test for heterogeneity: Q= 28.198 on 4 degrees of freedom (p= 0.000) 123 HSIL, Outcome=CIN3+ HSIL, Outcome=CIN3+ Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined 0 .05 .1 .15 .2 0 .05 .1 .15 .2 Prevalence of disease Detection rate of disease HSIL, Outcome=CIN3+ Kolkata 1 Jaipur Mumbai Trivandrum 1 Trivandrum 2 Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (CIN3+) in cytology studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.012 0.011 0.013 17.381 0.000 5 Random | 0.013 0.010 0.017 7.335 0.000 Test for heterogeneity: Q= 24.887 on 4 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (CIN3+) using cytology (HSIL+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.004 0.004 0.005 10.622 0.000 5 Random | 0.007 0.003 0.010 3.895 0.000 Test for heterogeneity: Q= 53.359 on 4 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of cyto (HSIL+)(outcome CIN3+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.994 0.993 0.995 1936.774 0.000 5 Random | 0.990 0.984 0.995 346.288 0.000 Test for heterogeneity: Q= 105.136 on 4 degrees of freedom (p= 0.000) 124 HSIL, Outcome=Cancer HSIL, Outcome=Cancer Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Specificity HSIL, Outcome=Cancer HSIL, Outcome=Cancer Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV NPV Meta-analysis of the sensitivity of cyto (HSIL+)(outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.681 0.593 0.770 15.109 0.000 6 Random | 0.698 0.491 0.904 6.617 0.000 Test for heterogeneity: Q= 24.699 on 5 degrees of freedom (p= 0.000) Meta-analysis of the specificity of cyto (HSIL+)(outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.989 0.988 0.990 1560.807 0.000 6 Random | 0.982 0.975 0.990 250.472 0.000 Test for heterogeneity: Q= 155.563 on 5 degrees of freedom (p= 0.000) Meta-analysis of the PPV of cyto (HSIL+)(outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.082 0.059 0.106 6.793 0.000 6 Random | 0.104 0.049 0.158 3.745 0.000 Test for heterogeneity: Q= 22.918 on 5 degrees of freedom (p= 0.000) Meta-analysis of the NPV of cyto (HSIL+)(outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.000 0.999 1.000 6705.381 0.000 6 Random | 0.999 0.998 1.000 2590.917 0.000 Test for heterogeneity: Q= 25.558 on 5 degrees of freedom (p= 0.000) 125 HSIL, Outcome=Cancer HSIL, Outcome=Cancer Kolkata 1 Kolkata 1 Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined 0 .025 .05 .075 .1 0 .025 .05 .075 .1 Prevalence of disease Detection rate of disease HSIL, Outcome=Cancer Kolkata 1 Jaipur Mumbai Trivandrum 1 Trivandrum 2 Harare Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Approximative specificity Meta-analysis of the prevalence of disease (cancer) in cytology studies | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.002 0.001 0.002 7.002 0.000 6 Random | 0.003 0.001 0.005 3.595 0.000 Test for heterogeneity: Q= 42.456 on 5 degrees of freedom (p= 0.000) Meta-analysis of the detection rate of disease (cancer) using cytology (HSIL+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.001 0.000 0.001 4.070 0.000 6 Random | 0.002 0.001 0.003 3.024 0.002 Test for heterogeneity: Q= 40.614 on 5 degrees of freedom (p= 0.000) Meta-analysis of the approximate specificity of cyto (HSIL+)(outcome cancer) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.989 0.988 0.990 1562.420 0.000 6 Random | 0.982 0.975 0.990 250.901 0.000 Test for heterogeneity: Q= 155.405 on 5 degrees of freedom (p= 0.000) 126 6.1.6. Summary of the meta-analyses of the accuracy for CIN1+ Table 25. Summary of the meta-analyses of screening tests considering CIN1+ as outcome: pooled estimates and confidence intervals, model of pooling (F=fixed; R=random effect model), range, number of study locations and identification of studies include (studies reporting lowest and highest parameters are respectively single and double underlined). Test VIA Outcome CIN1+ Included study Parameter Estimate 95% CI Model Range Number locations Ba,Br,Co,Ja,Ko1,Ko2, Sensitivity 61.8 (53.2 - 70.3) R 42.5 - 90.0 12 Mu,Ni,Ou,Tr1,Tr2,Ha Ba,Br,Co,Ja,Ko1,Ko2, Specificity 84.9 (81.2 - 88.7) R 67.9 - 95.1 12 Mu,Ni,Ou,Tr1,Tr2,Ha Ba,Br,Co,Ja,Ko1,Ko2, Mu,Ni,Ou,Tr1,Tr2,Ha, PPV 30.6 (21.1 - 40.0) R 13.0 - 62.8 14 Am,Bu Ba,Br,Co,Ja,Ko1,Ko2, NPV 95.7 (94.5 - 96.9) R 83.1 - 99.5 12 Mu,Ni,Ou,Tr1,Tr2,Ha Ba,Br,Co,Ja,Ko1,Ko2, Mu,Ni,Ou,Tr1,Tr2,Ha, Am,Bu,Ch,Ka,Pa,Ph, Test+ 16.2 (13.4 - 19.1) R 6.0 - 39.9 20 Ac,Am Test VILI Outcome CIN1+ Included study Parameter Estimate 95% CI Model Range Number locations Ba,Br,Co,Ja,Ko2,Mu, Sensitivity 73.7 (63.0 - 84.5) R 50.3 - 94.1 10 Ni,Ou,Tr1,Tr2 Ba,Br,Co,Ja,Ko2,Mu, Specificity 86.6 (83.4 - 89.8) R 74.1 - 92.8 10 Ni,Ou,Tr1,Tr2 Ba,Br,Co,Ja,Ko2,Mu, PPV 28.9 (19.9 - 37.9) R 13.3 - 50.8 10 Ni,Ou,Tr1,Tr2 Ba,Br,Co,Ja,Ko2,Mu, NPV 97.5 (96.6 - 98.4) R 90.1 - 99.9 10 Ni,Ou,Tr1,Tr2 Ba,Br,Co,Ja,Ko2,Mu, Test+ 17.2 (13.9 - 20.5) R 13.1 - 28.7 11 Ni,Ou,Tr1,Tr2,Bu Test VIAM Outcome CIN1+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 58.5 (43.2 - 73.9) R 42.5 - 68.4 3 Ko1,Ko2,Mu Specificity 88.1 (85.8 - 90.4) R 86.4 - 90.1 3 Ko1,Ko2,Mu PPV 23.1 (13.0 - 33.1) R 14.9 - 31.1 3 Ko1,Ko2,Mu NPV 97.3 (96.4 - 98.2) R 96.3 - 98.1 3 Ko1,Ko2,Mu Test+ 14.8 (10.5 - 19.0) R 11.2 - 18.0 3 Ko1,Ko2,Mu Test HPV Outcome CIN1+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 34.5 (16.4 - 52.6) R 21.5 - 63.8 5 Ko1,Ko2,Mu,Tr2,Ha Specificity 88.3 (83.9 - 92.8) R 64.8 - 95.1 5 Ko1,Ko2,Mu,Tr2,Ha PPV 25.2 (15.5 - 34.8) R 16.4 - 39.4 5 Ko1,Ko2,Mu,Tr2,Ha NPV 93.2 (90.87 - 95.5) R 83.30 - 96.7 5 Ko1,Ko2,Mu,Tr2,Ha Test+ 14.3 (7.9 - 20.6) R 6.0 - 42.8 5 Ko1,Ko2,Mu,Tr2,Ha Test Cytology Outcome CIN1+ Threshold ASCUS+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 36.3 (20.8 - 51.9) R 23.0 - 65.5 6 Ja,Ko1,Mu,Tr1,Tr2,Ha Specificity 92.1 (88.7 - 95.5) R 79.0 - 98.7 6 Ja,Ko1,Mu,Tr1,Tr2,Ha PPV 43.6 (21.4 - 65.8) R 14.5 - 72.7 6 Ja,Ko1,Mu,Tr1,Tr2,Ha NPV 92.2 (89.6 - 94.8) R 80.5 - 97.0 6 Ja,Ko1,Mu,Tr1,Tr2,Ha Test+ 11.6 (6.7 - 16.4) R 2.8 - 27.8 6 Ja,Ko1,Mu,Tr1,Tr2,Ha 127 Test Cytology Outcome CIN1+ Threshold LSIL+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 30.3 (14.8 - 45.7) R 17.2 - 63.3 6 Ja,Ko1,Mu,Tr1,Tr2,Ha Specificity 96.1 (94.3 - 97.9) R 92.9 - 99.3 6 Ja,Ko1,Mu,Tr1,Tr2,Ha PPV 50.7 (27.9 - 73.5) R 19.4 - 74.1 6 Ja,Ko1,Mu,Tr1,Tr2,Ha NPV 91.7 (88.8 - 94.6) R 78.4 - 96.9 6 Ja,Ko1,Mu,Tr1,Tr2,Ha Test+ 7.3 (4.2 - 10.3) R 2.1 - 13.7 6 Ja,Ko1,Mu,Tr1,Tr2,Ha 128 6.1.7. Summary of the meta-analyses of the accuracy for CIN2+ Table 26. Summary of the meta-analyses of screening tests considering CIN2+ as outcome: pooled estimates and confidence intervals, model of pooling (F=fixed; R=random effect model), range, number of study locations and identification of studies include (studies reporting lowest and highest parameters are respectively single and double underlined). Test VIA Outcome CIN2+ Included study Parameter Estimate 95% CI Model Range Number locations Ba,Br,Co,Ja,Ko1,Ko2,Mu, Sensitivity 79.0 (73.6 - 84.3) R 61.5 - 91.1 12 Ni,Ou,Tr1,Tr2,Ha Ba,Br,Co,Ja,Ko1,Ko2,Mu, Specificity 83.0 (78.8 - 87.2) R 64.1 - 94.5 12 Ni,Ou,Tr1,Tr2,Ha Ba,Br,Co,Ja,Ko1,Ko2,Mu, PPV 11.8 (9.1 - 14.6) R 3.8 - 22.5 14 Ni,Ou,Tr1,Tr2,Ha,Am,Bu Ba,Br,Co,Ja,Ko1,Ko2,Mu, NPV 99.3 (99.1 - 99.6) R 96.3 - 99.8 12 Ni,Ou,Tr1,Tr2,Ha Test VILI Outcome CIN2+ Included study Parameter Estimate 95% CI Model Range Number locations Ba,Br,Co,Ja,Ko1,Ko2,Mu, Sensitivity 91.2 (87.8 - 94.6) R 74.1 - 98.0 10 Ni,Ou,Tr1,Tr2 Ba,Br,Co,Ja,Ko1,Ko2,Mu, Specificity 84.5 (81.3 - 87.8) R 73.0 - 91.6 10 Ni,Ou,Tr1,Tr2 Ba,Br,Co,Ja,Ko1,Ko2,Mu, PPV 12.9 (8.0 - 17.9) R 3.5 - 33.9 10 Ni,Ou,Tr1,Tr2 Ba,Br,Co,Ja,Ko1,Ko2,Mu, NPV 99.8 (99.7 - 99.9) R 99.4 - 99.9 10 Ni,Ou,Tr1,Tr2 Test VIAM Outcome CIN2+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 67.0 (61.8 - 72.2) F 64.6 - 73.2 3 Ko1,Ko2,Mu Specificity 86.2 (82.4 - 90.0) R 83.3 - 89.3 3 Ko1,Ko2,Mu PPV 8.0 (5.4 - 10.7) R 5.7 - 9.8 3 Ko1,Ko2,Mu NPV 99.3 (98.7 - 99.8) R 99.2 - 99.8 3 Ko1,Ko2,Mu Test HPV Outcome CIN2+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 65.8 (54.5 - 77.0) R 48.4 - 80.2 5 Ko1,Ko2,Mu,Tr2,Ha Specificity 87.3 (81.9 - 92.6) R 61.2 - 94.6 5 Ko1,Ko2,Mu,Tr2,Ha PPV 14.1 (10.2 - 18.0) R 9.2 - 18.1 5 Ko1,Ko2,Mu,Tr2,Ha NPV 99.1 (98.6 - 99.5) R 96.7 - 99.7 5 Ko1,Ko2,Mu,Tr2,Ha Test Cytology Outcome CIN2+ Threshold ASCUS+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 58.4 (43.5 - 73.2) R 33.3 - 81.9 6 Ja,Ko1,Mu,Tr1,Tr2,Ha Specificity 90.0 (85.7 - 94.4) R 76.2 - 98.5 6 Ja,Ko1,Mu,Tr1,Tr2,Ha PPV 21.2 (13.1 - 29.3) R 5.2 - 48.2 6 Ja,Ko1,Mu,Tr1,Tr2,Ha NPV 98.6 (98.1 - 99.2) R 95.3 - 99.3 6 Ja,Ko1,Mu,Tr1,Tr2,Ha Test Cytology Outcome CIN2+ Threshold LSIL+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 50.1 (33.2 - 66.9) R 23.8 - 77.9 6 Ja,Ko1,Ja,Mu,Tr1,Tr2,Ha Specificity 94.2 (91.5 - 96.9) R 88.6 - 99.1 6 Ja,Ko1,Ja,Mu,Tr1,Tr2,Ha PPV 26.6 (16.3 - 37.0) R 5.9 - 57.3 6 Ja,Ko1,Ja,Mu,Tr1,Tr2,Ha NPV 98.3 (97.7 - 99.0) R 93.9 - 99.3 6 Ja,Ko1,Ja,Mu,Tr1,Tr2,Ha 129 Test Cytology Outcome CIN2+ Threshold HSIL+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 38.1 (22.4 - 53.7) R 16.4 - 61.7 6 Ja,Ko1,Mu,Tr1,Tr2,Ha Specificity 99.1 (98.6 - 99.5) R 97.7 - 99.7 6 Ja,Ko1,Mu,Tr1,Tr2,Ha PPV 56.1 (35.4 - 76.8) R 26.7 - 79.4 6 Ja,Ko1,Mu,Tr1,Tr2,Ha NPV 97.8 (97.0 - 98.7) R 91.7 - 99.3 6 Ja,Ko1,Mu,Tr1,Tr2,Ha 130 6.1.8. Summary of the meta-analyses of the accuracy for CIN3+ Table 27. Summary of the meta-analyses of screening tests considering CIN3+ as outcome: pooled estimates and confidence intervals, model of pooling (F=fixed; R=random effect model), range, number of study locations and identification of studies include (studies reporting lowest and highest parameters are respectively single and double underlined). Test VIA Outcome CIN3+ Included study Parameter Estimate 95% CI Model Range Number locations Ba,Br,Co,Ja,Ko1,Ko2,Mu, Sensitivity 82.9 (77.1 - 88.7) R 58.3 - 94.6 11 Ni,Ou,Tr1,Tr2 Ba,Br,Co,Ja,Ko1,Ko2,Mu, Specificity 84.2 (80.0 - 88.3) R 73.8 - 94.3 11 Ni,Ou,Tr1,Tr2 Ba,Br,Co,Ja,Ko1,Ko2,Mu, PPV 6.8 (5.2 - 8.5) R 2.7 - 16.6 13 Ni,Ou,Tr1,Tr2,Am,Bu Ba,Br,Co,Ja,Ko1,Ko2,Mu, NPV 99.7 (99.7 - 99.8) R 99.3 - 99.9 11 Ni,Ou,Tr1,Tr2 Test VILI Outcome CIN3+ Included study Parameter Estimate 95% CI Model Range Number locations Ba,Br,Co,Ja,Ko2,Mu, Sensitivity 93.8 (90.6 - 97.1) R 72.9 - 100.0 10 Ni,Ou,Tr1,Tr2 Ba,Br,Co,Ja,Ko2,Mu, Specificity 83.8 (80.5 - 87.1) R 72.6 - 91.4 10 Ni,Ou,Tr1,Tr2 Ba,Br,Co,Ja,Ko2,Mu, PPV 7.7 (5.0 - 10.3) R 2.4 - 17.6 10 Ni,Ou,Tr1,Tr2 Ba,Br,Co,Ja,Ko2,Mu, NPV 99.9 (99.9 - 100.0) R 99.5 - 100.0 10 Ni,Ou,Tr1,Tr2 Test VIAM Outcome CIN3+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 68.2 (61.8 - 74.7) F 65.7 - 74.4 3 Ko1,Ko2,Mu Specificity 85.9 (82.0 - 89.8) R 82.8 - 89.1 3 Ko1,Ko2,Mu PPV 5.2 (3.0 - 7.5) R 3.2 - 6.6 3 Ko1,Ko2,Mu NPV 99.5 (99.2 - 99.9) R 99.3 - 99.9 3 Ko1,Ko2,Mu Test HPV Outcome CIN3+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 68.4 (61.5 - 75.4) F 62.3 - 73.5 4 Ko1,Ko2,Mu,Tr2 Specificity 93.4 (92.2 - 94.6) R 91.4 - 94.4 4 Ko1,Ko2,Mu,Tr2 PPV 9.1 (5.5 - 12.7) R 5.8 - 13.1 4 Ko1,Ko2,Mu,Tr2 NPV 99.7 (99.5 - 99.9) R 99.5 - 99.9 4 Ko1,Ko2,Mu,Tr2 Test Cytology Outcome CIN3+ Threshold ASCUS+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 63.0 (37.9 - 88.2) R 35.6 - 96.4 5 Ja,Ko1,Mu,Tr1,Tr2 Specificity 92.3 (88.1 - 96.6) R 86.3 - 98.2 5 Ja,Ko1,Mu,Tr1,Tr2 PPV 13.6 (8.1 - 19.0) R 3.9 - 37.3 5 Ja,Ko1,Mu,Tr1,Tr2 NPV 99.5 (99.2 - 99.8) R 98.9 - 99.9 5 Ja,Ko1,Mu,Tr1,Tr2 Test Cytology Outcome CIN3+ Threshold LSIL+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 56.1 (32.7 - 79.6) R 26.7 - 89.3 5 Ja,Ko1,Mu,Tr1,Tr2 Specificity 94.5 (91.6 - 97.5) R 87.3 - 98.8 5 Ja,Ko1,Mu,Tr1,Tr2 PPV 16.4 (9.6 - 23.2) R 4.7 - 42.7 5 Ja,Ko1,Mu,Tr1,Tr2 NPV 99.4 (99.1 - 99.7) R 98.8 - 99.8 5 Ja,Ko1,Mu,Tr1,Tr2 131 Test Cytology Outcome CIN3+ Threshold HSIL+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 51.6 (32.0 - 71.1) R 22.2 - 76.8 5 Ja,Ko1,Mu,Tr1,Tr2 Specificity 99.0 (98.4 - 99.5) R 97.5 - 99.7 5 Ja,Ko1,Mu,Tr1,Tr2 PPV 42.7 (26.0 - 59.4) R 20.6 - 57.4 5 Ja,Ko1,Mu,Tr1,Tr2 NPV 99.4 (99.1 - 99.6) R 98.8 - 99.7 5 Ja,Ko1,Mu,Tr1,Tr2 132 6.1.9. Summary of the meta-analyses of the accuracy for cancer Table 28. Summary of the meta-analyses of screening tests considering cervical cancer as outcome: pooled estimates and confidence intervals, model of pooling (F=fixed; R=random effect model), range, number of study locations and identification of studies include (studies reporting lowest and highest parameters are respectively single and double underlined). Test VIA Outcome Cancer Included study Parameter Estimate 95% CI Model Range Number locations Ba,Br,Co,Ja,Ko1,Ko2,Mu, Sensitivity 89.1 (83.7 - 94.5) R 60.0 - 100.0 12 Ni,Ou,Tr1,Tr2,Ha Ba,Br,Co,Ja,Ko1,Ko2,Mu, Specificity 81.7 (77.1 - 86.3) R 60.2 - 94.1 12 Ni,Ou,Tr1,Tr2,Ha Ba,Br,Co,Ja,Ko1,Ko2,Mu, PPV 2.5 (1.7 - 3.5) R 0.3 - 12.4 14 Ni,Ou,Tr1,Tr2,Ha,Am,Bu Ba,Br,Co,Ja,Ko1,Ko2,Mu, NPV 99.9 (99.9 - 100.0) R 99.9 - 100.0 12 Ni,Ou,Tr1,Tr2,Ha Test VILI Outcome Cancer Included study Parameter Estimate 95% CI Model Range Number locations Ba,Br,Co,Ja,Ko2,Mu, Sensitivity 95.7 (91.8 - 99.7) R 66.7 - 100.0 10 Ni,Ou,Tr1,Tr2 Ba,Br,Co,Ja,Ko2,Mu, Specificity 83.2 (79.8 - 86.5) R 71.9 - 91.1 10 Ni,Ou,Tr1,Tr2 Ba,Br,Co,Ja,Ko2,Mu, PPV 2.7 (1.7 - 3.7) R 0.3 - 9.0 10 Ni,Ou,Tr1,Tr2 Ba,Br,Co,Ja,Ko2,Mu, NPV 100.0 (100.0 - 100.0) F 100.0 - 100.0 10 Ni,Ou,Tr1,Tr2 Test VIAM Outcome Cancer Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 81.4 (72.3 - 90.5) F 74.1 - 100.0 3 Ko1,Ko2,Mu Specificity 85.5 (81.5 - 89.6) R 82.4 - 88.9 3 Ko1,Ko2,Mu PPV 2.2 (0.6 - 3.8) R 0.9 - 3.3 3 Ko1,Ko2,Mu NPV 99.9 (99.7 - 100.0) R 99.8 - 100.0 3 Ko,Ko2,Mu Test HPV Outcome Cancer Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 71.8 (60.3 - 83.3) F 61.5 - 85.7 5 Ko1,Ko2,Mu,Tr2,Ha Specificity 85.9 (79.7 - 92.2) R 57.3 - 94.0 5 Ko1,Ko2,Mu,Tr2,Ha PPV 2.2 (0.7 - 3.8) R 0.2 - 5.2 5 Ko1,Ko2,Mu,Tr2,Ha NPV 99.9 (99.8 - 100.0) R 99.8 - 100.0 5 Ko1,Ko2,Mu,Tr2,Ha Test Cytology Outcome Cancer Threshold ASCUS+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 75.8 (59.1 - 92.4) R 40.0 - 100.0 6 Ja,Ko1,Mu,Tr1,Tr2,Ha Specificity 88.6 (83.8 - 93.5) R 72.3 - 97.7 6 Ja,Ko1,Mu,Tr1,Tr2,Ha PPV 2.0 (0.7 - 3.3) R 0.5 - 17.3 6 Ja,Ko1,Mu,Tr1,Tr2,Ha NPV 99.9 (99.8 - 100.0) R 99.6 - 100.0 6 Ja,Ko1,Mu,Tr1,Tr2,Ha 133 Test Cytology Outcome Cancer Threshold LSIL+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 74.8 (52.2 - 97.5) R 20.0 - 100.0 6 Ja,Ko1,Mu,Tr1,Tr2,Ha Specificity 93.1 (89.8 - 96.5) R 86.5 - 98.3 6 Ja,Ko1,Mu,Tr1,Tr2,Ha PPV 2.4 (0.6 - 4.3) R 0.4 - 20.7 6 Ja,Ko1,Mu,Tr1,Tr2,Ha NPV 99.9 (99.9 - 100.0) R 99.6 - 100.0 6 Ja,Ko1,Mu,Tr1,Tr2,Ha Test Cytology Outcome Cancer Threshold HSIL+ Included study Parameter Estimate 95% CI Model Range Number locations Sensitivity 69.8 (49.1 - 90.4) R 20.0 - 100.0 6 Ja,Ko1,Mu,Tr1,Tr2,Ha Specificity 98.2 (97.5 - 99.0) R 96.6 - 99.5 6 Ja,Ko1,Mu,Tr1,Tr2,Ha PPV 10.4 (4.9 - 15.8) R 3.6 - 28.3 6 Ja,Ko1,Mu,Tr1,Tr2,Ha NPV 99.9 (99.8 - 100.0) R 99.6 - 100.0 6 Ja,Ko1,Mu,Tr1,Tr2,Ha 134 6.2. Meta-analyses of the relative sensitivity and specificity of cervical cancer screening tests We will pool the ratio of the sensitivity for a given degree of cervical intraepithelial neoplasia of a screen test over that of another test. The same will be done for the ratio of specificities. Below follow a series of 18 possible comparisons of screen tests. 6.2.1. Relative sensitivity and specificity of VILI compared to VIA Outcome CIN1+ Outcome CIN2+ VILI/VIA VILI/VIA Bamako Bamako Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Kolkata 2 Kolkata 2 Mumbai Mumbai Niamey Niamey Ouagadougou Ouagadougou Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .5 .67 1 1.5 2 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Outcome CIN3+ Outcome Cancer VILI/VIA VILI/VIA Bamako Bamako Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Kolkata 2 Kolkata 2 Mumbai Mumbai Niamey Niamey Ouagadougou Ouagadougou Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .5 .67 1 1.5 2 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Meta-analysis of the sensitivity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.143 1.110 1.177 8.931 0.000 10 Random | 1.185 1.082 1.297 3.675 0.000 Test for heterogeneity: Q= 73.540 on 9 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.112 1.081 1.144 7.415 0.000 10 Random | 1.105 1.048 1.165 3.709 0.000 Test for heterogeneity: Q= 23.790 on 9 degrees of freedom (p= 0.005) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.074 1.043 1.106 4.793 0.000 10 Random | 1.074 1.043 1.106 4.793 0.000 Test for heterogeneity: Q= 7.123 on 9 degrees of freedom (p= 0.624) 135 Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.013 0.989 1.036 1.060 0.289 10 Random | 1.013 0.989 1.036 1.060 0.289 Test for heterogeneity: Q= 6.952 on 9 degrees of freedom (p= 0.642) 136 Specificity ratio (VILI/VIA) Outcome CIN1+ Outcome CIN2+ VILI/VIA VILI/VIA Bamako Bamako Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Kolkata 2 Kolkata 2 Mumbai Mumbai Niamey Niamey Ouagadougou Ouagadougou Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome CIN3+ Outcome Cancer VILI/VIA VILI/VIA Bamako Bamako Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Kolkata 2 Kolkata 2 Mumbai Mumbai Niamey Niamey Ouagadougou Ouagadougou Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Meta-analysis of the specificity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.994 0.990 0.998 -2.655 0.008 10 Random | 1.001 0.967 1.036 0.032 0.975 Test for heterogeneity: Q= 546.787 on 9 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.988 0.983 0.992 -5.253 0.000 10 Random | 0.996 0.962 1.031 -0.229 0.819 Test for heterogeneity: Q= 485.280 on 9 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.986 0.981 0.990 -6.087 0.000 10 Random | 0.994 0.962 1.027 -0.342 0.732 Test for heterogeneity: Q= 410.871 on 9 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.984 0.979 0.989 -6.609 0.000 10 Random | 0.993 0.962 1.025 -0.410 0.682 Test for heterogeneity: Q= 372.574 on 9 degrees of freedom (p= 0.000) 137 6.2.2. Relative sensitivity and specificity of VIAM compared to VIA Sensitivity ratio (VIAM/VIA) Outcome CIN1+ Outcome CIN2+ VIAM/VIA VIAM/VIA Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined .5 .67 1 1.5 2 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Outcome CIN3+ Outcome Cancer VIAM/VIA VIAM/VIA Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined .5 .67 1 1.5 2 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Meta-analysis of the sensitivity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.050 0.977 1.127 1.327 0.184 3 Random | 1.050 0.977 1.127 1.327 0.184 Test for heterogeneity: Q= 0.606 on 2 degrees of freedom (p= 0.739) Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.036 0.926 1.159 0.622 0.534 3 Random | 1.036 0.926 1.159 0.622 0.534 Test for heterogeneity: Q= 0.185 on 2 degrees of freedom (p= 0.912) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.028 0.897 1.180 0.401 0.688 3 Random | 1.028 0.897 1.180 0.401 0.688 Test for heterogeneity: Q= 0.196 on 2 degrees of freedom (p= 0.906) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.999 0.966 1.034 -0.039 0.969 3 Random | 0.999 0.966 1.034 -0.039 0.969 Test for heterogeneity: Q= 0.074 on 2 degrees of freedom (p= 0.964) 138 Specificity ratio (VIAM/VIA) Outcome CIN1+ Outcome CIN2+ VIAM/VIA VIAM/VIA Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome CIN3+ Outcome Cancer VIAM/VIA VIAM/VIA Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Meta-analysis of the specificity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.999 0.991 1.007 -0.238 0.812 3 Random | 0.998 0.979 1.018 -0.179 0.858 Test for heterogeneity: Q= 11.544 on 2 degrees of freedom (p= 0.003) Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.998 0.990 1.006 -0.498 0.619 3 Random | 0.997 0.978 1.016 -0.330 0.742 Test for heterogeneity: Q= 10.303 on 2 degrees of freedom (p= 0.006) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.998 0.990 1.006 -0.544 0.586 3 Random | 0.997 0.978 1.016 -0.363 0.716 Test for heterogeneity: Q= 9.703 on 2 degrees of freedom (p= 0.008) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.998 0.989 1.006 -0.601 0.548 3 Random | 0.996 0.977 1.015 -0.399 0.690 Test for heterogeneity: Q= 9.753 on 2 degrees of freedom (p= 0.008) 139 6.2.3. Relative sensitivity and specificity of HC2 compared to VIA Sensitivity ratio (HPV/VIA) Outcome CIN1+ Outcome CIN2+ HPV/VIA HPV/VIA Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined .3 .5 .67 1 1.5 2 .3 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Outcome CIN3+ Outcome Cancer HPV/VIA HPV/VIA Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Combined Combined .3 .5 .67 1 1.5 2 .3 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Meta-analysis of the sensitivity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.764 0.712 0.820 -7.509 0.000 5 Random | 0.569 0.363 0.890 -2.468 0.014 Test for heterogeneity: Q= 119.271 on 4 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.971 0.900 1.047 -0.768 0.442 5 Random | 0.933 0.820 1.061 -1.062 0.288 Test for heterogeneity: Q= 8.553 on 4 degrees of freedom (p= 0.073) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.940 0.825 1.071 -0.928 0.353 4 Random | 0.956 0.781 1.169 -0.440 0.660 Test for