Fifteen-Year Survival in Prostate Cancer

Document Sample
Fifteen-Year Survival in Prostate Cancer Powered By Docstoc
					 Fifteen-Year Survival in Prostate Cancer
A       Prospective, Population-Based Study in Sweden
Jan-Erik Johansson, MD, PhD; Lars Holmberg, MD, PhD; Sara Johansson;
Reinhold Bergstr\l=o"\m, PhD; Hans-Olov Adami, MD, PhD

  Objective.\p=m-\To            describe the natural history of initially untreated early-stage                      mortality rate from prostate cancer, both
prostate cancer. A key secondary objective was to calculate long-term survival rates                                 at 5 years6 and 10 years7 of follow-up,
by stage, grade, and age at diagnosis.                                                                               challenged the use of aggressive initial
                                                                                                                     treatment for all patients with early-
  Design.\p=m-\Prospectivecohort study.
  Seting.\p=m-\Population-based            in 1 county of Sweden, without screening for pros-                        stage disease. Similarly high survival
                                                                                                                     rates were revealed in a pooled analysis
tate    cancer.
                                                                                                                     of cohorts from different countries.8
    Patients.\p=m-\A group of 642 patients with prostate cancer of any stage, consecu-
tively diagnosed between 1977 and 1984 at a mean age of 72 years with complete                                         For editorial comment       see     497.
follow-up to 1994.
    Main Outcome Measures.\p=m-\Proportionof patients who died from prostate can-
cer, and 15-year survival (with 95% confidence interval [CI]), corrected for causes                                     The generalizability of our results has
of death other than prostate cancer.                                                                                 been questioned, mainly on the ground
    Results.\p=m-\Inthe entire cohort, prostate cancer accounted for 201 (37%) of all 541                            that some patients who received initial
deaths. Among 300 patients with a diagnosis of localized disease (T0-T2), 33 (11%)                                   treatment were excluded from the pub¬
died of prostate cancer. In this group, the corrected 15-year                                                        lished analyses.9·10 Recently, a Swedish
                                                                 survival rate was simi-
lar in 223 patients with deferred treatment (81%; 95% CI, 72%-89%) and in 77 who                                     study10—widely cited1113 but profoundly
                                                                                                                     criticized for its flawed design11·14·15—sug¬
received initial treatment (81%; 95% CI, 67%-95%). The corrected 15-year survival
was 57% (95% CI, 45%-68%) in 183 patients with locally advanced cancer (T3-T4)
                                                                                                                     gested much lower survival rates even in
and 6% (95% CI, 0%-12%) in those 159 who had distant metastases at the time
                                                                                                                     early-stage disease. To strengthen the
                                                                                                                     validity of our data, this article is based
of diagnosis.                                                                                                        on 15 years of complete follow-up of the
   Conclusion.\p=m-\Patientswith localized prostate cancer have a favorable outlook                                  entire population-based cohort compris¬
following watchful waiting, and the number of deaths potentially avoidable by radi-                                  ing 642 patients in whom prostate cancer
cal initial treatment is limited. Without reliable prognostic indicators, an aggressive                              was  consecutively diagnosed during          a

approach to all patients with early disease would entail substantial overtreatment.                                  7-year period.
In contrast, patients with locally advanced or metastatic disease need trials of ag-
gressive therapy to improve their poor prognosis.                                                                    PATIENTS AND METHODS
                                                                                            JAMA. 1997;277:467-471   Patients
                                                                                                                        From March 1977 to February 1984,
STUDIES of the natural history of pros¬                                 evidence of tumor growth and dissemi¬        648 consecutive cases of cancer of the
tate cancer have revealed a highly vari¬                                nation, while others experience rapid        prostate were diagnosed at Orebro Medi¬
able clinical course. A substantial pro¬                                progression and eventually die of pros¬      cal Centre, a hospital with a strictly de¬
portion of the patients have early-stage                                tate cancer. Given the present knowl¬        fined catchment area. No screening for
disease. Some of them show little or no                                 edge, the prognosis of the individual pa¬    prostate cancer took place during the
                                                                        tient is impossible to predict. Yet,         period of this study. All suspected cases
  From the Department of Urology, \l=O"\rebroMedical
                                                                        screening and aggressive treatment of        of prostate cancer were referred to the
Centre, \l=O"\rebro,Sweden (Dr Johansson and Ms Jo-                     early disease are becoming wide¬             Department of Urology for diagnosis,
hansson); the Department of Cancer Epidemiology,                        spread—a development that will esca¬         treatment, and follow-up. The diagnoses
Uppsala University Hospital (Drs Johansson, Holm-$                      late the risk of overdiagnosis and over-     were confirmed by fine-needle aspira¬
berg, Adami, and Bergstr\l=o"\m),          and the Department of
                                                                        treatment.1"5                                tion biopsy16 of palpable prostate tumors
Statistics (Dr Bergstr\l=o"\m),       Uppsala University, Uppsala,
Sweden; and the Center for Cancer Prevention and                           We previously published estimates of      in 542 (84%) of the 648 cases, while in
Department of Epidemiology, Harvard University                          progression-free and disease-specific        the remaining 106 cases (16%), the can¬
School of Public Health, Boston, Mass (Dr Adami).                       survival based on a population-based co¬     cer was detected by histopathological
  Reprints: Jan-Erik Johansson, MD, Department of                       hort of 223 patients with initially un¬      examinations of specimens obtained at
Urology, \l=O"\rebro            Medical Centre, S-701 85 \l=O"\rebro,
Sweden.                                                                 treated, early prostate cancer. The low      operations for benign prostatic hyper-
                                                    Downloaded from by guest on May 7, 2011
Table 1.—Distribution of 642 Consecutive Patients Diagnosed as               Having   Prostate Cancer at the   Örebro Medical Centre 1977-1984, by Age at Diagnosis,                   and
M Categories, Grade, and Treatment at the Time of Diagnosis*

