- Email: [email protected]
Population-based study of long-term survival in patients with clinically localised prostate cancer
THE LANCET 5
Beck RW, Cleary PA, Trobe JD, et al. The effect of corticosteroids in acute optic neuritis on the subsequent development of multiple sclerosis. N Engl J Med 1993; 329: 1764–69. 6 Kurtzke JF. Rating neurological impairment in multiple sclerosis: an expanded disability status scale (EDSS). Neurology 1983; 33: 1444–52. 7 Thompson AJ, Kennard C, Swash M, et al. Relative efficacy of intravenous methylprednisolone and ACTH in the treatment of acute relapse in MS. Neurology 1989; 39: 969–71. 8 Matthews JNS, Altman DG, Campbell MJ, Royston P. Analysis of serial measurements in medical research. BMJ 1990; 300: 230–35. 9 La Mantia L, Eoli M, Milanese C, et al. Double-blind trial of dexamethasone versus methylprednisolone in multiple sclerosis acute relapses. Eur Neurol 1994; 34: 199–203. 10 Barnes MP, Bateman DE, Cleland PG, et al. Intravenous methylprednisolone for multiple sclerosis. J Neurol Neurosurg Psychiatry 1985; 48: 157–59. 11 Kupersmith MJ, Kaufman D, Paty DW, et al. Megadose corticosteroids in multiple sclerosis. Neurology 1994; 44: 1–4.
12 Miller DH, Thompson AJ, Morrisey SP, et al. High dose steroids in acute relapses of multiple sclerosis: MRI evidence for a possible mechanisms of therapeutic effect. J Neurol Neurosurg Psychiatry 1992; 55: 450–53. 13 Barkhof F, Hommes OR, Scheltens P, Valk J. Quantitative MRI changes in gadolinium DTPA enhancement after high-dose intravenous methylprednisolone in multiple sclerosis. Neurology 1991; 41: 1219–22. 14 Beck RW, Cleary PA, Anderson MM Jr, et al. A randomisation controlled trial of corticosteroids in the treatment of acute optic neuritis. N Engl J Med 1992; 326: 581–88. 15 Beck RW. The Optic Neuritis Treatment Trial: three-year follow-up results. Arch Opthalmol 1995; 113: 136–37. 16 Defer G-L, Barré J, Ledudal P, Tillement J-P, Degos J-D. Methylprednisolone infusion during acute exacerbation of MS: plasma and CSF concentrations. Eur Neurol 1995; 35: 143–48. 17 Lyons PR, Newman PK, Saunders M. Methylprednisolone therapy in multiple sclerosis. A profile of adverse events. J Neurol Neurosurg Psychiatry 1988; 51: 285–87.
Population-based study of long-term survival in patients with clinically localised prostate cancer
Grace L Lu-Yao, Siu-Long Yao
Summary Background Choice of treatment in localised prostate cancer has been hampered by a lack of unbiased, representative data on outcome. Most existing data have come from small cohorts at specialised academic centres; precise overall and cancer-grade-specific data are not available, and the data are subject to differential staging bias. Randomised clinical trials have been undertaken, but the results w ill not be available for another decade. We have carried out a large population-based study to ascertain overall and prostatecancer-specific survival in men treated by prostatectomy, radiotherapy, or conservative management. Methods Data for 59 876 cancer-registry patients aged 50–79 w er e anal y sed. W e ex ami ned t he effec t of differential staging of prostate cancer by analysing the data both by intention to treat and by treatment received. Estimated survival w as calculated by the Kaplan-Meier method. Findings B y t he int ent ion-t o-t reat approac h, 10-y ear prostate-cancer-specific survival for grade 1 cancer was 94% (95% CI 91–95) after prostatectomy, 90% (87–92) after radiotherapy, and 93% ( 91–94) after conservative management. The corresponding survival figures in grade 2 c anc ers w ere 87% ( 85–89) , 76% ( 72–79) , and 77% (74–80); those in grade 3 cancer were 67% (62–71), 53% (47–58), and 45% (40–51). Although the intention-to-treat and treatment-received analyses yielded similar results for radiotherapy and conservative management, the 10-year Health Care Financing Administration, Office of Research and Demonstrations, Division of Health Information and Outcomes, 7500 Security Boulevard, Mail Stop C3-24-07, Baltimore, MD 21244-1850 (G L Lu-Yao PhD), and Johns Hopkins Oncology Center, Baltimore (S-L Yao MD), USA Correspondence to: Dr Siu-Long Yao, Merck Research Laboratories, RY33-64O, PO Box 2000, Rahway, NJ 07065-900, USA
906
disease-spec ific survival aft er prost at ec t omy differed substantially (83% [81–84] by intention to treat vs 89% [87–91] by treatment received). Interpretation The overall and cancer-grade-specific survival found in t his st udy differ subst ant ially from t hose in previous studies. Previous studies that used a treatmentrec eived approac h have g enerally overest imat ed t he benefits of radical prostatectomy. We found that grade 3 tumours are highly aggressive irrespective of stage.
Lancet 1997; 349: 906–910 See Commentary page 892
Introduction This year, about 334 500 men in the USA will be diagnosed as having prostate cancer, the most common non-skin cancer in American men.1 The optimum management of this cancer remains controversial.2,3 In an attempt to address the controversies, large randomised trials have been initiated, but their results will not be available for another decade.4,5 In the absence of such evidence, several organisations have developed practice guidelines for patients with prostate cancer. They have identified overall and causespecific mortality as crucial elements in treatment decision-making.6,7 Structured literature reviews based on meta-analysis techniques have been undertaken by the American Urological Association6 and the Prostate Disease Patient Outcomes Research Team,3 to ascertain estimates of these endpoints after different treatments. These reviews, however, were hampered by the small amount of published data on overall and disease-specific survival both in, and especially in settings other than, academic medical centres.8–11 Although there are population-based estimates of these endpoints for conservative management, no such information is available for radical prostatectomy and radiotherapy.
Vol 349 • March 29, 1997
THE LANCET
Another obstacle to meta-analyses and their interpretation is differential staging.12 The staging of surgical patients includes lymph-node dissection, and most patients found to have positive lymph nodes intraoperatively do not proceed to complete radical prostatectomy.13 By contrast, only a small percentage of patients who receive radiotherapy or conservative management undergo lymph-node dissection.14 Such patients may therefore have positive lymph nodes and distant disease, but they are classified as having localised cancer. Almost all published studies have used a treatment-received rather than an intention-to-treat analysis. Consequently, patients given radiotherapy or conservative management are understaged relative to those who undergo prostatectomy; this factor results in an exaggeration of the benefit of prostatectomy.3 A final difficulty is the lack of data on outcomes in relation to cancer grade, the most important predictor of progression in prostate cancer15–18 and a key determinant of the net treatment benefit in several decision models.19,20 Although data from the individual studies in the metaanalyses3,6 could be combined, statistically meaningful data could not be generated because of the small number of patients in each cohort. In an effort to address these limitations, two studies analysed pooled data assembled from many institutions to generate grade-specific data.18,21 Although limited data are available for conservative management and prostatectomy, there are no comparable data for radiotherapy in these analyses. Furthermore, whether these data, derived from selected institutions and not uniformly complete, are representative of outcomes in the general population is not clear. We undertook a population-based study to ascertain overall and prostate-cancer-specific survival in men treated for clinically localised prostate cancer by conservative management, prostatectomy, and radiotherapy in diverse clinical settings; to provide empirical data that can be used to examine the potential impact of differential staging on outcomes; and to provide survival estimates stratified by the key prognostic factor of cancer grade.