heterogeneity: Q= 7.019 on 3 degrees of freedom (p= 0.071) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.850 0.711 1.016 -1.786 0.074 5 Random | 0.850 0.711 1.016 -1.786 0.074 Test for heterogeneity: Q= 1.500 on 4 degrees of freedom (p= 0.827) 140 Specificity ratio (HPV/VIA) Outcome CIN1+ Outcome CIN2+ HPV/VIA HPV/VIA Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome CIN3+ Outcome Cancer HPV/VIA HPV/VIA Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Meta-analysis of the specificity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.049 1.043 1.055 16.206 0.000 5 Random | 1.043 1.022 1.064 3.999 0.000 Test for heterogeneity: Q= 40.864 on 4 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.067 1.061 1.074 20.933 0.000 5 Random | 1.059 1.033 1.085 4.580 0.000 Test for heterogeneity: Q= 53.891 on 4 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.070 1.063 1.076 21.271 0.000 4 Random | 1.075 1.051 1.099 6.369 0.000 Test for heterogeneity: Q= 36.219 on 3 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.068 1.062 1.075 20.585 0.000 5 Random | 1.059 1.031 1.087 4.249 0.000 Test for heterogeneity: Q= 58.963 on 4 degrees of freedom (p= 0.000) 141 6.2.4. Relative sensitivity and specificity of cervical cytology (at ASCUS+) compared to VIA Sensitivity ratio (Cytology at ASCUS+/VIA) Outcome CIN1+ Outcome CIN2+ Cytology at ASCUS+/VIA Cytology at ASCUS+/VIA Jaipur Jaipur Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined .3 .5 .67 1 1.5 2 .3 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Outcome CIN3+ Outcome Cancer Cytology at ASCUS+/VIA Cytology at ASCUS+/VIA Jaipur Jaipur Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Combined Combined .3 .5 .67 1 1.5 2 .3 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Meta-analysis of the sensitivity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.748 0.707 0.791 -10.217 0.000 6 Random | 0.531 0.338 0.834 -2.748 0.006 Test for heterogeneity: Q= 280.537 on 5 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.849 0.796 0.905 -5.003 0.000 6 Random | 0.770 0.640 0.926 -2.776 0.006 Test for heterogeneity: Q= 32.518 on 5 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.950 0.887 1.017 -1.472 0.141 5 Random | 0.795 0.613 1.032 -1.726 0.084 Test for heterogeneity: Q= 29.929 on 4 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.993 0.954 1.034 -0.327 0.744 6 Random | 0.988 0.935 1.044 -0.439 0.661 Test for heterogeneity: Q= 5.622 on 5 degrees of freedom (p= 0.345) 142 Specificity ratio (Cytology at ASCUS+/VIA) Outcome CIN1+ Outcome CIN2+ Cytology at ASCUS+/VIA Cytology at ASCUS+/VIA Jaipur Jaipur Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome CIN3+ Outcome Cancer Cytology at ASCUS+/VIA Cytology at ASCUS+/VIA Jaipur Jaipur Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Meta-analysis of the specificity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.101 1.095 1.107 33.104 0.000 6 Random | 1.119 1.066 1.175 4.517 0.000 Test for heterogeneity: Q= 332.786 on 5 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.120 1.113 1.127 35.682 0.000 6 Random | 1.134 1.085 1.186 5.589 0.000 Test for heterogeneity: Q= 230.268 on 5 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.120 1.113 1.127 34.743 0.000 5 Random | 1.127 1.073 1.183 4.817 0.000 Test for heterogeneity: Q= 220.608 on 4 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.123 1.116 1.131 35.148 0.000 6 Random | 1.139 1.087 1.192 5.525 0.000 Test for heterogeneity: Q= 231.980 on 5 degrees of freedom (p= 0.000) 143 6.2.5. Relative sensitivity and specificity of cervical cytology (at LSIL+) compared to VIA Sensitivity ratio (Cytology at LSIL+/VIA) Outcome CIN1+ Outcome CIN2+ Cytology at LSIL+/VIA Cytology at LSIL+/VIA Jaipur Jaipur Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined .25 .5 .67 1 1.5 2 .25 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Outcome CIN3+ Outcome Cancer Cytology at LSIL+/VIA Cytology at LSIL+/VIA Jaipur Jaipur Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Combined Combined .25 .5 .67 1 1.5 2 .25 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Meta-analysis of the sensitivity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.661 0.621 0.703 -13.152 0.000 6 Random | 0.433 0.246 0.761 -2.909 0.004 Test for heterogeneity: Q= 338.140 on 5 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.752 0.699 0.809 -7.651 0.000 6 Random | 0.645 0.500 0.833 -3.358 0.001 Test for heterogeneity: Q= 47.900 on 5 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.849 0.778 0.926 -3.680 0.000 5 Random | 0.708 0.535 0.937 -2.416 0.016 Test for heterogeneity: Q= 27.119 on 4 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.991 0.952 1.032 -0.425 0.671 6 Random | 0.961 0.878 1.052 -0.858 0.391 Test for heterogeneity: Q= 8.523 on 5 degrees of freedom (p= 0.130) 144 Specificity ratio (Cytology at LSIL+/VIA) Outcome CIN1+ Outcome CIN2+ Cytology at LSIL+/VIA Cytology at LSIL+/VIA Jaipur Jaipur Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome CIN3+ Outcome Cancer Cytology at LSIL+/VIA Cytology at LSIL+/VIA Jaipur Jaipur Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Meta-analysis of the specificity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.123 1.117 1.129 41.699 0.000 6 Random | 1.171 1.110 1.234 5.820 0.000 Test for heterogeneity: Q= 436.029 on 5 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.147 1.140 1.154 45.090 0.000 6 Random | 1.190 1.128 1.254 6.440 0.000 Test for heterogeneity: Q= 364.619 on 5 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.143 1.136 1.150 42.516 0.000 5 Random | 1.154 1.099 1.211 5.806 0.000 Test for heterogeneity: Q= 237.531 on 4 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.153 1.145 1.160 44.761 0.000 6 Random | 1.198 1.132 1.268 6.283 0.000 Test for heterogeneity: Q= 383.756 on 5 degrees of freedom (p= 0.000) 145 6.2.6. Relative sensitivity and specificity of cervical cytology (at HSIL+) compared to VIA Sensitivity ratio (Cytology at HSIL+/VIA) Outcome CIN2+ Outcome CIN3+ Cytology at HSIL+/VIA Cytology at HSIL+/VIA Jaipur Jaipur Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Harare Trivandrum 2 Combined Combined .25 .5 .67 1 1.5 2 .25 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Outcome Cancer Cytology at HSIL+/VIA Jaipur Kolkata 1 Mumbai Trivandrum 1 Trivandrum 2 Harare Combined .25 .5 .67 1 1.5 2 Sensitivity ratio Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.562 0.509 0.621 -11.350 0.000 6 Random | 0.462 0.310 0.688 -3.805 0.000 Test for heterogeneity: Q= 71.082 on 5 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.747 0.671 0.831 -5.344 0.000 5 Random | 0.660 0.499 0.872 -2.922 0.003 Test for heterogeneity: Q= 21.887 on 4 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.991 0.952 1.032 -0.425 0.671 6 Random | 0.961 0.878 1.052 -0.858 0.391 Test for heterogeneity: Q= 8.523 on 5 degrees of freedom (p= 0.130) 146 Specificity ratio (Cytology at HSIL+/VIA) Outcome CIN2+ Outcome CIN3+ Cytology at HSIL+/VIA Cytology at HSIL+/VIA Jaipur Jaipur Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Harare Trivandrum 2 Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome Cancer Cytology at HSIL+/VIA Jaipur Kolkata 1 Mumbai Trivandrum 1 Trivandrum 2 Harare Combined .67 .8 1 1.25 1.5 Specificity ratio Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.190 1.184 1.197 61.282 0.000 6 Random | 1.252 1.171 1.340 6.520 0.000 Test for heterogeneity: Q= 696.712 on 5 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.185 1.178 1.192 57.974 0.000 5 Random | 1.209 1.133 1.291 5.728 0.000 Test for heterogeneity: Q= 499.212 on 4 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.198 1.191 1.205 60.942 0.000 6 Random | 1.268 1.178 1.365 6.319 0.000 Test for heterogeneity: Q= 759.400 on 5 degrees of freedom (p= 0.000) 147 6.2.7. Relative sensitivity and specificity of VIAM compared to VILI Sensitivity ratio (VIAM/VILI) Outcome CIN1+ Outcome CIN2+ VIAM/VILI VIAM/VILI Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined .5 .67 1 1.5 2 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Outcome CIN3+ Outcome Cancer VIAM/VILI VIAM/VILI Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined .5 .67 1 1.5 2 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Meta-analysis of the sensitivity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.872 0.792 0.961 -2.760 0.006 2 Random | 0.872 0.792 0.961 -2.760 0.006 Test for heterogeneity: Q= 0.224 on 1 degrees of freedom (p= 0.636) Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.892 0.780 1.021 -1.655 0.098 2 Random | 0.892 0.780 1.021 -1.655 0.098 Test for heterogeneity: Q= 0.017 on 1 degrees of freedom (p= 0.896) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.903 0.767 1.064 -1.218 0.223 2 Random | 0.903 0.767 1.064 -1.218 0.223 Test for heterogeneity: Q= 0.116 on 1 degrees of freedom (p= 0.733) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.000 0.966 1.035 0.000 1.000 2 Random | 1.000 0.966 1.035 0.000 1.000 Test for heterogeneity: Q= 0.000 on 1 degrees of freedom (p= 1.000) 148 Specificity ratio (VIAM/VILI) Outcome CIN1+ Outcome CIN2+ VIAM/VILI VIAM/VILI Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome CIN3+ Outcome Cancer VIAM/VILI VIAM/VILI Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Meta-analysis of the specificity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.027 1.018 1.037 5.596 0.000 2 Random | 1.027 1.017 1.037 5.483 0.000 Test for heterogeneity: Q= 1.027 on 1 degrees of freedom (p= 0.311) Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.030 1.020 1.040 5.909 0.000 2 Random | 1.030 1.020 1.040 5.909 0.000 Test for heterogeneity: Q= 0.749 on 1 degrees of freedom (p= 0.387) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.030 1.020 1.041 5.935 0.000 2 Random | 1.030 1.020 1.041 5.935 0.000 Test for heterogeneity: Q= 0.588 on 1 degrees of freedom (p= 0.443) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.031 1.021 1.041 5.999 0.000 2 Random | 1.031 1.021 1.041 5.999 0.000 Test for heterogeneity: Q= 0.537 on 1 degrees of freedom (p= 0.464) 149 6.2.8. Relative sensitivity and specificity of HC2 compared to VILI Sensitivity ratio (HPV/VILI) Outcome CIN1+ Outcome CIN2+ HPV/VILI HPV/VILI Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Combined Combined .25 .5 .67 1 1.5 2 .25 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Outcome CIN3+ Outcome Cancer HPV/VILI HPV/VILI Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Combined Combined .25 .5 .67 1 1.5 2 .25 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Meta-analysis of the sensitivity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.408 0.360 0.462 -13.995 0.000 3 Random | 0.413 0.339 0.504 -8.684 0.000 Test for heterogeneity: Q= 4.976 on 2 degrees of freedom (p= 0.083) Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.834 0.740 0.939 -3.006 0.003 3 Random | 0.834 0.740 0.939 -3.006 0.003 Test for heterogeneity: Q= 0.403 on 2 degrees of freedom (p= 0.818) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.855 0.748 0.977 -2.301 0.021 3 Random | 0.857 0.714 1.028 -1.659 0.097 Test for heterogeneity: Q= 3.707 on 2 degrees of freedom (p= 0.157) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.877 0.718 1.071 -1.289 0.197 3 Random | 0.877 0.718 1.071 -1.289 0.197 Test for heterogeneity: Q= 0.916 on 2 degrees of freedom (p= 0.633) 150 Specificity ratio (HPV/VILI) Outcome CIN1+ Outcome CIN2+ HPV/VILI HPV/VILI Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome CIN3+ Outcome Cancer HPV/VILI HPV/VILI Kolkata 2 Kolkata 2 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Meta-analysis of the specificity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.074 1.067 1.082 20.233 0.000 3 Random | 1.075 1.052 1.097 6.733 0.000 Test for heterogeneity: Q= 17.338 on 2 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.096 1.088 1.104 24.671 0.000 3 Random | 1.097 1.085 1.110 16.132 0.000 Test for heterogeneity: Q= 4.480 on 2 degrees of freedom (p= 0.106) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.096 1.088 1.105 24.509 0.000 3 Random | 1.098 1.084 1.112 14.720 0.000 Test for heterogeneity: Q= 5.327 on 2 degrees of freedom (p= 0.070) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.097 1.089 1.106 24.229 0.000 3 Random | 1.099 1.086 1.111 16.175 0.000 Test for heterogeneity: Q= 4.292 on 2 degrees of freedom (p= 0.117) 151 6.2.9. Relative sensitivity and specificity of cytology (at ASCUS+) compared to VILI Sensitivity ratio (Cytology at ASCUS+/VILI) Outcome CIN1+ Outcome CIN2+ Cytology at ASCUS+/VILI Cytology at ASCUS+/VILI Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .25 .5 .67 1 1.5 2 .25 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Outcome CIN3+ Outcome Cancer Cytology at ASCUS+/VILI Cytology at ASCUS+/VILI Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .25 .5 .67 1 1.5 2 .25 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Meta-analysis of the sensitivity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.808 0.752 0.868 -5.854 0.000 4 Random | 0.495 0.219 1.117 -1.695 0.090 Test for heterogeneity: Q= 284.509 on 3 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.848 0.787 0.914 -4.324 0.000 4 Random | 0.735 0.570 0.948 -2.374 0.018 Test for heterogeneity: Q= 21.633 on 3 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.949 0.897 1.005 -1.799 0.072 4 Random | 0.796 0.618 1.024 -1.772 0.076 Test for heterogeneity: Q= 19.809 on 3 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.998 0.954 1.044 -0.076 0.939 4 Random | 0.998 0.954 1.044 -0.076 0.939 Test for heterogeneity: Q= 0.939 on 3 degrees of freedom (p= 0.816) 152 Specificity ratio (Cytology at ASCUS+/VILI) Outcome CIN1+ Outcome CIN2+ Cytology at ASCUS+/VILI Cytology at ASCUS+/VILI Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome CIN3+ Outcome Cancer Cytology at ASCUS+/VILI Cytology at ASCUS+/VILI Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Meta-analysis of the specificity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.136 1.128 1.143 37.525 0.000 4 Random | 1.152 1.091 1.216 5.116 0.000 Test for heterogeneity: Q= 190.222 on 3 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.157 1.148 1.165 38.570 0.000 4 Random | 1.162 1.092 1.236 4.760 0.000 Test for heterogeneity: Q= 200.978 on 3 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.159 1.151 1.168 38.512 0.000 4 Random | 1.164 1.094 1.239 4.776 0.000 Test for heterogeneity: Q= 197.259 on 3 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.162 1.153 1.171 38.186 0.000 4 Random | 1.166 1.094 1.242 4.731 0.000 Test for heterogeneity: Q= 195.714 on 3 degrees of freedom (p= 0.000) 153 6.2.10. Relative sensitivity and specificity of cytology (at LSIL+) compared to VILI Sensitivity ratio (Cytology at LSIL+/VILI) Outcome CIN1+ Outcome CIN2+ Cytology at LSIL+/VILI Cytology at LSIL+/VILI Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .25 .5 .67 1 1.5 2 .25 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Outcome CIN3+ Outcome Cancer Cytology at LSIL+/VILI Cytology at LSIL+/VILI Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .25 .5 .67 1 1.5 2 .25 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Meta-analysis of the sensitivity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.783 0.727 0.844 -6.444 0.000 4 Random | 0.443 0.182 1.076 -1.799 0.072 Test for heterogeneity: Q= 293.712 on 3 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.802 0.739 0.870 -5.323 0.000 4 Random | 0.668 0.500 0.892 -2.731 0.006 Test for heterogeneity: Q= 23.683 on 3 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.852 0.785 0.926 -3.780 0.000 4 Random | 0.721 0.550 0.945 -2.370 0.018 Test for heterogeneity: Q= 17.397 on 3 degrees of freedom (p= 0.001) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.996 0.952 1.042 -0.162 0.871 4 Random | 0.996 0.952 1.042 -0.162 0.871 Test for heterogeneity: Q= 2.083 on 3 degrees of freedom (p= 0.555) 154 Specificity ratio (Cytology at LSIL+/VILI) Outcome CIN1+ Outcome CIN2+ Cytology at LSIL+/VILI Cytology at LSIL+/VILI Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome CIN3+ Outcome Cancer Cytology at LSIL+/VILI Cytology at LSIL+/VILI Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Meta-analysis of the specificity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.146 1.139 1.154 41.073 0.000 4 Random | 1.165 1.092 1.242 4.659 0.000 Test for heterogeneity: Q= 276.628 on 3 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.170 1.161 1.178 42.409 0.000 4 Random | 1.176 1.096 1.261 4.513 0.000 Test for heterogeneity: Q= 271.368 on 3 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.173 1.164 1.181 42.352 0.000 4 Random | 1.178 1.097 1.265 4.519 0.000 Test for heterogeneity: Q= 267.468 on 3 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.176 1.168 1.185 42.130 0.000 4 Random | 1.180 1.098 1.269 4.488 0.000 Test for heterogeneity: Q= 265.510 on 3 degrees of freedom (p= 0.000) 155 6.2.11. Relative sensitivity and specificity of cytology (at HSIL+) compared to VILI Sensitivity ratio (Cytology at HSIL+/VILI) Outcome CIN2+ Outcome CIN3+ Cytology at HSIL+/VILI Cytology at HSIL+/VILI Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .1 .5 .67 1 1.5 2 .1 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Outcome Cancer Cytology at HSIL+/VILI Jaipur Mumbai Trivandrum 1 Trivandrum 2 Combined .1 .5 .67 1 1.5 2 Sensitivity ratio Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.627 0.562 0.700 -8.315 0.000 4 Random | 0.552 0.392 0.777 -3.400 0.001 Test for heterogeneity: Q= 24.066 on 3 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.749 0.672 0.835 -5.205 0.000 4 Random | 0.676 0.518 0.882 -2.886 0.004 Test for heterogeneity: Q= 13.487 on 3 degrees of freedom (p= 0.004) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.996 0.952 1.042 -0.162 0.871 4 Random | 0.996 0.952 1.042 -0.162 0.871 Test for heterogeneity: Q= 2.083 on 3 degrees of freedom (p= 0.555) 156 Specificity ratio (Cytology at HSIL+/VILI) Outcome CIN2+ Outcome CIN3+ Cytology at HSIL+/VILI Cytology at HSIL+/VILI Jaipur Jaipur Mumbai Mumbai Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome Cancer Cytology at HSIL+/VILI Jaipur Mumbai Trivandrum 1 Trivandrum 2 Combined .67 .8 1 1.25 1.5 Specificity ratio Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.211 1.203 1.219 55.425 0.000 4 Random | 1.228 1.144 1.318 5.669 0.000 Test for heterogeneity: Q= 322.490 on 3 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.215 1.207 1.224 55.297 0.000 4 Random | 1.233 1.145 1.327 5.578 0.000 Test for heterogeneity: Q= 331.461 on 3 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.220 1.211 1.228 54.766 0.000 4 Random | 1.236 1.147 1.332 5.538 0.000 Test for heterogeneity: Q= 326.532 on 3 degrees of freedom (p= 0.000) 157 6.2.12. Relative sensitivity and specificity of HC2 compared to VIAM Sensitivity ratio (HPV/VIAM) Outcome CIN1+ Outcome CIN2+ HPV/VIAM HPV/VIAM Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined .25 .5 .67 1 1.5 2 .25 .5 .67 1 1.5 2 Sensitvity ratio Sensitvity ratio Outcome CIN3+ Outcome Cancer HPV/VIAM HPV/VIAM Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined .25 .5 .67 1 1.5 2 .25 .5 .67 1 1.5 2 Sensitvity ratio Sensitvity ratio Meta-analysis of the sensitivity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.497 0.435 0.568 -10.262 0.000 3 Random | 0.497 0.435 0.568 -10.262 0.000 Test for heterogeneity: Q= 1.976 on 2 degrees of freedom (p= 0.372) Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.901 0.783 1.036 -1.461 0.144 3 Random | 0.898 0.770 1.048 -1.365 0.172 Test for heterogeneity: Q= 2.391 on 2 degrees of freedom (p= 0.302) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.025 0.877 1.199 0.314 0.753 3 Random | 1.025 0.877 1.199 0.314 0.753 Test for heterogeneity: Q= 0.244 on 2 degrees of freedom (p= 0.885) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.892 0.726 1.095 -1.096 0.273 3 Random | 0.892 0.726 1.095 -1.096 0.273 Test for heterogeneity: Q= 0.622 on 2 degrees of freedom (p= 0.733) 158 Specificity ratio (HPV/VIAM) Outcome CIN1 Outcome CIN2+ HPV/VIAM HPV/VIAM Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome CIN3+ Outcome Cancer HPV/VIAM HPV/VIAM Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Sensitvity ratio Meta-analysis of the specificity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.059 1.052 1.066 16.195 0.000 3 Random | 1.061 1.050 1.072 10.928 0.000 Test for heterogeneity: Q= 4.208 on 2 degrees of freedom (p= 0.122) Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.074 1.066 1.082 19.521 0.000 3 Random | 1.082 1.055 1.109 6.130 0.000 Test for heterogeneity: Q= 21.463 on 2 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.075 1.067 1.083 19.588 0.000 3 Random | 1.083 1.055 1.112 5.874 0.000 Test for heterogeneity: Q= 23.771 on 2 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.075 1.067 1.083 19.311 0.000 3 Random | 1.083 1.054 1.114 5.648 0.000 Test for heterogeneity: Q= 25.097 on 2 degrees of freedom (p= 0.000) 159 6.2.13. Relative sensitivity and specificity of cytology at (ASCUS+) compared to VIAM Sensitivity ratio (Cytology at ASCUS+/VIAM) Outcome CIN1+ Outcome CIN2+ Cytology at ASCUS+/VIAM Cytology at ASCUS+/VIAM Kolkata 1 Kolkata 1 