                                                                      Without Métastases                            With Métastases              Death From Prostate    Cancerf
                                                                     at   Diagnosis (n=483)                    at    Diagnosis (n=159)
                                                                   -                                    I      I                       I      Underlying              Contributory
                   Category                       No.       Local, No. (%)       Métastases, No. (%)               Progressive, No. (%)      Cause, No. (%)          Cause, No. (%)
All                                              642             122(25)                 83(17)                           126(79)               201 (31)                  35   (5.5)
Age,     y
      ==60                                         43          13/33 (39)               12/33(36)                        8/10(80)                 19(44)                   1(2)
      61-70                                      210          54/160(34)               38/160(24)                       42/50 (84)                73 (35)                 11(5)
      71-80                                      294          43/217(20)               29/217(13)                       57/77 (74)                85 (29)                 16(5)
      >80                                          95          12/73(16)                 4/73 (5)                       19/22(86)                 24   (25)                7(7)
      T01                                          76          14/76(14)                 6/76 (8)                                                  5(7)                    1(1)
      TOd                                          38          13/36(36)                 7/36(19)                         2/2(100)                 9(24)                   8(21)
      T1-T2                                      213          55/188(29)               24/188(13)                       18/25(72)                 36(17)                  11(5)
      T3-T4                                      315          40/183(22)               46/183(25)                     106/132(80)                151 (48)                 15(5)
M category
      M0                                         483         122/483(25)               83/483(17)                                                 77(15)                  25(5)
      M1                                         159                                                                  126/159(79)                124(78)                  10(6)
Grade, differentiation
      I, highly                                  212          43/195(22)               19/195(10)                       13/17(76)                 28(13)                   7(3)
      II, moderately                             306          60/224 (27)              44/224 (20)                      64/82 (78)               101 (33)                 15(5)
      III, poorly                                              19/64(30)                20/64(31)                       49/60 (82)                72 (58)                 13(10)
      Radical (surgery   and/or irradiation)       39           5/39(13)                 5/39(13)                                                  7(18)                   3(8)
      Orchiectomy                                 138          19/88(22)                26/88 (30)                      40/50   (80)              62 (45)                  5(4)
      Estrogen                                    102          11/64(17)                11/64(17)                       29/38   (76)              40 (39)                 10(10)
      Estramustine                                              5/42(12)                 6/42(14)                       45/58   (78)              51 (51)                 12(12)
      Deferred treatment                         223          73/223 (33)              29/223(13)                                                 25(11)                   4(2)
      Other   (antiandrogen)                       40           9/27 (33)                6/27 (22)                      12/13(92)                 16(40)                   1(3)
   *For each category, the number (and % within the category) of patients whose tumors progressed locally or systemically (to M1 ) is given. Ellipses indicate data are not applicable.
    (/Number of deaths as of September 1994.
        indicates the tumor is clinically occult and incidental; T01=T0pT1ok, <25% of the total specimen; T0d=T0pT diff, >25% of the total specimen; T1-T2, tumor is confined
to prostate gland; T1, nodule surrounded by normal prostate tissue; T2, large nodule or multiple nodules; T3, growth through the capsule; and T4, growth to surrounding organs.

plasia. At follow-up, the medical records                       no initial treatment. Those in whom the                          Follow-up
of 5 patients could not be found, and 1                         cancer  progressed to symptomatic dis¬
further patient could not be located.                           ease were treated with exogenous es¬                               All patients were followed up until
Thus, a total of 642 patients were in¬                          trogens or orchiectomy. The remaining                           death or until the end of the observation
cluded in the study. The baseline evalu¬                        77 patients were excluded from the                              period on September 1, 1994. The ob¬
ation at diagnosis included physical                            group subjected to expectant treatment                          servation period ranged from 126 to 210
examination, chest radiography, intra¬                          because of temporary additional restric¬                        months, the average being 168 months
venous pyelography, bone scan, and skel¬                        tions applied to patients with moder¬                           (14 years). At 2- to 12-month intervals,
etal radiography (of suspected lesions                          ately and poorly differentiated palpable                        the patients were reassessed. A bone
on bone scan). Digital rectal examina¬                          tumors (T1-T2); 75 received initial treat¬                      scan was performed every 6 to 12
tion was performed for clinical staging                         ment with local irradiation, estrogen,                          months. In patients with initially local¬
of palpable tumors. Nodal staging was                           estramustine, orchiectomy, or a combi¬                          ized disease, local progression was de¬
not carried out.                                                nation of these, while (in 1984) 2 pa¬                          fined as palpable tumor growth through
   The mean age at diagnosis was 72                             tients underwent radical prostatectomy.                         the prostate capsule (T3). Among pa¬
years. According to the TNM and World                              Altogether, 342 patients with locally                        tients with distant métastases (Ml), pro¬
Health Organization (WHO) classifica¬                           advanced (T3-T4 and MO) or metastatic                           gression was defined as an increase in
tions for staging,17·18 300 patients (47%)                      (T0-T4 and Ml) cancer were treated hor-                         size of the measurable lesions by 25% or
had localized tumors (T0-T2), and 183                           monally, predominantly with estrogen                            more, a definite increase in existing le¬
(28%) had locally advanced tumors (T3-                          or estramustine, during the first 2 years.                      sions on a bone scan, or the occurrence
T4) without detectable métastases (MO).                         From March 1979 to February 1982, pa¬                           of new lesions.
Metastatic disease (Ml) was found in                            tients newly diagnosed with prostate                               The medical records of all deceased
159 patients (25%) (Table 1). As all pa¬                        cancer and without cardiovascular con¬                          patients were reviewed. In most in¬
tient records were reviewed and a few                           traindications were randomly allocated                          stances, the cause of death was obvious
cases         reclassified, there       are    minor dis¬       to   undergo orchiectomy               or   estrogen            on clinical grounds alone. An autopsy
crepancies in stage distribution com¬                           treatment.19 Thereafter,             some   patients            was performed if the cause of death was
pared with earlier published data.6                             with métastases were randomly allo¬                             not clear. Prostate cancer was recorded
                                                                cated to receive estramustine or the                            as the underlying cause of death, a con¬
                                                                antiandrogen flutamide.20 Second-line                           tributory cause of death, or unrelated to
  Among the 300 patients with tumor                             treatment mostly consisted of high-dose                         death. If the treatment of the prostate
growth confined to the prostate gland                           medroxyprogesterone acetate or estra¬                           cancer was related to the death (eg, car¬
(T0-T2 and MO), a total of 223 were given                       mustine.21                                                      diovascular complications following es-