Methods The data used for this study were compiled by the Surveillance, Epidemiology, and End Results (SEER) Program, 14 which collects information on all cancer cases diagnosed in Connecticut, Hawaii, New Mexico, Iowa, Utah, San FranciscoOakland, Detroit, Atlanta, and Seattle. Data elements recorded include cancer stage and grade at diagnosis, initial cancer treatment, and vital status.22 Because rigorous quality control procedures are applied, 23 the data are nearly complete with respect to site-specific prostatectomy and radiotherapy and there is negligible misclassification of cancer-directed treatments.22 Vital status and underlying causes of death were ascertained from death certificates supplied by the state health department. The accuracy of death certificates for prostate cancer is high.24 Cancers were classified into four grades: grade 1 (Gleason scores 2–4), grade 2 (scores 5–7), grade 3 (scores 8–10), and grade unknown.15,25 Before 1988, cancers were classified into four SEER categories—localised, regional, distant, and unknown. Localised stage is defined as invasive neoplasm confined entirely to the prostate. From 1988 onwards, the standard American Joint Commission on Cancer staging 26 was included in the database. All men aged 50–79 who had prostate cancer diagnosed between Jan 1, 1983, and Dec 31, 1992, were included in this study. Data from before 1983 were excluded because specific types of surgery were not recorded before that year. Patients who
Vol 349 • March 29, 1997
Number of patients (% of grade category)
Mean (SE) age at diagnosis
Grade I Radical prostatectomy Radiotherapy Conservative management All
3854 (21·8%) 4065 (22·9%) 9804 (55·3%) 17 723
65·2(0·10) 69·8 (0·10) 69·9 (0·06) 68·9 (0·05)
Grade 2 Radical prostatectomy Radiotherapy Conservative management All
14 287 (50·3%) 7939 (27·9%) 6198 (21·8%) 28 424
65·7 (0·05) 70·6 (0·06) 71·3 (0·08) 68·3 (0·04)
Grade 3 Radical prostatectomy Radiotherapy Conservative management All
5133 (51·5%) 2596 (26·1%) 2236 (22·4%) 9965
66·4 (0·08) 70·7 (0·11) 72·3 (0·12) 68·9 (0·06)
All grades* Radical prostatectomy Radiotherapy Conservative management All
24 257(40·5%) 15 721 (26·3%) 19 898 (33·2%) 59 876
65·8 (0·04) 70·4 (0·05) 70·7 (0·04) 68·6 (0·03)
*3764 patients with cancer of unknown grade are not shown separately, but are included in “all grades”.
Table 1: Average age of patients and intended initial treatment according to cancer grade in patients with clinically localised prostate cancer had other types of cancer besides prostate cancer were excluded from the analyses. The SEER database uses the best available evidence for cancer staging and grading. Pathological evidence (samples taken at prostatectomy) is used for those with radical prostatectomy, and clinical evidence (including histological biopsy evidence) for those without prostatectomy. Since some patients in the prostatectomy group were thought to have clinically localised cancer before prostatectomy, but were subsequently classified as having regional or metastatic disease, analyses of only surgically treated patients classified as having localised cancer would result in differential staging and would favour the surgical group (since patients found to have regional or distant cancers would not be included in the dataset under this ascertainment strategy). We included all patients undergoing radical prostatectomy in the prostatectomy group irrespective of cancer stage, but limited the radiotherapy and conservative management groups to patients with localised cancer only. An intraoperative finding that the lymph nodes have been seeded with cancer usually results in the decision to abandon prostatectomy, since the patient can no longer be classified as having localised disease. Exclusion of patients with positive nodes would thus make the outcomes of the surgical group appear more favourable. To assess the impact of this differential staging and selection process, we analysed the data by two different approaches—intention to treat and treatment received. For the Intention to treat
Treatment received
n
% survival (95% CI)
n
% survival (95% CI)
Grade I Prostatectomy* Radiotherapy Conservative
3854 4065 9804
94 (91–95) 90 (87–92) 93 (91–94)
3402 4188 10 133
98 (97–99) 89 (87–92) 92 (90–93)
Grade 2 Prostatectomy* Radiotherapy Conservative
14 287 7939 6198
87 (85–89) 76 (72–79) 77 (74–80)
12 922 8456 7046
91 (89–93) 74 (71–77) 76 (73–78)
Grade 3 Prostatectomy* Radiotherapy Conservative
5133 2596 2236
67 (62–71) 53 (47–58) 45 (40–51)
4154 2977 2834
76 (71–80) 52 (46–57) 43 (38–48)
*Intention-to-treat and treatment-received results differed significantly (p<0·0001).
Table 2: 10-year disease-specific survival in patients with clinically localised prostate cancer by intention to treat and by treatment received
907
THE LANCET
Grade and treatment
% overall survival (95% CI) Prostate-cancer patients
Grade I Prostatectomy Radiotherapy Conservative
Endpoint
Age-matched cohort
77 (73–80) 63 (60–66) 54 (52–56)
10 year relative survival in agematched cohort
65 54 53
1·17 1·17 1·01
Grade 2 Prostatectomy Radiotherapy Conservative
71 (68–74) 48 (45–51) 38 (36–41)
64 52 49
1·11 0·93 0·78
Grade 3 Prostatectomy Radiotherapy Conservative
54 (50–58) 33 (28–38) 17 (14–20)
62 52 47
0·87 0·63 0·36
Radiotherapy
Conservative management
Overall survival (%) Cancer patients Age-matched cohort
81 (78–83) 83
70 (66–74) 76
45 (41–50) 74
Relative survival
0·98
0·92
0·61
86 (79–91) 84 (74–90)
69 (56–78) 63 (44–77)
Prostate-cancer-specific survival (%) AJCC stage I 87% (85–90)† AJCC stage II
}
*Intention to treat. †AJCC=American Joint Commission on Cancer. Since some of the patients in the prostatectomy group were initially thought to have clinically localised cancer before prostatectomy, but were later upstaged to regional or metastatic disease after prostatectomy, analyses of only those surgical patients classified as having localised cancer would result in differential staging bias. Therefore, all patients undergoing radical prostatectomy were included in the prostatectomy group irrespective of cancer stage.
*Intention-to-treat.