Mumbai Mumbai Combined Combined .25 .5 .67 1 1.5 2 .25 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Outcome CIN3+ Outcome Cancer Cytology at ASCUS+/VIAM Cytology at ASCUS+/VIAM Kolkata 1 Kolkata 1 Mumbai Mumbai Combined Combined .25 .5 .67 1 1.5 2 .25 .5 .67 1 1.5 2 Sensitvity ratio Sensitvity ratio Meta-analysis of the sensitivity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.389 0.338 0.446 -13.357 0.000 2 Random | 0.396 0.313 0.501 -7.709 0.000 Test for heterogeneity: Q= 2.707 on 1 degrees of freedom (p= 0.100) Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.800 0.683 0.936 -2.785 0.005 2 Random | 0.797 0.486 1.307 -0.899 0.368 Test for heterogeneity: Q= 9.851 on 1 degrees of freedom (p= 0.002) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.876 0.734 1.044 -1.478 0.139 2 Random | 0.863 0.573 1.301 -0.702 0.483 Test for heterogeneity: Q= 5.393 on 1 degrees of freedom (p= 0.020) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.868 0.688 1.095 -1.194 0.233 2 Random | 0.859 0.598 1.235 -0.822 0.411 Test for heterogeneity: Q= 2.420 on 1 degrees of freedom (p= 0.120) 160 Specificity ratio (Cytology at ASCUS+/VIAM) Outcome CIN1 Outcome CIN2+ Cytology at ASCUS+/VIAM Cytology at ASCUS+/VIAM Kolkata 1 Kolkata 1 Mumbai Mumbai Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome CIN3+ Outcome Cancer Cytology at ASCUS+/VIAM Cytology at ASCUS+/VIAM Kolkata 1 Kolkata 1 Mumbai Mumbai Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Meta-analysis of the specificity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.074 1.063 1.084 14.479 0.000 2 Random | 1.061 0.948 1.188 1.033 0.302 Test for heterogeneity: Q= 131.356 on 1 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.100 1.089 1.112 18.493 0.000 2 Random | 1.091 0.990 1.202 1.755 0.079 Test for heterogeneity: Q= 89.243 on 1 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.102 1.090 1.113 18.489 0.000 2 Random | 1.093 0.995 1.200 1.865 0.062 Test for heterogeneity: Q= 80.622 on 1 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.101 1.090 1.113 18.050 0.000 2 Random | 1.093 0.999 1.197 1.931 0.053 Test for heterogeneity: Q= 73.535 on 1 degrees of freedom (p= 0.000) 161 6.2.14. Relative sensitivity and specificity of cytology (at LSIL+) compared to VIAM Sensitivity ratio (Cytology at LSIL+/VIAM) Outcome CIN1+ Outcome CIN2+ Cytology at LSIL+/VIAM Cytology at LSIL+/VIAM Kolkata 1 Kolkata 1 Mumbai Mumbai Combined Combined .25 .5 .67 1 1.5 2 .25 .5 .67 1 1.5 2 Sensitvity ratio Sensitvity ratio Outcome CIN3+ Outcome Cancer Cytology at LSIL+/VIAM Cytology at LSIL+/VIAM Kolkata 1 Kolkata 1 Mumbai Mumbai Combined Combined .25 .5 .67 1 1.5 2 .25 .5 .67 1 1.5 2 Sensitvity ratio Sensitvity ratio Meta-analysis of the sensitivity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.315 0.269 0.369 -14.396 0.000 2 Random | 0.325 0.218 0.484 -5.528 0.000 Test for heterogeneity: Q= 6.228 on 1 degrees of freedom (p= 0.013) Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.689 0.579 0.821 -4.182 0.000 2 Random | 0.681 0.386 1.202 -1.324 0.185 Test for heterogeneity: Q= 10.591 on 1 degrees of freedom (p= 0.001) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.740 0.607 0.902 -2.981 0.003 2 Random | 0.732 0.478 1.119 -1.442 0.149 Test for heterogeneity: Q= 4.590 on 1 degrees of freedom (p= 0.032) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.775 0.600 1.001 -1.953 0.051 2 Random | 0.770 0.539 1.101 -1.433 0.152 Test for heterogeneity: Q= 1.942 on 1 degrees of freedom (p= 0.163) 162 Specificity ratio (Cytology at LSIL+/VIAM) Outcome CIN1+ Outcome CIN2+ Cytology at LSIL+/VIAM Cytology at LSIL+/VIAM Kolkata 1 Kolkata 1 Mumbai Mumbai Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome CIN3+ Outcome Cancer Cytology at LSIL+/VIAM Cytology at LSIL+/VIAM Kolkata 1 Kolkata 1 Mumbai Mumbai Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Meta-analysis of the specificity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.105 1.095 1.114 21.854 0.000 2 Random | 1.103 1.050 1.158 3.915 0.000 Test for heterogeneity: Q= 29.888 on 1 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.134 1.124 1.145 26.046 0.000 2 Random | 1.133 1.095 1.173 7.165 0.000 Test for heterogeneity: Q= 12.952 on 1 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.136 1.125 1.147 26.048 0.000 2 Random | 1.135 1.102 1.170 8.248 0.000 Test for heterogeneity: Q= 9.816 on 1 degrees of freedom (p= 0.002) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.137 1.126 1.148 25.785 0.000 2 Random | 1.137 1.105 1.169 9.010 0.000 Test for heterogeneity: Q= 8.101 on 1 degrees of freedom (p= 0.004) 163 6.2.15. Relative sensitivity and specificity of cytology (at HSIL+) compared to VIAM Sensitivity ratio (Cytology at HSIL+/VIAM) Outcome CIN2+ Outcome CIN3+ Cytology at HSIL+/VIAM Cytology at HSIL+/VIAM Kolkata 1 Kolkata 1 Mumbai Mumbai Combined Combined .5 .67 1 1.5 2 .5 .67 1 1.5 2 Sensitvity ratio Sensitvity ratio Outcome Cancer Cytology at HSIL+/VIAM Kolkata 1 Mumbai Combined .5 .67 1 1.5 2 Sensitvity ratio Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.618 0.512 0.746 -5.027 0.000 2 Random | 0.606 0.324 1.132 -1.571 0.116 Test for heterogeneity: Q= 11.066 on 1 degrees of freedom (p= 0.001) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.706 0.575 0.866 -3.332 0.001 2 Random | 0.692 0.428 1.118 -1.505 0.132 Test for heterogeneity: Q= 5.415 on 1 degrees of freedom (p= 0.020) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.775 0.600 1.001 -1.953 0.051 2 Random | 0.770 0.539 1.101 -1.433 0.152 Test for heterogeneity: Q= 1.942 on 1 degrees of freedom (p= 0.163) 164 Specificity ratio (Cytology at HSIL+/VIAM) Outcome CIN2+ Outcome CIN3+ Cytology at HSIL+/VIAM Cytology at HSIL+/VIAM Kolkata 1 Kolkata 1 Mumbai Mumbai Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome Cancer Cytology at HSIL+/VIAM Kolkata 1 Mumbai Combined .67 .8 1 1.25 1.5 Specificity ratio Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.166 1.156 1.176 33.879 0.000 2 Random | 1.166 1.152 1.180 24.364 0.000 Test for heterogeneity: Q= 1.928 on 1 degrees of freedom (p= 0.165) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.169 1.158 1.179 33.955 0.000 2 Random | 1.169 1.150 1.187 19.180 0.000 Test for heterogeneity: Q= 3.121 on 1 degrees of freedom (p= 0.077) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.171 1.160 1.181 33.646 0.000 2 Random | 1.170 1.149 1.192 16.979 0.000 Test for heterogeneity: Q= 3.902 on 1 degrees of freedom (p= 0.048) 165 6.2.16. Relative sensitivity and specificity of cytology (at ASCUS+) compared to HC2 Sensitivity ratio (Cytology at ASCUS+/HPV) Outcome CIN1+ Outcome CIN2+ Cytology at ASCUS+/HPV Cytology at ASCUS+/HPV Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined .5 .67 1 1.5 2 .5 .67 1 1.5 2 Sensitvity ratio Sensitvity ratio Outcome CIN3+ Outcome Cancer Cytology at ASCUS+/HPV Cytology at ASCUS+/HPV Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Combined Combined .5 .67 1 1.5 2 .5 .67 1 1.5 2 Sensitvity ratio Sensitvity ratio Meta-analysis of the sensitivity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.791 0.723 0.864 -5.166 0.000 4 Random | 0.854 0.704 1.036 -1.603 0.109 Test for heterogeneity: Q= 9.239 on 3 degrees of freedom (p= 0.026) Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.865 0.787 0.950 -3.019 0.003 4 Random | 0.883 0.780 0.999 -1.971 0.049 Test for heterogeneity: Q= 4.223 on 3 degrees of freedom (p= 0.238) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.929 0.793 1.088 -0.913 0.361 3 Random | 0.915 0.742 1.130 -0.824 0.410 Test for heterogeneity: Q= 3.406 on 2 degrees of freedom (p= 0.182) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.076 0.854 1.355 0.618 0.537 4 Random | 1.076 0.854 1.355 0.618 0.537 Test for heterogeneity: Q= 1.537 on 3 degrees of freedom (p= 0.674) 166 Specificity ratio (Cytology at ASCUS+/HPV) Outcome CIN1+ Outcome CIN2+ Cytology at ASCUS+/HPV Cytology at ASCUS+/HPV Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome CIN3+ Outcome Cancer Cytology at ASCUS+/HPV Cytology at ASCUS+/HPV Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Meta-analysis of the specificity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.030 1.024 1.035 10.472 0.000 4 Random | 1.052 0.996 1.110 1.815 0.070 Test for heterogeneity: Q= 229.935 on 3 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.028 1.022 1.034 9.646 0.000 4 Random | 1.057 1.000 1.117 1.957 0.050 Test for heterogeneity: Q= 231.817 on 3 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.024 1.018 1.030 8.046 0.000 3 Random | 1.008 0.956 1.064 0.311 0.756 Test for heterogeneity: Q= 151.373 on 2 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.027 1.021 1.033 8.569 0.000 4 Random | 1.061 1.000 1.125 1.966 0.049 Test for heterogeneity: Q= 229.069 on 3 degrees of freedom (p= 0.000) 167 6.2.17. Relative sensitivity and specificity of cytology (at LSIL+) compared to HC2 Sensitivity ratio (Cytology at LSIL+/HPV) Outcome CIN1+ Outcome CIN2+ Cytology at LSIL+/HPV Cytology at LSIL+/HPV Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined .5 .67 1 1.5 2 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Outcome CIN3+ Outcome Cancer Cytology at LSIL+/HPV Cytology at LSIL+/HPV Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Combined Combined .5 .67 1 1.5 2 .5 .67 1 1.5 2 Sensitivity ratio Sensitivity ratio Meta-analysis of the sensitivity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.569 0.511 0.635 -10.116 0.000 4 Random | 0.663 0.456 0.964 -2.152 0.031 Test for heterogeneity: Q= 28.786 on 3 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.712 0.635 0.798 -5.845 0.000 4 Random | 0.755 0.559 1.019 -1.838 0.066 Test for heterogeneity: Q= 18.654 on 3 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.842 0.711 0.997 -1.991 0.046 3 Random | 0.816 0.609 1.093 -1.366 0.172 Test for heterogeneity: Q= 5.744 on 2 degrees of freedom (p= 0.057) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.013 0.796 1.291 0.108 0.914 4 Random | 1.013 0.796 1.291 0.108 0.914 Test for heterogeneity: Q= 2.532 on 3 degrees of freedom (p= 0.470) 168 Specificity ratio (Cytology at LSIL+/HPV) Outcome CIN1+ Outcome CIN2+ Cytology at LSIL+/HPV Cytology at LSIL+/HPV Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Harare Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome CIN3+ Outcome Cancer Cytology at LSIL+/HPV Cytology at LSIL+/HPV Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Meta-analysis of the specificity ratio for outcome CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.046 1.041 1.052 17.114 0.000 4 Random | 1.116 1.054 1.182 3.780 0.000 Test for heterogeneity: Q= 296.609 on 3 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.048 1.042 1.053 17.110 0.000 4 Random | 1.127 1.058 1.200 3.701 0.000 Test for heterogeneity: Q= 351.989 on 3 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.040 1.035 1.046 14.024 0.000 3 Random | 1.037 1.014 1.060 3.177 0.001 Test for heterogeneity: Q= 30.599 on 2 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.049 1.043 1.055 16.166 0.000 4 Random | 1.136 1.059 1.218 3.585 0.000 Test for heterogeneity: Q= 373.645 on 3 degrees of freedom (p= 0.000) 169 6.2.18. Relative sensitivity and specificity of cytology (at HSIL+) compared to HC2 Sensitivity ratio (Cytology at HSIL+/HPV) Outcome CIN2+ Outcome CIN3+ Cytology at HSIL+/HPV Cytology at HSIL+/HPV Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Combined Combined .1 .25 .5 .67 1 1.5 2 .1 .25 .5 .67 1 1.5 2 Sensitvity ratio Sensitvity ratio Outcome Cancer Cytology at HSIL+/HPV Kolkata 1 Mumbai Trivandrum 2 Harare Combined .1 .25 .5 .67 1 1.5 2 Sensitvity ratio Meta-analysis of the sensitivity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.629 0.545 0.725 -6.390 0.000 4 Random | 0.560 0.286 1.096 -1.693 0.090 Test for heterogeneity: Q= 64.346 on 3 degrees of freedom (p= 0.000) Meta-analysis of the sensitivity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.822 0.692 0.976 -2.231 0.026 3 Random | 0.786 0.562 1.099 -1.408 0.159 Test for heterogeneity: Q= 7.301 on 2 degrees of freedom (p= 0.026) Meta-analysis of the sensitivity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.013 0.796 1.291 0.108 0.914 4 Random | 1.013 0.796 1.291 0.108 0.914 Test for heterogeneity: Q= 2.532 on 3 degrees of freedom (p= 0.470) 170 Specificity ratio (Cytology at HSIL+/HPV) Outcome CIN2+ Outcome CIN3+ Cytology at HSIL+/HPV Cytology at HSIL+/HPV Kolkata 1 Kolkata 1 Mumbai Mumbai Trivandrum 2 Trivandrum 2 Harare Combined Combined .67 .8 1 1.25 1.5 .67 .8 1 1.25 1.5 Specificity ratio Specificity ratio Outcome Cancer Cytology at HSIL+/HPV Kolkata 1 Mumbai Trivandrum 2 Harare Combined .67 .8 1 1.25 1.5 Specificity ratio Meta-analysis of the specificity ratio for outcome CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.067 1.062 1.072 26.029 0.000 4 Random | 1.170 1.092 1.253 4.478 0.000 Test for heterogeneity: Q= 497.486 on 3 degrees of freedom (p= 0.000) Meta-analysis of the specificity ratio for outcome CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.060 1.054 1.065 22.298 0.000 3 Random | 1.061 1.052 1.070 13.944 0.000 Test for heterogeneity: Q= 5.013 on 2 degrees of freedom (p= 0.082) Meta-analysis of the specificity ratio for outcome cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 1.070 1.065 1.076 24.988 0.000 4 Random | 1.187 1.096 1.286 4.194 0.000 Test for heterogeneity: Q= 584.614 on 3 degrees of freedom (p= 0.000) 171 6.2.19. Summary of the meta-analyses of the relative accuracy of screening tests Table 29. Pooled relative sensitivity and specificity of 5 screening tests relative to each other with respect to detection of CIN1+, CIN2+, CIN3+ or cancer (* indicates ratios which do not include unity, indicating statistically significant differences between tests). Relative sensitivity & CI Relative specificity & CI Number of study Test combination Outcome locations Model Ratio lower upper Model Ratio lower upper VILI/VIA CIN1+ 10 R 1.19 1.08 1.30 * R 1.00 0.97 1.04 CIN2+ R 1.11 1.05 1.17 * R 1.00 0.96 1.03 CIN3+ F 1.07 1.04 1.11 * R 0.99 0.96 1.03 Cancer F 1.01 0.99 1.04 R 0.99 0.96 1.03 VIAM/VIA CIN1+ 3 F 1.05 0.98 1.13 R 1.00 0.98 1.02 CIN2+ F 1.04 0.93 1.16 R 1.00 0.98 1.02 CIN3+ F 1.03 0.90 1.18 R 1.00 0.98 1.02 Cancer F 1.00 0.97 1.03 R 1.00 0.98 1.02 HC2/VIA CIN1+ 4 R 0.57 0.36 0.89 * R 1.04 1.02 1.06 * CIN2+ R 0.93 0.82 1.06 R 1.06 1.03 1.09 * CIN3+ R 0.96 0.78 1.17 R 1.08 1.05 1.10 * Cancer R 0.85 0.71 1.02 R 1.06 1.03 1.09 * Pap (ASCUS+)/VIA CIN1+ 5 R 0.53 0.34 0.83 * R 1.12 1.07 1.18 * CIN2+ R 0.77 0.64 0.93 * R 1.13 1.09 1.19 * CIN3+ R 0.80 0.61 1.03 R 1.13 1.07 1.18 * Ca R 0.99 0.94 1.04 R 1.14 1.09 1.19 * Pap (LSIL+)/VIA CIN1+ 5 R 0.43 0.25 0.76 * R 1.17 1.11 1.23 * CIN2+ R 0.65 0.50 0.83 * R 1.19 1.13 1.25 * CIN3+ R 0.71 0.54 0.94 * R 1.15 1.10 1.21 * Ca R 0.96 0.88 1.05 * R 1.20 1.13 1.27 * Pap (HSIL+)/VIA CIN2+ 5 R 0.46 0.31 0.69 * R 1.25 1.17 1.34 * CIN3+ R 0.66 0.50 0.87 * R 1.21 1.13 1.29 * Ca R 0.96 0.88 1.05 * R 1.27 1.18 1.37 * VIAM/VILI CIN1+ 2 F 0.87 0.79 0.96 * F 1.03 1.02 1.04 * CIN2+ F 0.89 0.78 1.02 F 1.03 1.02 1.04 * CIN3+ F 0.90 0.77 1.06 F 1.03 1.02 1.04 * Ca F 1.00 0.97 1.04 F 1.03 1.02 1.04 * HC2/VILI CIN1+ 3 R 0.41 0.34 0.50 * R 1.08 1.05 1.10 * CIN2+ F 0.83 0.74 0.94 * R 1.10 1.09 1.11 * CIN3+ R 0.86 0.71 1.03 R 1.10 1.08 1.11 * Ca F 0.88 0.72 1.07 R 1.10 1.09 1.11 * Pap (ASCUS+)/VILI CIN1+ 4 R 0.50 0.22 1.12 R 1.15 1.09 1.22 * CIN2+ R 0.74 0.57 0.95 * R 1.16 1.09 1.24 * CIN3+ R 0.80 0.62 1.02 R 1.16 1.09 1.24 * Ca F 1.00 0.95 1.04 R 1.17 1.09 1.24 * Pap (LSIL+)/VILI CIN1+ 4 R 0.44 0.18 1.08 R 1.17 1.09 1.24 * CIN2+ R 0.67 0.50 0.89 * R 1.18 1.10 1.26 * CIN3+ R 0.72 0.55 0.95 * R 1.18 1.10 1.27 * Ca F 1.00 0.95 1.04 R 1.18 1.10 1.27 * Pap (HSIL+)/VILI CIN2+ 4 R 0.55 0.39 0.78 * R 1.23 1.14 1.32 * CIN3+ R 0.68 0.52 0.88 * R 1.23 1.15 1.33 * 172 Ca F 1.00 0.95 1.04 R 1.24 1.15 1.33 * HC2/VIAM CIN1+ 3 F 0.50 0.44 0.57 R 1.06 1.05 1.07 CIN2+ F 0.90 0.78 1.04 R 1.08 1.06 1.11 CIN3+ F 1.03 0.88 1.20 R 1.08 1.06 1.11 Ca R 0.89 0.73 1.10 R 1.08 1.05 1.11 Pap (ASCUS+)/VIAM CIN1+ 2 R 0.40 0.31 0.50 R 1.06 0.95 1.19 CIN2+ R 0.80 0.49 1.31 R 1.09 0.99 1.20 CIN3+ R 0.86 0.57 1.30 R 1.09 1.00 1.20 Ca R 0.86 0.60 1.24 R 1.09 1.00 1.20 Pap (LSIL+)/VIAM CIN1+ 2 R 0.33 0.22 0.48 R 1.10 1.05 1.16 CIN2+ R 0.68 0.39 1.20 R 1.13 1.10 1.17 CIN3+ R 0.73 0.48 1.12 R 1.14 1.10 1.17 Ca R 0.77 0.54 1.10 R 1.14 1.11 1.17 Pap (HSIL+)/VIAM CIN2+ 2 R 0.61 0.32 1.13 R 1.17 1.15 1.18 CIN3+ R 0.69 0.43 1.12 R 1.17 1.15 1.19 Ca R 0.77 0.54 1.10 R 1.17 1.15 1.19 Pap (ASCUS+)/HC2 CIN1+ 4 R 0.85 0.70 1.04 R 1.05 1.00 1.11 CIN2+ F 0.87 0.79 0.95 R 1.06 1.00 1.12 CIN3+ R 0.92 0.74 1.13 R 1.01 0.96 1.06 Ca F 1.08 0.85 1.36 R 1.06 1.00 1.13 Pap (LSIL+)/HC2 CIN1+ 4 R 0.66 0.46 0.96 R 1.12 1.05 1.18 CIN2+ R 0.76 0.56 1.02 R 1.13 1.06 1.20 CIN3+ R 0.82 0.61 1.09 R 1.04 1.01 1.06 Ca F 1.01 0.80 1.29 R 1.14 1.06 1.22 Pap (HSIL+)/HC2 CIN2+ 4 R 0.56 0.29 1.10 R 1.17 1.09 1.25 CIN3+ R 0.79 0.56 1.10 R 1.06 1.05 1.07 Ca F 1.01 0.80 1.29 R 1.19 1.10 1.29 173 6.3. Summary ROC curves The simultaneous variation of sensitivity and specificity over the different study locations is assessed by plotting the false positive rate (complement of the specificity) against the sensitivity. Using sROC regression [Moses, 1987; Arbyn, 2002] we can obtain a smoothed curve through the observed points. We have limited the sROC regression analyses to VIA and VILI for which we have respectively 12 and 10 study points. Figure 7 shows the observed sensitivity against the observed FPR and the smoothed sROC over a limited proportion of the ROC-space considering CIN2+ as outcome. It clearly shows a few exceptional points. In Conakry, a simultaneously outlying high sensitivity (91%) and specificity (94%) were observed; whereas in Kolkata 1, a very low sensitivity (62%) and an intermediate specificity (82%) were found. In Harare, the lowest specificity (64%) was found in spite of only an intermediate sensitivity (76%). VIA Outcome=CIN2+ 1 Co Tr1 Ou Ja Ba Tr2 Br .8 Ko2 Ha Ni Sensitivity Mu Ko1 .6 .4 .2 0 0 .1 .2 .3 .4 FPR = 1- Specificity Figure 7. Partial summary ROC curve (sROC): observed sensitivity against observed false positive rate (indicated by the first 3 letters of the 12 included study locations); fitted sensitivity and 95% CI obtained by an SROC regression; for VIA considering CIN2+ as outcome. Figure 8 shows the complete sROC curve running from (0,0) to (1,1). The area under this curve occupies 89.2% (CI: 84.9% to 92.4%) of the total ROC space. Figure 9 and Figure 10 display respectively the partial and complete sROC curve describing the accuracy of VILI for CIN2. Figure 11 contains the accuracy points and partial sROC curves for both, VIA and VILI in one plot. The red sROC curve, describing the accuracy of VILI jumps abruptly under the blue VIA curve at the left part of the plot (in the vertical zone where FPR≈ 0.1). This downward part of the curve is located beyond observed data points. Therefore the area under the curve (AUC) is not really representative for the observed range of accuracy. For this reason, the area under the VILI curve is nearly equal to the area under the VIA curve, in spite of the fact that most VILI points are situated at left and above the 174 VIA points. We prefer using the partial AUC considering only observed data points. The partial AUC was 16.9 % for VIA and the 95% CI (15.0% to 18.4%) includes the value for VILI (17.5%). In this particular case, even the difference in partial AUC can not describe appropriately the contrast in accuracy between both tests appropriately, since most VILI data points are situated in a narrow range near the point where the sROC jumps downward and no points are located around the downward part of the curve. We constructed forest plots for the other outcomes. Interested readers can contact us if they wish to obtain them. VIA Outcome=CIN2+ 1 .8 Sensitivity .6 .4 .2 0 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 FPR = 1- Specificity Figure 8. Complete sROC curve describing the relation between the FPR and sensitivity for the same 12 study locations as in Figure 7. VILI Outcome=CIN2+ 1 Co Br Ba Ou Ni Tr1 Ja Ko2 Tr2 .8 Mu Sensitivity .6 .4 .2 0 0 .1 .2 .3 FPR = 1- Specificity Figure 9. Partial sROC curve describing the relation between the FPR and the sensitivity of VILI considering CIN2+ as outcome. 175 VILI Outcome=CIN2+ 1 .8 Sensitivity .6 .4 .2 0 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 FPR = 1- Specificity Figure 10. Complete sROC curve describing the describing the same relation between the FPR and sensitivity for the 10 study locations plotted in Figure 9. VIA VILI 1 .8 Sensitivity .6 .4 .2 0 0 .1 .2 .3 .4 FPR = 1- Specificity Figure 11. Relation between sensitivity and FPR of VIA (blue x-points and interrupted line) and VILI (red circles and full line) for CIN2+. The points represent respectively 12 VIA studies and 10 VILI studies. The smoothed curves are obtained using sROC regression. 