                                                 Downloaded from by guest on May 7, 2011
trogen administration), prostate cancer
was  recorded as a contributory cause.                                                                                     Localized Disease (Untreated)
As a validation, we compared our own                                                                                       Localized Disease (Treated)
classifications ofthe causes of death with                                                                                 Locally Advanced Disease
                                                                                                                           Metastatic Disease
those recorded in the death register.                                                                        100
There was agreement in 90% of the pa¬
tients and no evidence of systematic                                                                         80
overascertainment or underascertain-                                                                    ra
ment of prostate cancer as cause of death
in our data.
                                                                                                        CO 40
Statistical Methods                                                                                     T3
                                                      O                5               10          15
  Progression-free survival and ob¬          No. of           Time Since         Diagnosis, y
                                                                                                        Ë    20
served survival for all causes of death      Patients                                                   o
                                             at Risk 642   473   362       286   221   167   99   38    O
were calculated by means of the actu¬
arial (life-table) method.22 Survival cor¬                                                                                  5             10               15
rected for causes of death other than        Figure 1.—Observed      survival (curve 1), corrected                    Time Since    Diagnosis, y
                                             survival taking into account prostate cancer as an
prostate  cancer was analyzed by con¬
                                             underlying cause of death (curve 2), and corrected
sidering only mortality with prostate        survival taking into account prostate cancer both as       Figure 2.—Corrected survival (prostate cancer as
cancer as the underlying cause of death,     an underlying and a contributory cause of death            an underlying cause of death) among patients with
thus treating deaths from other causes       (curve 3) in a cohort of 642 patients with prostate        localized disease with (77 patients) and without
as censored observations (corrected sur¬     cancer.                                                    (223 patients) initial treatment, and those with
                            we calculated
                                                                                                        locally advanced (183 patients) and metastatic dis¬
vival). As a complement,                                                                                ease (159 patients).
the relative survival.23·24 We also calcu¬   relative survival rates at 10 and 15 years,
lated corrected survival rates, taking       which were 56.4% and 44.4%, respec¬                        survival rates after 10 and 15 years were
prostate cancer into account both as an      tively.                                                    85.6% (95% CI, 79.8%-91.4%) and 80.9%
underlying and a contributory cause of       Survival by Stage at Diagnosis
                                                                                                        (95% CI, 72.4%-89.4%), respectively-
death. To study the effect of different                                                                 similar to those in the entire group with
variables on survival while taking other        The corrected 15-year survival rate                     localized disease (Table 2).
variables into account, the Cox propor¬      was 71.8% among patients without dis¬                         The 77 patients treated at diagnosis
tional hazards model was used.25             tant métastases detected at diagnosis                      were also analyzed separately. Their ob¬
                                             and 5.7% among those with such mé¬                         served survival rate after 10 and 15 years
RESULTS                                      tastases. In the former category, 300                      of follow-up was lower than among the
Overall Results                              patients had a clinically localized dis¬                   223 patients given no initial treatment,
                                             ease. During follow-up, 37 (12%) of the                    partly because 5 of them died within 2
   At the end of the observation period,     300 developed métastases, and 33 (11%)                     years from cardiovascular complications
541 (84%) of all 642 patients in the study   died of prostate cancer (Table 3). The                     during estrogen treatment. However,
cohort had died. Prostate cancer was         corrected survival rates among these                       the corrected survival rates were ap¬
considered the underlying cause ofdeath      300 patients at 10 and 15 years were                       proximately the same for the 2 catego¬
in 201 patients, while in 35 patients, the   85.3% and 80.9%, respectively (Table 2).                   ries (Table 2).
malignancy or the treatment was con¬            As previously reported,6·7 223 of the                      Among the 183 patients with locally
sidered a contributory cause. Prostate       300 patients with localized disease re¬                    advanced disease at diagnosis, 44 (24%)
cancer accounted for a higher propor¬        ceived no initial treatment, which per¬                    died of prostate cancer, while 13 pa¬
tion of all deaths in patients younger       mitted a study of its natural history.                     tients died of causes related to the dis¬
than 61 years (44%) than in those older      Altogether, 29 (13%) of these 223 pa¬                      ease or its treatment. The survival rates
than 80 years (25%) at diagnosis. More       tients developed métastases; 25 of them                    are shown in Table 2. Only 4 (3%) of the
patients with poorly differentiated tu¬      died of prostate cancer, and 4 died with                   159 patients who had métastases at di¬
mors and/or an advanced local tumor          prostate cancer as a contributory cause                    agnosis were alive at the end of the ob¬
stage died of prostate cancer. As hor¬       of death (Table 3). At the end of the                      servation period, compared with 97
mone treatment was used mostly for           observation period, only 2 of the 57 pa¬                   (20%) of the 483 with no signs of mé¬
palliation in advanced stages, more pa¬      tients still alive have shown distant mé¬                  tastases. Among these 483 patients, 83
tients receiving this treatment died of      tastases. The distribution, by age, grade,                 developed métastases during follow-up.
prostate cancer compared with those          and tumor stage, of the patients who                       Their distribution by age, grade, and
given other treatment modalities. In the     experienced progression and died is                        category is shown in Table 1. Only 5% of
entire cohort, local progression and/or      shown in Table 3. The survival rates                       the patients older than 80 years at di¬
generalization occurred in 280 patients      among these patients are shown in Table                    agnosis developed métastases, compared
(44%) (Table 1).                             2, and the corrected survival rates are                    with 36% of those younger than 61 years.
   The survival rates among all 642 pa¬      further illustrated in Figure 2. The rates                    The prognostic outlook was drastically
tients are illustrated in Figure 1. The      of survival without métastases were                        different for patients with metastatic
progression-free, overall, and corrected     similar to the rates for the whole group                   disease vs those with localized disease
survival rates are further specified in      of 300 patients with localized disease                     (Figure 2). In the former group, the cor¬
Table 2. The corrected rates with pros¬      (86.6% at 10 years and 81.6% at 15 years).                 rected 15-year survival rate was only
tate cancer taken into account both as       In a subgroup of 58 patients (34 with                      about 5% (Table 2).
an underlying and a contributory cause       highly and 24 with moderately differ¬
of death were 55.2% (95% confidence          entiated tumors) who met current indi¬                     Multivariate Analyses
interval [CI], 50.5%-59.9%) at 10 years      cations for radical prostatectomy, 11 de¬                    We used a proportional hazards model
and 49.6% (95% CI, 43.9%-55.3%) at 15        veloped métastases, and 8 (14%) died of                    to quantify the differences in corrected
years. These figures were similar to the     prostate cancer (Table 3). The corrected                   survival rate between the main catego-

