1166 Lung Cancer in Telephone Linemen • Meguellati-Hakkas et al Lung Cancer Mortality and Occupational Exposure to Asbestos Among Telephone Linemen: A Historical Cohort Study in France M A Djamila Meguellati-Hakkas, MS sbestos dust has well-known carcino- O Diane Cyr, MS genic properties.1,2 Most studies on asbestos and lung cancer risk have .C ¨ Isabelle Stucker, PhD been carried out in populations with ¨ ´ Joelle Fevotte, MS elevated occupational exposures such PE Corinne Pilorget, PhD as asbestos miners and millers,3,4 tex- tile workers,5 or asbestos– cement fac- ` Daniele Luce, PhD tory workers,6 but the carcinogenic A ´ Pascal Guenel, MD, PhD effects of intermittent exposures to low -T levels of asbestos has not been inves- Objective: The authors studied the mortality by lung cancer in tigated extensively. It is generally as- N telephone linemen exposed to asbestos at low levels during installation of sumed that the risk of lung cancer telephone cables. Methods: Three hundred eight lung cancers deaths associated with low exposure to asbes- .C were identified in the cohort. Exposure to asbestos and to other tos can be predicted from a downward occupational carcinogens was assessed using a job-exposure matrix. linear extrapolation of the risk ob- W Results: The relative risk for lung cancer death associated with an served in highly exposed cohorts.7 estimated exposure of approximately 2 f/cc-years was 2.1 (95% con- Several groups of workers operating in W fidence interval 1.1– 4.0) as compared with workers exposed to less asbestos-insulated areas such as tele- than 0.5 f/cc-years. Mean annual exposure or exposure duration were phone linemen may provide empirical W not clearly related to lung cancer. Adjustment for other occupational observation of lung cancer risk due to sporadic exposure to asbestos. To our lung carcinogens did not change this finding. Conclusion: The observed knowledge, however, no epidemio- t mortality by lung cancer associated with asbestos exposure at low levels ne logic study on respiratory cancer risk is higher than the prediction based on linear downward extrapolations has been specifically carried out in from highly exposed occupational cohorts. (J Occup Environ Med. telephone linemen, possibly because y 2006;48:1166 –1172) occupational exposure levels have lth been lower and hence more difficult to assess than in other industries and ea because the variety of exposures to potential lung carcinogens at the work- H place makes difficult the study of a particular agent separately. We conducted a study in the French national telephone utility, the only tele- phone company operating in France at From Inserm, U754, IFR69 (Dr Meguellati-Hakkas, Dr Stucker, Dr Guenel), Villejuif, France; ¨ ´ the time of the study. Because of con- ´ Inserm, U687, IFR69 (Dr Cyr, Dr Luce), Saint-Maurice, France; and Universite Claude Bernard (Dr cern due to a cluster of deaths from ´ Fevotte, Dr Pilorget), Lyon, France. respiratory cancers and other cancers ´´ The study was financially supported by Fondation de France and by France Telecom. Address correspondence to: Pascal Guenel, MD, PhD, INSERM U754, 16 avenue Paul Vailllant- ´ sites in a group of telephone linemen, a Couturier, 94807 Villejuif Cedex, France; E-mail: email@example.com. nationwide historical cohort study of Copyright © 2006 by American College of Occupational and Environmental Medicine cancer mortality was initiated in this DOI: 10.1097/01.jom.0000243357.70143.47 large company. Past exposures to oc- JOEM • Volume 48, Number 11, November 2006 1167 cupational hazards were assessed International Classification of Dis- TABLE 1 based on recollection of workers with eases, 9th Revision.8 The record of Occupational Exposures Included in longstanding experience to investigate the cause of death was found by the Job-Exposure Matrix—Telephone cancer risks in relation to occupational matching on gender, date and place Linemen Cohort Study exposures. We report on lung cancer of birth, and date and place of death Acid mortality associated with intermittent of the deceased employees. In total, Isopropanol asbestos exposure occurring mainly 2917 matches were obtained (97.6% Asbestos during telephone cable