176 Table 30. Area under the sROC curve (AUC) obtained by integration of the surface under the complete sROC-regression curve. Test Outcome AUC 95% CI VIA CIN2+ 0.892 0.849 0.924 CIN3+ 0.912 0.878 0.937 VILI CIN2+ 0.906 0.863 0.936 CIN3+ 0.895 0.85 0.93 Table 31. Area under the partial sROC curves for respectively VIA obtained by integration over the range of observed FPR values for VILI (between 8.3% and 30.0%). Partial Test Outcome AUC 95% CI VIA CIN2+ 0.169 0.150 0.184 VILI CIN2+ 0.175 0.134 0.201 177 6.4. Influence of study characteristics on the prevalence of disease The overall prevalence of cervical disease by category of severity is documented in Table 32. The relation between the prevalence of cervical disease and age group is plotted in Figure 12. The relation between CIN2+ and age is further detailed by study setting in Figure 13. Overall, the prevalence of CIN2+, CIN3+ and cancer increased modestly with age. After the age of 60 the slope of the curve increased substantially. A marked increase of CIN2+ prevalence, in the older age groups, could be observed in Brazzaville, Conakry and Kolkata 1. In Harare the prevalence of CIN2+ was higher than in the other settings. Here, the prevalence decreased until the age group 50-54 and increased thereafter. There was no important change by study phase (see Table 33). Table 32. Prevalence of cervical intra-epithelial neoplasia and invasive cervical cancer and 95% CI. Category Prevalence 95% CI CIN1+ 9.0% 8.8% 9.2% CIN2+ 2.5% 2.4% 2.7% CIN3+ 1.3% 1.2% 1.4% Cancer 0.48% 0.43% 0.53% Table 33. Prevalence of CIN2+ by study phase. Period Prevalence 95% CI 1 2.8% 2.6% 3.0% 2 2.2% 2.1% 2.4% 3 2.6% 2.4% 2.8% CIN 1 CIN 2 CIN 3 Cancer .2 .15 Prevalence .1 .05 0 30 35 40 45 50 55 60 65 Age group Figure 12. Relation between prevalence of cervical neoplasia by category of disease and by 5-year age group. 178 CIN 2+ . Ko1 .3 Br .2 Prevalence Br Br Co Ko1 .1 Br Co Ba Br Br Ko1 Br Br Ba Br Ko1 Co Ba Ko1 Co Ko1 Ba Ba Ko1 Ba Ko2 Ja Ko1 Co Ja Ja Ja Ja Ko2 Co Ko2 Ko2 Ko2 Co Ba Co Ja Ko2 Ja Ko2 Ko2 0 25 30 35 40 45 50 55 60 65 Age group CIN 2+ . .3 .2 Prevalence Ha Ha Ha .1 Tr2 Ha Ha Mu Ou Tr1 Ha Tr1 Ou Ou Tr1 Ha Tr1 Ou Tr1 Mu Tr2 Tr1 Tr1 Tr1 Ou Tr1 Mu Tr2 Tr2 Tr2 Mu Tr2 Mu Tr2 Mu Tr2 Mu Ni Ou Ou Ni Tr2 Ni Ni Mu Ni Ni Ni Ni Ni 0 25 30 35 40 45 50 55 60 65 Age group Figure 13. Relation between the prevalence of CIN2+ and age group by setting. 179 6.5. Influence of study covariates on test characteristics 6.5.1. Variation of the test positivity rate by age group The variation of test results by age is illustrated for all five studied screening tests in a series of line plots (see Figure 14-Figure 16) and assessed statistically using Pearson's chi square test and a trend test [Sasieni, 1996]. The variation by age in VIA-test positivity rate was not homogenous over the different study locations. The variation by age was significant in most settings, with the exception of Niamey, Harare and Busia. A significant increase in test positivity with increasing age was noticed (p value for trend test <0.05) in Brazzaville, Conakry and Trivandrum. In Conakry in particular, a remarkable age trend was found, with VIA test positivity increasing from 2% in the age group 25-29 to 33% among women 60-64 years old. On the other hand, a significant decrease was found in Kolkata 1 and 2, Mumbai, Harare, Ambilikai and Accra. VIA VIA . . .4 .4 Br Co Ou Br Tr1 Ou Tr1 Test positiviity rate Test positiviity rate Br Ou .3 Ja Br .3 Ou Tr1 Ja Br Ja Br Tr1 Ja Br Br Tr1 Ko1 Ja Tr1 Ko1 Ou Tr1 Ja .2 Co .2 Tr1 Mu Ja Mu Ou Ko1 Mu Ou Co Ko2 Ba Ko1 Ba Ko2 Ba Ba Ko1 Co Ko1 Ko2 Ko1 Mu Ko2 Mu .1 Ba .1 Ko2 Mu Ni Co Co Ba Mu Ni Mu Ba Ni Ni Ni Ni Ni Ni Ko2 Ko2 Co Co 0 0 25 30 35 40 45 50 55 60 65 25 30 35 40 45 50 55 60 65 Age group Age group VIA VIA . . Ha Ha .4 Ha .4 Ha Ha Test positiviity rate Test positiviity rate Ha .3 .3 Ha Bu Tr2 Bu Bu .2 .2 Tr2 Ac Ch Ch Ka Ch Ac Ka Tr2 Ac Ch Ph Ka Ph Pa Am Am Tr2 Pa Pa Ac Ac Ph Ka Tr2 Tr2 Tr2 Tr2 .1 .1 Bu Am Pa Ph Am Am Am Am 0 0 25 30 35 40 45 50 55 60 65 25 30 35 40 45 50 55 60 65 Age group Age group Figure 14. Variation in test positivity rate of VIA by age group and study setting. Settings are identified by the first 2 characters. The changes by age in VILI positivity rate were generally smaller than for VIA, but these changes were still significant in Brazzaville, Conakry, Mumbai, Ouagadougou and Trivandrum 1. In Conakry the same spectacular increase over age could be discerned as already remarked for VIA. The change in VIAM positivity rate was homogenous in the three locations where it was evaluated, decreasing slightly by age until the age group 50-59 and increasing among women older than 60 years. The variation in HC2 positivity rate was limited in the four Indian sites (Kolkata 1 & 2, Mumbai and Trivandrum): between 7 and 9% between 30-54 years. In Harare, the HPV 180 positivity rate was very high. Half of the women between 25 and 29 years old were HC2- positive. The rate decreased to 23% among women of 45-49 years. After that age an increase was noticed. Cytological positivity changed significantly over age groups in all settings where it was evaluated and for all cytological cutoffs, with the exception of the HSIL rate in Trivandrum 2 and Harare. In general, cytological rates increased slightly with age. The continuous increase by age observed in Kolkata 1 is notable. Table 34. Variation of the VIA test positivity rate by 5-year age group and by study location (computed only for the groups representing at least 50 women). Variation is assessed statistically by Pearson's chi-square test and a trend test. 5-year age group (first year of interval) chi2 Trend Direction Setting Total 25 30 35 40 45 50 55 60 65 (Pearson) test trend Bamako 5552 % 11.2% 12.9% 12.1% 9.1% 7.4% 12.7% 0.02 0.08 N 1687 1346 1106 822 393 181 Brazzaville 6935 % 24.3% 26.4% 24.6% 26.4% 28.8% 30.0% 33.5% 32.0% 0.02 0.00 + N 1008 1498 1413 1321 878 483 230 97 Conakry 8627 % 2.3% 4.6% 7.9% 7.7% 11.7% 14.5% 19.1% 32.9% 0.01 0.00 + N 880 2283 1902 1519 1175 557 241 70 Jaipur 5786 % 24.4% 27.0% 26.6% 27.7% 22.7% 20.2% 16.8% 0.00 0.06 N 1126 1696 1223 859 498 253 131 Kolkata 1 5894 % 22.8% 20.6% 13.6% 12.2% 11.3% 10.7% 15.4% 0.00 0.00 - N 2697 1299 801 576 238 169 104 Kolkata 2 8080 % 12.6% 12.7% 11.4% 10.6% 6.3% 7.0% 17.7% 0.00 0.00 - N 3709 1938 1072 749 337 183 73 Mumbai 4004 % 16.2% 16.7% 11.1% 10.6% 7.2% 9.8% 0.00 0.00 - N 616 1081 937 698 405 215 Niamey 2534 % 6.4% 5.8% 6.3% 5.8% 5.6% 5.9% 0.99 0.96 N 141 676 646 600 302 118 Ouagadougou 2051 % 21.4% 30.2% 33.3% 29.9% 27.7% 15.9% 17.2% 0.00 0.60 N 556 430 429 318 191 69 58 Trivandrum 1 4457 % 18.1% 23.8% 21.0% 22.1% 25.8% 28.1% 32.2% 30.5% 24.1% 0.00 0.00 + N 166 513 1098 894 841 445 301 141 58 Trivandrum 2 4759 % 18.1% 14.7% 11.6% 12.7% 10.4% 10.8% 10.0% 22.2% 0.02 0.07 N 72 715 1111 1067 891 548 289 54 Harare 2206 % 40.0% 44.5% 39.5% 35.1% 33.6% 29.0% 0.06 0.04 - N 880 488 392 242 110 62 Ambilikai 31317 % 12.7% 12.0% 8.5% 5.2% 2.6% 1.9% 5.6% 0.00 0.00 - N 12946 5886 4267 3648 2532 1949 89 Busia 2295 % 21.4% 20.8% 21.9% 0.67 0.92 N 1074 946 128 Roi-Et 1990 % 12.8% 13.7% 14.3% 8.3% 7.7% 0.03 0.98 N 2257 1987 1572 169 Accra 3665 % 16.1% 14.0% 10.6% 11.3% 13.7% 0.00 0.00 - N 1016 953 928 635 124 181 VILI VILI . . .4 Co .4 Ou Mu Ou Ou Test positiviity rate Test positiviity rate .3 Ja Ja .3 Ou Tr1 Tr1 Ja Ja Ni Ja Ja Br Br Tr1 Tr1 Co Ko2 Mu Ja Ou Tr1 Mu .2 .2 Tr1 Tr2 Mu Tr1 Tr1 Mu Br Br Co Ou Ko2 Br Br Tr2 Tr2 Ba Br Co Ko2 Ko2 Tr2 Ou Tr2 Ba Br Ba Ba Tr2 Ko2 Ko2 Mu Ba Co Co Ba Tr2 Mu Ba Ni Ni Ko2 Mu Ni Tr2 .1 Co .1 Ni Ni Ni Ni Co 0 0 25 30 35 40 45 50 55 60 25 30 35 40 45 50 55 60 Age group Age group VIAM HC2 . . Ha .5 Ha Ha .4 .4 Test positiviity rate Test positiviity rate Ha .3 .3 Ha Mu Ha Ha Ko1 Ko1 Mu .2 Mu .2 Mu Ko1 Ko2 Mu Ko1 Ko1 Ko2 Ko2 Mu Ko2 Ko2 Ko2 Ko1 Mu Ko1 .1 Mu Ko1 Mu .1 Mu Tr2 Mu Ko2 Tr2 Ko2 Mu Ko2 Ko2 Tr2 Ko1 Tr2 Ko1 Mu Mu Tr2 Ko1 Tr2 Tr2 Ko1 Ko2 Ko2 Ko2 Ko1 Ko2 Mu Ko1 Tr2 Mu 0 0 25 30 35 40 45 50 55 60 25 30 35 40 45 50 55 60 Age group Age group Figure 15. Variation in test positivity rate of VILI, VIAM and HPV detection using the HC2 assay, by age group and study setting. Settings are identified by the first 2 characters. Cytology, ASCUS+ Cytology, LSIL+ . . .4 Ko1 .4 Test positiviity rate Test positiviity rate .3 Ha Ha Ha Ha Ko1 .3 Ko1 Ko1 Ha Tr1 .2 Ko1 Tr1 .2 Ko1 Tr1 Ko1 Tr1 Tr1 Ko1 Tr1 Tr1 Tr1 Tr1 Tr1 Ha Ha Ha Tr1 Ha Tr1 Ko1 Ha Tr1 Tr1 Ko1 Ha .1 Ko1 Ja Ja Ja Ja Ja Ja Tr2 .1 Tr1 Tr1 Ko1 Ha Ja Tr2 Mu Ja Ja Mu Tr2 Ko1 Mu Tr2 Ja Ko1 Ja Ja Tr2 Mu Tr2 Mu Tr2 Tr2 Tr2 Mu Tr2 Mu Ja Tr2 Mu Tr2 Mu Mu Tr2 Mu Ha Mu Mu Tr2 Mu Tr2 Mu Ja Tr2 Mu Ha 0 0 25 30 35 40 45 50 55 60 25 30 35 40 45 50 55 60 Age group Age group Cytology, HSIL+ . .4 Test positiviity rate .3 .2 Ko1 .1 Ko1 Ko1 Mu Tr1 Ha Tr2 Ha Ha Ko1 Ko1 Tr1 Ha Tr1 Ha Mu Tr1 Mu Tr1 Mu Tr2 Tr1 Tr2 Ko1 Ko1 Tr1 Tr2 Mu Tr1 Tr2 Ja Tr2 Mu Ha Ja Tr2 Ja Ja Mu Tr2 Ja Mu Ja Ja Ha 0 25 30 35 40 45 50 55 60 Age group Figure 16. Variation in test positivity rate of the Pap smear, defined at 3 different cutoffs (ASCUS+, LSIL+ and HSIL+) by age group and study setting. Settings are identified by the first 2 characters. 182 Table 35. Variation of the prevalence of HC2 positivity by 5-year age group in four Indian locations and in Harare (computed only for groups representing at least 50 women). Variation is assessed statistically by Pearson's chi-square test and a trend test. 5-year age group (first year of interval) chi2 Trend Direction Setting Total 25 30 35 40 45 50 55 60 (Pearson) test trend Kolkata 1 & 2 22737 % 4.8% 7.1% 6.7% 6.7% 6.7% 7.1% 8.6% 12.2% 0.10 0.13 Mumbai, Trivandrum 2 N 84 5979 4411 3302 2494 1346 746 189 Harare 2206 % 49.7% 46.6% 39.9% 26.1% 22.7% 33.3% 0.00 0.00 - N 884 489 398 245 110 63 6.5.2. Variation of the test positivity rate by time period We subdivided the study subjects in three period groups based on the tertiles of the screening date or the rank number in the Zimbabwe data file. The change of the test positivity rate by time period and study location is shown in Figure 17. Important significant period effects can be noticed for the visual screening methods. In general, the positivity rate of VIA and VILI dropped over time. The Hybrid Capture II positivity rate remained remarkably constant over time (see Table 36). In the Indian trials a significant but relatively weak period effect was observed. The time trend of HC2 positivity in Harare was insignificant. Changes in cytological positivity rates were less important than for the direct visual methods but, nevertheless, the period variation was most often statistically significant. Table 36. Change in prevalence of HC2 positivity by period in the Indian study location and Harare. Period chi2 Direction Setting Total 1 2 3 (Pearson) Trend test trend Kolkata 1 & 2, Mumbai, Trivandrum 2 18554 6.2% 7.3% 7.3% 0.03 0.03 + Harare 2199 44.5% 43.4% 40.1% 0.20 0.08 183 VIA VIA . . Br Ou .4 Tr1 Test positiviity rate Test positiviity rate .4 .3 Ko1 Ou Tr1 Ja .3 Ja .2 Ou Tr1 Ja Ko2 .2 Ba Mu Mu Co Ko1 Mu Ko2 Br Ko1 Br .1 Ko2 .1 Ba Ba Ni Ni Ni Co Co 0 0 1 2 3 1 2 3 Period Period VIA VILI . . Ha Ha .4 .4 Ja Test positiviity rate Test positiviity rate Ha .3 Ja .3 Ja Bu Bu Ba .2 Co .2 Br Br Bu Br Tr2 Tr2 Am Tr2 Co Ba .1 Am Am .1 Ba Co 0 0 1 2 3 1 2 3 Period Period VILI VIAM . . Ou .5 .4 .4 Test positiviity rate Test positiviity rate .3 Tr1 Ou .3 Ko1 .2 Mu Ou Tr1 Mu Ko2 Tr2 Tr1 Tr2 .2 Mu Ko2 Mu Mu Ko2 Tr2 Mu Ko2 Ko1 Ko1 Ni Ni Ko2 .1 Ni .1 Ko2 0 0 1 2 3 1 2 3 Period Period HC2 Cytology, LSIL+ . . Ha Ha .4 Ha .4 Test positiviity rate Test positiviity rate .3 .3 .2 .2 Ha Tr1 Tr1 Tr1 Ha Ja Ko1 Ha .1 Mu .1 Tr2 Ko1 Ko1 Mu Ko2 Tr2 Tr2 Ko1 Ko1 Mu Ko2 Ko2 Tr2 Tr2 Mu Ko1 Ja Mu Mu Tr2 Ja 0 0 1 2 3 1 2 3 Period Period Figure 17. Variation of the test positivity rate by period and by study location. 184 6.5.3. Influence of age, period and study setting on the PPV We used a logistic model to assess the impact of age group, time period (defined by the tertiles of the screening date), and study location on the positive predictive value, considering CIN2 or worse disease as outcome. In Table 37, we show the results of the logistic regression describing the relation of the PPV of VIA for CIN2+. It shows, first of all, the PPV0 for the base case, this means the fitted PPV in age group 30-34, in the first period in Bamako is 13%. Table 37. Multivariate variation of the PPV for CIN2+ of VIA by age group, period and study location. Basic PPV 13.3% Age groupa RR 95% CI 30 1 35 1.07 0.93 1.22 40 1.26 1.09 1.46 45 1.41 1.20 1.65 50 1.61 1.32 1.94 55 2.54 2.12 3.00 60 2.64 2.00 3.35 Period RR 95% CI 1 1 2 1.49 1.33 1.68 3 1.87 1.68 2.07 Setting RR 95% CI Bamako 1 Brazzaville 0.77 0.62 0.95 Conakry 1.00 0.79 1.26 Jaipur 0.16 0.11 0.23 Kolkata 1 0.42 0.32 0.55 Kolkata 2 0.22 0.16 0.31 Mumbai 0.36 0.26 0.51 Niamey 0.36 0.20 0.64 Ouagadougou 0.42 0.29 0.59 Trivandrum 1 0.43 0.33 0.55 Trivandrum 2 0.41 0.30 0.56 Harare 0.67 0.50 0.88 Ambilikai 0.41 0.33 0.51 The PPV increases with age and period. The change by period was significant. The PPV of VIA was highest in Bamako, Brazzaville and Conakry. The interaction between setting and age was significant. However for simplicity we only showed the main effects. We can compare the relative risks for the study locations with the forest plot showing the meta-analysis of the PPV of VIA for CIN2+ presented in chapter 6.1.1. To assist the reader we redisplay this forest plot below. a Women of 20-24 years were excluded from the logistic regression since this age group was not enrolled in all studies. 185 VIA, Outcome=CIN2+ Bamako Brazzaville Conakry Jaipur Kolkata 1 Kolkata 2 Mumbai Niamey Ouagadougou Trivandrum 1 Trivandrum 2 Harare Ambilikai Busia Combined 0 .1 .2 .3 .4 .5 .6 PPV Figure 18. Forest plot showing the PPV of VIA for CIN2+ observed in 14 study locations. The results from the logistic regression targeting the PPV of the other tests are shown in Table 38,Table 39 and Table 40. Remark that the PPV of the HC2, contrary to the other tests, does not vary by period. 186 Table 38. Multivariate variation of the PPV for CIN2+ of VILI by age group, period and study location Basic PPV 16.0% Age group RR 95% CI 30 1 35 1.16 0.97 1.36 40 1.32 1.11 1.55 45 1.41 1.18 1.68 50 1.45 1.16 1.79 55 2.13 1.73 2.57 60 2.08 1.50 2.75 Period RR 95% CI 1 1 2 1.20 1.05 1.37 3 1.40 1.22 1.59 Setting RR 95% CI Bamako 1 Brazzaville 1.45 1.22 1.72 Conakry 0.70 0.56 0.87 Jaipur 0.14 0.10 0.20 Kolkata 2 0.20 0.14 0.27 Mumbai 0.33 0.24 0.45 Niamey 0.33 0.20 0.53 Ouagadougou 0.40 0.29 0.56 Trivandrum 1 0.52 0.41 0.67 Trivandrum 2 0.35 0.26 0.48 Table 39. Multivariate variation of the PPV for CIN2+ of the HC2 test by age group, period and study location Basic PPV 7.9% Age group RR Upper lower 30 1 35 1.04 0.74 1.45 40 1.73 1.23 2.39 45 1.17 0.74 1.81 50 1.14 0.65 1.92 55 1.78 0.97 3.08 60 3.05 1.35 5.81 Period RR Upper lower 1 1 2 0.98 0.71 1.34 3 0.95 0.69 1.30 Setting RR Upper lower Kolkata 1 1 Kolkata 2 1.14 0.71 1.77 Mumbai 1.62 1.02 2.50 Trivandrum 2 1.80 1.15 2.72 Harare 2.18 1.47 3.13 187 Table 40. Multivariate variation of the PPV for CIN2+ of cytology (at cutoff LSIL+) by age group, period and study location. Basic PPV 4.6% Age group RR 95% CI 30 1 35 0.81 0.53 1.24 40 1.51 1.01 2.23 45 1.17 0.74 1.82 50 1.08 0.64 1.77 55 1.68 0.97 2.85 60 1.73 0.89 3.25 Period RR 95% CI 1 1 2 1.10 0.81 1.50 3 1.19 0.87 1.61 Setting RR 95% CI Jaipur 1 Kolkata 1 2.02 1.11 3.55 Mumbai 10.67 6.90 14.48 Trivandrum 1 3.24 1.86 5.36 Trivandrum 2 6.78 4.05 10.28 Harare 7.00 4.22 10.50 188 6.5.4. Influence of study characteristics on the sensitivity and specificity of screen tests The effect of the 5-year age group (restricted to women between 30 and 64), the study phase (1, 2 or 3) or the study location on sensitivity and specificity was explored using logistic regression. The sensitivity did not vary by age. A period effect was noticed for VIA, with higher sensitivity in the third study phase. The sensitivity always differed by study location. Almost all effects are significant when prediction of absence of disease (specificity) is explored. There was one exception, the specificity of HC2 was not influenced by study phase. As already noticed by exploring the forest plots, we notice also here a simultaneously outlying high sensitivity and specificity in Conakry and the opposite low sensitivity and specificity are found in Kolkata 1 and Mumbai. For VILI, we remark simultaneous low sensitivity and specificity in Kolkata 1 & 2, Mumbai and Trivandrum II. The sensitivity of HC2 was considerably higher in Harare, compared with the Indian settings. Table 41. Variation of the sensitivity (at left) and specificity (at right) of VIA for presence or absence of CIN2+, according to age group, period or study setting, assessed by multivariate logistic regression. Non- significant effects are omitted. VIA, accuracy for CIN2+ Basic sensitivity 75.8% Basic specificity 84.8% Age group RR 95% CI Age group RR 95% CI 30 30 1.00 35 35 1.01 1.00 1.02 40 40 1.02 1.01 1.03 45 45 1.02 1.01 1.03 50 50 1.03 1.02 1.04 55 55 1.03 1.01 1.05 60 60 0.98 0.94 1.02 Period RR 95% CI Period RR 95% CI 1 1 1 1 2 1.03 0.96 1.10 2 1.09 1.09 1.10 3 1.14 1.08 1.18 3 1.09 1.08 1.09 Setting RR 95% Setting RR 95% CI Bamako 1 Bamako 1 Brazzaville 0.98 0.85 1.08 Brazzaville 0.75 0.72 0.78 Conakry 1.17 1.07 1.24 Conakry 1.05 1.03 1.06 Jaipur 1.10 0.89 1.22 Jaipur 0.75 0.72 0.78 Kolkata 1 0.77 0.61 0.92 Kolkata 1 0.85 0.83 0.88 Kolkata 2 0.87 0.67 1.04 Kolkata 2 0.98 0.96 1.00 Mumbai 0.71 0.52 0.90 Mumbai 0.95 0.92 0.97 Niamey 0.79 0.47 1.06 Niamey 1.07 1.05 1.09 Ouagadougou 1.20 1.01 1.28 Ouagadougou 0.70 0.66 0.74 Trivandrum 1 1.13 1.00 1.21 Trivandrum 1 0.77 0.74 0.80 Trivandrum 2 0.99 0.80 1.13 Trivandrum 2 0.96 0.94 0.99 Harare 0.88 0.70 1.03 Harare 0.57 0.53 0.62 189 Table 42. Variation of the sensitivity and specificity of VILI for presence or absence of CIN2+, according to age group, period or study setting, assessed by multivariate logistic regression. Sensitivity and specificity of VILI for CIN2+ VILI Basic sensitivity 97.0% Basic specificity 86.7% Age group RR 95% CI Age group RR 95% CI 30 30 1 35 35 0.99 0.98 1.00 40 40 1.00 0.99 1.01 45 45 1.01 0.99 1.02 50 50 1.01 1.00 1.03 55 55 1.01 0.99 1.02 60 60 0.92 0.87 0.96 Period RR 95% CI Period RR 95% CI 1 1 1.00 2 2 1.04 1.04 1.05 3 3 1.04 1.04 1.05 Setting RR 95% CI Setting RR 95% CI Bamako 1 Bamako 1 Brazzaville 0.98 0.90 1.01 Brazzaville 1.00 0.98 1.01 Conakry 1.00 0.93 1.02 Conakry 1.01 0.99 1.02 Jaipur 0.89 0.68 0.98 Jaipur 0.79 0.76 0.81 Kolkata 2 0.84 0.62 0.96 Kolkata 2 0.96 0.94 0.98 Mumbai 0.76 0.51 0.91 Mumbai 0.93 0.90 0.95 Niamey 0.93 0.64 1.01 Niamey 1.03 1.01 1.05 Ouagadougou 1.01 0.86 1.03 Ouagadougou 0.75 0.71 0.79 Trivandrum 1 0.94 0.80 1.00 Trivandrum 1 0.89 0.86 0.91 Trivandrum 2 0.82 0.60 0.95 Trivandrum 2 0.96 0.94 0.98 Table 43. Variation of the sensitivity and specificity of the Hybrid Capture II assay for presence or absence of CIN2+, according to age group, period or study setting, assessed by multivariate logistic regression. Sensitivity and specificity of HC2 for CIN2+ HC2 Basic sensitivity 48.4% Basic specificity 91.3% RR RR 95% CI Age group RR 95% CI 30 30 1 35 35 1.01 1.00 1.02 40 40 1.03 1.01 1.04 45 45 1.02 1.01 1.03 50 50 1.01 0.99 1.03 55 55 1.00 0.97 1.02 60 60 0.97 0.90 1.01 Period RR 95% CI Period RR 95% CI 1 1 2 2 3 3 Setting RR 95% CI Setting RR 95% CI Kolkata 1 1.00 Kolkata 1 1 Kolkata 2 1.15 1.02 1.23 Kolkata 2 1.03 1.02 1.04 Mumbai 1.13 0.99 1.22 Mumbai 1.02 1.00 1.03 Trivandrum 2 1.13 1.00 1.22 Trivandrum 2 1.03 1.02 1.04 Harare 1.24 1.17 1.28 Harare 0.70 0.66 0.74 190 Table 44. Variation of the sensitivity and specificity of the Pap smear (at cutoff LSIL+) for presence or absence of CIN2+, according to age group, period or study setting, assessed by multivariate logistic regression. Sensitivity and cytology at LSIL+ for CIN2+ Cytology, LSIL+ Basic sensitivity 16.6% Basic specificity 94.8% Age RR 95% CI Age group RR 95% CI 30 1 30 1 35 0.46 17.24 0.90 35 0.98 0.97 0.99 40 1.01 0.82 1.10 40 0.97 0.95 0.98 45 0.96 0.56 1.09 45 0.96 0.95 0.98 50 1.01 0.65 1.12 50 0.94 0.92 0.96 55 1.00 0.55 1.12 55 0.95 0.92 0.97 60 1.07 0.61 1.17 60 0.88 0.83 0.93 Period RR 95% CI Period RR 95% CI 1 1 1 2 2 1.03 1.02 1.03 3 3 1.03 1.02 1.03 Setting RR 95% CI Setting RR 95% CI Jaipur 1 Jaipur 1 Kolkata 1 0.72 13.43 1.03 Kolkata 1 0.97 0.96 0.98 Mumbai 1.05 0.78 1.14 Mumbai 1.05 1.04 1.05 Trivandrum 1 1.15 1.07 1.19 Trivandrum 1 0.94 0.92 0.96 Trivandrum 2 1.07 0.85 1.15 Trivandrum 2 1.04 1.03 1.04 Harare 1.05 0.79 1.14 Harare 0.96 0.93 0.98 191 6.5.5. Influence of study characteristics on the diagnostic odds ratio using multi-variate SROC-regression analysis The logistic regression allowed us to assess the influence of study characteristics on each dichotomous diagnostic parameter separately. By using sROC-regression we can evaluate the impact of covariates simultaneously on sensitivity, specificity and overall accuracy. The sROC regression analysis is restricted to the same study covariates assessed in the previous sub-chapters, which are available for all included studies: 5-year age group, the study phase (study period subdivided in tertiles), and the study location. In Table 45 we show the coefficients of significant terms included in the linear sROC regression equation describing the relation between D and S (which are the difference [D] and the sum [S] of the logits of the true and false