Population Based Study of Hormonal Therapy and Survival in Men With Metastatic Prostate Cancer Grace Lu-Yao, Dirk F. Moore, John U. Oleynick, Robert S. DiPaola and Siu-Long Yao*,† From the Department of Environmental and Occupational Medicine (GLY) and Cancer Institute of New Jersey (GLY, DFM, RSD, SLY), University of Medicine and Dentistry of New Jersey, Robert Wood Johnson Medical School, Dean and Betty Gallo Prostate Cancer Center (GLY, DFM, RSD, SLY) and Cancer Institute of New Jersey (DFM, JUO), New Brunswick and Department of Biostatistics, School of Public Health, University of Medicine and Dentistry of New Jersey (DFM, JUO), Piscataway, New Jersey
Purpose: Although the palliative benefits of hormonal therapy for metastatic prostate cancer are widely recognized, little information is available regarding the effect of hormonal therapy on cancer specific and overall survival, and the types of patients who might benefit the most or least from hormonal therapy. Materials and Methods: Prostate cancer specific and overall survival according to hormonal therapy use was determined by the Kaplan-Meier method in 6,098 men 65 years or older diagnosed with metastatic prostate cancer in 1991 to 1999 who were identified through the population based Surveillance, Epidemiology, and End Results, and Medicare linked database. Cox proportional hazards and propensity score methods were used to adjust for potential confounders, such as disease status and patient comorbidity. Results: Propensity score adjusted median overall survival was 26 months in men who received hormonal therapy compared with 13 months in those who did not (HR 0.66, 95% CI 0.17– 0.70, p ⬍0.0001). The benefit of hormonal therapy was observed across all comorbidity strata and races. Effects were most evident in patients with poorly differentiated cancer (cancer specific mortality in favor of treatment HR 0.60, 95% CI 0.53– 0.69, p ⬍0.001). Benefit was not found in patients with well differentiated cancer (cancer specific mortality in favor of no treatment HR 1.92, 95% CI 0.90 – 4.10, p ⫽ 0.09). Conclusions: Hormonal therapy is associated with improved prostate cancer specific and overall survival in men with poorly differentiated cancer. Improved survival does not appear evident in men with well differentiated disease. Key Words: prostate; prostatic neoplasms; Medicare; SEER program; antineoplastic agents, hormonal
potential biases due to improving patient care with time.3,4 Another open question with HT has been the delineation of the types of patients who might benefit the most or least from this therapy. Because of these limitations, a randomized study was ultimately performed by the VACURG to determine the effects of HT.5 However, interpretation of this study has been limited because patients assigned placebo crossed over to HT, thus, confounding survival results. Perhaps because of the lack of data demonstrating a survival advantage with treatment, many patients and health care professionals may have come to consider HT to not be required. Such a conclusion would explain results in a United Kingdom clinical trial of early vs delayed HT, which showed that the majority of patients assigned delayed therapy died without ever receiving HT.6 Similar findings were recently observed in a population based study in the United States, in which it was also found that a substantial proportion of the patients who died of prostate cancer died without ever receiving HT.7 HT is one of the most potent and best tolerated therapies available for advanced prostate cancer. Since the number of therapies for prostate cancer are limited, it is important to be sure that the effects of HT are not underestimated and patients who could easily benefit are not missing an important, readily available and tolerable therapy. Because a randomized study would not be feasible, we performed a
lthough definitive hormonal therapy, ie diethylstilbestrol, orchiectomy or LHRH agonist, has documented palliative benefits and is generally considered standard therapy for men with metastatic disease,1 there has been some debate as to whether the benefits of HT include prolongation of cancer specific or overall survival.2 Initial observational studies used historical rather than concurrent controls and they were not formally adjusted for
Submitted for publication March 28, 2006. Study received approval from the institutional review boards of the University of Medicine and Dentistry of New Jersey, the SEER program and the Center for Medicare and Medicaid Services, and the performance and design of this study were reviewed and approved by the National Cancer Institute, and Center for Medicare and Medicaid Services. Supported by Award DAMD17-01-1-0755 from the United States Army Medical Research Acquisition Activity, Fort Detrick, Maryland and Cancer Institute of New Jersey. This study used the linked SEER-Medicare Database. The interpretation and reporting of these data are the sole responsibility of the authors. The content of the information does not necessarily reflect the position or the policy of the Government and no official endorsement should be inferred. * Correspondence and requests for reprints: Cancer Institute of New Jersey, University of Medicine and Dentistry of New Jersey, Robert Wood Johnson Medical School, and Dean and Betty Gallo Prostate Cancer Center, 195 Little Albany St., Room 5544, New Brunswick, New Jersey (telephone: 732-235-8830; FAX: 732-2358808; e-mail: [email protected]
). † Financial interest and/or other relationship with ScheringPlough.