installation in of all deaths). Benzene asbestos-insulated areas. Adhesives (PVC material or epoxy) Occupational History Ceramic fibers Manmade mineral fibers Materials and Methods Individual work histories since the Welding fumes The cohort included workers in start of employment in the company Car exhaust fumes were obtained from electronic records Herbicide M activity in a technical branch of the kept by the personnel department. Jobs Polycyclic aromatic hydrocarbons (PAH) telephone company on January 1, were defined by both an occupation Isocyanates O 1978, as well as men newly hired Plastics between January 1, 1978, and Decem- and a sector of activity. Occupations Electromagnetic fields (ELF or .C ber 31, 1994. Only men fulfilling these were coded according to one of the radiofrequency) criteria and employed as telephone line- following six groups: lineman, su- Paints PE pervisor, technician, special skill Lead men at any time during their work his- worker, garage worker, or office Wood dust tory were included in the present cohort, Wood preservatives representing a total of 34,305 workers. worker. All cohort members were Ionizing radiations A employed as linemen during at least Solvents (chlorinated, oxygenated, or Vital Status and Cause of Death one job period but may have been -T petroleum solvents) Follow up of individual cohort employed in any other occupation Styrene during their work history in the com- Ultraviolet radiation members started on January 1, 1978, N and terminated at the date of death or pany. Sectors of activity were one of .C on December 31, 1996, whichever the following: construction of tele- came first. The deceased workers, and phone lines, maintenance of tele- quency (F: proportion of exposed work W the corresponding date of death, were phone lines, construction of national time), and exposure intensity (I: es- identified from the National Mortality telephone network, maintenance of the timate of the number of fibers per W database collecting information on all national telephone network, interven- cubic centimeters for asbestos or ap- deaths in France. This database was tion units, garages, or other sectors. propriate exposure score for other provided by the National Institute of For each job period, the date of start carcinogens). W Statistics for the years 1978 through and the date of termination of employ- Exposure assessment was super- 1996 and permitted the identification ment were obtained. vised by occupational hygienists and t of the deceased workers in the cohort epidemiologists according to a stan- ne by matching on name, sex, date of Assessment of dardized protocol. First, a list of work birth, and place of birth. To overcome Occupational Exposures tasks entailing exposure to the occupa- y possible problems due to inaccuracies Past exposures were assessed by tional carcinogens was established. or missing data in the national mortal- experts using a job-exposure matrix Work tasks entailing exposure to as- lth ity database, each cohort member was (JEM) approach, ie, a crosstabulation bestos were as follows: using and cut- also sought in the files implemented by of jobs and occupational exposures. ting asbestos sheets as protection ea the personnel department of the tele- Jobs were coded in the JEM accord- equipment during welding, cutting, or phone company that keeps track of ing to both the occupation and the drilling asbestos material for insulation H deaths among active and retired work- sector of activity, as described pre- purposes, installation of telephone ers. The two mortality registers were viously. Exposures were well-estab- lines, or cable splicing in asbestos- first used as independent sources of lished or suspected occupational insulated buildings. Experts were information and were then compared carcinogens that occurred at the work- asked to assess the proportion of work- for identifying duplicates. Based on place (Table 1). Different time periods ers within each job who carried out a these two files, a total of 2989 deaths could be defined in the JEM for a particular work task and the time du- were identified in the cohort. given job to account for changes in ration when the work task was per- The National