positivity rates). We only document the accuracy of VIA, VILI, HC2 and cytology (at the cutoff LSIL) for CIN2+. Interested readers can contact us to obtain tables concerning the accuracy of VIAM, cytology at other cutoffs, and the accuracy for other outcomes. The prediction of presence or absence of CIN2+ was influenced significantly by the study phase and study location when VIA, VILI, VIAM or cytology was used. The age group never influenced test accuracy. The HC2 assay was the only screening test whose accuracy did not vary significantly by study period. Table 45. Coefficients of terms included in the sROC regression equation, t-test for significance of the hypothesis that coefficients are zero and 95% confidence interval around the estimate of the coefficient. VIA, reference setting = Bamako Term Coef. Std.er t P>t (95% CI) S 0.26 0.05 4.75 0.00 0.15 0.37 Period 2 0.83 0.16 5.05 0.00 0.50 1.15 Period 3 1.04 0.16 6.69 0.00 0.74 1.35 Brazzaville -1.22 0.37 -3.28 0.00 -1.96 -0.48 Conakry 0.98 0.43 2.29 0.02 0.13 1.84 Jaipur -2.21 0.50 -4.39 0.00 -3.20 -1.21 Kolkata 1 -1.51 0.37 -4.10 0.00 -2.24 -0.78 Kolkata 2 -0.94 0.41 -2.29 0.03 -1.76 -0.12 Mumbai -1.34 0.37 -3.65 0.00 -2.07 -0.61 Niamey 0.11 0.54 0.20 0.85 -0.97 1.18 Ouagadougou -1.34 0.63 -2.14 0.04 -2.59 -0.10 Trivandrum 1 -1.71 0.42 -4.02 0.00 -2.55 -0.86 Trivandrum 2 -0.50 0.44 -1.12 0.26 -1.37 0.38 Harare -2.69 0.43 -6.22 0.00 -3.55 -1.83 Intercept 3.24 0.30 10.70 0.00 2.64 3.85 192 VILI, reference country = Burkina Faso Term Coef. Std.er t P>t 95% CI S 0.38 0.10 3.65 0.00 0.17 0.59 Period 2 0.46 0.28 1.64 0.11 -0.10 1.02 Period 3 0.73 0.25 2.87 0.01 0.22 1.24 Congo 2.41 0.88 2.75 0.01 0.65 4.17 Guinea 2.10 0.94 2.22 0.03 0.20 3.99 India 0.67 0.89 0.75 0.46 -1.12 2.45 Mali 2.44 0.91 2.66 0.01 0.60 4.27 Niger 2.30 1.10 2.08 0.04 0.08 4.51 Constant 1.74 0.85 2.06 0.04 0.05 3.44 VIAM, reference setting = Kolkata 1 Term Coef. Std.er t P>t 95% CI S 0.11 0.11 0.99 0.33 -0.12 0.34 Period 2 0.46 0.24 1.96 0.06 -0.02 0.95 Period 3 0.59 0.25 2.38 0.02 0.08 1.10 Kolkata 2 0.63 0.26 2.42 0.02 0.10 1.16 Mumbai -0.03 0.23 -0.15 0.88 -0.50 0.43 Constant 2.15 0.25 8.60 0.00 1.64 2.66 HC2 (reference setting = Kolkata 1) Term Coef. Std.er t P>t 95% CI S Kolkata 2 0.94 0.34 2.78 0.01 0.25 1.62 Mumbai 0.40 0.33 1.22 0.23 -0.27 1.08 Trivandrum 2 0.88 0.31 2.81 0.01 0.25 1.52 Harare -2.69 0.53 -5.09 0.00 -3.75 -1.62 Constant 4.48 0.43 10.39 0.00 3.61 5.35 Cytology at cutoff LSIL+ (reference setting=Jaipur) Term Coef. Std.er t P>t 95% CI S 0.19 0.10 1.91 0.06 -0.01 0.38 Period 2 0.55 0.28 1.92 0.06 -0.02 1.12 Period 3 0.76 0.28 2.68 0.01 0.19 1.33 Kolkata 1 -1.41 0.45 -3.10 0.00 -2.31 -0.50 Mumbai 1.87 0.47 3.97 0.00 0.92 2.81 Trivandrum 1 -0.39 0.51 -0.77 0.44 -1.42 0.63 Trivandrum 2 1.59 0.46 3.49 0.00 0.68 2.51 Harare -0.85 0.49 -1.73 0.09 -1.83 0.14 Constant 3.31 0.51 6.52 0.00 2.29 4.33 The interpretation of the sROC-regression coefficients is not straightforward. Let us take the example of the first table concerning VIA. The coefficient of the S term is different from zero. This means that the accuracy (diagnostic odds ratio=DOR) depends on a latent cutoff even when in theory VIA is a dichotomous test. The accuracy of VIA in the first period in Bamako is taken as reference. The diagnostic odds ratio increases in the second and third period. This means that the area under the sROC curve is greater in last two 193 periods than in the first period. The accuracy of VIA (DOR) is greater in Conakry and Niamey than in Bamako. In the other settings, the DOR is lower. The relative size of an effect can be computed by taking the anti-log of the effect. For instance: the DOR in Conakry is e0.98= 2.7 (95% CI: e0.13 or 1.1 to e1.84 or 6.3). 6.6. Correlations between tests The rank correlation between the respective evaluated screening tests is shown in Table 46. Table 46. Correlation between screening tests and colposcopy assessed with the rank correlation coefficient of Spearman. (*Spearman coefficient not significantly different from 0). VIA VILI VIAM HC2 Cytology Colposcopy VIA - 0.83 0.86 0.07* 0.18* 0.36 VILI - - 0.63 -0.02* 0.11 0.74 VIAM - - - 0.04 0.08 0.64 HC2 - - - - 0.16 0.21 Cytology - - - - - 0.11 Colposcopy - - - - - - The degree of correlation is high to very high for all visual testing methods: VIAM, VILI, VIAM and colposcopy. The correlations between the other tests are substantially lower. The coefficients of correlation between HC2 and VIA, between HC2 and VILI, and between cytology (LSIL+) and VIA were not significantly different from zero. 6.7. PPV of colposcopy triage Using a "see & treat strategy", requires treatment of all eligible VIA-positive women. This strategy was applied in Ghana and Thailand. In all other settings a "see-triage-treat" strategy was applied using colposcopy to decide which women to treat. The forest plot of the colposcopy positivity rate among VIA-positive women ,defined at low- and high-grade colposcopy, is shown in Figure 19. The pooled "low-grade+" rate was 64.7% (CI: 53.4% to 76.1%). The rate varied over a broad range. The two outlying high rates were at Jaipur (92%) and Ambilikai (93%). The positivity rate defined at high-grade or worse was substantially lower and varied less; its pooled value was 18.4% (CI: 14.7% to 22.0%). The PPVs of a low-grade colposcopy result for the four outcomes CIN1+, CIN2+, CIN3+ and cancer are shown in Figure 20. For comparison, we also show the PPV of VIA without colposcopy triage. By adding colposcopy triage the PPV is increased by 59%, 13% and 62% when the outcomes CIN2+, CIN3+ or cancer respectively are considered. 194 Table 47. PPV of VIA and of VIA followed by colposcopy, where the cutoff for a positive colposcopy result is determined at low-grade or worse. VIA VIA + & colposcopy triage Outcome PPV 95% CI PPV 95% CI Ratio 95% CI CIN1+ 30.6 21.1 40.0 45.3 32.5 18.1 1.48 1.54 0.45 CIN2+ 11.8 9.1 14.6 18.8 13.5 24.0 1.59 1.48 1.64 CIN3+ 5.3 4.0 6.6 6.0 4.1 7.8 1.13 1.03 1.18 Cancer 2.1 1.6 2.5 3.4 2.4 4.5 1.62 1.50 1.80 195 Figure 19. Meta-analysis of the test positivity rate for colposcopy, applied to VIA-positive women, defined as low-grade colposcopy or worse (at left) or high-grade colposcopy (at right). Colposcopy at cut-off low-grade Colposcopy at cut-off high-grade Among VIA+ women Among VIA+ women Bamako Bamako Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Niamey Niamey Ouagadougou Ouagadougou Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Ambilikai Ambilikai Busia Busia Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Positivity rate Positivity rate Meta-analysis of the colposcopy test positivity rate (low-grade+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.792 0.786 0.798 254.401 0.000 14 Random | 0.647 0.534 0.761 11.206 0.000 Test for heterogeneity: Q= 3703.764 on 13 degrees of freedom (p= 0.000) Meta-analysis of the colposcopy test positivity rate (high-grade+) | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.163 0.157 0.170 51.762 0.000 14 Random | 0.184 0.147 0.220 9.790 0.000 196 Figure 20. PPV of low-grade colposcopy applied to VIA-positive women for presence of CIN1+, CIN2+, CIN3+ or cancer. Colposcopy at cut-off low-grade, outcome=CIN1+ Colposcopy at cut-off low-grade, outcome=CIN2+ Among VIA+ women Among VIA+ women Bamako Bamako Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Niamey Niamey Ouagadougou Ouagadougou Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Harare Ambilikai Ambilikai Busia Busia Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV PPV Colposcopy at cut-off low-grade, outcome=CIN3+ Colposcopy at cut-off low-grade, outcome=Cancer Among VIA+ women Among VIA+ women Bamako Bamako Brazzaville Brazzaville Conakry Conakry Jaipur Jaipur Kolkata 1 Kolkata 1 Kolkata 2 Kolkata 2 Mumbai Mumbai Niamey Niamey Ouagadougou Ouagadougou Trivandrum 1 Trivandrum 1 Trivandrum 2 Trivandrum 2 Harare Ambilikai Ambilikai Busia Busia Combined Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 PPV PPV Meta-analysis of the PPV of low-grade colposcopy for CIN1+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.436 0.427 0.445 94.046 0.000 14 Random | 0.453 0.325 0.581 6.928 0.000 Test for heterogeneity: Q= 2355.179 on 13 degrees of freedom (p= 0.000) Meta-analysis of the PPV of low-grade colposcopy for CIN2+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.103 0.097 0.109 33.758 0.000 14 Random | 0.188 0.135 0.240 7.038 0.000 Test for heterogeneity: Q= 800.461 on 13 degrees of freedom (p= 0.000) Meta-analysis of the PPV of low-grade colposcopy for CIN3+ | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.034 0.031 0.038 17.868 0.000 13 Random | 0.060 0.041 0.078 6.435 0.000 Test for heterogeneity: Q= 203.365 on 12 degrees of freedom (p= 0.000) Meta-analysis of the PPV of low-grade colposcopy for cancer | Pooled 95% CI Asymptotic No. of Method | Est Lower Upper z_value p_value studies -------+---------------------------------------------------- Fixed | 0.010 0.008 0.012 9.889 0.000 14 Random | 0.034 0.024 0.045 6.343 0.000 Test for heterogeneity: Q= 207.181 on 13 degrees of freedom (p= 0.000) 197 6.8. Treatment of VIA-positive cases The data file containing results from the Zimbabwe trial did not contain treatment information. Some treatment data were available for the multi-center cross-sectional study, but these were incomplete and lacked for some settings. Therefore, we considered only four studies (Ambilikai, Busia, Roi-Et and Accra) to evaluate treatment rate and compliance. Table 48 shows the number of women per category of final cervical status and the percentage of them that was treated. More than 80% of women with known CIN2+ status in Ambilikai and Busia got treatment. In Thailand and Ghana more than 90% of VIA- positive women were treated. In Ambilikai 91% of treated women received cryotherapy. Table 48. Percentage of women treated per final cervical status and per setting. Ambilikai Busia Roi-Et Accra Final status N % treated N % treated N % treated N % treated CIN 1 1745 72.1% 68 80.9% CIN 2 153 81.7% 19 94.7% CIN 3 71 81.7% 10 80.0% Cancer 67 80.6% 2 0.0% VIA+ 798 92.5% 484 90.3% Table 49 shows the proportion of women treated with cryotherapy who received screening and treatment on the same day. This percentage was very low in Kenya, since women had to be referred to the hospital for further diagnosis, which was sometimes at a considerable distance from the health centers where VIA screening took place. In the other settings between 71% and 82% received cryotherapy the same day. Table 49. Percentage of women receiving cryotherapy, treated the same day as VIA screening % same N treated day Ambilikai 2355 76.7% Busia 116 3.4% Roi-Et 738 82.2% Accra 437 70.5% The experiences of women and the occurrence of secondary effects of cryotherapy are documented in detail in the Thai and Ghanaian demonstration projects, summarized in Table 17 and Table 19. The main issues of the interviews are further summarized in Table 50. It can be concluded that cryotherapy offered to VIA-positive women the same day as screening is highly accepted and quite safe. Pain during or just after cryotherapy was the most noticed secondary effect (40% in Thailand, 23% in Ghana), which was most often mild (79% in Roi-Et and 86% in Accra). Only in 2% in Roi-Et and 3% in Accra was pain considered severe. In 2% of treated women some minor blood loss occurred. In Thailand, 83% and 93% of treated women presented at the follow-up visits at 3 and 12 months, respectively, after treatment. In Ghana these proportions were 77% and 55%. 198 199 Table 50. Main answers extracted from the Thai and Ghanaian treatment questionnaire (see Table 17 and Table 19). Thailand Ghana Question Definition of the denominator (N) N Percent N Percent Experience about testing Satisfaction about being tested Tested woman 5742 99.9% 3621 79.8% Woman will recommend VIA testing to others Women with negative VIA result 5122 99.8% 3115 99.2% Experience at the moment of treatment Pain reported by provider Treated women 756 39.6% 412 22.6% Pain was mild Treated women reporting pain 290 79.0% 90 85.6% Pain was moderate Treated women reporting pain 290 18.3% 90 12.2% Pain was severe Treated women reporting pain 290 2.8% 90 2.2% An analgesic drug was given Treated women reporting pain 746 6.2% 422 1.7% Bleeding after cryo Treated women 748 1.7% 400 2.0% Bleeding was mild Treated women reporting bleeding 11 100.0% 5 100.0% Problem visit of a health structure Women contacted health service after cryo Treated women 756 4.4% 439 1.1% Women having contacted a health The visit: related to the cryo service 33 21.2% - - Reason of problem visit: Women having contacted a health bleeding related to menses service 33 9.1% - - Reason of problem visit: Women having contacted a health bleeding unrelated to menses service 33 12.1% - - Reason of problem visit: pain/cramping unrelated to Women having contacted a health menses service 33 27.3% - - Reason of problem visit was: Women having contacted a health abnorm vaginal discharge service 34 44.1% - - Reason of problem visit: Women having contacted a health chills/hot sweats/fever service 35 2.9% - - Reason of problem visit: another Women having contacted a health problem service 36 2.8% - - Planned follow-up visit at 3 months Attended at the visit Treated women 756 83.2% 439 77.2% Had pain or cramping not Treated women presenting at 3M FU associated with menses visit 626 13.6% 316 10.1% Less menstrual blood loss since cryo Treated women at 3M FU visit 552 5.6% 337 5.9% More menstrual blood loss since cryo Treated women at 3M FU visit 552 4.3% 337 14.5% Shorter menses Treated women at 3M FU visit 552 4.3% 337 5.0% Longer menses Treated women at 3M FU visit 552 3.1% 337 11.0% More cramping Treated women at 3M FU visit 552 10.0% 337 8.0% Less cramping Treated women at 3M FU visit 552 1.4% 337 9.5% Abnormal vaginal discharge Treated women at 3M FU visit 628 0.5% 310 5.5% Woman will recommend VIA testing to others Treated women at 3M FU visit 627 97.9% 335 98.5% 200 Thailand Ghana Question Definition of the denominator (N) N Percent N Percent Compliance with recommendation concerning sexual hygiene Woman had sex within 4W post- treatment Treated women at 3M FU visit 459 30.9% 229 14.4% No condom use if sex within 4W post-treatment Having had sex <4W post-treatment 142 10.6% 32 3.1% Condom use all the time if sex within 4W post-treatment Having had sex <4W post-treatment 142 79.6% 30 86.7% Planned follow-up at 12 months after treatment Attended at follow-up Treated women 756 93.5% 439 54.9% Suspect for cancer at VIA Treated women, attending at 12 examination month visit 704 0.1% 232 0.0% VIA positive Treated women, at 12 month visit 704 5.7% 232 2.6% 201 7. Discussion The ACCP trials have provided a mass of new data, which we have tried to pool and to synthesize using meta-analytical methods. We can conclude the following: VILI was 10% more sensitive as a screening test than VIA, but had the same specificity; therefore, it seems to be the preferred visual inspection method to detect high-grade CIN in developing countries. This quite spectacular result was only documented in one study, but it was a very large one involving multiple centers. The good triage results of VILI, compared to VIAM and cytology, reported in the Kenyan study give some corroboration. Nevertheless, it is recommended to conduct more research, by unrelated researchers to confirm the findings of the cross-sectional multi-center study conducted by IARC. The inter-study variation of VIA accuracy parameters was very wide, which might reflect its low reproducibility . This is further suggested by the inconsistent relation between VIA positivity and age. The accuracy of VIA increased significantly by study period. It also increased in Trivandrum II and Kolkata II, where the examinations were done by the same teams as in Trivandrum 1 and Kolkata 1. These findings suggest an impact of experience, training and supervision. VIAM requires an additional device but does not show better test characteristics than VIA. Therefore, in low-resource settings, VILI and VIA appear to be preferable screening methods. The significant correlation between all visual screening methods (VIA, VIAM, VILI) with colposcopy, which is intrinsically integrated in the gold standard definition, may have contributed to inflating their sensitivity and specificity. The hybrid capture II showed an unexpectedly low sensitivity (66%) for high-grade CIN in the five Indian settings. There are not many publications where the accuracy of HC2 is documented without verification bias in developing countries. In all retrieved studies, a substantially higher sensitivity was reported. In a Chinese rural area, Belinson [2001] found a sensitivity of HC2 for CIN2+ of 95% and a specificity of 89%. In a study conducted by De Vuyst in Nairobi (Kenya), a PCR system was used to detect HPV, followed by typing for high-risk types, which yielded a sensitivity of 94% (and a specificity of 69%). Also in the Harare study, which is included in our meta-analysis, considerably higher sensitivity (80%) was reported for HC2 (the specificity was 61%) [University of Zimbabwe, 2001]. In a small study conducted in South Kivu, with complete histological gold standard verification, all histologically confirmed CIN2 and CIN3 were detected by HPV testing using GP5+/6+ PCR followed by enzyme-immunoassay [Arbyn, 2000]. Table 51 shows the accuracy parameters of HC2 and PCR tests for the detection of CIN2+ and CIN3+ in the framework of primary screening for cervical cancer. It is derived from a Cochrane review that is currently being finalized where the test performance of HPV testing is compared with cytology screening [Koliopoulos, 2006; Arbyn, 2005]. Figure 1 shows the forest plot representing the variation of the sensitivity of HC2 for CIN2+. From 202 both the table and the forest plot the outlying low sensitivity values reported for the Indian studies become clear. Table 51. Sensitivity, specificity, positive predictive value, negative predictive value of HC2 or HPV DNA tests based on PCR to detect CIN2+ or CIN3+; prevalence of CIN2+ and CIN4+, assessed in the framework of primary screening for cervical cancer. Study Country HPV test Outcome Sense Specie PPV NPV Prev Kuhn, 2000 South-Africa HC2 CIN2+ 0.88 0.80 0.12 1.00 0.033 Ratnam, 2000 Canada HC2 CIN2+ 0.83 0.88 0.17 0.99 0.029 Clavel, 2001 (C. Cyt arm) France HC2 CIN2+ 1.00 0.90 0.18 1.00 0.021 Clavel, 2001 (LBC arm) France HC2 CIN2+ 1.00 0.91 0.15 1.00 0.016 Coste, 2003 France HC2 CIN2+ 0.95 0.85 0.13 1.00 0.023 Cuzick, 2003 UK HC2 CIN2+ 0.97 0.93 0.11 1.00 0.009 Belinson, 2003 China HC2 CIN2+ 0.97 0.80 0.18 1.00 0.044 Petry, 2003 Germany HC2 CIN2+ 0.98 0.95 0.11 1.00 0.006 Salmeron, 2003 Mexico HC2 CIN2+ 0.93 0.93 0.15 1.00 0.013 Sankaranarayanan, K1 2004 India HC2 CIN2+ 0.46 0.92 0.10 0.99 0.018 Sankaranarayanan, M 2004 India HC2 CIN2+ 0.66 0.94 0.16 0.99 0.018 Sankaranarayanan, T2 2004 India HC2 CIN2+ 0.64 0.95 0.17 0.99 0.017 Blumenthal, 2001 Zimbabwe HC2 CIN2+ 0.81 0.61 0.19 0.87 0.100 Sankaranarayanan, 2005 India HC2 CIN2+ 0.11 Sarion, 2005 Brazil-Argentina HC2 CIN2+ 0.83 0.84 0.07 1.00 0.015 Bigras,2005 Switzerland HC2 CIN2+ 0.97 0.92 0.09 1.00 0.006 Pooled HC2 CIN2+ 0.90 0.87 Syrjanen, 2002 Russia HC2 CIN3+ 0.97 0.009 Petry, 2003 Germany HC2 CIN3+ 0.97 0.95 0.09 1.00 0.005 Sherman, 2003 Costa Rica HC2 CIN3+ 0.64 0.86 0.04 1.00 0.008 Sankaranarayanan, K1 2004 India HC2 CIN3+ 0.64 0.92 0.06 1.00 0.009 Sankaranarayanan, M 2005 India HC2 CIN3+ 0.74 0.94 0.13 1.00 0.013 Sankaranarayanan, T2 2005 India HC2 CIN3+ 0.62 0.94 0.13 1.00 0.013 Sarion, 2005 Brazil-Argentina HC2 CIN3+ 0.97 0.84 0.04 1.00 0.008 Pan, 2003 China HC2 CIN3+ 0.98 0.84 0.12 1.00 0.022 Pooled 0.85 0.90 Cuzick, 1995 UK PCR CIN2+ 0.75 0.96 0.42 0.99 0.040 Cuzick, 1999 UK PCR CIN2+ 0.64 0.99 0.42 1.00 0.015 Schneider, 2000 Germany PCR CIN2+ 0.95 0.96 0.36 1.00 0.025 Oh, 2001 South-Korea PCR CIN2+ 0.78 0.99 0.78 0.99 0.039 Paraskevaidis, 2001 Greece PCR CIN2+ 0.89 0.79 0.11 1.00 0.029 Agorastos, 2005 Greece PCR CIN2+ 0.75 0.97 0.08 1.00 0.003 Pooled 0.81 0.95 Cuzick 1999 UK PCR CIN2+ 0.74 0.99 0.42 1.00 0.012 Kulasingam 2002 US PCR CIN2+ 0.89 0.86 0.14 1.00 0.021 Pooled 0.83 0.89 203 HC2, testcutoff >1pg/mL Outcome: CIN2+ Kuhn, 2000 Ratnam, 2000 Belinson, 2001 Blumenthal, 2001 Clavel, 2001 Belinson, 2003 Coste, 2003 Cuzick, 2003 Pan, 2003 Petry, 2003 Sankaranarayanan, K1, 2004 Sankaranarayanan, M, 2004 Sankaranarayanan, T2, 2004 Bigras, 2005 Sarian, 2005 Combined 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Sensitivity Figure 21. Estimation of the sensitivity of the Hybrid Capture II assay for detection of CIN2+ in the framework of primary screening, pooled from 12 studies using a random effect model. Pooled sensitivity: 90% (95% CI: 86-94%). Q test for heterogeneity=139,5, 14 df, p=0.000. Possible explanations for this low sensitivity could be: contamination of the sample by acetic acid or Lugol’s iodine or deterioration of the sample due to exposure at high temperature. Contamination of the sample by acetic acid or Lugol’s iodine is unlikely since, according to the protocol, the sample for HC2 was collected before application of vinegar or iodine