Table 3: 10-year overall survival in patients with clinically localised prostate cancer and age-matched cohorts
Table 4: Survival in patients with grade 3 disease 5 years after diagnosis
intention-to-treat analyses, patients with localised cancer, lymphnode dissection, and radiotherapy, or radiotherapy alone were assigned to the radiotherapy group. Otherwise, patients were classified in the prostatectomy group if a lymph-node dissection or radical prostatectomy was documented. Any remaining patients were included in the conservative management group. For the treatment-received analyses, patients were classified in the prostatectomy group only if radical prostatectomy was documented in their SEER records. The classification of the radiotherapy and conservative management groups was the same as in the intention-to-treat analysis. Major endpoints ascertained in this study were prostatecancer-specific survival, overall survival, and relative survival (ratio of survival in cancer patients to that of an age-matched cohort). The Kaplan-Meier method was used to estimate survival at 5 years and 10 years. In estimating prostate-cancer-specific survival, follow-up was censored at the time of death if it was due to other causes. We used the log-rank test to look for significant differences between the intention-to-treat and treatment-received approaches. The Cox proportional-hazards model was used to examine the effects of cancer grade on prostate-cancer-specific survival after adjustment for differences in age, race, years of diagnosis, and place of diagnosis. In the calculation of expected survival, 1990 Social Security Administration period life tables were applied to age-matched cohorts with age distributions identical to those of the cancer cohorts.27
Results Our study population consisted of 59 876 men (table 1). The mean length of follow-up was 44·5 months; 10% of patients were followed up for 92 months or longer. Radical prostatectomy was the most common treatment modality. Patients who underwent prostatectomy were, on average, 5 years younger than those undergoing radiotherapy or conservative management (table 1). Of the 24 257 men scheduled to undergo prostatectomy, 89% underwent lymph-node dissection, 51% had pathologically localised cancer, and 80% actually proceeded to complete radical prostatectomy. Although
the proportion undergoing lymph-node dissection in the prostatectomy group is somewhat lower than one might expect, it is similar to the proportion in the meta-analysis by the American Urological Association.28 Our analyses showed that differential staging can significantly influence observed outcomes and that the benefits of prostatectomy are generally overestimated when analyses are done by the treatment-received approach (table 2). For prostatectomy, 10-year prostatecancer-specific survival rates were significantly higher by treatment-received analysis than by intention-to-treat analysis (p<0·0001) for all three grades of disease. By contrast, because only a small proportion of patients receiving radiotherapy or conservative management undergo lymph-node dissection, outcomes derived by the two analytical approaches are similar for these groups of patients (table 2). We found, as expected, that cancer grade significantly affects overall survival (table 3). All patients with grade 1 disease had similar or even better overall survival than the age-matched cohort, whatever the initial treatment. By contrast, patients with grade 3 disease had much lower overall survival than their age-matched cohorts in all treatment groups. The difference in survival in patients with grade 1 or 2 prostate cancers between those who underwent prostatectomy and their age-matched cohorts may be due to an inherent selection bias, whereby healthier patients are more likely to be selected for aggressive treatments. Accordingly, adequate adjustment for comorbidity is essential in comparison of overall mortality across treatment groups. In contrast to overall mortality, however, the impact of comorbidity on prostate-cancer-specific survival is modest.29 Differences in this variable are therefore more likely to be attributable to the nature of the cancers and the impact of interventions than to comorbidity. The prostate-cancer-specific survival was consistently poor in
Our study
Radical prostatectomy Radiotherapy Conservative management
5-year survival (95% CI)* Prostatectomy
Intention-to-treat
Treatment received
n
% survival
n
% survival
24 257 15 721 19 898
83 (81–84) 76 (74–78) 82 (81–84)
21 122 16 869 21 885
89 (87–91) 74 (72–76) 80 (79–81)
PORT*
AUA†
% survival
n
% survival
91 (88–93) 79 (74–90) 91 (87–94)
747 222 126
90 (88–92) 67 (59–75) 87 (81–93)
*Prostate Disease Patient Outcomes Research Team. Number of patients in each treatment group was not reported. Annual mortality rate reported was converted to 10-year survival by formula: 10-year survival=e⫺10⫻annual mortality rate. †American Urological Association. Results specific to stage B prostate cancer. 95% CI derived from formula: 95% CI=1·96⫻SE, where SE=mortality rate divided by the square root of the number of deaths.
Table 5: 10-year prostate cancer-specific survival (with 95% CI) in patients with clinically localised prostate cancer
908
Vol 349 • March 29, 1997
THE LANCET
patients with grade 3 cancers in all treatment groups (table 2). The risk of dying of grade 3 prostate cancer within 10 years of diagnosis (by Cox analysis) was ten times that of dying of grade 1 disease. Because the category of localised disease includes both stage I and II disease, we further refined our analysis to delineate outcomes separately for each stage. Grade 3 disease uniformly resulted in a poor outcome even among patients with stage I disease (table 4). Furthermore, the effect of high-grade prostate cancer on survival is rapid. 5 years after diagnosis, patients with grade 3 disease managed conservatively had relative survival of only 0·61 and disease-specific survival of only 63–69%. Patients with grade 3 disease treated by prostatectomy or radiotherapy had higher relative and prostate-cancer-specific survival (table 4). For comparison with previous studies, overall 10-year survival by treatment is given in table 5 with calculated 10-year survival from the two meta-analyses.
Discussion For any medical treatment, informed decisions about treatment choice can be made only if unbiased, representative data on outcomes for each option are available. Ideally, such data are derived from randomised clinical trials. If such trials have not been done, however, population-based studies and meta-analyses are often used to provide suggestive estimates about outcomes. Interpretation and application of meta-analyses based on study-level data, however, can be difficult because the inclusion and exclusion criteria vary among studies, and data on many recognised modifiers of outcomes are incomplete. Although cancer grade is recognised to be the most important predictor of survival in prostate cancer,15–18 reliable meta-analytical estimates of cancer-grade-specific survival could not be made because few published data were available.3.6 The only meta-analysis able to provide cancer-grade-specific estimates of survival was based on individual-level data; however, that study was limited to patients who received conservative management and there were few patients with grade 3 disease.18 Population-based studies, on the other hand, are derived from individual-level data for all patients in defined geographic areas. Thus, the same ascertainment criteria can be applied for every patient. Results from population-based studies can be generalised more readily than those from case series or meta-analyses because they include patients treated in all clinical settings, not just those treated in selected medical centres. Previous population-based analyses have focused on patients receiving conservative management,17,30 and there are no comparable data for patients receiving radiotherapy or undergoing prostatectomy. Our results from this population-based database showed some differences from those of previous studies (table 5).3,6 In general, our results suggest a poorer outcome for both conservative management and prostatectomy than the previous studies, but the outcome after radiotherapy is similar.3,7 There are many possible explanations for these differences. However, when the prostatectomy data in our study were analysed by a treatment-received instead of an intention-to-treat approach (table 5), the results after prostatectomy were nearly identical to those in the previous meta-analyses, which suggests that the discrepancies for prostatectomy may be due partly to the difference in analytical approach.
Vol 349 • March 29, 1997
We believe that, compared with studies that used the treatment-received approach, results derived from the intention-to-treat approach more closely approximate the general clinical situation, in which the lymph-node status of a patient is not known at the time of decision-making (before prostatectomy). Although it is well recognised that differential staging resulting from a treatment-received as opposed to an intention-to-treat analysis can profoundly influence observed outcomes, few studies have provided empirical data to quantify the effects. We found that 10-year prostate-cancer-specific mortality for prostatectomy was more than 60% greater by intention to treat than by treatment received. This finding suggests that results from previous studies that used the treatment-received approach should be interpreted with greater caution. The cancer-grade-specific data obtained in this study emphasise the often under-recognised heterogeneity of clinical prostate cancer. Although this disorder generally has a protracted course, we found that grade 3 prostate cancers progress rapidly. The outlook for patients with grade 3 disease was poor irrespective of stage (table 4). Conclusions drawn from data pooled across cancer grades are therefore too optimistic for patients with grade 3 cancer and too pessimistic for patients with grade 1 disease. Grade-specific data may help to facilitate informed decision-making about treatment. As in previous studies,17,18,30 patients with grade 1 prostate cancer have overall mortality similar to that of an age-matched cohort at 10 years, even when managed conservatively. Given the negligible effect of grade 1 cancer on mortality and the absence of evidence demonstrating the benefit of aggressive therapies, conservative management seems to be a viable option for such patients. In this study, about one in six patients with clinically localised prostate cancer had grade 3 disease. These patients have the most to gain from effective treatment. Although this study was not randomised and therefore cannot fairly compare the efficacies of the various treatment options, several observations suggest that active interventions (prostatectomy or radiotherapy) might improve long-term survival in patients with grade 3 (stage I and II) cancers. Because localised prostate cancer is, in general, a slowly progressive disease, many of the potential therapeutic gains are lost because patients die of other causes. Since grade 3 disease is associated with low disease-specific survival, intervention would provide a more immediate gain, reducing the effects of other causes of death. Differential grading is less likely to operate in the grade 3 group than in the lower grades, thus increasing the possibility of potential gains to be achieved with interventional therapy in this group. Grading of patients who do not undergo surgery is based on biopsy samples, which may not always come from the most advanced part of the tumour. These patients may therefore be given a lower grading than is appropriate. Complete removal of the prostate allows more complete pathological ascertainment of cancer grade. Since grade 3 is the worst category, whatever the therapeutic route chosen, this bias cannot preferentially benefit any specific treatment group. Although most clinicians agree that the benefits of aggressive therapy are generally not apparent until 10 years later, we observed a large difference in prostatecancer-specific survival in grade 3 disease between 909
THE LANCET
patients receiving conservative management (63–69%) and interventions (84–86% for radiotherapy and 87% for prostatectomy) within 5 years of diagnosis. This magnitude of difference was not observed in patients with grade 1 or grade 2 disease even after 10 years of follow-up. The 10-year life expectancy rule may need to be reevaluated for patients with grade 3 disease. We did not attempt to compare the relative effects of different treatments because information on many factors associated with selection of patients (comorbidity, pretreatment concentrations of prostate specific antigen, tumour size, and ploidy) was not available. Although the independent prognostic value of these variables beyond cancer stage and grade has not been demonstrated,12 a combination of these, or as yet unknown, factors may affect outcomes. In addition, the favourable outcome of prostatectomy might have resulted from differential cancer grading or other potential biases. Whereas comparisons between the radiotherapy and conservative management groups may be more valid because of their similarities in terms of age, grading (biopsy based), and staging (clinically based), the comorbidity burden might have played a part in selection of patients for these treatments. Finally, the lack of Gleason grades in the SEER database may result in incomplete cancer-grade adjustments, since the Gleason grade distribution may be uneven across different treatment groups within the same broad category (grade 1, 2, and 3) defined by SEER. We could not adjust for all confounding factors, therefore we did not attempt to draw definite conclusions about treatment efficacies. Clearly, large, randomised, multicentre trials are needed to provide definitive information about the relative efficacy of prostate-cancer treatments. Such trials are under way, but their results will not be available for some time. Our aim was to provide information from a large database on the current controversies. We thank Peter Albertsen, Michael Barry, Maria Friedman, Marian Gornick, Barnett Kramer, Jim Lubitz, Arnold Potosky, William Shipley, Joan Warren, John Wasson, John Wennberg, Yulin Yao, and Anthony Zietman for review and comments. This study was supported by the Agency for Health Care Policy and Research (grant number HS 08397) and the Health Care Financing Administration.