                               Downloaded from by guest on May 7, 2011
Table 2.—The 10- and 15-Year Progression-Free Survival, Observed          Survival, and Corrected Survival with 95% Confidence Intervals (in Parentheses) for 642
Patients With Prostate Cancer, by Stage of Disease and Diagnosis

                                             Progression-Free Survival,       %                 Observed Survival, %                 Corrected Survival, %

         Category                No.         10-y                     15-y                    10-y                 15-y15-y    10-y
Without métastases_483                  57.0(51.2-62.8)          46.1(36.8-55.4)         32.9(28.6-37.2)                 78.0(73.2-82.8)
Localized (T0-T2)   300                 59.3(52.3-66.3)          48.0(37.3-58.7)         37.7(32.1-43.3)         80.9(73.6-88.2)
   Without initial treatment 223        55.1(47.1-63.1)          43.4(31.2-55.6)         40.8(34.2-47.4)                 85.6(79.8-91.4)
   With initial treatment_77            76.0(63.2-88.8)          65.9(44.0-87.8)         29.9(19.5-40.3)         80.7 (66.8-94.6)
                                                                                                         17.0(7.6-26.4)_84.7 (72.6-96.8)
Locally advanced (T3-T4)     183        57.5(48.2-66.8)          46.6(32.7-60.5)                                 56.5 (44.6-68.4)
                                                                                         25.1(18.7-31.5)_3.4 (-2.6-9.4)  65.7 (56.5-74.8)
With métastases_159_9.5(3.8-15.2)_6.2(0.8-10.6)_5.0(1.5-8.5)_1.5 (-0.8-3.8)_8.5(3.2-11.8)_5.7 (-0.1-11.5)
All                          642 46.1(41.3-50.9) 35.7(28.4-43.0) 26.0(22.5-29.5) 11.8 (8.6-15.2) 59.6(54.8-64.4) 54.0(48.1-59.9)