solution. A laboratory testing problem, is also unlikely because of the high concordance between the Indian results and those performed in a specialized virological French laboratory on a random subsample [Sankaranarayanan, 2004d]. Existence of other HPV types not included in the hrHPV DNA HC2 cocktail is another possibility which merits further examination. Finally, misclassification of the outcome could explain the low accuracy of HC2. It is known that colposcopy followed by biopsy taken from colposcopically suspect lesions is not a perfect gold standard. In this study, all visual inspection methods yielded highly correlated results, which might explain – as mentioned earlier - the high apparent accuracy (sensitivity and specificity) of the visual methods. The test sensitivity of colposcopy itself was not evaluated in the ACCP trials. Belinson [2001] found a sensitivity of colposcopy for CIN2+ of 81%. The inter-study and inter-period variation in HC2 test positivity and accuracy was narrow and most often non-significant, which most probably reflects high reproducibility, independent of training or experience. Among all evaluated tests, cytology showed the lowest sensitivity, even at the lowest cytological cutoff (58% for CIN2+). We have inserted the sensitivity/specificity values in a sROC curve containing the data of a previously published meta-analyses of the accuracy of 204 the Pap smear [Fahey, 1995; Nanda, 2000]. We see that the data points from the multi- center and the Zimbabwe trial all belong to the cloud of points from Fahey's meta-analysis. VIA Outcome=CIN2+ 1 .8 .6 Sensitivity .4 .2 0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 FPR = 1- Specificity Figure 22. Summary ROC curve displaying the relation between the false positive rate and the sensitivity of cytology for the outcome of CIN2+ (derived from the meta-analysis of Fahey [1995]). The values from the five Indian settings of the multi-center cross-sectional study are added as blue squares; the point corresponding with the Harare study is added as a green triangle. There are still a few issues which we mention here for future exploration: • Inclusion of the data from three other ACCP trials, after primary publication of the results: o The See and Treat randomized trial, conducted in South Africa by EngenderHealth in collaboration with Columbia University (New York) and the University of Cape Town. o The trial conducted by PAHO, in San Martin (Peru) comparing different screening test and triage combinations. o Multi-arm RCT, conducted in Osmanabad (India) by IARC. • Analysis of the second see & treat demonstration project, conducted in Amasaman (Ghana), and further analysis of the completed Busia study. • Assessment of the accuracy of combinations of screening tests. • Multi-level and Bayesian hierarchical meta-analyses to differentiate better between individual, provider, study setting, country and supervising agency effects. • Cost-effectiveness analyses: of different screening options, combinations of tests, triage and treatment procedures. • The impact of possible gold standard misclassification using external data documenting the accuracy of colposcopy and histology of biopsies. 205 • Pooling of other published studies targeting cervical cancer screening in developing countries. We obtained recently the registration of the title of a future systematic Cochrane review. Moreover a grant was awarded to the Scientific Institute of Public Health in Brussels by the Gynaecological Cancer Cochrane Collaboration to finalize a meta-analysis on screening methods applicable in developing countries. • Continued surveillance of longitudinal indicators from the Ambilikai and Osmanabad RCTs. In fact, the results of this type of study will give more clear answers on screening and treatment efficacy in terms of incidence and mortality reduction, under the condition that registration of cancer cases and deaths is sufficiently complete and that cause of death certification is accurate and balanced according to the screening arms. Contamination of the randomized groups (other screening/treatment/vaccination interventions affecting incidence of cervix cancer) can become another danger that can bias study results. The outcome of five European randomized screening trials comparing cytology screening with HPV or combined cytology/HPV screening will be pivotal for deciding on a possible shift from the Pap smear to the HPV test as first choice screening method in Europe. All these trials will allow observation of a reduced incidence of CIN2, CIN3 or cancer among screen-negatives in the HPV arm 3 to 5 years after initial screening (Davies, 2005). • Definition of future screening strategies in a context where probably HPV vaccination will be introduced progressively, which is expected to protect vaccinated teenagers against anogenital cancer within a few decades but which meanwhile will not affect current trends in the cohorts of women who are already 15 years and older. An enormous quantity of new knowledge has become available from the studies which we have analyzed in the current report, but the definitive answers on how to define the best screening and treatment policies can not yet be given. We hope therefore that researchers will continue their work in cervical cancer screening to complete this picture in the near future. 8. Acknowledgements We acknowledge all the researchers and data managers for the provision of the individual data files, in particular, the collaborators of Dr. R. Sankaranarayanan, Mr. R. Muwonge and Dr. C. Mahé for the data from the trials coordinated by IARC (Lyon), Mrs. Joyce Erickson Kristen Lewis (PATH, Seattle) for the data from Kenya and Mrs. Robbyn Lewis (JHPIEGO, Baltimore) who provided data from Zimbabwe, Thailand and Ghana. Moreover, we want to express our gratitude to Ms. Jacqueline Sherris (PATH) who invited us to conduct this pooled analysis and who facilitated the contacts with the data providers . All studies, with the exception of the Zimbabwean trial, were funded by the Bill & Melinda Gates Foundation through the Alliance for Cervical Cancer Prevention. Financial support for the Zimbabwe study came from the JHPIEGO Corporation and an affiliate of The Johns Hopkins University through funds from the US Agency for International Development (USAID Office of Population, Center for Population, Health and Nutrition, Global Bureau, under the terms of Cooperative Agreement Number CCP-3069-A-00-3020-00) . 206 207 9. Abbreviations AA: acetic acid ACCP: Alliance for Cervical Cancer Prevention AdenoCa: adenocarcinoma AGC: atypical glandular cells AGUS: atypical squamous cells of unspecified significance AIS: Adenocarcinoma in situ ASC: atypical squamous cells ASC-US: atypical squamous cells of unspecified significance ASC-H: atypical squamous cells, cannot rule out HSIL AUC: area under the curve AW: acetowhite Ca: cancer CI: 95% confidence interval CIN: cervical intra-epithelial lesion CP: conventional Papanicolaou smear CT: cytotechnologist DNA: Desoxyribo-nucleic acid DOR: diagnostic odds ratio DVI: direct visual inspection ECC: endocervical curettage FN: false negative FNR: false negative rate (= complement of sensitivity; FNR = 1 - SE) FP: false positive FU: follow-up HC2: Hybrid Capture-2 assay HPV: human papillomavirus hrHPV: high risk HPV type HSIL: high-grade squamous intra-epithelial lesion IARC: International Agency for Research in Cancer JHPIEGO: John Hopkins Program for International Education of Gynecologists and Obstetricians LBC: liquid based cytology LEEP: loop electrosurgical excision procedure lrHPV: low risk HPV type LSIL: low-grade squamous intra-epithelial lesion NPV: negative predictive value OR: odds ratio PAHO: PanAmerican Health Organization PATH: Program for Appropriate Technology in Health RCT: randomized controlled trial RLU: Relative Light Units ROC-curve: Receiver Operational Characteristic curve RR: relative risk SBLB: satisfactory but limited by SCJ: squamocolumnar junction SE: sensitivity 208 SIL: squamous intra-epithelial lesion SP: specificity sROC-curve: summary Receiver Operational Characteristic curve T+ rate: proportion of test positives TBS: The Bethesda System TBS91: The Bethesda System (version 1991) TBS01: The Bethesda System (version 2001) TN: true negative TP: true positive TZ: transformation zone VIA: visual inspection after application of acetic acid VIAM: visual inspection after application of acetic acid with magnification VILI: visual inspection after application of Lugol’s iodine WNL: within normal limits 10. References Anttila A, Ronco G, Clifford G, Bray F, Hakama M, Arbyn M et al. Cervical cancer screening programmes and policies in 18 European countries. Br J Cancer 2004; 91: 935- 41. Arbyn, M., Buntinx, F., Van Ranst, M., and Cortiñas Abrahantes, J. Triage of women with atypical or low-grade cytological abnormalities of the cervix by HPV testing: systematic review and meta-analysis. IPH/EPI-REPORTS Nr. 2001-019, Scientific Institute of Public Health, Brussels, 2002, pp 1-240. Arbyn M, Buntinx F, Van Ranst M, Paraskevaidis E, Martin-Hirsch P, Dillner J. Virologic versus cytologic triage of women with equivocal Pap smears: a meta-analysis of the accuracy to detect high-grade intraepithelial neoplasia. J Natl Cancer Inst 2004; 96: 280-93. Arbyn M, Dillner J, Van Ranst M, Buntinx F, Martin-Hirsch P, Paraskevaidis E. Re: Have we resolved how to triage equivocal cervical cytology? J Natl Cancer Inst 2004; 96: 1401- 2. Arbyn M. Keynote lecture: Cervical screening in Europe. 31st European Congress of Cytology, organised by the European Federation of Cytology Societies and the French Society of Clinical Cytology, Paris (France), 2-5 October, 2005. Arbyn M, Hovland S. Test characteristics of multiple cytological and HPV nucleic acid assays to detect high-grade CIN. A study on 343 women from South-Kivu (Democratic Republic of Congo). IPH/EPI-REPORT 2006-0xx. Brussels, Scientific Institute of Public Health. 209 Belinson JL, Qiao YL, Pretorius R, Zhang WH, Elson P, Li L et al. Shanxi province cervical cancer screening study: a cross-sectional comparative trial multiple techniques to detect cervical neoplasia. Gynecol Oncol 2001; 83: 439-44. Blumenthal PD, Graffikin L, Chirenje ZM. Adjunctive testing for cervical cancer in low resource settings with visual inspection, HPV, and the Pap smear. Int J Gynaecol Obstet 2001; 72: 47-53. Bosch FX, Lorincz A, Munoz N, Meijer CJ, Shah KV. The causal relation between human papillomavirus and cervical cancer. J Clin Pathol 2002; 55: 244-65. Bosch FX, de Sanjose S. Chapter 1: Human papillomavirus and cervical cancer--burden and assessment of causality. J Natl Cancer Inst Monogr 2003; 3-13. Bray F, Loose JH, McCarron P, Weiderpass E, Arbyn M, Moller H et al. Time trends in cervical squamous cell carcinoma incidence in 13 European countries: the relative contributions of screening and changing risk. Cancer Epidemiol Biomarkers Prev 2004. Davies P, Arbyn M, Dillner J, Kitchener H, Ronco G, Hakama M. A report on the current status of European research on the use of human papillomavirus testing for primary cervical cancer screening. Int J Cancer 2005; 118: 791-6. Devesa SS, Silverman DT, Young JL, Pollack ES, Brown CC, Horm JW et al. Cancer incidence and mortality trends among whites in the United States,1947-84. J Natl Cancer Inst 1987; 79: 701-45. De Vuyst H, Claeys P, Njiru S, Muchiri L, Steyaert S, De Sutter P et al. Comparison of pap smear, visual inspection with acetic acid, human papillomavirus DNA-PCR testing and cervicography. Int J Gynaecol Obstet 2005; 89: 120-6. IARC. Cervix Cancer Screening. IARC Handbooks of Cancer Prevention. Vol. 10. Lyon: 2005. Fahey MT, Irwig L, Macaskill P. Meta-analysis of Pap test accuracy. Am J Epidemiol 1995; 141: 680-9. Ferlay J, Bray F, Pisani P, Parkin DM. GLOBOCAN 2002. Cancer Incidence, Mortality and Prevalence Worldwide. IARC CancerBase No. 5, version 2.0. IARCPress, Lyon, 2004. Franco EL. Chapter 13: Primary screening of cervical cancer with human papillomavirus tests. J Natl Cancer Inst Monogr 2003; 89-96. Gaffikin L, Blumenthal PD, Emerson M, Limpaphayom K. Safety, acceptability, and feasibility of a single-visit approach to cervical-cancer prevention in rural Thailand: a demonstration project. Lancet 2003; 361: 814-20. Gaffikin L, Lauterbach M, Emerson M, Lewis R. Safety, acceptability, and feasibility of a single visit approach to cervical cancer prevention: results from a demonstration project in Ghana. JHPIGO Publication, Baltimore, November 2004, pp 1-42. 210 Glasziou P, Irwig L, Bain C, Colditz G. Systematic reviews in health care: a practical guide. Cambridge University Press, Cambridge, 2001; pp 1-137. Hosmer DW, Lemeshow S. Applied Logistic Regression. New York: John Wiley & Sons, 1989, pp 1-317. Irwig L, Tosteson ANA, Gastonis C et al. Guidelines for meta-analyses evaluation of diagnostic tests. Ann Int Med 1994; 120: 667-676. Irwig L, Macaskill P, Gasziou P et al. Meta-analytic methods for diagnostic test accuracy. J Clin Epidemiol 1995; 48: 119-130. Lörincz AT. Methods of DNA hybridisation and their clinical applicability to human papillomavirus detection. In: New Developments in Cervical Cancer Screening and Prevention. Ed: Franco E & Monsonego J, Blackwell Science, Oxford 1997, pp 325-337. Lorincz AT, Richart RM. Human papillomavirus DNA testing as an adjunct to cytology in cervical screening programs. Arch Pathol Lab Med 2003; 127: 959-68. Lundberg GD, National Cancer Institute. The 1988 Bethesda System for Reporting Cervical/Vaginal Cytologic Diagnoses. JAMA 1989; 262: 931-4. Luff RD, National Cancer Institute. The Bethesda System for Reporting Cervical/Vaginal Cytologic diagnoses: a Report of the 1991 Bethesda Workshop. Hum Pathol 1992; 23: 719- 21. Martin-Hirsch PL, Paraskevaidis E, Kitchener H. Surgery for cervical intraepithelial neoplasia. Cochrane Database Syst Rev 2000; niet vermeld: niet vermeld. McNutt LA, Wu C, Xue X, Hafner JP. Estimating the relative risk in cohort studies and clinical trials of common outcomes. Am J Epidemiol 2003; 157: 940-3 Miller AB. Programmes de dépistage du cancer du col: guide du gestionnaire. Genève: OMS, 1993, pp 1-52. Miller AB, Nazeer S, Fonn S, Brandup-Lukanow A, Rehman R, Cronje H et al. Report on consensus conference on cervical cancer screening and management. Int J Cancer 2000; 86: 440-7. Miller A, Sankaranarayanan R, Bosch FX. Cervical Cancer Screening in Developing Countries: Report of a WHO consultation. Geneva: World Health Organisation, 2002, pp 1- 75 Miller AB. The (in)efficiency of cervical screening in Europe. Eur J Cancer 2002; 38: 321- 6. Mitchell MF, Tortolero-Luna G, Cook E, Whittaker L, Rhodes-Morris H, Silva E. A randomized clinical trial of cryotherapy, laser vaporization, and loop electrosurgical excision for treatment of squamous intraepithelial lesions of the cervix. Obstet Gynecol 1998; 92: 737-44. 211 Moses LE, Shapiro D, Littenberg B. Combining independent studies of a diagnostic test into a summary ROC curve: data-analytic approaches and some additional considerations. Stat Med 1993; 12: 1293-316. Nanda K, McCrory DC, Myers ER, Bastian LA, Hasselblad V, Hickey JD et al. Accuracy of the Papanicolaou Test in Screening for and Follow-up of Cervical Cytologic Abnormalities: A Systematic Review. Ann Intern Med 2000; 132: 810-9. PATH. Western Kenya Cervical Cancer Prevention Program: Development and evaluation of a model for low-resource settings. Phase II Project Protocol. July 200. Pepe MS. The Statistical Evaluation of Medical Tests for Classification and Prediction . Oxford: Oxford University Press, 2003, pp 1-318. Richart RM. Cervical intra-epithelial neoplasia: a review. In: Pathology Annual, 1973; Vol. 8, Sommers SC, Appleton Century Crafts, Norwall, 301. Sankaranarayanan R, Rajkumar R, Arrossi S, Theresa R, Esmy PO, Mahe C et al. Determinants of participation of women in a cervical cancer visual screening trial in rural south India. Cancer Detect Prev 2003a; 27: 457-65. Sankaranarayanan R, Wesley R, Thara S, Dhakad N, Chandralekha B, Sebastian P et al. Test characteristics of visual inspection with 4% acetic acid (VIA) and Lugol's iodine (VILI) in cervical cancer screening in Kerala, India. Int J Cancer 2003b; 106: 404-8. Sankaranarayanan R, Wesley RS. A practical manual on visual screening for cervical neoplasia. IARC Technical Publication N° 41, IARC Press, Lyon, 2003c; pp 1-49. Sankaranarayanan R, Shastri SS, Basu P, Mahé C, Mandal R, Amin G. The role of low- level magnification in visual inspection with acetic acid for the early detection of cervical neoplasia. Cancer Detect Prev 2004a; 28: 345-51. Sankaranarayanan R, Rajkumar R, Theresa R, Esmy PO, Mahe C, Bagyalakshmi KR et al. Initial results from a randomized trial of cervical visual screening in rural south India. Int J Cancer 2004b; 109: 461-7. Sankaranarayanan R, Basu P, Wesley RS, Mahé C, Keita N, Mbalawa CCG et al. Accuracy of visual screening for cervical neoplasia: results from an IARC multi-center study in India and Africa. Int J Cancer 2004c; 110: 907-13. Sankaranarayanan R, Chatterji R, Shastri SS, Basu P, Mahé C, Muwonge R et al. Accuracy of human papillomavirus testing in primary screening of cervical neoplasia: results from a multi-center study in India. Int J Cancer 2004d; 112: 341-7. Sankaranarayanan R, Thara S, Sharma A, Roy C, Shastri S, Mahe C et al. Accuracy of conventional cytology: results from a multi-center screening study in India. J Med Screen 2004e; 11: 77-84. 212 Sasieni P, Stepniewska KA, Altman DG. Stratified test for trend across ordered groups. Stata Tech Bull Reprints 1996; 6: 193-196. Sharp S, Sterne J. Meta-analysis. Stata Technical Bulletin 1997; 38: 9-14. Sutton AJ, Abrams KR, Jones DR, Sheldon TA, Song F. Methods for Meta-Analysis in Medical Research. Chichester (UK): 2000, pp 1-317. University of Zimbabwe/JHPIEGO Cervical Cancer Project, Anonymous. Visual inspection with acetic acid for cervical-cancer screening: test qualities in a primary-care setting. Lancet 1999; 353: 869-73. van Ballegooijen M, van den Akker van Marle ME, Patnick J, Lynge E, Arbyn M, Anttila A et al. Overview of important cervical cancer screening process values in EU-countries, and tentative predictions of the corresponding effectiveness and cost-effectiveness. Eur J Cancer 2000; 36: 2177-88. van den Akker van Marle ME, van Ballegooijen M, van Oortmarssen GJ, Boer R, Habbema JDF. Cost-effectiveness of cervical cancer screening: comparison of screening policies. J Natl Cancer Inst 2002; 94: 193-204. Womack SD, Chirenje ZM, Blumenthal PD, Gaffikin L, McGrath JA, Chipato T et al. Evaluation of a human papillomavirus assay in cervical screening in Zimbabwe. BJOG 2000; 107: 33-8. Womack SD, Chirenje Z, Gaffikin L, Blumenthal PD, McGrath JA, Chipato T et al. HPV- based cervical cancer screening in a population at high risk for HIV infection. Int J Cancer 2000; 85: 206-10. Yang BH, Bray FI, Parkin DM, Sellors JW, Zhang Z-F. Cervical cancer as a priority for prevention in different world regions : an evaluation using years of life lost. Int J Cancer 2004; 109: 418-24. Zhang J, Yu KF. What is the relative risk: a method of correcting the odds ratio in cohort studies of common outcomes. JAMA 1998; 280: 1690-1. 11. Annexes 11.1. Data files and statistical syntax files • Directories: /xls/metaanalysis/via/sankar (1st multi-center cross-sectional study) and /xls/metaanalysis/via/accp/ (last 5 studies). • Primary data files: 1. Cross-sectional multi-center study: /xls/metaanalysis/via/sankar/crossdata.dta 2. Ambilikai RCT: /xls/metaanalysis/via/accp/ambi/ambidata.dta 3. Zimbabwe study: /xls/metaanalysis/via/accp/zimb/zimb.dta 4. Kenya study: (3 files) /xls/metaanalysis/via/accp/kenya/kenya10nov04/metaanalysis_WKCCPP_Nov8.dta; ellie_meta_nov_04.dta ; /xls/metaanalysis/via/accp/kenya/providers.dta 213 5. Thai study: /xls/metaanalysis/via/accp/thai via.init.cry.ref.prob1_2.ffol.oy.12_11_01.sav 6. Ghana study: /xls/metaanalysis/via/accp/ghana/ghana i_v_c_f_o_r_p_08-04- 04_5.sav 11.2. Abstracts from the included separate published studies 11.2.1. Cross-sectional multi-country study (5 publications) Sankaranarayanan R, Wesley R, Thara S, Dhakad N, Chandralekha B, Sebastian P et al. Test characteristics of visual inspection with 4% acetic acid (VIA) and Lugol's iodine (VILI) in cervical cancer screening in Kerala, India. Int J Cancer 2003b; 106: 404-8. Simple and inexpensive methods based on visual examination of the cervix are currently being investigated as alternative methods of cervical screening. The test characteristics of visual inspection with 4% acetic acid (VIA), and Lugol's iodine (VILI) and conventional cytology were investigated in a cross-sectional study involving 4,444 women aged 25 to 65 years in Kerala, India. While detection of any acetowhite area constituted a low-threshold positive VIA, detection of well-defined, opaque acetowhite lesions close to or touching the squamocolumnar junction constituted a high-threshold positive VIA test. Detection of definite yellow iodine non-uptake areas in the transformation zone close to or touching the squamocolumnar junction constituted a positive VILI test. Cytology was considered positive if reported as atypia or worse lesions. All screened women were evaluated by colposcopy and biopsies were directed in 1,644 women (37.0%), which allowed the direct estimation of sensitivity, specificity and predictive values. The reference diagnosis was based on a combination of histology and/or colposcopy. True disease status was defined as CIN 2 and worse lesions. A total of 149 (3.4%) women had CIN 2 or worse lesions. The sensitivities of low-threshold VIA, high-threshold VIA, VILI and cytology to detect CIN 2 or worse disease were 88.6%, 82.6%, 87.2% and 81.9%, respectively; the corresponding specificities were 78.0%, 86.5%, 84.7% and 87.8%. Our results indicate that VIA and VILI are suitable alternate screening tests to cytology for detecting cervical neoplasia in low- resource settings Sankaranarayanan R, Shastri SS, Basu P, Mahé C, Mandal R, Amin G. The role of low-level magnification in visual inspection with acetic acid for the early detection of cervical neoplasia. Cancer Detect Prev 2004a; 28: 345-51. Several studies have investigated the accuracy of naked eye visual inspection with acetic acid (VIA) in the early detection of cervical neoplasia. It is not clear whether low-level (2- 4x) magnification (VIAM) can improve the sensitivity and specificity of VIA. The accuracy of both VIA and VIAM, provided by independent health workers, were evaluated in three cross-sectional studies involving 18,675 women aged 25-65 years in Kolkata and Mumbai in India. All screened women were investigated with colposcopy and biopsies were obtained based on colposcopy findings. The final disease status was based on the reference standard of histology (if biopsies had been taken) or colposcopy. Data from the studies were pooled to calculate the test characteristics for the detection of high-grade squamous intraepithelial lesions (HSIL). 