0022-5347/07/1772-0535/0 THE JOURNAL OF UROLOGY® Copyright © 2007 by AMERICAN UROLOGICAL ASSOCIATION
Vol. 177, 535-539, February 2007 Printed in U.S.A. DOI:10.1016/j.juro.2006.09.049
HORMONAL THERAPY FOR METASTATIC PROSTATE CANCER
large, population based, concurrent cohort study to determine the effects of HT on survival and better define the types of patients who might benefit the most or least from this treatment. MATERIALS AND METHODS Data Sources The SEER program database was linked with Medicare files to obtain data for this study. Demographics, diagnostic data and cause of death were compiled from SEER data. Treatment information was obtained from Medicare claims. The SEER database covers approximately 14% of the population in the United States. The Medicare database covers approximately 97% of individuals 65 years or older in the United States and linkage to the SEER database was approximately 94% complete. Study Participants Patients with metastatic prostate cancer (M1) at diagnosis during 1991 to 1999 were included. Patients younger than 65 years were excluded since they were not covered by Medicare and, therefore, data on HT were not available. Patients had to have Medicare Part A (hospitalization coverage) plus Part B (physician services, laboratory and x-ray services, durable medical equipment, and outpatient and other service coverage) as the primary health insurance to ensure that claims data provided an accurate representation of patient care. Health maintenance organization enrollees and those with Veterans Administration coverage were excluded. The study was approved by the institutional review boards of the University of Medicine and Dentistry of New Jersey, the SEER program and the Center for Medicare and Medicaid Services. Use of HT Medicare physician, inpatient and outpatient claims were used to identify orchiectomy (Healthcare Common Procedure Coding System codes 54520, 54521, 54522, 54530 or 54535, or International Classification of Diseases, version 9 codes 623 and 624) and LHRH agonist use (Healthcare Common Procedure Coding System codes J1950, J9202, J9217, J9218 and J9219) through a previously described algorithm.8 Only 2 patients received diethylstilbestrol in the absence of orchiectomy or LHRH agonist. Therefore, we chose not to modify the previously described algorithm. Overall and Prostate Cancer Specific Survival The date of death was available in the Medicare database through December 31, 2002 and it was used to assess overall survival. The Medicare database does not contain information on cause of death. Therefore, prostate cancer specific death was determined through the SEER database, which had information available through December 31, 2000. Other Variables for Risk Adjustment The median income in 2000 for the ZIP code of patient residence was used to assess socioeconomic status since it is well correlated with self-reported income.9 Medicare Provider Analysis and Review claims during the first 12 months after cancer diagnosis were used to calculate the Charlson score, which is a validated measure of comorbidity.