Register of Causes exposures. For a given exposure in a formed in hours per week or month. ´ of Death (CepiDC–INSERM) was given job period, semiquantitative in- Exposure probability and exposure fre- used to obtain the cause of death of dices were used to define exposure quency were independently obtained the deceased subjects as coded from probability (P: proportion of exposed from three information sources: a the death certificate according to the workers in a given job), exposure fre- group constituted of occupational phy- 1168 Lung Cancer in Telephone Linemen • Meguellati-Hakkas et al sicians, epidemiologists, and industrial Statistical Analyses 1.05) (figures not shown). In internal hygienists; a group of experienced or The expected number of lung cancer analyses, relative risks (RRs) were retired workers with good knowledge deaths was calculated using the French calculated by comparing lung cancer of the work practices in the company; national age- and period-specific mor- mortality between exposure groups and a large random sample of current tality rates for lung cancer in men as a within the cohort using different salaried workers who completed a self- standard to obtain standardized mortal- variables for estimating asbestos ex- administered questionnaire on the ity ratios (SMRs). Most analyses pre- posure (Table 3). The RR in the work tasks. Whenever a difference oc- sented in this article were carried out highest cumulative exposure cate- curred between exposure indices, a internally by comparing asbestos ex- gory (estimate 1.7 f/cc-years) was final consensus was reached during posure groups within the cohort with 2.1, and it was statistically signifi- meetings of occupational physicians, no reference to external mortality cant. Slightly increased RRs were experienced workers, and industrial rates. These analyses were conducted observed in the intermediate catego- hygienists. The intensity of exposure ries of cumulative exposure, but M using Cox proportional hazards mod- associated with each task (eg, semi- els,9,10 in which age was used as the there was no evidence of a linear quantitative exposure indices or esti- dose–response relationship. Because O time scale and asbestos exposure mated fibers per cubic centimeters for scores were treated as time-depen- the range of exposure was small, .C asbestos) was determined by the indus- dent variables. Adjustment for calen- only three exposure categories were trial hygienists from the detailed dar period (in 5-year periods) was used for the mean annual exposure. PE description of the work task and a used in all the models. Because the No significant increase in lung can- literature review. When more than one exposure that occurs in the years cer mortality was observed (RR work task caused exposure to the same before lung cancer death may be 1.3) when comparing workers with A carcinogen in a given job, the expo- etiologically irrelevant to the disease, mean annual exposure above the 75th percentile with workers with -T sures indices attributed to the job were cumulative exposure was also calcu- calculated as a weighted average of the lated by applying a lag time period of mean exposure below the 25th per- exposure indices for the different work centile of exposure distribution. N 5, 10, or 20 years and was also fitted tasks. as a time-dependent variable. Adjust- However, the exposure gradient be- .C ment for other occupational lung car- tween exposure groups was too small cinogens was carried out in separate for a meaningful analysis. Duration Individual Exposure Scores W analyses to evaluate a potential con- of asbestos exposure was associated Annual exposure scores were cal- founding effect. All analyses were with some increase in risk, particu- W culated as the product P F I of carried out using STATA software.11 larly in workers exposed more than the exposure indices in the JEM. 32 years, but none of the RRs were Cumulative exposure to an occupa- statistically significant. W tional carcinogen was then calcu- Results Cumulative exposures were calcu- lated for each worker as the sum of A total of 589,162 person-years of lated allowing for latency periods of t the annual exposure scores over the follow up were