14.1% and 14.2% were positive on testing with VIA and VIAM respectively. Two hundred twenty-nine were diagnosed with HSIL and 68 with 214 invasive cancer. The pooled sensitivity, specificity, positive and negative predictive values for VIA in detecting high-grade squamous intraepithelial lesions (HSIL) were 60.3% (95% CI: 53.6-66.7), 86.8% (95% CI: 86.3-87.3), 5.9% (95% CI: 5.0-7.0), and 99.4% (95% CI: 99.2-99.5), respectively. The values were 64.2% (95% CI: 57.6-70.4), 86.8% (95% CI: 86.2-87.3), 6.3% (95% CI: 5.3-7.3) and 99.4% (95% CI: 99.3-99.6), respectively, for VIAM. Low-level magnification did not improve the test performance of naked eye visualization of acetic acid impregnated uterine cervix. Sankaranarayanan R, Basu P, Wesley RS, Mahé C, Keita N, Mbalawa CCG et al. Accuracy of visual screening for cervical neoplasia: results from an IARC multi- center study in India and Africa. Int J Cancer 2004c; 110: 907-13. Visual inspection-based screening tests, such as visual inspection with 4% acetic acid (VIA) and with Lugol's iodine (VILI), have been proposed as alternatives to cytology in mass screening programs. To date, there is only limited information on the accuracy of these tests in detecting High-grade Squamous Intraepithelial Lesions (HSIL). Eleven cross- sectional studies involving 56,939 women aged 25-65 years were conducted in Burkina Faso, Congo, Guinea, India, Mali and Niger to evaluate the accuracy of VIA and VILI performed by health workers. A common protocol and questionnaire was used. For final diagnosis, all women were investigated with colposcopy and biopsies were taken when necessary. Data from the studies were pooled to calculate sensitivity, specificity and predictive values of the tests for the detection of HSIL. Of the screened women, 16.1% and 16.4% were positive on examination using, respectively, VIA and VILI; 1,063 were diagnosed with HSIL. The pooled sensitivity, specificity, positive and negative predictive values for VIA were 76.8% (95% CI: 74.2-79.4%), 85.5% (95% CI: 85.2-85.8%), 9.4% (95% CI:8.8-10.8%) and 99.5% (95% CI:99.4-99.6%), respectively. The values were 91.7% (95% CI: 89.7-93.4%), 85.4% (95% CI: 85.1-85.7%), 10.9% (95% CI: 10.2-11.6%) and 99.8% (95% CI:99.7-99.9%), respectively for VILI. The range of sensitivity and specificity for VIA was 56.1-93.9% and 74.2-93.8%, respectively, between studies and were 76.0-97.0 % and 73.0-91.3% for VILI. VILI had a significantly higher sensitivity than VIA in detecting HSIL, but specificity was similar. VILI appears to be a more accurate visual test for use in screening and treatment programs in low-resource settings. Sankaranarayanan R, Chatterji R, Shastri SS, Basu P, Mahé C, Muwonge R et al. Accuracy of human papillomavirus testing in primary screening of cervical neoplasia: results from a multi-center study in India. Int J Cancer 2004d; 112: 341-7. The knowledge that cervical neoplasia are caused by human papillomavirus (HPV) infection has led to the evaluation of its role in screening. We evaluated the accuracy of HPV testing by Hybrid capture II (HC II) method in detecting cervical intraepithelial neoplasia grade 2 and 3 (CIN 2 and 3) lesions in 4 cross-sectional studies with common protocol and questionnaire in 3 different locations (Kolkata, Mumbai and Trivandrum) in India. These studies involved 18,085 women aged 25-65 years. The reference standard for final diagnosis was a combination of colposcopy/biopsy. All women were investigated with colposcopy and 3,116 received directed biopsy. The sensitivity of HPV testing for detecting CIN 2-3 lesions varied from 45.7% to 80.9% across the study sites; the specificity varied from 91.7% to 94.6% and the positive predictive value from 6.7% to 13.7%. Retesting of 298 randomly chosen denatured samples in France revealed an agreement rate of 85.9% and a kappa-value of 0.72. Although HPV testing seems to be a promising approach for cervical cancer prevention, a large range in sensitivity was observed in our study, possibly due to 215 variations in the quality of specimen collection and reference standards. A higher sensitivity was associated with the center performing the test well. Further developments in terms of more reproducible, less expensive and less sophisticated testing are essential to make the test feasible and effective in low-resource settings. Sankaranarayanan R, Thara S, Sharma A, Roy C, Shastri S, Mahe C et al. Accuracy of conventional cytology: results from a multi-center screening study in India. J Med Screen 2004e; 11: 77-84. OBJECTIVE: We conducted a multi-center cross-sectional study in India to evaluate the accuracy of conventional cytology to detect high-grade squamous intraepithelial lesions (HSIL). SETTING: Cross-sectional studies in Jaipur, Kolkata, Mumbai and Trivandrum, India, during 1999-2003. METHODS: A common protocol and questionnaire were used to test 22,663 women aged 25-65 years with conventional cytology in five cross-sectional studies. Three thresholds were used to define test positivity: atypical squamous cells of uncertain significance (ASCUS), low-grade squamous intra-epithelial lesion (LSIL), or HSIL. All screened women were investigated with colposcopy, and biopsies were taken when necessary. The reference standard for final disease status was histology or negative colposcopy. Data from the studies were pooled to evaluate the test characteristics for the detection of histologically confirmed HSIL. RESULTS: The test positivity rates of cytology were 8.8% at ASCUS, 6.2% at LSIL and 1.8% at HSIL thresholds, and 355 women had histologically confirmed HSIL while 74 had invasive cancer. The pooled sensitivity, specificity, positive and negative predictive values at ASCUS threshold were 64.5%, 92.3%, 11.8% and 99.4% respectively. The corresponding values at LSIL threshold were 58.0%, 94.9%, 15.2% and 99.3%, while at the HSIL threshold they were 45.4%, 99.2%, 46.3% and 99.1%. The sensitivity varied between 37.8-81.3% at ASCUS, 28.9-76.9% at LSIL and 24.4-72.3% at HSIL thresholds. A significantly low sensitivity was observed in women aged 25-39 years (p<0.001). The wide variation in sensitivity across study sites persisted even after age standardization. CONCLUSION: The sensitivity of cytology varied widely between the study sites. Findings from our study and other reviews indicate that sustained efforts in improving sampling, preparation and reading of cytological specimens and improvements in clinical judgment are essential to achieve concurrently high sensitivity and specificity. 11.2.2. Zimbabwe study (3 publications) University of Zimbabwe/JHPIEGO Cervical Cancer Project, Anonymous. Visual inspection with acetic acid for cervical-cancer screening: test qualities in a primary- care setting. Lancet 1999; 353: 869-73. BACKGROUND: Naked-eye visual inspection of the cervix with acetic-acid wash (VIA), or cervicoscopy, is an alternative to cytology in screening for cervical cancer in poorly resourced locations. We tested the sensitivity, specificity, and predictive value of VIA done by nurse-midwives in a less-developed country. METHODS: Women were screened by six trained nurse-midwives in a two-phase, cross-sectional study at 15 primary-care clinics in Zimbabwe. VIA and Pap smears were done concurrently, and their sensitivity and 216 specificity compared. Colposcopy, with biopsy as indicated, was used as the reference test to allow a direct comparison of the test unaffected by verification bias. FINDINGS: 10934 women were screened. In phase II, 2148 (97.5%) of the 2203 participants for whom there was a screening result also had a reference test result. Also in phase II, VIA was more sensitive but less specific than cytology. Sensitivity (95% CI) was 76.7% (70.3-82.3) for VIA and 44.3% (37.3-51.4) for cytology. Specificity was 64.1% (61.9-66.2) for VIA and 90.6% (89.2-91.9) for cytology. INTERPRETATION: The high sensitivity of VIA shows that the test could be valuable in detection of precancerous lesions of the cervix. However, there are costs to the patient and system costs associated with high numbers of false- positive results, so attention should be given to improving the specificity of VIA. Womack SD, Chirenje ZM, Blumenthal PD, Gaffikin L, McGrath JA, Chipato T et al. Evaluation of a human papillomavirus assay in cervical screening in Zimbabwe. BJOG 2000; 107: 33-8. OBJECTIVE: To determine the utility of an assay for high risk genital human papillomavirus (HPV) in cervical screening in Zimbabwe, Africa. DESIGN: Cross- sectional study. SETTING: Harare, Zimbabwe. POPULATION: Zimbabwe women (n = 2,140), 25 to 55 years old, recruited in clinics in Chitungwiza and Greater Harare. METHODS: Genital specimens were assessed for HPV, using the HPV DNA test Hybrid Capture II (probe B). Further assessment of the women was conducted using colposcopy and biopsy as indicated. High grade squamous intraepithelial lesions were diagnosed in 215 women. Colposcopy and/or biopsy showed low grade lesions in 346 women. RESULTS: The overall prevalences were: 42 x 7% for HPV, 10% for high grade squamous intraepithelial lesions and 16% for low grade lesions. Prevalence for HPV decreased significantly with increase in age (P for trend < 0 x 0001) and increased significantly with increasing disease severity, from 35% in normal women, to 53% in women with low grade lesions and 81% in women with high grade lesions (P for trend < 0 x 001). In specimens positive for HPV, the amount was 14-fold higher in women with high grade lesions compared with normal women. In screening for high grade lesions the assay for HPV had a sensitivity of 81% (CI 75%-86%); sensitivity for low grade lesions was 64% (CI 60%- 68%). Specificity was 62% (CI 59%-64%) for high grade lesions and 65% (CI 62%-67%) for low grade lesions. The positive predictive value was 19% (CI 17%-22%) for high grade lesions and 39% (CI 36%-42%) for low grade lesions. CONCLUSION: For high grade squamous intraepithelial lesions the sensitivity of the Hybrid Capture II HPV DNA test was high, but the specificity was relatively low. The test may therefore be most useful in conjunction with other screening tests. Blumenthal PD, Graffikin L, Chirenje ZM. Adjunctive testing for cervical cancer in low resource settings with visual inspection, HPV, and the Pap smear. Int J Gynaecol Obstet 2001; 72: 47-53. OBJECTIVE: To test whether the performance of visual inspection using acetic acid (VIA) could be improved through adjunctive testing and to determine whether the combination of visual inspection of the cervix and HPV testing could prove useful for identifying those at highest risk of cervical precancer. METHODS: Between October 1995 and August 1997, 2199 women willing to be screened for cervical cancer in peri-urban clinics in Harare, Zimbabwe received VIA, Pap smear and HPV as screening tests. The presence or absence 217 of (pre)cancer was confirmed via colposcopy with biopsy as indicated for >97% of all women. Computerized simulations of sequential testing scenarios provided estimates of the joint (net) test qualities of different paired combinations of the three tests and allowed for comparisons with the individual test qualities. RESULTS: Using HGSIL/CIN II-III as the reference threshold of disease, the net sensitivity and specificity of VIA and HPV when used sequentially were 63.6 and 81.9%, respectively, compared to 43.3 and 91%, respectively, when Pap smears were followed by HPV testing. VIA followed by the Pap smear yielded a net sensitivity of 37.5% and net specificity of 94.3%. CONCLUSIONS: For programs with limited resources but with the capacity for HPV testing, sequential testing involving the use of VIA followed by HPV could yield fewer false positives than the use of VIA alone at a cost of relatively few additional false negatives. 11.2.3. Ambilikai RCT (2 publications) Sankaranarayanan R, Rajkumar R, Arrossi S, Theresa R, Esmy PO, Mahe C et al. Determinants of participation of women in a cervical cancer visual screening trial in rural south India. Cancer Detect Prev 2003a; 27: 457-65. The efficacy of a single round of screening of visual inspection with acetic acid (VIA) on cervical cancer incidence and mortality is investigated in a cluster randomized controlled trial in south India. Women aged 30-59 years in 113 clusters in Dindigul District were randomized to VIA screening by nurses (57 clusters, 48,225 eligible women) and to a control group (56 clusters, 30,167 women). 30,577 (63.4%) eligible women participated in screening. Younger, educated, married, multiparous, low-income women and those who have had tubal sterilization had a higher compliance with screening. Of the 2069 women diagnosed with CIN and invasive cancer, 1498 (72.4%) received treatment. Young women, those who practiced contraception and women with high-grade precursor lesions and invasive cancers were more likely to comply with treatment. In summary, our study indicates that women accept screening with VIA by nurses and a moderate level of compliance with screening and treatment can be reached through appropriate service delivery systems including health education activities, personal invitations, clinics in proximity to the target women, and testing and treatment in the same session. Our results imply that integration of screening activities with primary health services seems to have the potential to replicate most of these service delivery conditions in routine programs Sankaranarayanan R, Rajkumar R, Theresa R, Esmy PO, Mahe C, Bagyalakshmi KR et al. Initial results from a randomized trial of cervical visual screening in rural south India. Int J Cancer 2004b; 109: 461-7. The impact of a single round of screening of visual inspection with acetic acid (VIA) on cervical cancer incidence and mortality was investigated in a cluster randomized trial in south India. Women 30-59 years of age in 113 clusters in Dindigul District were randomized to VIA screening (57 clusters, 48,225 women) by nurses and to a control group (56 clusters, 30,167 women). 30,577 eligible women were screened between May 2000 and April 2003; 2,939 (9.6%) screen-positive women were investigated with colposcopy by nurses and 2,777 (9.1%) women had biopsy. CIN 1 was diagnosed in 1,778 women, CIN 2- 3 lesions were found in 222, and there were 69 screen detected invasive cervical cancers. The detection rates of lesions per 1,000 screened women were 58.2 for CIN 1, 7.3 for CIN 218 2-3, and 2.3 for invasive cancer. The detection rate of high-grade lesions in our study was 2-3-fold higher than those observed in repeatedly screened populations in developed countries. 71% of women with CIN 1 and 80% of those with CIN 2-3 lesions accepted cryotherapy provided by nurses and surgical treatment by mid-level clinicians. Overall, 97 and 34 incident cervical cancer cases were observed in the intervention and control arms, respectively. The intervention arm accrued 124,144 person years and the control arm accrued 90,172 during the study period. The age standardized cervical cancer incidence rates were 92.4/100,000 person-years in the intervention and 43.1/100,000 in the control arms. In the screened arm, 35.0% of cases were in Stage I as opposed to none in the control arm. The preliminary findings from our study indicate that not only is a VIA-based screening program feasible, safe and acceptable to a population in rural settings, it also results in early detection of cervical neoplasia. 11.2.4. Kenya study Not published yet. 11.2.5. Thailand demonstration project Gaffikin L, Blumenthal PD, Emerson M, Limpaphayom K. Safety, acceptability, and feasibility of a single-visit approach to cervical-cancer prevention in rural Thailand: a demonstration project. Lancet 2003; 361: 814-20. BACKGROUND: To increase screening and treatment coverage, innovative approaches to cervical-cancer prevention are being investigated in rural Thailand. We assessed the value of a single-visit approach combining visual inspection of the cervix with acetic acid wash (VIA) and cryotherapy. METHODS: 12 trained nurses provided services in mobile (village health center-based) and static (hospital-based) teams in four districts of Roi-et Province, Thailand. Over 7 months, 5999 women were tested by VIA. If they tested positive, after counseling about the benefits, potential risks, and probable side-effects they were offered cryotherapy. Data measuring safety, acceptability, feasibility, and effort to implement the program were gathered. FINDINGS: The VIA test-positive rate was 13.3% (798/5999), and 98.5% (609/618) of those eligible accepted immediate treatment. Overall, 756 women received cryotherapy, 629 (83.2%) of whom returned for their first follow-up visit. No major complications were recorded, and 33 (4.4%) of those treated returned for a perceived problem. Only 17 (2.2%) of the treated women needed clinical management other than reassurance about side-effects. Both VIA and cryotherapy were highly acceptable to the patients (over 95% expressed satisfaction with their experience). At their 1-year visit, the squamocolumnar junction was visible to the nurses, and the VIA test-negative rate was 94.3%. INTERPRETATION: A single-visit approach with VIA and cryotherapy seems to be safe, acceptable, and feasible in rural Thailand, and is a potentially efficient method of cervical-cancer prevention in such settings. 219 11.2.6. Ghana demonstration project Gaffikin L, Lauterbach M, Emerson M, Lewis R. Safety, acceptability, and feasibility of a single visit approach to cervical cancer prevention: results from a demonstration project in Ghana. JHPIEGO Publication, Baltimore, November 2004, pp 1-42. A demonstration project to assess a single visit approach (SVA) to cervical cancer prevention was implemented in Ghana from March 2001 to July 2003. The SVA approach links visual inspection of the cervix with acetic acid (VIA) with an offer of immediate treatment or referral, as indicated. The objective of this demonstration project was to assess the safety, acceptability, feasibility and programmatic effort (SAFE) of this approach. VIA has been established as a viable testing option for low-resource settings. Cryotherapy was selected as the treatment of choice for this project because it: 1) has a cure rate comparable to other commonly performed procedures; 2) is easy to learn, does not require electricity, requires few consumables, and has a long history in the scientific literature of low complication rates; and 3) has an established, safe, and effective performance record with non-physicians in developed countries. The project was implemented in a phased approach, beginning in an urban setting and then expanding to a rural setting. The first phase of this demonstration project was implemented in an urban regional hospital, located in the capital city, Accra. Beginning the project in an urban center helped ensure the presence of a functional referral network for cases requiring advanced diagnosis and management prior to commencing screening services. The second phase of this demonstration project, involving expansion of services to a rural site, will be described in a subsequent report. The project protocol required that test-positive, treatment-eligible women were offered immediate cryotherapy and counseled about the benefits, potential risks, and likely side effects of treatment. Women ineligible for treatment, that is, whose lesions exceeded 75% of the cervical surface area or who had suspect cervical cancer, were referred. Data measuring safety, acceptability and feasibility of the single visit approach were collected. During the 18-month recruitment period at Ridge Hospital, 3,665 women were tested for cervical precancer using VIA. The overall VIA test-positive rate at Ridge Hospital was 3.2%. Among test-positive, treatment-eligible women for whom data are available, 70.1 % accepted the offer of immediate cryotherapy. In total, 439 women accepted the offer of cryotherapy and, of those, 76.8% returned for their first follow-up visit. There were no major complications, and fewer than 6% of those treated returned to the project facility for any perceived problem. Only 4% of those treated required any management other than reassurance for side effects. Both VIA and cryotherapy were highly acceptable to women-- more than 98% were satisfied with the experience--and 1 year after treatment the squamocolumnar junction (SCJ) was clearly visible for the majority of the women who returned, yielding a 1-year VIA test-negative rate of 96.7%. A single visit approach linking VIA testing with the offer of immediate treatment (or referral) of test-positive cases is safe, acceptable and feasible in low resource settings such as Ghana. This report describes the key results of this demonstration project involving an alternative, field-based, resource-appropriate approach to cervical cancer prevention. 