References 1 2 3
4
5 6
Parker SL, Tong T, Bolden S, Wingo PA. Cancer statistics. CA Cancer J Clin 1997; 47: 5–27. Scardino P. Problem of prostate cancer. J Urol 1994; 152: 1677–78. Wasson J, Cushman C, Bruskewitz R, et al. A structured literature review of treatment for localized prostate cancer. Arch Family Med 1993; 2: 487–93. Wilt TJ, Brawer MK. The prostate cancer intervention versus observation trial: a randomized trial comparing radical prostatectomy versus expectant management for the treatment of clinically localized prostate cancer. J Urol 1994; 152: 1910–14. Norlén BJ. Swedish randomized trial of radical prostatectomy versus watchful waiting. Can J Oncol 1994; 4 (suppl): 38–40. Austenfeld MS, Thompson IM, Middleston RG. Meta-analysis of the literature: guideline development for prostate cancer treatment. J Urol 1994; 152: 1866–69.
910
7
Middleton RG, Thompson IM, Austenfeld MS, et al. Prostate cancer clinical guidelines panel summary report on the management of clinically localized prostate cancer. J Urol 1995; 154: 2144–48. 8 Walsh PC, Partin AW, Epstein JI. Cancer control and quality of life following anatomical radical retropubic prostatectomy: results at 10 years. J Urol 1994; 152: 1831–36. 9 Zinche H, Oesterling JE, Blute ML, Bergstralh EJ, Myers RP, Barrett DM. Long-term (15 years) results after radical prostatectomy for clinically localized (stage T2c or lower) prostate cancer. J Urol 1994; 152: 1850–57. 10 Zietman AL, Coen JJ, Dallow KC, Shipley WU. The treatment of prostate cancer by conventional radiation therapy: an analysis of longterm outcome. Int J Radiat Oncol Biol Phys 1995; 32: 287–92. 11 Bagshaw MA, Cox RS, Hancock SL. Control of prostate cancer with radiotherapy: long-term results. J Urol 1994; 152: 1781–85. 12 Barry MJ, Fleming C, Coley CM, Wasson JH, Fahs MC, Osterling J. Should Medicare provide reimbursement for prostate-specific antigen testing for early detection of prostate cancer? Urology 1995; 46: 2–13. 13 Narayan P, Gajendran V, Taylor SP, et al. The role of transrectal ultrasound-guided biopsy-based staging, preoperative serum prostatespecific antigen, and biopsy Gleason score in prediction of final pathologic diagnosis in prostate cancer. Urology 1995; 46: 205–12. 14 Surveillance Epidemiology and End Results Program. Case diagnosed in 1973–1992, public-use database. Bethesda, MD: National Cancer Institute, 1995. 15 Gleason DF, Mellinger GT. Prediction of prognosis for prostatic adenocarcinoma by combined histological grading and clinical staging. J Urol 1974; 11: 58–64. 16 Zincke H, Bergstralh EJ, Blute ML, et al. Radical prostatectomy for clinically localized prostate cancer: long-term results of 1,143 patients from a single institution. J Clin Oncol 1994; 12: 2254–63. 17 Johansson J-E, Adami H-O, Andersson SO, Bergstrom R, Holmberg L, Krusenmo UB. High 10-year survival rate in patients with early, untreated prostatic cancer. JAMA 1992; 267: 2191–96. 18 Chodak GW, Thisted RA, Gerber GS, et al. Results of conservative management of clinically localized prostate cancer. N Engl J Med 1994; 330: 242–48. 19 Fleming C, Wasson J, Albertsen P, Barry M, Wennberg JE. Prostate PORT: a decision analysis of alternative treatment for clinically localized prostate cancer. JAMA 1993; 269: 2650–58. 20 Beck JR, Kattan MW, Miles BJ. A critique of the decision analysis for clinically localized prostate cancer. J Urol 1994; 152: 1894–99. 21 Gerber GS, Thisted RA, Scardino PT, et al. Results of radical prostatectomy in men with clinically localized prostate cancer-multiinstitutional pooled analysis. JAMA 1996; 276: 615–19. 22 Potosky AL, Riley GF, Lubitz JD, Mentnech RM, Kessler LG. Potential for cancer related health services research using a linked Medicare-tumor registry database. Med Care 1993; 31: 732–48. 23 Horm JW, Asire AJ, Young JLJ. Surveillance, Epidemiology, and End Results Program: cancer incidence and mortality in the United States, 1973–91. Bethesda, MD: National Cancer Institute, 1984. 24 Percy C, Stanek E, Gloeckler L. Accuracy of cancer death certificates and its effect on cancer mortality statistics. Am J Public Health 1981; 71: 242–50. 25 Shambaugh EM, Weiss MA. Summary staging guide for the Cancer Surveillance, Epidemiology and End Results Reporting (SEER) Program: US Department of Health and Human Services Public Health Service National Institutes of Health, 1986. 26 American Joint Committee on Cancer. Manual for staging of cancer, 4th edn. Philadelphia: Lipponcott, 1992. 27 Bell FC, Wade AH, Goss SC. Life tables for the United States Social Security Area 1900–2080. Baltimore: Social Security Administration, 1992. 28 Middleton RG. The management of clinically localized prostate cancer: guidelines from the American Urological Association. CA Cancer J Clin 1996; 46: 249–53. 29 Albertsen PC, Fryback DG, Storer BE, Kolon TF, Fine J. The impact of comorbidity on life expectancy among men with localized prostate cancer. J Urol 1996; 156: 127–32. 30 Albertsen PC, Fryback DG, Storer BE, Kolon TF, Find J. Long-term survival among men with conservatively treated localized prostate cancer. JAMA 1995; 274: 626–31.