accumulated by co- 5, 10, or 20 years before cancer ne entire work history. Mean annual ex- hort members. Among the 2989 death, ie, not accruing cumulative posure was also calculated as the deaths from all causes that occurred exposure by annual exposure scores cumulative exposure divided by expo- y in the cohort, 308 were from lung during these periods (Table 4). Ele- sure duration in years. cancer. Descriptive data on employ- vated exposure to asbestos was more lth A priori exposure cut points were ment, follow up, and estimated as- strongly associated with lung cancer defined for the exposure variables to bestos exposure are shown in Table 2 mortality when accounting for a la- ea take apart the highest possible expo- for all study subjects and for lung tency period of 5 years. Some in- sure levels. We defined a reference cancer cases separately. The mean crease in risk of lung cancer death H exposure category that also included year of birth was 1931 for the cases was still apparent after allowing for a workers with low exposure, because of lung cancer and 1945 for the latency period of 20 years. only few linemen had never been whole cohort. Indices of asbestos Results of the multivariate analysis exposed to asbestos. For cumulative exposure at the end of the follow up are shown in Table 5. Exposures to exposure to asbestos, the 10th, 25th, were slightly higher in cases of lung asbestos and to other lung carcinogens 50th, 75th, and 95th percentiles were cancer deaths than in the rest of the selected among occupational expo- chosen as exposure cut points. For cohort. sures that occurred in company are mean annual exposure, the variabil- The expected number of lung can- included in the models. Adjusting for ity of exposure was very low, and cer deaths in the cohort based on the PAH, engine exhaust and arc welding only the 25th and the 75th percen- French national mortality rates was fumes did not modify substantially the tiles of exposure distribution were 328.2, leading to a SMR of 0.94 RR associated with asbestos exposure, used as cut points. (95% confidence interval 0.84 – although it was no longer statistically JOEM • Volume 48, Number 11, November 2006 1169 TABLE 2 Selected Characteristics of Employment, Follow Up, and Asbestos Exposure—Telephone Linemen Cohort Study, 1978 –1996 All Cohort Subjects (n 34,305) Lung Cancer Deaths (n 308) Mean (SD) Minimum Maximum Mean (SD) Minimum Maximum Year of birth 1945 (12.4) 1913 1974 1931 (7.6) 1917 1959 Mortality follow up Age at start of follow up (yr)* 34 (11.3) 17 65 47 (7.5) 19 60 Age at end of follow up (yr)† 51 (12.1) 19 84 59 (8.0) 34 74 Duration of follow up (yr) 17 (3.6) 1 19 12 (4.6) 1 19 Employment Year at start of employment 1970 (10.1) 1937 1994 1959 (8.2) 1943 1985 Age at start of employment (yr) 25 (5.1) 15 59 29 (5.5) 19 52 Duration of employment (yr) 23 (7.4) 1 45 26 (6.1) 4 40 M Proportion of work time as a 85 (30) 1 100 93 (21) 1 100 lineman (%) O Asbestos exposure‡ Mean annual exposure (f/cc)§ 0.05 (0.01) 0.01 0.42 0.05 (0.01) 0.03 0.06 .C Cumulative exposure (f/cc-yr) 0.87 (0.54) 0.01 12.53 1.13 (0.43) 0.01 2.04 Exposure duration (yr) 17.0 (9.5) 0.1 42.0 22.0 (7.9) 0.3 37.0 PE *Start of follow up: January 1, 1978, or date of start of employment if it occurred later. †End of follow up: date of death or December 31, 1996, at the latest. ‡Among workers ever exposed to asbestos (30,625 in the total cohort and 305 lung cancer deaths). Exposure values are estimated as A explained in the text. §Excluding periods with no exposure to asbestos. -T SD indicates standard deviation. N TABLE 3 .C Relative Risk of Lung Cancer Death According to Exposure Scores for Asbestos—Telephone Linemen Cohort Study, 1978 –1996 Mean Exposure No. of Lung W in the Cancer Relative 95% Confidence Exposure Percentile Exposure Limits Exposure Group Deaths Risk Interval W Cumulative asbestos 0% to 10% 0.00 – 0.50 0.36 32 1.0 — exposure (f/cc-yr*) 10% to 25% 0.50 – 0.86 0.76 48 1.2 0.7–1.8 W 25% to 50% 0.86 –1.23 1.10 76 1.4 0.9 –2.1 50% to 75% 1.23–1.46 1.39 75 1.3 0.8 –2.1 75% to 95% 1.46 –1.73 1.59 61 1.3 0.8 –2.0 t 95% 1.73 1.91 16 2.1 1.1– 4.0 ne Mean annual asbestos 0% to 25% 0.00 – 0.05 0.048 45 1.0 — exposure (f/cc*) 25% to 