Table 3.—Local and Systemic Disease Progression and Deaths From Prostate Cancer Among 300 Patients                  survival may arise as a result of varia¬
With Initially Localized (T0-T2, MO) Disease, by Age, Tumor Stage, and Grade at Time of Diagnosis*                  tion in diagnostic intensity.31 However,
                                                                                                                    the general pattern of agreement indi¬
                                                                Progression                                         cates that our cohort is representative
                                                                                                 Death From
                                                    Local,             Métastases,             Prostate Cancer,     for prostate cancer diagnosed in the ab¬
                     Category          No.          No. (%)              No.      (%)_             No. (%)
                                                                                                                    sence of population screening, a situation
Total                                  300          82 (27)               37(12)                    33 (11)         which prevailed in most European set¬
  With initial treatment                              1(12)                 1(10)                     1(10)
  Without initial treatment                         73 (33)
                                                                                                                    tings at the time of this study. In con¬
                                       223                                29(13)                    25(11)          trast to the European data, higher sur¬
    Age,         y                                                                                                  vival rates have been reported from the
        <70                             85          38   (45)             16(19)                    12(14)          United States.32 Widespread use of pros¬
        ï70                            138          35   (25)             13(9)                     13(9)           tate-specific antigen (PSA), a diagnostic
    Tumor            staget                                                                                         blood test, and detection of asymptom¬
      T01                                           11(15)                    6(8)                   5(7)           atic disease can readily explain the fa¬
      TOd                               34          13(38)                 7(21)                     7(21)          vorable mortality to incidence ratio, as
        T1-T2                          117          49 (42)               16(14)                    13(11)          well as the recent unprecedented upsurge
   Grade, differentiation                                                                                           in the incidence of prostate cancer in the
      I, highly                        148          37   (25)              2(8)                      9(6)           United States.5
         ,                              66          33 (50)                2(18)                    11 (17)            Radical surgery has been compared
             ,                                       3(33)                 6(67)                     5(56)          with deferred treatment in only 1 ran¬
Candidates for radical surgen/          58          27 (47)               11 (19)                     1(14)         domized study which, after 23 years of
  Those who experienced both systemic and local progression are included in both categories.                        follow-up, still shows no difference in
  tSee third footnote to Table 1 for explanation of stages.                                                         survival.33·34 However, cautious interpre¬
  íAged <70 years at diagnosis, T0d-T2, highly or moderately differentiated tumors.                                 tation of this study is necessary in view
                                                                                                                    of its limited size and flawed design. In
ries of patients using the 223 patients                  patients with deferred treatment and                       3 recent reviews of nonrandomized stud¬
without initial treatment as a reference                 patients with initial treatment. The in¬                   ies, the investigators found little evi¬
category. Separate models were fitted                    ternal validity of our study should be                     dence of survival benefit after radical
without and with adjustments for age,                    high, as the cohort was strictly popula¬                   surgery compared with deferred treat¬
category, and tumor grade. The relative                  tion based, and the routines for diagno¬                   ment and irradiation.3587 Similarly, in a
hazard for the 77 patients who received                  sis, staging, grading, follow-up, and clas¬                population-based cohort from the United
treatment was 1.2 (95% CI, 0.6-2.7) with¬                sification of the causes of death were                     States of patients aged 65 to 75 years
out and 1.0 (95% CI, 0.4-2.3) with ad¬                   strictly standardized. Longer observa¬                     with conservatively treated localized
justment (P=.91 and .62 for unadjusted                   tion is unlikely to change our overall                     cancer, men with Gleason score 2 to 4
and adjusted analysis, respectively). In                 findings, since only 16% of the patients                   tumors had a survival rate not signifi¬
contrast, the death rate was 2 to 3 times                were still alive at the end of follow-up,                  cantly different from that of the general
higher in patients with locally advanced                 and few of them have shown signs of                        population, while those with tumors with
cancer and 12- to 20-fold (with and with¬                metastatic disease.                                        a score of 5 to 10 had a much poorer
out adjustment) higher in patients who                       Our analyses revealed an overall, long-                outlook.38
already had distant métastases at diag¬                  term, corrected survival rate that was a                      The favorable survival rate among un¬
nosis compared with those without ini¬                   little higher than contemporary data from                  treated patients with early-stage dis¬
tial treatment.                                          the whole of Sweden,26 the north of Swe¬                   ease indicates that at least 80% of them

                                                         den,15 and most other Scandinavian2729                     would be treated without survival ben¬
                                                         as well as other European countries.30 In                  efit if an aggressive policy were gener¬
   Our most salient finding was the fa¬                  Denmark, physicians and urologists "have                   ally adopted (Table 2). The real propor¬
vorable outlook among initially un¬                      questioned the benefit of diagnosing tu¬                   tion of overtreatment would, however,
treated patients with highly or moder¬                   mors that do not cause local or general                    be higher because of competing causes
ately differentiated, localized disease.                 symptoms."29 As a corollary, both inci¬                    of death. Indeed, figures given in Table
Claims that, in a previous report, the                   dence and survival rates are substan¬                      3 show that only about 10% of all pa¬
survival rate in this category was over¬                 tially lower in Denmark than in other                      tients with early-stage disease at diag¬
estimated because 77 patients received                   Nordic countries, while mortality rates                    nosis died of prostate cancer. Although
initial treatment9 were not supported.                   appear virtually identical.29 These find¬                  some of them might be cured by ag¬
Indeed, the corrected survival rate of                   ings provide powerful supporting evi¬                      gressive treatment, some might die of
about 80% at 15 years was identical in                   dence for the view that differences in                     the disease despite such treatment. Like