220 11.3. Comprehensive table with characteristics of the included studies In the framework of an ongoing meta-analysis of published studies on cervical cancer screening in developing countries, we constructed comprehensive tables containing all study characteristics. From those tables we extracted the rows containing data from the ACCP studies (see Table 52 to Table 57). 221 Table 52. Characteristics of the study population. Study Publications Country Study Study population Inclusion criteria Exclusion criteria Study Age range location size Cross- Sankaranarayanan et al. Int J B-Fasso Ouagadougou, Apparently healthy, Consecutive series recruited Pregnancy, hysterectomy, 58679 25-65y. sectional Cancer 2003; 106: 404-408 Congo Brazzaville, ambulant, symptomatic opportunistically from previous treatment for multi- Sankaranarayanan et al. Int J Guinea Conakry, women. Women were resident populations. cervical cancer or CIN. country Cancer 2004;110: 907-913 Mali Bamako, encouraged to attend open Women in Kolkata B & Study Sankaranarayanan et al. Int J Niger Niamey. access clinics via publicity Thiruvananthpuram were Cancer 2004; 112: 341-347 India Jaipur, campaigns. recruited actively from Sankaranarayanan et al. J Med Kolkata, Urban populations. circumscribed communities. Screen 2004; 11: 77-84 Mumbai Sankaranarayanan et al. Thiruvananthp Cancer Det Prev 2004; 8: 345- uram (India). 351 Cross- UZ. Lancet 1999: 353: 869- Zimbabw 15 1ary care Women invited to a health Absence of inclusion criteria. Pregnancy, history of Ph I: 25-55 y; sectional 873 e clinics in education talk on cervical cervical cancer, 8731 mean 32: Zimbabwe Blumenthal et al. Int J Gyn Ob Chitungwiza cancer. hysterectomy, missing Ph II: (SD: 6.6). Study 2001; 72: 47-53 & greater Urban populations. identifying information. 2203 Harare area. Ambilikai Sankaranarayanan et al. India 113 Women motivated by local Apparently healthy women Absence of uterus. 30577 30-59 y. RCT Cancer Det Prev 2003; 27: panchayaths, health workers to with an intact uterus. & 457-465 Ambilikai participate in the study No prior screening. 30167 Sankaranarayanan et al. Int J area, Dindigul (study group). Control Cancer 2004; 109:461-467 district. group was educated on cervical cancer. Rural population. WKCCPP No publications yet Kenya 3 divisions in Rural population, residing Not further explained. Previous history of cancer, 2295 29-40 y. the Busia in one of the 3 target hysterectomy. district divisions. Thai Royal Thai College. Lancet Thailand 4 districts in Women recruited by Women consulting at the Total hysterectomy, history 5999 30-45y: demonstrat 2003; 361:814-820 Roi-et mobile (village health health services for whatever of cervical ca, pregnancy mean: 37y ion trial Province. center based) & static reason or specifically for (>20W). (sd 4.4 y). (hospital-based) teams. cervical cancer screening Rural populations. were invited to participate in the study. Ghanaian Gaffikin et al. JHPIEGO Ghana Ridge Hospital Urban population attending Women consulting at the Hysterectomy, pregnancy, 3665 Range: 25- demonstrat Report, 2004, 1-42 (Accra the family planning clinic, health services for whatever history of cervical cancer. 45; ion trial Metropolitan after campaigns. reason or specifically for Median: 34. Area). cervical cancer screening were invited to participate in the study. 222 Table 53. Characteristics of the study population (continued). Study Coordinating Agency Period Applied screening tests Cross-sectional IARC, Lyon, France 1999-2003. VIA, VILI, VIAM, Conventional Pap and HC2 multi-country Study Cross-sectional JHPIEGO, Baltimore, US October 1995-August Conventional Pap, VIA, HC2 Zimbabwe 1997 Study Ambilikai IARC, Lyon, France Enrolment between March VIA versus no intervention. RCT 2000-April 2003. Follow- up until 2007. WKCCPP PATH, Seattle, US November 2002-March VIA (VILI at the end of the study) 204 Thai JHPIEGO, Baltimore, US Enrolment between VIA demonstration February -November 2000. trial Follow-up during 1 year after treatment Ghanaian JHPIEGO, Baltimore, US Women were enrolled VIA demonstration during an 18-month period trial (March 2001-August 2002). Follow-up during 1 year after treatment 223 Table 54. Main characteristics of the visual inspection tests being evaluated. Study Description of the VIA test Illumination Qualification/training of "visual Definition of positive VIA result. Categorization of inspector" VIA result Cross-sectional After the CP, freshly prepared 4% Halogen focus lamp. A variety of health workers: auxiliary nurse Negative: multi-country AA was applied to the cervix with a midwives, qualified graduated nurses, CTs, VIA+: faint, translucent, ill defined, irregular AW Study cotton swab, inspection after 1 high school or university graduates. All lesions on the cervix. Definite, geographic AW lesions, minute. received intensive courses using the IARC far from the SCJ. manual, photographs & practical exercises; VIA++: opaque, dense, dull, well-defined AA touching instructions in cryotherapy. Refresher the SCJ or close to the external os. Large, courses of 1-2d during the study. circumorificial, well-defined, thick, dense AW lesions. AW growth. Cross-sectional After collection of a smear, excess Handheld flash light. Nurse-midwife, trained during 3 days in Normal: smooth,pink,uniform,featureless. Zimbabwe mucus was cleansed away with a VIA. Atypical: cervicitis (inflammation,red spots); Study saline-soaked swab. AA (4%) was discharge, ectropion, polyp. applied to the cervix with a cotton- Abnormal: white plaques, ulcer, AW epithelium. tipped applicator. Cervical cancer: cauliflower-like growths, fungating mass. Ambilikai Application of 4% AA to the cervix Bright halogen focus Not documented. Opaque dull well defined confluent AW lesions RCT with a cotton swab & excess mucus lamp. touching the SCJ or close to the external os. Large was cleared. VIA findings were Opaque dull well defined AW lesions surrounding the noted after 1’. os. Warts & leukoplakia turning AW. Dense, opaque AW visible ulceroproliferative growth WKCCPP Use dry cotton swab to wipe away Flashlight. Family planning nurses or clinical officers Negative: No AW lesions; faint patchy lesions without any discharge, blood or mucus. previously trained in both visual cervical definite margins; polyp; Nabothian follicles; faint line- Application of 5% AA to the cervix inspection techniques. like AW at SCJ; AW lesions far from TZ; streaky AW; with a cotton swab. After 1’ the dot-like AW in endocervix cervix was inspected. VIA+: sharp, distinct, well-defined, dense (opaque) AW areas, with or without raised margins, close to SCJ; dense AW lesions in columnar epithelium or near os; condyloma and leukopakia close to SCJ that turn intensely AW. Suspicious for cancer: visible ulcerative-proliferative growth; oozing and/or bleeding on touch. Indeterminate: early category based on uncertainty. Dropped after initial training phase. Thai After assessment of presence of Flash light 12 nurses with some experience in Positive: raised & thickened AW plaques. demonstration gross lesions consistent with cancer, reproductive & 4 skilled colposcopists health Cancer: cauliflower-like growth or ulcer; fungating trial 5% AA was applied to the cervix. were trained in VIA & cryotherapy. The mass. After 1’ the cervix was inspected colposcopists supervised the nurses. Special attention was given to the observation of the entire SCJ. 224 Study Description of the VIA test Illumination Qualification/training of "visual Definition of positive VIA result. Categorization of inspector" VIA result Ghanaian Application 3-5% AA, inspection of Hand-held flashlight or Nurse, midwife or other health-worker with Negative: smooth, pink, uniform, and featureless; demonstration the cervix after 1’. torch. experience in pelvic examinations with ectropion, polyp, cervicitis, inflammation, Nabothian trial adequate eyesight, trained in VIA & cysts. cryotherapy. Competency training of 2W, 1 Positive: raised & thickened AW plaques. M pilot with proficiency assessment. Cancer: cauliflower-like growth or ulcer, fungating Supervision by ObGyn physicians during the mass. whole study. 225 Table 55. Main characteristics of the other tests being evaluated (continued). Study Description of other tests Cross- CP: Cervex-Brush in Jaipur, Kolkata & Mumbai; Ayre’s spatula & thin cotton swab. The Cervex or spatula were placed at the os sectional and rotated gently over 2x360°. Material was uniformly spread on a slide, which was immediately fixed in 95% ethyl alcohol. multi-country Smears were processed in cytopathology labs in India. Interpretation according to the Bethesda System or CIN. CIN was Study converted into the Bethesda system. A sample of slides was reviewed in French laboratories. HC2: after collection of the cytology sample, a brush was rotated 3 full turns in the os and place into STM by breaking the tip of the sampler. The specimens were stored at –20°C until processing. Local laboratories were set up and the personnel trained in use of HC2 (prove B for high-risk HPV types). Cutoff at RLU>1 with +control containing 1pg HPV/mL. 300 samples were retested in France. VIAM: following VIA a 2nd health worker applied AA an visualized the cervix with a hand lens (4x) in Mumbai & Kolkata 2; in Kolkata 1 the Aviscope device, including an in-built rechargeable light source. VILI: after VIA and colposcopy, LI was applied to the cervix by an uninformed 2nd health worker. Interpretation using color charts. VILI positivity was defined as: presence of well-defined, dense, yellow non-iodine uptake areas touching the SCJ or circumorificial occupying large portion of the cervix. Growth on the cervix turning yellow. Cross- A cytology specimen was obtained with a wooden Ayre spatula. LSIL+ was the cytological cut-off. ASCUS & AGUS were sectional considered as cytologically negative. Interpretation by cytologists from the Univ Harare. CTs took part in a review course in Zimbabwe cytology. All + slides & 10% of negative slides were reviewed by a cytopathologist of Harare and also by a cytopathologist from J Study Hopkins, Baltimore (in phase I). In phase II: review only in J Hopkins. HC2 Ambilikai No intervention in the control group, where only cancers were detected via the normal existing health care and cancer registration RCT system. WKCCPP VILI: applied as screening test at the end of the study (137 cases). After VIA was done to identify cervical landmarks, LI was applied to the cervix with cotton swab. Cervix was examined with flashlight. VILI+ was defined as thick bright mustard- or saffron-yellow iodine-nonuptake areas seen touching the SCJ or os; circumferential yellow area surrounds os; or entire cervix appears mustard-yellow. VILI neg was normal mahogany brown or black cervix and columnar epithelium does not change color; patchy areas of no or partial uptake in the TZ; polyps or Nabothian follicles; “leopard skin” appearance; dots or patches far from SCJ. Suspicious for cancer: visible ulcerative-proliferative growth; oozing and/or bleeding on touch; densely yellow irregular surface. Thai No other tests. demonstration trial 226 Study Description of other tests Ghanaian No other tests. demonstration trial 227 Table 56. Main study characteristics of the applied gold standard used to validate screen test results. Author, year Gold standard Criteria for gold standard application Masking of screeners, colposcopists, histologist Categorization of the gold standard outcome Cross-sectional The colposcopist was unaware of the VIA findings. Colposcopy was performed on all women. Cytology, VIA and VILI were interpreted CIN1+ multi-country Colposcopy was done after VIA and completed after VILI. Punch biopsies were taken from colposcopically abnormal independently. CIN2/3 Study A 2nd colposcopy was done after VILI, by the same doctor. areas. Biopsies were assessed ii national laboratories. Histologists received colposcopy information. Cancer Colposcopists were medical staff being trained intensively Pathologists received colposcopy reports. Special internal during 15d using a IARC manual. Refreshment courses every and external QA procedures were introduced (not 4-6M during 1d. described). A sample of histology samples was reviewed in Biopsies were fixed in formaldehyde and processed & a reference laboratory. All personal received a training interpreted in the pathology department of RCC. In absence course. In absence of biopsies, colposcopy was used for of biopsies, colposcopy was accepted as gold standard. reference diagnosis. Cross-sectional Colposcopy at Harare University by faculty members. Phase I: Abnormal VIA result, cytology LSIL+, and every Screen tests were interpreted independently, also CIN1 Zimbabwe Biopsy if suspect colposcopy. (Treatment according to 10th women with normal/atypical VIA. Potential colposcopy was done without knowledge of CIN2 Study colposcopy/biopsy). verification bias. Phase II: All women. No verification screen test results. Biopsy interpretation was also bias. masked to colposcopy. Ambilikai Colposcopy with a 9-12x magnification by a nurse under VIA positive women. The colposcopist knew that she examined women CIN1 RCT supervision by a medical officer, who was consulted who were VIA+. CIN2-3 whenever in doubt. Categories: normal, inflammation, Cancer probable CIN1, probable CIN2-3, probable & frank invasive Ca. Punch biopsies from every abnormal area, which were examined in the Inst. Med. Science & Research of Coimbatore. WKCCPP Colposcopy targeted punch biopsies, in absence of biopsies VIA+ cases and VIA indeterminate cases. Triage tests (VIAM) were performed by different CIN1 expert colposcopy performed by Dr. E. Kauffman was At the end of the study (for 137) women: VILI was applied providers blinded to the results of the other triage CIN2 accepted as gold standard and in absence of expert as well and the VILI result were determinant for referral. tests. CIN3 colposcopy, nurse colposcopy was considered as pseudo-gold Cancer standard. Thai No confirmation of screen negatives. Outcome VIA+ cases (not suspicious for cancer, lesions not Not applicable. Not applicable. demonstration (complications, failure) after treatment was observed at 3M extending onto the vaginal wall, occupying <75% of the trial and 12M after treatment. cervix, not extending >2mm beyond the cryoprobe) were treated with cryotherapy immediately. Cryotherapy was delayed: menses, menses expected <7days, known HIV (referred to physician), polyp in cervical os (referred to physician for removal), fibroid tumor (>12W size) (referred to physician). 228 Author, year Gold standard Criteria for gold standard application Masking of screeners, colposcopists, histologist Categorization of the gold standard outcome Ghanaian Colposcopy and biopsy were offered only to women with No confirmation of screen negatives. Outcome VIA+ cases (not suspicious for cancer, lesions Not applicable. demonstration suspicion of cancer or with AW areas greater than 2/3 of the (complications, failure) after treatment was observed at 3M not extending onto the vaginal wall, occupying trial cervix or extending to the vaginal wall. and 12M after treatment. <75% of the cervix, not extending >2mm beyond the cryoprobe) were treated with cryotherapy immediately. Cryotherapy was delayed: menses, menses expected <7days, known HIV (referred to physician), polyp in cervical os (referred to physician for removal), fibroid tumor (>12W size) (referred to physician). 229 Table 57. Study design and remarks. Study Study design Remarks Cross-sectional multi- The test accuracy of five cervical cancer screening methods were evaluated in more than 58000 It is not clear how 1st colposcopy after VIA interfered with the 2nd country study women from eleven urban settings in India and five African countries. The assessed tests were: colposcopy. 1st and 2nd colposcopy interpretation was performed by naked eye visual inspection of the uterine cervix after application of acetic acid (VIA) or after the same colposcopist. Further question: was there a residual VIA application of Lugol’s iodine (VILI), VIA with a low level magnifying loop (VIAM), the effect on the VILI examination. conventional Papanicolaou smear and the HC2 test. The number of women examined and tests The contrast in sensitivity between VIA & VILI could be enhanced assessed per center is shown in Table 1. by the fact that in Kolkata, where sensitivity of VIA was lowest, VILI was not performed. Multiple screening tests were applied independently on the same women by different examiners, who were blinded towards the results of the other tests. All participants were subsequently inspected with colposcopy on the same day and, if a colposcopically suspect or abnormal lesion was identified, punch biopsies were taken. Colposcopists and histologists examining biopsies were masked with respect to the screening test results. The histopatho- logical diagnosis was used as gold standard for women from whom biopsies were taken; otherwise the colposcopic impression was used for determination of the final cervical status. This final outcome was categorized in five classes: normal or non-neoplastic changes, cervical intra-epithelial neoplasia grade 1 (CIN1) including HPV changes, CIN2, CIN3 and invasive cancer. Cross-sectional Zimbabwe Phase I: only VIA+ (AW lesions, or worse aspect) and LSIL+ cases and 10% VIA- women 10% random sample of VIA-/cyto- did not participate at random at Study were referred for colposcopy. colposcopy, therefore this sample could not be used to correct for Phase II: all women were referred for colposcopy. verification bias. Unfortunately this non random participation is not documented. Was it significantly not random? Ambilikai RCT Cluster randomized study. 57 panchayaths (villages) with 48,225 women among whom 30,577 Biopsies were taken from colposcopically abnormal areas. participated in the study arm, receiving a single round of VIA versus 56 villages with 30,172 Immediate cryotherapy on low-grade & high-grade lesions if <3/4 of women in the control arm receiving usual care. the ectocervix, without extension in vagina/endocervix. More extensive lesions were referred for LEEP/conization. Suspect cancer cases were referred to a cancer treatment center. WKCCPP Women were examined with VIA; VIA+ and VIA-indeterminate cases were referred at the Attendance at colposcopy referral and follow-up of treated women district hospital for colposcopy. Biopsies were taken from colposcopic suspect areas. VIAM was limited. The study protocol changed over the period. The gold and cytological examination were performed as well as alternative triage methods. Decision standard definition was not homogenous. for treatment with cryotherapy was based on the colposcopic impression. Women having received cryotherapy or screen+/colposcopic negative women were followed at 1-3M and at 12M. Thai demonstration trial Testing of 5999 women with VIA, immediate cryotherapy of VIA+ (referral of cases suspect This study only provides test-positivity rates for the meta-analysis. for cancer). Observation of complications, failure, satisfaction after treatment. Nevertheless, it provides information on safety, acceptability, feasibility of immediate treatment without confirmation of the screen test result. Ghanaian demonstration Testing of 3665 women with VIA, immediate cryotherapy of VIA+ (referral of cases suspect This study only provides test-positivity rates for the meta-analysis. trial for cancer). Observation of complications, failure, satisfaction after treatment. Nevertheless, it provides information on safety, acceptability, feasibility of immediate treatment without confirmation of the screen test result. 230