Vol 349 • March 29, 1997
Beck RW, Cleary PA, Trobe JD, et al. The effect of corticosteroids in acute optic neuritis on the subsequent development of multiple sclerosis. N Engl J Med 1993; 329: 1764–69. 6 Kurtzke JF. Rating neurological impairment in multiple sclerosis: an expanded disability status scale (EDSS). Neurology 1983; 33: 1444–52. 7 Thompson AJ, Kennard C, Swash M, et al. Relative efficacy of intravenous methylprednisolone and ACTH in the treatment of acute relapse in MS. Neurology 1989; 39: 969–71. 8 Matthews JNS, Altman DG, Campbell MJ, Royston P. Analysis of serial measurements in medical research. BMJ 1990; 300: 230–35. 9 La Mantia L, Eoli M, Milanese C, et al. Double-blind trial of dexamethasone versus methylprednisolone in multiple sclerosis acute relapses. Eur Neurol 1994; 34: 199–203. 10 Barnes MP, Bateman DE, Cleland PG, et al. Intravenous methylprednisolone for multiple sclerosis. J Neurol Neurosurg Psychiatry 1985; 48: 157–59. 11 Kupersmith MJ, Kaufman D, Paty DW, et al. Megadose corticosteroids in multiple sclerosis. Neurology 1994; 44: 1–4.
12 Miller DH, Thompson AJ, Morrisey SP, et al. High dose steroids in acute relapses of multiple sclerosis: MRI evidence for a possible mechanisms of therapeutic effect. J Neurol Neurosurg Psychiatry 1992; 55: 450–53. 13 Barkhof F, Hommes OR, Scheltens P, Valk J. Quantitative MRI changes in gadolinium DTPA enhancement after high-dose intravenous methylprednisolone in multiple sclerosis. Neurology 1991; 41: 1219–22. 14 Beck RW, Cleary PA, Anderson MM Jr, et al. A randomisation controlled trial of corticosteroids in the treatment of acute optic neuritis. N Engl J Med 1992; 326: 581–88. 15 Beck RW. The Optic Neuritis Treatment Trial: three-year follow-up results. Arch Opthalmol 1995; 113: 136–37. 16 Defer G-L, Barré J, Ledudal P, Tillement J-P, Degos J-D. Methylprednisolone infusion during acute exacerbation of MS: plasma and CSF concentrations. Eur Neurol 1995; 35: 143–48. 17 Lyons PR, Newman PK, Saunders M. Methylprednisolone therapy in multiple sclerosis. A profile of adverse events. J Neurol Neurosurg Psychiatry 1988; 51: 285–87.
Population-based study of long-term survival in patients with clinically localised prostate cancer
Grace L Lu-Yao, Siu-Long Yao
Summary Background Choice of treatment in localised prostate cancer has been hampered by a lack of unbiased, representative data on outcome. Most existing data have come from small cohorts at specialised academic centres; precise overall and cancer-grade-specific data are not available, and the data are subject to differential staging bias. Randomised clinical trials have been undertaken, but the results w ill not be available for another decade. We have carried out a large population-based study to ascertain overall and prostatecancer-specific survival in men treated by prostatectomy, radiotherapy, or conservative management. Methods Data for 59 876 cancer-registry patients aged 50–79 w er e anal y sed. W e ex ami ned t he effec t of differential staging of prostate cancer by analysing the data both by intention to treat and by treatment received. Estimated survival w as calculated by the Kaplan-Meier method. Findings B y t he int ent ion-t o-t reat approac h, 10-y ear prostate-cancer-specific survival for grade 1 cancer was 94% (95% CI 91–95) after prostatectomy, 90% (87–92) after radiotherapy, and 93% ( 91–94) after conservative management. The corresponding survival figures in grade 2 c anc ers w ere 87% ( 85–89) , 76% ( 72–79) , and 77% (74–80); those in grade 3 cancer were 67% (62–71), 53% (47–58), and 45% (40–51). Although the intention-to-treat and treatment-received analyses yielded similar results for radiotherapy and conservative management, the 10-year Health Care Financing Administration, Office of Research and Demonstrations, Division of Health Information and Outcomes, 7500 Security Boulevard, Mail Stop C3-24-07, Baltimore, MD 21244-1850 (G L Lu-Yao PhD), and Johns Hopkins Oncology Center, Baltimore (S-L Yao MD), USA Correspondence to: Dr Siu-Long Yao, Merck Research Laboratories, RY33-64O, PO Box 2000, Rahway, NJ 07065-900, USA
906
disease-spec ific survival aft er prost at ec t omy differed substantially (83% [81–84] by intention to treat vs 89% [87–91] by treatment received). Interpretation The overall and cancer-grade-specific survival found in t his st udy differ subst ant ially from t hose in previous studies. Previous studies that used a treatmentrec eived approac h have g enerally overest imat ed t he benefits of radical prostatectomy. We found that grade 3 tumours are highly aggressive irrespective of stage.
Lancet 1997; 349: 906–910 See Commentary page 892
Introduction This year, about 334 500 men in the USA will be diagnosed as having prostate cancer, the most common non-skin cancer in American men.1 The optimum management of this cancer remains controversial.2,3 In an attempt to address the controversies, large randomised trials have been initiated, but their results will not be available for another decade.4,5 In the absence of such evidence, several organisations have developed practice guidelines for patients with prostate cancer. They have identified overall and causespecific mortality as crucial elements in treatment decision-making.6,7 Structured literature reviews based on meta-analysis techniques have been undertaken by the American Urological Association6 and the Prostate Disease Patient Outcomes Research Team,3 to ascertain estimates of these endpoints after different treatments. These reviews, however, were hampered by the small amount of published data on overall and disease-specific survival both in, and especially in settings other than, academic medical centres.8–11 Although there are population-based estimates of these endpoints for conservative management, no such information is available for radical prostatectomy and radiotherapy.
Vol 349 • March 29, 1997
THE LANCET
Another obstacle to meta-analyses and their interpretation is differential staging.12 The staging of surgical patients includes lymph-node dissection, and most patients found to have positive lymph nodes intraoperatively do not proceed to complete radical prostatectomy.13 By contrast, only a small percentage of patients who receive radiotherapy or conservative management undergo lymph-node dissection.14 Such patients may therefore have positive lymph nodes and distant disease, but they are classified as having localised cancer. Almost all published studies have used a treatment-received rather than an intention-to-treat analysis. Consequently, patients given radiotherapy or conservative management are understaged relative to those who undergo prostatectomy; this factor results in an exaggeration of the benefit of prostatectomy.3 A final difficulty is the lack of data on outcomes in relation to cancer grade, the most important predictor of progression in prostate cancer15–18 and a key determinant of the net treatment benefit in several decision models.19,20 Although data from the individual studies in the metaanalyses3,6 could be combined, statistically meaningful data could not be generated because of the small number of patients in each cohort. In an effort to address these limitations, two studies analysed pooled data assembled from many institutions to generate grade-specific data.18,21 Although limited data are available for conservative management and prostatectomy, there are no comparable data for radiotherapy in these analyses. Furthermore, whether these data, derived from selected institutions and not uniformly complete, are representative of outcomes in the general population is not clear. We undertook a population-based study to ascertain overall and prostate-cancer-specific survival in men treated for clinically localised prostate cancer by conservative management, prostatectomy, and radiotherapy in diverse clinical settings; to provide empirical data that can be used to examine the potential impact of differential staging on outcomes; and to provide survival estimates stratified by the key prognostic factor of cancer grade.