                                         Downloaded from by guest on May 7, 2011
others,26·38 we found no clear association                         tients with highly differentiated tumors                       make historical comparisons unreliable
between age and prognosis. Neverthe¬                               is larger in our cohort than in a series of                    during the foreseeable future.31
less, younger patients are subjected to                            patients treated with radical surgery or                          In our earlier communications, we con¬
the risk of dying from prostate cancer                             irradiation following histopathological                        cluded that radical early treatment needs
for a longer period of time, as competing                          confirmation of the diagnosis. It also im¬                     assessment in randomized trials with
causes of death are less prevalent.                                plies that, in the group of highly differ¬                     initially untreated controls, and the room
   In our study, a majority of the pa¬                             entiated tumors, some have a higher                            for therapeutic improvement is limited
tients were diagnosed by means of fine-                            Gleason score.                                                 in early-stage disease.7 We also empha¬
needle aspiration biopsy. This diagnos¬                               Recent data suggest that serum lev¬                         sized that the large gaps in our knowl¬
tic tool has a high specificity and                                els of PSA may be elevated even 5 to 10                        edge prevent any reliable prediction of
sensitivity,39·40 and during the study, all                        years before prostate cancer becomes                           the consequences of prostate cancer
living patients had biopsies updated af¬                           clinically evident.41"44 Such a long lead                      screening.6 Given the risk of doing more
ter 2 and 4 years. Among those 117 pa¬                             time could have important implications.                        harm than good, even randomized trials
tients with palpable tumors diagnosed                              Patient cohorts in which most tumors                           to assess the full impact of screening
with fine-needle biopsy and left without                           are detected through elevated PSA lev¬                         may be unethical.45·46 Our present ex¬
initial treatment, the malignant diagno¬                           els will need many years of follow-up                          tended analyses after prolonged follow-
sis was confirmed in all instances. Some                           before their mortality rates can be re¬                        up encourage us to reemphasize these
studies have shown undergrading when                               liably compared with those in patients                         conclusions.
cytology specimens have been compared                              with localized prostate cancer diagnosed
with histological grading based on core                            clinically in the pre-PSA period. In re¬                         This study was supported by grants from the
biopsies.39·40 Such misclassification may                          ality, differences in patient mix with re¬                     örebro County Council Research Committee and
partly explain why the proportion of pa-                           gard to diagnostic method are likely to                        the Swedish Cancer             Society.
1. Catalona WJ, Smith DS, Ratliff TL, Baslar JW.                   18. World Health Organization. International His-              34. IversenP, Madsen \l=A%o\P, Corle DK. Radical pros-
Detection of organ-confined prostate cancer is in-                 tological Classification of Tumours: Histological              tatectomy          expectant treatment for early
creased through prostate-specific antigen-based                     Typing of Prostate Tumours. Geneva, Switzerland:              carcinoma ofthe prostate: twenty-three year follow\x=req-\
screening. JAMA. 1993;270:948-954.                                 World Health Organization; 1980; No. 22.                       up of   a prospective randomized study. Scand J
2. Littrup PJ, Goodman AC, Mettlin CJ. The ben-                    19. Johansson J-E, Andersson S-O, Holmberg L,                  Urol Nephrol. 1995;172(suppl):65-72.
efit and cost of prostate cancer early detection. CA               Bergstr\l=o"\m  R. Primary orchiectomy versus estro-           35. Adolfsson J, Steineck G, Whitmore WF Jr. Re-
Cancer J Clin. 1993;43:134-149.                                    gen therapy in advanced prostatic cancer: a ran-               cent results of management of palpable clinically
3. Chodak GW. Questioning the value of screening                   domized study: results after 7 to 10 years of follow\x=req-\   localized prostate cancer. Cancer. 1993;72:310-322.
for prostate cancer in asymptomatic men. Urology.                  up. J Urol. 1991;145:519-523.                                  36. Austenfeld MS, Thompson IM Jr, Middleton
1993;42:116-118.                                                   20. Johansson J-E, Andersson S-O, Beckman KW,                  RG, for the American Urological Association Pros-
4. Fracchia JA. Costs and other aspects of screen-                 Ling\l=a%o\rdh
                                                                                  G, Zador G. Clinical evaluation of flu-         tate Cancer Guideline Panel. Meta-analysis of the
ing vs early detection of prostate cancer. Prim                    tamide and estramustine as initial treatment of                literature: guideline development for prostate can-
Care Cancer. 1993;13:31-32.                                        metastatic carcinoma ofthe prostate. Urology. 1987;            cer treatment. J Urol. 1994;152:1866-1869.
5. Waterbor JW, Beuschen AJ. Prostate cancer                       29:55-59.                                                      37. Wasson JH, Cushman CC, Bruskewitz RC, Lit-
screening (United States). Cancer Causes Control.                  21. Johansson J-E, Andersson S-O, Holmberg L.                  tenberg B, Mulley AG Jr, Wennberg JE, for the
1995;6:267-274.                                                    High-dose medroxyprogesterone acetate versus es-               Prostate Disease Patient Outcome Research Team.
6. Johansson J-E, Adami H-O, Andersson S-O, Berg-                  tramustine in therapy-resistant prostatic cancer: a            A structured literature review of treatment for
str\l=o"\mR, Krusemo UB, Kraaz W. Natural history                  randomised study. Br J Urol. 1991;68:67-73.                    localized prostate cancer. Arch Fam Med. 1993;2:
oflocalized prostate cancer: a population based study              22. Cutler SJ, Ederer F. Maximum utilization of                487-493.
in 223 untreated patients. Lancet. 1989;1:799-803.                 the life table method in analyzing survival. J Chronic         38. Albertsen P, Fryback DG, Storer BE, Kolon
7. Johansson J-E, Adami H-O, Andersson S-O, Berg-                  Dis. 1958;8:699-710.                                           TF, Fine J. Long-term survival among men with
str\l=o"\mR, Holmberg L, Krusemo UB. High 10-year                  23. Ederer F, Axtell IM, Cutler SJ. The relative               conservatively treated localized prostate cancer.
survival rate in patients with early, untreated pros-              survival rate: a statistical methodology. Natl Can-            JAMA. 1995;274:626-631.
tatic cancer. JAMA. 1992;267:2191-2196.                            cer Inst Monogr. 1961;6:101-121.                               39. Chodak GW, Bibbi M, Straus FH, Wed GL.
8. Chodak GW, Thisted RA, Gerber GS, et al. Results                24. Hakulinen T. On long-term relative survival                Transrectal aspiration biopsy versus transperineal
of conservative management of clinically localized                 rates. J Chronic Dis. 1977;30:431-443.                         core biopsy for the diagnosis of carcinoma of the