75% 0.05– 0.055 0.051 187 1.2 0.9 –1.6 75% 0.055 0.057 76 1.3 0.9 –1.8 y Exposure duration (yr) 0% to 10% 0.0 –10.0 7.5 35 1.0 — lth 10% to 25% 10.0 –17.1 15.2 45 1.0 0.6 –1.6 25% to 50% 17.1–23.5 21.5 75 1.4 0.9 –2.1 50% to 75% 23.5–27.8 26.6 77 1.2 0.8 –1.9 ea 75% to 95% 27.8 –32.3 30.3 60 1.1 0.7–1.8 95% 32.3 33.7 16 1.6 0.9 –3.0 H *Number of fibers per cubic centimeter estimated from semiquantitative exposure indices as explained in the text. significant. There was no indication of phone linemen exposed to asbestos response relationship between cu- an association between occupational mainly from telephone cable instal- mulative exposure and lung cancer exposures to these carcinogens and lation in asbestos-insulated build- mortality was not apparent. Exposure lung cancer death. ings. We found that the risk of lung to other lung occupational carcinogens cancer increased twofold in workers did not confound the results. Discussion with an estimated cumulative expo- The study was based on a large In this historical cohort study, lung sure of approximately 2 f/cc-years, a cohort of telephone linemen with a cancer risk was investigated in tele- low exposure value. A linear dose– long duration of follow up (median 1170 Lung Cancer in Telephone Linemen • Meguellati-Hakkas et al 19 years), including over 300 lung TABLE 4 cancer deaths. All workers were Relative Risk of Lung Cancer Death According to Cumulative Exposure to Asbestos Allowing for Different Lag Time Periods Between Exposure and state-employed, representing a stable Death—Telephone Linemen Cohort Study, 1978 –1996 population in which the follow up in Cumulative Exposure No. of Relative 95% Confidence mortality registers was highly facili- Lag Time (f/cc-yr*) Deaths Risk Interval tated. Because we used two mortal- 5 yr 0 – 0.43 34 1.0 — ity registers, one national and one 0.43– 0.77 46 1.0 0.6 –1.6 specific to the company, a crossvali- 0.77–1.11 76 1.1 0.7–1.8 dation could be carried out and vir- 1.11–1.38 76 1.3 0.8 –2.0 tually no worker was lost to follow 1.38 –1.72 61 1.0 0.6 –1.6 1.72 15 2.3 1.2– 4.5 up for mortality. Moreover, the 10 yr 0 – 0.35 38 1.0 — causes of deaths were obtained from another national database for 98% of M 0.35– 0.62 44 1.1 0.7–1.7 0.62– 0.93 76 1.1 0.7–1.6 the deceased workers. 0.93–1.23 75 1.0 0.7–1.6 O 1.23–1.60 60 0.9 0.6 –1.4 Exposure Misclassification 1.60 15 1.4 0.7–2.7 .C 20 yr 0 – 0.16 69 1.0 — Because measurement data of past 0.16 – 0.32 40 1.0 0.6 –1.5 occupational exposures were only PE 0.32– 0.57 66 0.8 0.6 –1.2 very sparse or inexistent in the com- 0.57– 0.88 68 0.7 0.4 –1.0 pany, exposure to carcinogens was 0.88 –1.26 51 0.7 0.5–1.2 assessed using the expertise of com- A 1.26 14 1.4 0.7–2.8 pany workers, occupational physi- *Number of fibers per cubic centimeter estimated from semiquantitative exposure indices cians and occupational hygienists -T as explained in the text. with longstanding experience. These work groups rated the proportion of N workers within each job who had .C executed well-defined work tasks TABLE 5 and the frequency with which the W Multivariate Analysis of Relative Risk for Lung Cancer Death According to work tasks were carried out. Three Cumulative Exposure to Asbestos After Adjustment for Exposure to Arc Welding independent sources of information Fumes, Engine Exhaust, and PAHs—Telephone Linemen Cohort Study, 1978 –1996 W permitted an internal validation of No. of Relative 95% Confidence the exposure assessment process, and Cumulative Exposure Deaths Risk Interval a final consensus was obtained. De- W Asbestos (f/cc-yr*) spite considerable effort to obtain 0 – 0.50 32 1.0 — 0.50 – 0.86 48 1.4 0.8 –2.5 valid exposure estimates, nondiffer- t ential exposure misclassification was ne 0.86 –1.23 76 1.6 0.8 –3.0 1.23–1.46 75 1.4 0.7–2.9 likely to occur. To examine the effect 1.46 –1.73 61 1.3 0.6 –2.8 of exposure classification on the re- 1.73 16 2.1 0.9 –5.3 y sults, we conducted sensitivity anal- Arc welding (yr of exposure) yses in which jobs were considered lth 0 54 1.0 — 0 – 0.03 127 1.2 0.8 –1.6 as nonexposed to asbestos if the ex- 0.03– 0.04 64 1.3 0.8 –2.2 posure probability and/or the expo- ea 0.04 63 1.4 0.7–2.8 sure frequency were below different Engine exhaust (ppm CO-yr†) thresholds. However, the findings H 0 –20.4 30 1.0 — 20.4 –39.0 124 0.8 0.4 –1.5 were not changed importantly (results 39.0 – 46.0 77 0.8 0.4 –1.7 not shown). 