Methods The data used for this study were compiled by the Surveillance, Epidemiology, and End Results (SEER) Program, 14 which collects information on all cancer cases diagnosed in Connecticut, Hawaii, New Mexico, Iowa, Utah, San FranciscoOakland, Detroit, Atlanta, and Seattle. Data elements recorded include cancer stage and grade at diagnosis, initial cancer treatment, and vital status.22 Because rigorous quality control procedures are applied, 23 the data are nearly complete with respect to site-specific prostatectomy and radiotherapy and there is negligible misclassification of cancer-directed treatments.22 Vital status and underlying causes of death were ascertained from death certificates supplied by the state health department. The accuracy of death certificates for prostate cancer is high.24 Cancers were classified into four grades: grade 1 (Gleason scores 2–4), grade 2 (scores 5–7), grade 3 (scores 8–10), and grade unknown.15,25 Before 1988, cancers were classified into four SEER categories—localised, regional, distant, and unknown. Localised stage is defined as invasive neoplasm confined entirely to the prostate. From 1988 onwards, the standard American Joint Commission on Cancer staging 26 was included in the database. All men aged 50–79 who had prostate cancer diagnosed between Jan 1, 1983, and Dec 31, 1992, were included in this study. Data from before 1983 were excluded because specific types of surgery were not recorded before that year. Patients who
Vol 349 • March 29, 1997
Number of patients (% of grade category)
Mean (SE) age at diagnosis
Grade I Radical prostatectomy Radiotherapy Conservative management All
3854 (21·8%) 4065 (22·9%) 9804 (55·3%) 17 723
65·2(0·10) 69·8 (0·10) 69·9 (0·06) 68·9 (0·05)
Grade 2 Radical prostatectomy Radiotherapy Conservative management All
14 287 (50·3%) 7939 (27·9%) 6198 (21·8%) 28 424
65·7 (0·05) 70·6 (0·06) 71·3 (0·08) 68·3 (0·04)
Grade 3 Radical prostatectomy Radiotherapy Conservative management All
5133 (51·5%) 2596 (26·1%) 2236 (22·4%) 9965
66·4 (0·08) 70·7 (0·11) 72·3 (0·12) 68·9 (0·06)
All grades* Radical prostatectomy Radiotherapy Conservative management All
24 257(40·5%) 15 721 (26·3%) 19 898 (33·2%) 59 876
65·8 (0·04) 70·4 (0·05) 70·7 (0·04) 68·6 (0·03)
*3764 patients with cancer of unknown grade are not shown separately, but are included in “all grades”.
Table 1: Average age of patients and intended initial treatment according to cancer grade in patients with clinically localised prostate cancer had other types of cancer besides prostate cancer were excluded from the analyses. The SEER database uses the best available evidence for cancer staging and grading. Pathological evidence (samples taken at prostatectomy) is used for those with radical prostatectomy, and clinical evidence (including histological biopsy evidence) for those without prostatectomy. Since some patients in the prostatectomy group were thought to have clinically localised cancer before prostatectomy, but were subsequently classified as having regional or metastatic disease, analyses of only surgically treated patients classified as having localised cancer would result in differential staging and would favour the surgical group (since patients found to have regional or distant cancers would not be included in the dataset under this ascertainment strategy). We included all patients undergoing radical prostatectomy in the prostatectomy group irrespective of cancer stage, but limited the radiotherapy and conservative management groups to patients with localised cancer only. An intraoperative finding that the lymph nodes have been seeded with cancer usually results in the decision to abandon prostatectomy, since the patient can no longer be classified as having localised disease. Exclusion of patients with positive nodes would thus make the outcomes of the surgical group appear more favourable. To assess the impact of this differential staging and selection process, we analysed the data by two different approaches—intention to treat and treatment received. For the Intention to treat
Treatment received
n
% survival (95% CI)
n
% survival (95% CI)
Grade I Prostatectomy* Radiotherapy Conservative
3854 4065 9804
94 (91–95) 90 (87–92) 93 (91–94)
3402 4188 10 133
98 (97–99) 89 (87–92) 92 (90–93)
Grade 2 Prostatectomy* Radiotherapy Conservative
14 287 7939 6198
87 (85–89) 76 (72–79) 77 (74–80)
12 922 8456 7046
91 (89–93) 74 (71–77) 76 (73–78)
Grade 3 Prostatectomy* Radiotherapy Conservative
5133 2596 2236
67 (62–71) 53 (47–58) 45 (40–51)
4154 2977 2834
76 (71–80) 52 (46–57) 43 (38–48)
*Intention-to-treat and treatment-received results differed significantly (p<0·0001).
Table 2: 10-year disease-specific survival in patients with clinically localised prostate cancer by intention to treat and by treatment received
907
THE LANCET
Grade and treatment
% overall survival (95% CI) Prostate-cancer patients
Grade I Prostatectomy Radiotherapy Conservative
Endpoint
Age-matched cohort
77 (73–80) 63 (60–66) 54 (52–56)
10 year relative survival in agematched cohort
65 54 53
1·17 1·17 1·01
Grade 2 Prostatectomy Radiotherapy Conservative
71 (68–74) 48 (45–51) 38 (36–41)
64 52 49
1·11 0·93 0·78
Grade 3 Prostatectomy Radiotherapy Conservative
54 (50–58) 33 (28–38) 17 (14–20)
62 52 47
0·87 0·63 0·36
Radiotherapy
Conservative management
Overall survival (%) Cancer patients Age-matched cohort
81 (78–83) 83
70 (66–74) 76
45 (41–50) 74
Relative survival
0·98
0·92
0·61
86 (79–91) 84 (74–90)
69 (56–78) 63 (44–77)
Prostate-cancer-specific survival (%) AJCC stage I 87% (85–90)† AJCC stage II
}
*Intention to treat. †AJCC=American Joint Commission on Cancer. Since some of the patients in the prostatectomy group were initially thought to have clinically localised cancer before prostatectomy, but were later upstaged to regional or metastatic disease after prostatectomy, analyses of only those surgical patients classified as having localised cancer would result in differential staging bias. Therefore, all patients undergoing radical prostatectomy were included in the prostatectomy group irrespective of cancer stage.
*Intention-to-treat.