Statistical Analyses The effect of HT on survival was evaluated by propensity score methodology and the Cox proportional hazards model. Propensity score methods10 were used to minimize selection and other biases when comparing survival. The propensity score is defined as the conditional probability of being treated, given all measured covariates. If 2 subjects have the same propensity score, they are matched and in theory equally likely to be treated or not treated. Under such conditions the actual decision to treat or not treat would be viewed as random with respect to all included covariates. Propensity score methods are generally regarded as more robust and accurate because they are subject to fewer restrictions and are able to more completely account for confounders.10 Propensity scores (the probability of receiving HT during the study period) were derived from a logistic regression model. These propensities were divided into 5 quintiles and a separate prostate cancer specific Kaplan-Meier survival curve was fitted for each treatment group (HT or no HT in the first 6 months) for each quintile. Estimated survival curves were averaged across the 5 quintiles to produce propensity adjusted estimates for the HT and no HT groups. Propensity adjusted overall survival curves were calculated in the same way using the R statistical programming language. We assessed the effect of HT on survival using a Cox model stratified on propensity quintile. The Cox proportional hazards model with covariates from table 1 was also used to assess multiple risk factors that affect overall and cancer specific survival. Assumptions of the proportional hazards model were verified. All tests of statistical significance were 2-sided at an ␣ level of 0.05. RESULTS Baseline Characteristics Table 1 lists baseline characteristics in 6,098 patients diagnosed with metastatic prostate cancer in 1991 to 1999. More than half of the patients were 75 years or older, most had aggressive disease (Gleason 8 –10) and most did not have significant comorbidity. No patients were lost to followup. Effect of HT on Overall and Cancer Specific Survival Propensity score adjusted median overall survival was 26 months in men who received HT compared with 13 months in those who did not receive it (HR 0.66, 95% CI 0.17– 0.70, p ⬍0.0001). HT was associated with a 34.1% decrease (95% CI 29.9%–38.1%) in overall mortality and a 28.8% decrease (95% CI 22.7%–34.3%) in prostate cancer specific mortality (figs. 1 and 2). Results were consistent within each quintile of propensity score and similar results were obtained when a conventional Cox proportional hazards model was used (data not shown). Median followup in censored patients was 9 years. Effects of Comorbidity, Race and Cancer Grade The effect of HT was similar across all Charlson scores (fig. 3). Decreases in overall and prostate cancer specific mortality, ie HR ⬍1, were evident irrespective of Charlson score and all decreases were statistically significant except prostate can-
HORMONAL THERAPY FOR METASTATIC PROSTATE CANCER
TABLE 1. Characteristics of patients ⱖ65 years or older diagnosed with M1 prostate cancer in SEER-Medicare Database Characteristic
No. HT (%)
No. No HT (%)
Overall age: 65–69 70–74 75 or Older Race: White Black Gleason score at diagnosis: 2–4 5–7 8–10 Unknown Median income in residence ZIP code: Lowest tertile Higher than tertile Unknown SEER region:* North central Northeast South West Diagnosis yr: 1991–1994 1995–1999 Highest Charlson score:† Not evaluated‡ 0 1 2 3 or Greater Marital status: Married Not married
4,471 722 (16.2) 1,128 (25.2) 2,621 (58.6)
1,627 225 (13.8) 345 (21.2) 1,057 (65.0)
3,877 (86.7) 594 (13.3)
1,230 (75.6) 397 (24.4)
131 (2.9) 1,319 (29.5) 2,101 (47.0) 920 (20.6)
48 (3.0) 377 (23.2) 585 (36.0) 617 (37.9)
1,463 (32.7) 2,837 (63.5) 171 (3.8)
653 (40.1) 899 (55.3) 75 (4.6)
1,649 (36.9) 645 (14.4) 237 (5.3) 1,940 (43.4)
680 (41.8) 192 (11.8) 101 (5.3) 654 (43.4)
2,659 (59.5) 2,112 (40.5)
970 (59.6) 657 (40.4)
1,894 (42.4) 1,435 (32.1) 677 (15.1) 254 (5.7) 211 (4.7)
599 (36.8) 508 (31.2) 243 (14.9) 135 (8.3) 142 (8.7)
2,892 (64.7) 1,579 (35.3)
873 (53.7) 754 (46.3)
* North central includes Detroit and Iowa registries; Northeast consists of Connecticut; South includes Atlanta; West consists of Hawaii, New Mexico, Seattle, Utah, and San Francisco/San Jose/Los Angeles. † Calculated only in patients hospitalized during the first 12 months after cancer diagnosis. ‡ Cohort for which comorbidity scores could not be calculated because patients were not hospitalized during the first 12 months following cancer diagnosis.
cer specific mortality in patients with Charlson scores 3 or greater (p ⫽ 0.25). Decreases in mortality were also evident irrespective of race (data not shown).