46.0 77 0.7 0.3–1.5 Besides employment in the com- PAH (unit-yr‡) pany, exposure to asbestos may have 0 –1.8 30 1.0 — occurred in other companies or out- 1.8 –15.2 124 0.7 0.4 –1.1 15.2–24.6 77 0.7 0.4 –1.2 side work, but this information was 24.6 77 1.0 0.5–2.0 not available. It is possible that some workers were exposed to asbestos in *Number of fibers per cubic centimeter estimated from semiquantitative exposure indices jobs held before, rather than after, as explained in the text. †Estimated cumulative exposure to carbon monoxide in ppm-yr used as an indicator of their employment in the telephone exposure to engine exhaust fumes. company, because almost all workers ‡Cumulative exposure values using arbitrary exposure scores. leave the company at the age of JOEM • Volume 48, Number 11, November 2006 1171 retirement. Although the unidentified (PAH), arc welding fumes, and en- that lung cancer mortality rises sources of asbestos exposure outside gine exhaust as potential confound- slowly with mean annual exposure. the company may contribute to ex- ers. For other lung carcinogens such For exposure duration, a more posure misclassification, we believe as ionizing radiations, only a small distinct, although not significant, that this problem is of relatively mi- number of subjects were exposed at pattern was observed in workers ex- nor importance because most of the low levels, and they were not intro- posed more than 32 years. It is there- workers’ careers were completed in duced in the multivariate models. fore possible that the conjunction of the company. Moreover, these addi- The adjustment for the potential con- both a high exposure duration and a tional sources of exposure are most founders did not modify the relation relatively high annual exposure is likely distributed equally across between exposure to asbestos and necessary for lung cancer risk to be cases and noncases of lung cancer lung cancer. In addition, none of these apparent. death, leading to nondifferential ex- exposures was independently associ- Our finding of an increased lung posure misclassification. ated with lung cancer mortality in our cancer mortality associated with cu- M data. mulative exposure contrasts with risk Confounding predictions based on downward ex- O Like most occupational cohort trapolations from highly exposed oc- Comparison With .C studies, individual data on tobacco cupational cohorts. In these studies, a smoking were not available in our Previous Studies wide range of exposure was ob- PE study on telephone linemen. Con- Asbestos is not known to be a served, from 1 f/cc14 to more than founding due to tobacco smoke oc- major occupational hazard in tele- 250 f/cc.15 These values are clearly curs when a marked difference in the phone linemen. These workers, how- distinct from the mean exposure of A prevalence of tobacco smoking ex- ever, may have been exposed during 0.05 f/cc estimated for all workers in ists between compared exposure installation or maintenance of tele- the present study, although this value groups. It has been recognized that relative risks above 1.4 between -T phone cables in asbestos-insulated buildings or areas or when using is based on experts’ judgment. Based on the findings reported in the highly N occupation and lung cancer are un- protective equipment containing as- exposed cohorts, most authors as- .C likely due to uncontrolled confound- bestos, eg, for welding. These activ- sume a linear relationship between ing from tobacco smoking.12 It has ities are not specific to telephone relative risk of lung cancer mortality W been demonstrated recently that sys- linemen, and similar exposures in and cumulative exposure to asbes- tematic or chance differences in life- other occupations performing com- tos.7,16 It can also be estimated from W style habits such as tobacco smoking parable work tasks may occur. We a literature review that 1-f/cc-year between exposure groups in occupa- found a significantly twofold in- increase in exposure is associated tional cohort