prostate cancer. N Engl J Med. 194;30:242-48.                      25. Cox DR. Regression model and life tables. J R              prostate. J Urol. 1984;132:480-482.
9. Scardino PT. Early detection of prostate cancer.                Stat Soc. 1972;B34:187-220.                                    40. Waisman J, Adolfsson J, L\l=o"\whagen T, Skoog L.
 Urol Clin North Am. 198 ;16: 35-65 .                              26. Stenb\l=a"\ck M, Ros\l=e'\n M. Cancer survival in Swe-     Comparison of transrectal prostate digital aspira-
10. Aus G, Hugosson J, Norl\l=e'\n    L. Long-term sur-            den in 1961-1991. Acta Oncol. 19 5;43):6(s2u-6p3.l             tion and ultrasound-guided core biopsies in 99 men.
vival and mortality in prostate cancer treated with                27. Cancer Registry of Norway. Survival of Can-                Urology. 1991;37:301-307.
noncurative intent. J Urol. 1995;154:460-465.                      cer Patients: Cases Diagnosed in Norway 1968\x=req-\           41. Stenman U-H, Hakama M, Knekt P, Teppo L,
11. Chodak G. What to expect from prostate can-                    1975. Oslo: The Norwegian Cancer Registry; 1980.               Leinonen J. Serum concentrations of prostate spe-
cer. J Urol. 1995;154:2132-2133.                                   28. Hakulinen T, Pukkala E, Hakama M, et al. Sur-              cific antigen and its complex with                   \g=a\1-antichymo-

 12. Nitti V, Lepor H. Controversies old and new.                  vival of cancer patients in Finland in 1953-1974.              trypsin before diagnosis of prostate cancer. Lancet.
Lancet. 1995;346(suppl 26).                                        Ann Clin Res. 1981;13(suppl 31):59-61.                         1994;344:1594-1598.
 13. Otnes B, Harvei S, Fossa SD. The burden of                    29. Tretli S, Engeland A, Haldorsen T, et al. Pros-            42. Pearson JD, Carter HB. Natural history of
prostate cancer from diagnosis until death. Br J                   tate cancer: look to Denmark. J Natl Cancer Inst.              changes in prostate specific antigen in early-stage
Urol. 1995;76:587-594.                                             1996;88:128.                                                   prostate cancer. J Urol. 1994;152:1743-1748.
14. Abrahamsson PA, Adami H-O, Kim KM, Zelen                       30. Levi F, Randimbison L, Te V-C, et al. Trends               43. Tibblin G, Welin L, Bergstr\l=o"\m                       R, Ronquist G,
M, Kulldorf M. Re: long-term survival and mortal-                  in cancer survival in Vaud, Switzerland. Eur J Can-            Norl\l=e'\n
                                                                                                                                                               B-J, Adami H-O. The value of prostate spe-
ity in prostate cancer treated with noncurative in-                cer.         1992;28A:1490-1495.                               cific antigen in early diagnosis of prostate cancer:
tent. J Urol. 1996; 155:296-298.                                   31.          Helgesen F, Holmberg L, Johansson J-E, Berg-      the study of men born in 1913. J Urol. 1995;154:
15. Gr\l=o"\nberg        H. Prostate Cancer: Epidemiologi-         str\l=o"\m     R, Adami H-O. Trends in prostate cancer         1386-1389.
cal Studies. Ume\l=a%o\,    Sweden: Ume\l=a%o\ University; 1995.   survival in Sweden 1960 through 1988: evidence of              44. Schr\l=o"\der                  FH. Detection of prostate cancer.
Medical dissertation No. 431.                                      increasing diagnosis of nonlethal tumors. J Natl               BMJ. 1995;310:140-141.
16. Franz\l=e'\n     S, Giertz G, Zaijcek J. Cytological di-       Cancer Inst. 1996;88:1216-1221.                                45. Adami H-O, Baron JA, Rothman KJ. Ethics of
agnosis of prostatic tumours by transrectal aspi-                  32. Wingo PA, Tong T, Bolden S. Cancer statistics.             a prostate cancer screening trial. Lancet. 1994;343:
ration biopsy: a preliminary report. J Urol. 1960;                 CA Cancer J Clin. 1995;45:8-30.                                958-960.
32:193-196.                                                        33. Graversen PH, Corle DK, Nielsen KT, Madsen                 46. Johansson J-E, Abrahamsson PA, Adami H-O,
17. International Union Against Cancer. TNM Clas-                  PO, Casser TC. Radical prostatectomy versus ex-                et al. Mas unders\l=o"\kning                 f\l=o"\r prostatacancer [Mass
sification of Malignant Tumours. Rev 3rd ed.                       pectant primary treatment in stage I and II pros-              screening for prostate cancer]. Stockholm: Swedish
Geneva, Switzerland: International Union Against                   tate cancer: a fifteen-year follow-up.     Urology. 1990;      Council on Technology Assessment in Health Care;
Cancer; 1978.                                                      36:493-498.                                                    1995. Report No. 126.