Table 3: 10-year overall survival in patients with clinically localised prostate cancer and age-matched cohorts
Table 4: Survival in patients with grade 3 disease 5 years after diagnosis
intention-to-treat analyses, patients with localised cancer, lymphnode dissection, and radiotherapy, or radiotherapy alone were assigned to the radiotherapy group. Otherwise, patients were classified in the prostatectomy group if a lymph-node dissection or radical prostatectomy was documented. Any remaining patients were included in the conservative management group. For the treatment-received analyses, patients were classified in the prostatectomy group only if radical prostatectomy was documented in their SEER records. The classification of the radiotherapy and conservative management groups was the same as in the intention-to-treat analysis. Major endpoints ascertained in this study were prostatecancer-specific survival, overall survival, and relative survival (ratio of survival in cancer patients to that of an age-matched cohort). The Kaplan-Meier method was used to estimate survival at 5 years and 10 years. In estimating prostate-cancer-specific survival, follow-up was censored at the time of death if it was due to other causes. We used the log-rank test to look for significant differences between the intention-to-treat and treatment-received approaches. The Cox proportional-hazards model was used to examine the effects of cancer grade on prostate-cancer-specific survival after adjustment for differences in age, race, years of diagnosis, and place of diagnosis. In the calculation of expected survival, 1990 Social Security Administration period life tables were applied to age-matched cohorts with age distributions identical to those of the cancer cohorts.27
Results Our study population consisted of 59 876 men (table 1). The mean length of follow-up was 44·5 months; 10% of patients were followed up for 92 months or longer. Radical prostatectomy was the most common treatment modality. Patients who underwent prostatectomy were, on average, 5 years younger than those undergoing radiotherapy or conservative management (table 1). Of the 24 257 men scheduled to undergo prostatectomy, 89% underwent lymph-node dissection, 51% had pathologically localised cancer, and 80% actually proceeded to complete radical prostatectomy. Although
the proportion undergoing lymph-node dissection in the prostatectomy group is somewhat lower than one might expect, it is similar to the proportion in the meta-analysis by the American Urological Association.28 Our analyses showed that differential staging can significantly influence observed outcomes and that the benefits of prostatectomy are generally overestimated when analyses are done by the treatment-received approach (table 2). For prostatectomy, 10-year prostatecancer-specific survival rates were significantly higher by treatment-received analysis than by intention-to-treat analysis (p<0·0001) for all three grades of disease. By contrast, because only a small proportion of patients receiving radiotherapy or conservative management undergo lymph-node dissection, outcomes derived by the two analytical approaches are similar for these groups of patients (table 2). We found, as expected, that cancer grade significantly affects overall survival (table 3). All patients with grade 1 disease had similar or even better overall survival than the age-matched cohort, whatever the initial treatment. By contrast, patients with grade 3 disease had much lower overall survival than their age-matched cohorts in all treatment groups. The difference in survival in patients with grade 1 or 2 prostate cancers between those who underwent prostatectomy and their age-matched cohorts may be due to an inherent selection bias, whereby healthier patients are more likely to be selected for aggressive treatments. Accordingly, adequate adjustment for comorbidity is essential in comparison of overall mortality across treatment groups. In contrast to overall mortality, however, the impact of comorbidity on prostate-cancer-specific survival is modest.29 Differences in this variable are therefore more likely to be attributable to the nature of the cancers and the impact of interventions than to comorbidity. The prostate-cancer-specific survival was consistently poor in
Our study
Radical prostatectomy Radiotherapy Conservative management
5-year survival (95% CI)* Prostatectomy
Intention-to-treat
Treatment received
n
% survival
n
% survival
24 257 15 721 19 898
83 (81–84) 76 (74–78) 82 (81–84)
21 122 16 869 21 885
89 (87–91) 74 (72–76) 80 (79–81)
PORT*
AUA†
% survival
n
% survival
91 (88–93) 79 (74–90) 91 (87–94)
747 222 126
90 (88–92) 67 (59–75) 87 (81–93)
*Prostate Disease Patient Outcomes Research Team. Number of patients in each treatment group was not reported. Annual mortality rate reported was converted to 10-year survival by formula: 10-year survival=e⫺10⫻annual mortality rate. †American Urological Association. Results specific to stage B prostate cancer. 95% CI derived from formula: 95% CI=1·96⫻SE, where SE=mortality rate divided by the square root of the number of deaths.
Table 5: 10-year prostate cancer-specific survival (with 95% CI) in patients with clinically localised prostate cancer
908
Vol 349 • March 29, 1997
THE LANCET
patients with grade 3 cancers in all treatment groups (table 2). The risk of dying of grade 3 prostate cancer within 10 years of diagnosis (by Cox analysis) was ten times that of dying of grade 1 disease. Because the category of localised disease includes both stage I and II disease, we further refined our analysis to delineate outcomes separately for each stage. Grade 3 disease uniformly resulted in a poor outcome even among patients with stage I disease (table 4). Furthermore, the effect of high-grade prostate cancer on survival is rapid. 5 years after diagnosis, patients with grade 3 disease managed conservatively had relative survival of only 0·61 and disease-specific survival of only 63–69%. Patients with grade 3 disease treated by prostatectomy or radiotherapy had higher relative and prostate-cancer-specific survival (table 4). For comparison with previous studies, overall 10-year survival by treatment is given in table 5 with calculated 10-year survival from the two meta-analyses.
Discussion For any medical treatment, informed decisions about treatment choice can be made only if unbiased, representative data on outcomes for each option are available. Ideally, such data are derived from randomised clinical trials. If such trials have not been done, however, population-based studies and meta-analyses are often used to provide suggestive estimates about outcomes. Interpretation and application of meta-analyses based on study-level data, however, can be difficult because the inclusion and exclusion criteria vary among studies, and data on many recognised modifiers of outcomes are incomplete. Although cancer grade is recognised to be the most important predictor of survival in prostate cancer,15–18 reliable meta-analytical estimates of cancer-grade-specific survival could not be made because few published data were available.3.6 The only meta-analysis able to provide cancer-grade-specific estimates of survival was based on individual-level data; however, that study was limited to patients who received conservative management and there were few patients with grade 3 disease.18 Population-based studies, on the other hand, are derived from individual-level data for all patients in defined geographic areas. Thus, the same ascertainment criteria can be applied for every patient. Results from population-based studies can be generalised more readily than those from case series or meta-analyses because they include patients treated in all clinical settings, not just those treated in selected medical centres. Previous population-based analyses have focused on patients receiving conservative management,17,30 and there are no comparable data for patients receiving radiotherapy or undergoing prostatectomy. Our results from this population-based database showed some differences from those of previous studies (table 5).3,6 In general, our results suggest a poorer outcome for both conservative management and prostatectomy than the previous studies, but the outcome after radiotherapy is similar.3,7 There are many possible explanations for these differences. However, when the prostatectomy data in our study were analysed by a treatment-received instead of an intention-to-treat approach (table 5), the results after prostatectomy were nearly identical to those in the previous meta-analyses, which suggests that the discrepancies for prostatectomy may be due partly to the difference in analytical approach.
Vol 349 • March 29, 1997
We believe that, compared with studies that used the treatment-received approach, results derived from the intention-to-treat approach more closely approximate the general clinical situation, in which the lymph-node status of a patient is not known at the time of decision-making (before prostatectomy). Although it is well recognised that differential staging resulting from a treatment-received as opposed to an intention-to-treat analysis can profoundly influence observed outcomes, few studies have provided empirical data to quantify the effects. We found that 10-year prostate-cancer-specific mortality for prostatectomy was more than 60% greater by intention to treat than by treatment received. This finding suggests that results from previous studies that used the treatment-received approach should be interpreted with greater caution. The cancer-grade-specific data obtained in this study emphasise the often under-recognised heterogeneity of clinical prostate cancer. Although this disorder generally has a protracted course, we found that grade 3 prostate cancers progress rapidly. The outlook for patients with grade 3 disease was poor irrespective of stage (table 4). Conclusions drawn from data pooled across cancer grades are therefore too optimistic for patients with grade 3 cancer and too pessimistic for patients with grade 1 disease. Grade-specific data may help to facilitate informed decision-making about treatment. As in previous studies,17,18,30 patients with grade 1 prostate cancer have overall mortality similar to that of an age-matched cohort at 10 years, even when managed conservatively. Given the negligible effect of grade 1 cancer on mortality and the absence of evidence demonstrating the benefit of aggressive therapies, conservative management seems to be a viable option for such patients. In this study, about one in six patients with clinically localised prostate cancer had grade 3 disease. These patients have the most to gain from effective treatment. Although this study was not randomised and therefore cannot fairly compare the efficacies of the various treatment options, several observations suggest that active interventions (prostatectomy or radiotherapy) might improve long-term survival in patients with grade 3 (stage I and II) cancers. Because localised prostate cancer is, in general, a slowly progressive disease, many of the potential therapeutic gains are lost because patients die of other causes. Since grade 3 disease is associated with low disease-specific survival, intervention would provide a more immediate gain, reducing the effects of other causes of death. Differential grading is less likely to operate in the grade 3 group than in the lower grades, thus increasing the possibility of potential gains to be achieved with interventional therapy in this group. Grading of patients who do not undergo surgery is based on biopsy samples, which may not always come from the most advanced part of the tumour. These patients may therefore be given a lower grading than is appropriate. Complete removal of the prostate allows more complete pathological ascertainment of cancer grade. Since grade 3 is the worst category, whatever the therapeutic route chosen, this bias cannot preferentially benefit any specific treatment group. Although most clinicians agree that the benefits of aggressive therapy are generally not apparent until 10 years later, we observed a large difference in prostatecancer-specific survival in grade 3 disease between 909
THE LANCET
patients receiving conservative management (63–69%) and interventions (84–86% for radiotherapy and 87% for prostatectomy) within 5 years of diagnosis. This magnitude of difference was not observed in patients with grade 1 or grade 2 disease even after 10 years of follow-up. The 10-year life expectancy rule may need to be reevaluated for patients with grade 3 disease. We did not attempt to compare the relative effects of different treatments because information on many factors associated with selection of patients (comorbidity, pretreatment concentrations of prostate specific antigen, tumour size, and ploidy) was not available. Although the independent prognostic value of these variables beyond cancer stage and grade has not been demonstrated,12 a combination of these, or as yet unknown, factors may affect outcomes. In addition, the favourable outcome of prostatectomy might have resulted from differential cancer grading or other potential biases. Whereas comparisons between the radiotherapy and conservative management groups may be more valid because of their similarities in terms of age, grading (biopsy based), and staging (clinically based), the comorbidity burden might have played a part in selection of patients for these treatments. Finally, the lack of Gleason grades in the SEER database may result in incomplete cancer-grade adjustments, since the Gleason grade distribution may be uneven across different treatment groups within the same broad category (grade 1, 2, and 3) defined by SEER. We could not adjust for all confounding factors, therefore we did not attempt to draw definite conclusions about treatment efficacies. Clearly, large, randomised, multicentre trials are needed to provide definitive information about the relative efficacy of prostate-cancer treatments. Such trials are under way, but their results will not be available for some time. Our aim was to provide information from a large database on the current controversies. We thank Peter Albertsen, Michael Barry, Maria Friedman, Marian Gornick, Barnett Kramer, Jim Lubitz, Arnold Potosky, William Shipley, Joan Warren, John Wasson, John Wennberg, Yulin Yao, and Anthony Zietman for review and comments. This study was supported by the Agency for Health Care Policy and Research (grant number HS 08397) and the Health Care Financing Administration.