studies is unlikely to creased RR of lung cancer mortality with approximately 1% increase in W cause a more than 20% change in for linemen in the highest category RR.16,17 Under this assumption, a relative risk.13 In the present study, of cumulative exposure estimate. RR of 1.02 should be observed for t the observed number of lung cancer The average asbestos exposure in workers in the highest cumulative ne deaths in the cohort was slightly this group was estimated to be ap- exposure group of our study, a value lower than the expected number proximately 2 f/cc-years. The associ- much lower than the RR of 2.1, based on national mortality rates y ation between asbestos exposure and which was actually observed in the (SMR 0.94), indicating that the lung cancer was strengthen after al- present study. lth tobacco consumption in telephone lowing for a 5-year latency period, Although it is possible that the linemen is similar or slightly lower indicating that the most recent expo- exposure to asbestos is slightly un- ea than in the general population. In sure was unimportant to cancer risk. derestimated in our study, it cannot addition, because we carried out in- These findings suggest a possible fit the prediction for lung cancer H ternal analyses within a group of causal association between intermit- mortality derived from studies in socially homogeneous workers with tent exposure to asbestos and lung highly exposed populations. In these identical lifestyles, the differences in cancer mortality. studies, asbestos was usually present prevalence of tobacco smoking be- Because cumulative exposure is permanently in the workers environ- tween exposure groups should be the product of the mean annual ex- ment,16 and hence the results provide modest, making a strong confound- posure dose and the exposure dura- information on lung cancer risk as- ing effect very unlikely. tion, these later exposure parameters sociated with continuous exposures Confounding from other occupa- were studied separately. The vari- only. Conversely, in the present co- tional lung carcinogens was investi- ability of the mean annual exposure hort, a sporadic pattern of asbestos gated using the exposure data collected was small and did not permit a com- exposure exists. For example, expo- for the JEM. We selected exposure to parison between clearly contrasted sure peaks may occur in workers polycyclic aromatic hydrocarbons exposure groups, but it is suggested installing telephone lines in asbestos- 1172 Lung Cancer in Telephone Linemen • Meguellati-Hakkas et al insulated buildings but will not for lar low asbestos exposures. These 9. Cox DR. Regression models and life the same task at other work places. It findings provide additional support tables (with discussion). J R Stat Soc Series B. 1972;34:187–220. is therefore possible that peak expo- that even low levels of exposure to 10. Breslow NE, Day NE. Statistical Meth- sures are the most etiologically rele- asbestos in occupational settings ods in Cancer Research. Volume II—The vant exposures for lung cancer in should be strictly avoided. Design and Analysis of Cohort Studies. workers who are intermittently ex- Lyon: IARC Scientific Publications; 1987: posed to asbestos. Unfortunately, Acknowledgments 1– 406. peak exposures could not be ob- The authors are indebted to the groups of 11. 2003 Stata Statistical Software, release served directly in the present study experts who participated in the occupa- 8.0. College Station, TX: Stata Corp; because of its retrospective design. tional exposure assessment process, in par- 2005. Several other studies in occupa- ticular the MATCOM group: Dr Bernard 12. Siemiatycki J, Wacholder S, Dewar R, et Siano, Dr Martine Aucomte, Dr Annie Bar- al. Degree of confounding bias related to tions with low exposure and lung reau, Dr Bruno Jahan, Dr Nicole Honorat, smoking, ethnic group, and socioeco- cancer point to the same direction. In M M. Marc Mougel, and Dr Maryse Ville- nomic status in estimates of the associa- a population-based case– control goureix; and the group of employees who tions between occupation and cancer. study carried out in Sweden,18 the O provided information of invaluable interest J Occup Med. 1988;30:617– 625. cumulative exposure dose of 4 f/cc- on occupational exposures and work prac- 13. Kriebel D, Zeka A, Eisen EA, et al. .C years was associated with a RR of tices: Alain Jubeau, Michel Robert, Yves Quantitative evaluation of the effects of Le Gohalen, Charles Mauzale, Daniel Du- 1.9 (95% confidence interval 1.3– uncontrolled confounding by alcohol clos, Robert Tressard, Philippe Perrault, PE 2.7). In a cohort study conducted and tobacco in occupational cancer Jean-Pierre Kolczik, Jacky Poingt, Alain studies. Int J Epidemiol. 