FIG. 1. Propensity score adjusted estimates of overall survival probability in patients 65 years or older with metastatic disease at diagnosis in SEER-Medicare database with and without HT. Date of death due to any cause was available through December 31, 2002.
FIG. 2. Propensity score adjusted estimates of prostate cancer specific survival probability in patients 65 years or older with metastatic disease at diagnosis in SEER-Medicare database with and without HT. Date of death due to prostate cancer was available through December 31, 2000.
Analysis by cancer grade revealed a strong trend in HT benefit according to Gleason score (p for trend ⬍0.0001, fig. 4). HT improved prostate cancer specific survival most in patients with Gleason 8 –10 cancer (HR 0.60, p ⬍0.001), followed by an intermediate effect in Gleason 5–7 cancer and a potentially adverse effect in Gleason 2– 4 cancer (HR 0.87, p ⫽ 0.16 and HR 1.92, p ⫽ 0.09, respectively). A similarly strong trend in HT benefit according to Gleason score was observed for overall survival (p for trend ⬍0.0001). HT improved overall survival most in patients with Gleason 8 –10 cancer (HR 0.59, p ⬍0.0001), followed by an intermediate effect in Gleason 5–7 cancer and a potentially adverse effect in Gleason 2– 4 cancer (HR 0.79, p ⫽ 0.0006 and HR 1.29, p ⫽ 0.30, respectively). The observation that HT did not benefit or may have been detrimental in patients with Gleason 2– 4 cancer was not an-
FIG. 3. Results in patients 65 years or older with metastatic disease at diagnosis in SEER-Medicare database with and without HT. HR ⬍1 indicates decreased mortality with HT.
HORMONAL THERAPY FOR METASTATIC PROSTATE CANCER
FIG. 4. Results in patients 65 years or older with metastatic disease at diagnosis in SEER-Medicare database with and without HT. HR ⬍1 indicates decreased mortality with HT. Increasing Gleason score was associated with increasing prostate cancer specific and overall survival benefit from HT (each p for trend ⬍0.0001).
ticipated. Further analyses were performed to determine if the effects could have been due to imbalances in patient comorbidity, ie whether treated patients may have been unhealthier, thus, negating potential survival benefits due to treatment. Table 2 shows that in Gleason 2– 4 patients comorbidity was actually worse in untreated rather than in treated patients. Similarly in patients with high comorbidity, ie Charlson score 3 or greater, HT was consistently less likely to have been prescribed compared with patients with no or low (Charlson score 0) comorbidity (Gleason score 2– 4, 5–7 and 8 –10 RR 0.16, p ⫽ 0.005; RR ⫽ 0.59, p ⫽ 0.019; and RR ⫽ 0.74, p ⫽ 0.10, respectively). Because healthier patients were more likely to receive HT, it is unlikely that the lack of benefit of HT in these patients can be attributable to baseline health status. DISCUSSION Information on the effect of HT on cancer specific and overall survival has been limited and one of the most definitive attempts, ie the VACURG study, to determine the effect of HT on survival was confounded by crossover.5 Because a randomized study would not be possible due to ethical considerations, we used a population based cohort design, which is often recognized as one of the best available study designs short of a randomized clinical trial,11 to address some of the limitations of previous studies. We found that modern HT was associated with a 34.1% decrease in overall mortality and an approximately 13-month increase in overall survival. Aside from the VACURG study, to our knowledge there have not been any large, randomized comparisons of HT to placebo. A method of gaining insight into the effect of a therapy under such circumstances would be a randomized comparison to an ineffective therapy. In fact, this unintentionally occurred when it was found that treatment with 50 mg bicalutamide was inferior to HT in men with metastatic prostate cancer. Analysis of combined data from 3 studies involving a total of 1,672 men showed a 1.44 overall survival HR (95% CI 1.20 –1.73) in favor of HT (p ⬍0.0001).12 Median overall survival in men treated with HT was 28 months. These results compare favorably with our results (HR 1.52 and overall survival 26 months). If single agent antiandrogen therapy is assumed to be equal to or better than placebo, these results would suggest