References 1 2 3
4
5 6
Parker SL, Tong T, Bolden S, Wingo PA. Cancer statistics. CA Cancer J Clin 1997; 47: 5–27. Scardino P. Problem of prostate cancer. J Urol 1994; 152: 1677–78. Wasson J, Cushman C, Bruskewitz R, et al. A structured literature review of treatment for localized prostate cancer. Arch Family Med 1993; 2: 487–93. Wilt TJ, Brawer MK. The prostate cancer intervention versus observation trial: a randomized trial comparing radical prostatectomy versus expectant management for the treatment of clinically localized prostate cancer. J Urol 1994; 152: 1910–14. Norlén BJ. Swedish randomized trial of radical prostatectomy versus watchful waiting. Can J Oncol 1994; 4 (suppl): 38–40. Austenfeld MS, Thompson IM, Middleston RG. Meta-analysis of the literature: guideline development for prostate cancer treatment. J Urol 1994; 152: 1866–69.
910
7
Middleton RG, Thompson IM, Austenfeld MS, et al. Prostate cancer clinical guidelines panel summary report on the management of clinically localized prostate cancer. J Urol 1995; 154: 2144–48. 8 Walsh PC, Partin AW, Epstein JI. Cancer control and quality of life following anatomical radical retropubic prostatectomy: results at 10 years. J Urol 1994; 152: 1831–36. 9 Zinche H, Oesterling JE, Blute ML, Bergstralh EJ, Myers RP, Barrett DM. Long-term (15 years) results after radical prostatectomy for clinically localized (stage T2c or lower) prostate cancer. J Urol 1994; 152: 1850–57. 10 Zietman AL, Coen JJ, Dallow KC, Shipley WU. The treatment of prostate cancer by conventional radiation therapy: an analysis of longterm outcome. Int J Radiat Oncol Biol Phys 1995; 32: 287–92. 11 Bagshaw MA, Cox RS, Hancock SL. Control of prostate cancer with radiotherapy: long-term results. J Urol 1994; 152: 1781–85. 12 Barry MJ, Fleming C, Coley CM, Wasson JH, Fahs MC, Osterling J. Should Medicare provide reimbursement for prostate-specific antigen testing for early detection of prostate cancer? Urology 1995; 46: 2–13. 13 Narayan P, Gajendran V, Taylor SP, et al. The role of transrectal ultrasound-guided biopsy-based staging, preoperative serum prostatespecific antigen, and biopsy Gleason score in prediction of final pathologic diagnosis in prostate cancer. Urology 1995; 46: 205–12. 14 Surveillance Epidemiology and End Results Program. Case diagnosed in 1973–1992, public-use database. Bethesda, MD: National Cancer Institute, 1995. 15 Gleason DF, Mellinger GT. Prediction of prognosis for prostatic adenocarcinoma by combined histological grading and clinical staging. J Urol 1974; 11: 58–64. 16 Zincke H, Bergstralh EJ, Blute ML, et al. Radical prostatectomy for clinically localized prostate cancer: long-term results of 1,143 patients from a single institution. J Clin Oncol 1994; 12: 2254–63. 17 Johansson J-E, Adami H-O, Andersson SO, Bergstrom R, Holmberg L, Krusenmo UB. High 10-year survival rate in patients with early, untreated prostatic cancer. JAMA 1992; 267: 2191–96. 18 Chodak GW, Thisted RA, Gerber GS, et al. Results of conservative management of clinically localized prostate cancer. N Engl J Med 1994; 330: 242–48. 19 Fleming C, Wasson J, Albertsen P, Barry M, Wennberg JE. Prostate PORT: a decision analysis of alternative treatment for clinically localized prostate cancer. JAMA 1993; 269: 2650–58. 20 Beck JR, Kattan MW, Miles BJ. A critique of the decision analysis for clinically localized prostate cancer. J Urol 1994; 152: 1894–99. 21 Gerber GS, Thisted RA, Scardino PT, et al. Results of radical prostatectomy in men with clinically localized prostate cancer-multiinstitutional pooled analysis. JAMA 1996; 276: 615–19. 22 Potosky AL, Riley GF, Lubitz JD, Mentnech RM, Kessler LG. Potential for cancer related health services research using a linked Medicare-tumor registry database. Med Care 1993; 31: 732–48. 23 Horm JW, Asire AJ, Young JLJ. Surveillance, Epidemiology, and End Results Program: cancer incidence and mortality in the United States, 1973–91. Bethesda, MD: National Cancer Institute, 1984. 24 Percy C, Stanek E, Gloeckler L. Accuracy of cancer death certificates and its effect on cancer mortality statistics. Am J Public Health 1981; 71: 242–50. 25 Shambaugh EM, Weiss MA. Summary staging guide for the Cancer Surveillance, Epidemiology and End Results Reporting (SEER) Program: US Department of Health and Human Services Public Health Service National Institutes of Health, 1986. 26 American Joint Committee on Cancer. Manual for staging of cancer, 4th edn. Philadelphia: Lipponcott, 1992. 27 Bell FC, Wade AH, Goss SC. Life tables for the United States Social Security Area 1900–2080. Baltimore: Social Security Administration, 1992. 28 Middleton RG. The management of clinically localized prostate cancer: guidelines from the American Urological Association. CA Cancer J Clin 1996; 46: 249–53. 29 Albertsen PC, Fryback DG, Storer BE, Kolon TF, Fine J. The impact of comorbidity on life expectancy among men with localized prostate cancer. J Urol 1996; 156: 127–32. 30 Albertsen PC, Fryback DG, Storer BE, Kolon TF, Find J. Long-term survival among men with conservatively treated localized prostate cancer. JAMA 1995; 274: 626–31.
Vol 349 • March 29, 1997