2004;33:1040 – among gas and electricity utility ´ ´ Guelode, Bernard Lerat, and Jean-Claude 1045. workers in France,19 an RR of 2.0 Martzolff. The authors also thank Alice A 14. Peto J. Lung cancer mortality in relation to (95% confidence interval 1.3–3.2) ´ Gueguen for her help in data analysis. measured dust levels in an asbestos textile -T for lung cancer was reported for factory. IARC Sci Publ. 1980:829 – 836. exposure as low as 8.1 f/cc-years. No References N 15. McDonald JC, Liddell FD. Mortality in further increase in risk was observed 1. Doll R. Mortality from lung cancer in Canadian miners and millers exposed to in higher exposure group. In another asbestos workers. Br J Ind Med. 1955;12: chrysotile. Ann N Y Acad Sci. 1979;330: .C recent cohort study among pulp and 81– 86. 1–9. paper industry workers,20 the RR for 2. Asbestos. Monographs on the Evaluation 16. Expertise Collective INSERM: Effets sur la W of Carcinogenic Risks to Humans, vol 14. ´ sante des principaux types d’exposition a ` lung cancer was 1.44 (95% confidence Lyon: International Agency for Research l’amiante. [Health Effects of the Main interval 0.85–2.45) for exposure to on Cancer; 1977. Types of Asbestos Exposure.] Paris: Les W asbestos greater or equal to 0.78 f/cc- 3. Nicholson WJ, Selikoff IJ, Seidman H, et Editions INSERM; 1997. years, and there was an indication of a al. Long-term mortality experience of 17. Nicholson WJ. Airborne Asbestos Health W dose–response relationship. chrysotile miners and millers in Thetford Assessment Update. Washington, DC: In the present study, it was sug- Mines, Quebec. Ann N Y Acad Sci. 1979; US Environmental Protection Agency, gested that lung cancer mortality in 330:11–21. EPA-600/8 – 84/003F, Office of Health t 4. McDonald JC, Liddell FD, Gibbs GW, et and Environmental Assessment; 1986. telephone linemen with low asbestos ne al. Dust exposure and mortality in chry- exposure may be higher than what 18. Gustavsson P, Nyberg F, Pershagen G, sotile mining, 1910 –75. Br J Ind Med. et al. Low-dose exposure to asbestos can be suspected from previous stud- 1980;37:11–24. and lung cancer: dose–response rela- y ies in highly exposed workers. This 5. Dement JM, Harris RL Jr, Symons MJ, tions and interaction with smoking in a lth finding was not likely due to uncon- et al. Exposures and mortality among population-based case-referent study in trolled confounding from tobacco chrysotile asbestos workers. Part II: Stockholm, Sweden. Am J Epidemiol. mortality. Am J Ind Med. 1983;4:421– smoking or from occupational car- ea 433. 2002;155:1016 –22. cinogens other than asbestos. The 6. Finkelstein MM. Mortality among long- 19. Imbernon E, Goldberg M, Bonenfant S, increased lung cancer mortality in et al. Occupational respiratory cancer and H term employees of an Ontario asbestos- the highest cumulative exposure cement factory. Br J Ind Med. 1983;40: exposure to asbestos: a case– control group was possibly related to peak 138 –144. study in a cohort of workers in the 7. Omenn GS, Merchant J, Boatman E, et electricity and gas industry. Am J Ind exposures to asbestos. This assump- al. Contribution of environmental fibers Med. 1995;28:339 –352. tion requires confirmation, but it is 20. Carel R, Boffetta P, Kauppinen T, et al. to respiratory cancer. Environ Health potentially important because it may Perspect. 1986;70:51–56. Exposure to asbestos and lung and pleu- concern several occupational groups 8. International Classification of Diseases. ral cancer mortality among pulp and pa- performing the same work tasks as Geneva: World Health Organization; per industry workers. J Occup Environ telephone linemen and having simi- 1977. Med. 2002;44:579 –584.