that HT would be associated with a survival benefit if it were compared to placebo. On the other hand, it is possible that the improved survival with HT in these studies could be due to worsened survival with antiandrogens. However, this seems less likely because higher doses of antiandrogens were associated with better outcomes in other studies and the addition of similar doses of antiandrogens to HT was shown to modestly improve survival. Despite these inferences it is not possible to rule out an adverse effect of low dose antiandrogen treatment, which in part prompted us to perform the current study. We addressed some of the limitations of previous studies. The use of concurrent controls minimized biases associated with prior observational studies that used historical controls. The use of Cox and propensity score statistical methods increased confidence that the groups being compared were as similar as possible short of randomization. Finally, the population based nature of the study helped increase the likelihood that the results are generally representative, rather than being limited to particular practice settings. However, because the study is observational and nonrandomized, there are important considerations. Since propensity score and Cox analyses rely on measured covariates, bias due to unmeasured confounders cannot be completely removed. The results of our study also might not be applicable to younger patients or those not in the United States since our study populations consisted of men 65 years or older in the United States. However, because greater than 70% of newly diagnosed prostate cancer occurs in men 65 years or older, the results should be generalizable to most patients in countries similar to the United States.13 A notable finding in our study is that there was a strong cancer grade-response type of effect with the best outcomes in poorly differentiated disease, intermediate outcomes in moderately differentiated disease and the most disquieting outcomes in well differentiated disease (p for trend ⬍0.0001), in which patients treated with HT were an average of 1.92-fold more likely to die of prostate cancer compared with untreated patients, although these results were not statistically significant. In retrospect this may be consistent with previous studies that showed little benefit for treatment in patients with well differentiated disease in other settings, such as combined androgen blockade14 and prostatectomy,15 perhaps because death from competing causes is so prevalent in these patients. In addition, prostate cancer has been observed to often be more aggressive in the presence of low testosterone16 –18 and
TABLE 2. Charlson score by cancer grade and HT Gleason Score (HT) 2–4: Yes No 5–7: Yes No 8–10: Yes No
% Charlson Score 0*
3 or Greater
* Includes patients with no Charlson score and Charlson score 0 with Charlson score calculated only in patients hospitalized during the first 12 months after cancer diagnosis.
HORMONAL THERAPY FOR METASTATIC PROSTATE CANCER the application of HT in an environment consisting predominantly of normal tissue or well differentiated disease could lead to suppression of indolent disease at the expense of promoting aggressive clones. These findings might also be consistent with preclinical studies demonstrating that androgen receptor mutations, eg arising from HT, can transform normal tissue and result in highly aggressive disease.19 However, irrespective of these observations it is important to note that our results in well differentiated disease were based on a subset analysis with a limited sample size and caution should be exercised when considering the results. Further study is critical for clarifying the risks and benefits of HT according to cancer grade.
3. 4. 5.
CONCLUSIONS Briefly, we found that HT is associated with an increase in prostate cancer specific and overall survival, especially in patients with high grade (Gleason 8 –10) cancer. These results imply that men with high grade cancer should be given a course of therapy, if at all possible. Because results from a previous Medical Research Council6 and SEER-Medicare7 study suggested incomplete use of HT, HT may be an unrealized option in many of these men. On the other hand, there did not appear to be a benefit for HT in men with well differentiated disease. In contrast, these men might consider avoiding HT unless treatment is absolutely necessary, eg for an acute disease related complication. Although ideally information on the clinical efficacy of a treatment should be derived from randomized studies, occasionally beliefs regarding a therapy can become pervasive and render the performance of such a study impractical or impossible. In the absence of randomized data results in population based cohort studies, such as the current study, have been shown to provide good estimates of results that are ultimately obtained from randomized clinical trials.20 The results of this study should provide guidance for patients and physicians as they try to understand who benefits the most or least from HT and choose among the increasingly many alternative approaches available for advanced disease.
APPENDIX Disclaimer The Applied Research Branch, Division of Cancer Prevention and Population Science, National Cancer Institute; Office of Information Services and Office of Strategic Planning, Health Care Financing Agency; Information Management Services, Inc.; and the SEER Program tumor registries provided effort in the creation of the SEER-Medicare database.