                                              Downloaded from by guest on May 7, 2011
2. National Centml 1rl11reau Statistics. Causes of Ih!f!nlh
                               of                              1877-1094 iii Sweden.' Om-   between relative survival rates (that are independent of clas-
cia1 Statist~c  Sumden. SL<r:kholm.  Sweden: Statistics Sweden: 187H-1996.
1. Jot~;~r.,son Ande~rson Knwmo LILt. Adami H-0.
                J.E.             SO;                              Berg~trrX H. K n ~ W
                                                                           I           u    sification of causes of death) and corrected survival rates.
Natur;ll h i s t ~ w
                   nllornlized prostate cancer. Irr,,ct t 19a9;l 799-803.                      As for the remaining points, the correspondence of Aus and
4. Cronberu H Pn,.rlr~h.('unreo- Epade,~nologtc,ai S l d d , ~ ~   Urncd. Sweden: UmrA      Hugosson would gain in clarity if women were not included in
~nivemit~~lW5.       Medical dissertivn-No.431.
5. Aclolfsson J, Steineck G, Hedlund PO. Symptom-guided treatment o f localin~~l            the denominator for mortality rates, if they refrained from
prostate cancer: uhsrrvcd 10-year and projected IS ,yrar survival. Scand J Urol             confusing mortality and survival rates (2 entirely different
Nqhrol. 1006;30(suppl1H):2:{.
                                                                                            measures), and if they accepted that a comparison with all of
                                                                                            Sweden may be more meaningful and stdtistically stable than
In Reply.-Drs Aus and Hugosson claim that we have under-                                    the comparison with 1 relatively small population in northern
ascertained deaths due to prostate cancer in our study. Their                               Sweden.
concern is based on a rather cavaIier approach to epidemio-                                    Finally, any effort to validate individual studies or indeed
logical methods; they attempt to validate the mortality data                                try to assess the impact of therapy through external compari-
we reported in a closed cohort study by means of cross-sec-                                 sons mav remain futile as long as the profound influence of
tional data from the entire population.                                                     diagnost"ic intensity on patien'i mix and on overdiagnosis of
   It is well established,' and further demonstrated in our ar-                             nonlethal disease"" cannot be accommodated.
ticle that patients with prostate cancer have excess mortality                                        Jan-Erik Johansson, MD, PhD
during many years. Hence, mortality rates during the 18-year                                           Lars Holmberg,MD, PhD
period 1977 through 1994 are generated by 3 distinguishable                                            Reinhold Bergstrom,PhD
cohorts of patients, namely, those who developed prostate                                              Hans-Olov Adami, MD, PhD
cancer (1) duringa numher of years prior to 1977;(2)between                                            Uppsala University Hospital
1977 and 1983 (those who were clinically diagnosed are in-                                             Uppsala, Sweden
cluded in our study cohort); and (3) between 1984 and 1994.                                            Sara Johansson
   Aus and Hugosson apparently overlook the need to sepa-                                              Orebro Medical Centre
rate mortality generated by our incidence cohort from that                                             Orebro, Sweden
among prevalent cases in the general population; there are                                  1. Stenback M, Roxa M.Cenewr sumval in Sweden in 1961-1Wl. ActaOftcol. 1%34
methods to accommodate this situation even in large-scale                                   (suppl4):62-63.
                                                                                            2. Yucn J, Persson I , Bergkvist L, et al. Hormonix replacement therapy and breaat
studies.' Our study, being of limited size, used a more ambi-                               cancer mortality in Swedish women: results aRer adjustment fur hedthy drug users
tious validation approach. For each deceased cohort member,                                 effect. Cancer Cati81,sC(nilw1. IWJ;4:369-374.
we compared our own classificationof cause of death with that                               3. Helgesen F, Holmberg L, Johanxsan J-F:, Iierjrstrom R, Adami H-0. Trends in
                                                                                            proatatic canccr survival in Sweden 19601!%8 evidence of increasing diagnosie of
of the Death Registry. It was possible to compare 526 out of                                nonlethal tumom. d Null Cancer hut. lW96:8:1216-1221.
541 who had died. In YO% of these cases there was complete                                  4. Tretli S , Engeland A, Haldorsen T, ot ul. Prostate cancer:look to Denmark.d Nrrtl
                                                                                            Cnnrer lrut. 1996;88:128.
agreement. A total of 28 patients (5%)were classified in the
Death Registry as deceased due to prostate cancer, while we
classified them as having died from other causes. A careful
record review revealed that most of them had early stage
cancer without any recorded evidence of progression. A ma-                                  CORRECTION
jority died in nursing homes where the established prostate
cancer diagnosismight have been added routinely to the death                                1ncorrectData.-In theOriginalContribution     entitled4'Fifteen-Year
certificates. In contrast, we had assigned prostate cancer as                               Survival in Prostate Cancer: A Prospective, Population-BasedStudy
the cause of death in 25 patients (5%)in whom prostate cancer                               in Sweden," hy Johansson et al, published in the February 12,1997,
                                                                                            issue of THEJOURNAL (1997;277:467-471), error occurred in Table
was not noted as the underlying cause of death in the Death                                 3. The number of metastasesfor Grade I and Grade I1 prostate cancer
Registry. Although we rely more on our own careful death                                    should be 12 and 12.
classification,the net difference between the 2 sources was
only 3 cases. As a corollary, we found excellent agreement

206 JAMA. July 16. 1997-Val 2 , No. 3
                                                                        Downloaded from by guest on May

hkksew3563rd hkksew3563rd http://