Abbreviations and Acronyms HT ⫽ hormonal therapy LHRH ⫽ luteinizing hormone-releasing hormone SEER ⫽ Surveillance, Epidemiology, and End Results VACURG ⫽ Veterans Administration Cooperative Urological Research Group REFERENCES 1.
Physician’s Data Query, Prostate Cancer. Bethesda: National Cancer Institute 2005.
Walsh PC, DeWeese TL and Eisenberger MA: A structured debate: immediate versus deferred androgen suppression in prostate cancer-evidence for deferred treatment. J Urol 2001; 166: 508. Vest S and Frazier T: Survival following castration for prostatic cancer. J Urol 1946; 56: 97. Nesbit R and Baum W: Endocrine control of prostatic carcinoma. JAMA 1950; 143: 1317. Treatment and survival of patients with cancer of the prostate. The Veterans Administration Co-operative Urological Research Group. Surg Gynecol Obstet 1967; 124: 1011. Immediate versus deferred treatment for advanced prostatic cancer: initial results of the Medical Research Council Trial. The Medical Research Council Prostate Cancer Working Party Investigators Group. Br J Urol 1997; 79: 235. Lu-Yao G, Moore DF, Oleynick MS, DiPaola RS and Yao SL: Use of hormonal therapy in men with metastatic prostate cancer. J Urol 2006; 176: 526. Shahinian VB, Kuo YF, Freeman JL and Goodwin JS: Risk of fracture after androgen deprivation for prostate cancer. N Engl J Med 2005; 352: 154. Bach PB, Schrag D, Brawley OW, Galaznik A, Yakren S and Begg CB: Survival of blacks and whites after a cancer diagnosis. JAMA 2002; 287: 2106. Nieto FJ and Coresh J: Adjusting survival curves for confounders: a review and a new method. Am J Epidemiol 1996; 143: 1059. Ball C, Sackett D, Phillips R, Haynes B, Straus S and Dawes M: Levels of evidence and grades of recommendation. Centre for Evidence-Based Medicine. Available at www.cebm. net/levels_of_evidence.asp. Accessed March 28, 2006. Bales GT and Chodak GW: A controlled trial of bicalutamide versus castration in patients with advanced prostate cancer. Urology 1996; 47: 38. Ries LAG, Eisner MP, Kosary CL, Hankey BF, Miller BA, Clegg L et al: SEER Cancer Statistics Review, 1973–2001. Bethesda: National Cancer Institute, 2004. Seidenfeld J, Samson DJ, Aronson N, Albertson PC, Bayoumi AM, Bennett C et al: Relative effectiveness and cost-effectiveness of methods of androgen suppression in the treatment of advanced prostate cancer. Evid Rep Technol Assess 1999; summer: i. Lu Yao GL and Yao SL: Population-based study of long-term survival in patients with clinically localised prostate cancer. Lancet 1997; 349: 906. Chodak GW, Vogelzang NJ, Caplan RJ, Soloway M and Smith JA: Independent prognostic factors in patients with metastatic (stage D2) prostate cancer. The Zoladex Study Group. JAMA 1991; 265: 618. Ishikawa S, Soloway MS, Van der Zwaag R and Todd B: Prognostic factors in survival free of progression after androgen deprivation therapy for treatment of prostate cancer. J Urol 1989; 141: 1139. Schatzl G, Madersbacher S, Haitel A, Gsur A, Preyer M, Haidinger G et al: Associations of serum testosterone with microvessel density, androgen receptor density and androgen receptor gene polymorphism in prostate cancer. J Urol 2003; 169: 1312. Han G, Buchanan G, Ittmann M, Harris JM, Yu X, Demayo FJ et al: Mutation of the androgen receptor causes oncogenic transformation of the prostate. Proc Natl Acad Sci U S A 2005; 102: 1151. Benson K and Hartz AJ: A comparison of observational studies and randomized, controlled trials. N Engl J Med 2000; 342: 1878.