Maximal androgen blockade for the treatment of metastatic prostate cancer--a systematic review.

INTRODUCTION
Maximal androgen blockade (MAB) versus castration alone in patients with metastatic prostate cancer has been extensively evaluated in randomized trials. The inconsistent results have led to the publication of multiple meta-analyses. The present review examines the evidence from meta-analytic reports to determine whether MAB using agents such as flutamide, nilutamide, and cyproterone acetate (CPA) is associated with a survival advantage.


METHODS
We conducted a systematic review of the literature (MEDLINE, EMBASE, and the Cochrane Library through July 2004; CANCERLIT through October 2002) for meta-analyses that compared MAB with castration alone in previously untreated men with metastatic prostate cancer (D1 or D2, N+/M0 or M1). Two reviewers selected papers for eligibility; disagreement was resolved by all the authors through consensus.


RESULTS
The literature search identified six meta-analyses that met the eligibility criteria of the review. Two of those reports were based on individual patient data (IPD), and four were based on data from the published literature. All six meta-analyses pooled data on overall survival. The best evidence came from the largest meta-analysis, conducted by the Prostate Cancer Trialists Collaborative Group and based on IPD (8725 patients) from 27 trials. That analysis detected no difference in overall survival between mab and castration alone at 2 or 5 years. However, a subgroup analysis showed that MAB with nonsteroidal anti-androgens (NSAAS) was associated with a statistically significant improvement in 5-year survival over castration alone (27.6% vs. 24.7%; p = 0.005). The combination of MAB with CPA, a steroidal anti-androgen, was associated with a statistically significant increased risk of death (15.4% vs. 18.1%; p = 0.04). Compared with castration alone, MAB was associated with more side effects (that is, gastrointestinal, endocrine function) and reduced quality of life in domains related to treatment symptoms and emotional functioning.


CONCLUSIONS
The small survival benefit conferred by MAB with NSAA is of questionable clinical significance given the added toxicity and concomitant decline in quality of life observed in patients treated with MAB. Therefore, combined treatment with flutamide or nilutamide should not be routinely offered to patients with meta-static prostate cancer beyond the purpose of blocking testosterone flare. Monotherapy, consisting of orchiectomy or the administration of a luteinizing hormone-releasing hormone agonist is recommended as standard treatment.


Introduction
Maximal androgen blockade (MAB) versus castration alone in patients with metastatic prostate cancer has been extensively evaluated in randomized trials. The inconsistent results have led to the publication of multiple meta-analyses. The present review examines the evidence from meta-analytic reports to determine whether MAB using agents such as flutamide, nilutamide, and cyproterone acetate (CPA) is associated with a survival advantage.

Methods
We conducted a systematic review of the literature (MEDLINE, EMBASE, and the Cochrane Library through July 2004; CANCERLIT through October 2002) for metaanalyses that compared MAB with castration alone in previously untreated men with metastatic prostate cancer (D1 or D2, N+/M0 or M1). Two reviewers selected papers for eligibility; disagreement was resolved by all the authors through consensus.

Results
The literature search identified six meta-analyses that met the eligibility criteria of the review. Two of those reports were based on individual patient data (IPD), and four were based on data from the published literature. All six meta-analyses pooled data on overall survival.
The best evidence came from the largest metaanalysis, conducted by the Prostate Cancer Trialists Collaborative Group and based on IPD (8725 patients) from 27 trials. That analysis detected no difference in overall survival between MAB and castration alone at 2 or 5 years. However, a subgroup analysis showed that MAB with nonsteroidal anti-androgens (NSAAs) was associated with a statistically significant improvement in 5-year survival over castration alone (27.6% vs. 24.7%; p = 0.005). The combination of MAB with CPA, a steroidal anti-androgen, was associated with a statistically significant increased risk of death (15.4% vs. 18.1%; p = 0.04). Compared with castration alone, MAB was associated with more side effects (that is, gastrointestinal, endocrine function) and reduced quality of life in domains related to treatment symptoms and emotional functioning.

Conclusions
The small survival benefit conferred by MAB with NSAA is of questionable clinical significance given the added toxicity and concomitant decline in quality of life observed in patients treated with MAB. Therefore, combined treatment with flutamide or nilutamide should not be routinely offered to patients with metastatic prostate cancer beyond the purpose of blocking testosterone flare. Monotherapy, consisting of orchiectomy or the administration of a luteinizing hormone-releasing hormone agonist is recommended as standard treatment.

KEY WORDS
Prostatic neoplasms, androgen antagonists, hormonal anti-neoplastic agents

INTRODUCTION
Prostate cancer is currently the most prevalent form of male cancer in Canada 1 . At diagnosis, 20%-30% of patients will present with advanced or metastatic disease. Of those men, approximately 25% will die from their disease within 2 years 2 . Therapeutic interventions seek not only to increase survival in those patients, but also to improve quality of life (QOL) 3 .
The mainstay of treatment for advanced or metastatic prostate cancer is to inhibit the biosynthesis of androgens, the hormones responsible for prostate cancer cell growth. Androgen suppression can be achieved through surgical (bilateral orchiectomy) or medical castration. Medical castration involves the long-term use of luteinizing hormone-releasing hormone (LHRH) agonists. The two methods of castration appear equally effective in removing testicular androgens 4 .
The testes are the major locale for testosterone production; however, the adrenal glands also produce a small but measurable quantity of androgens. It has been hypothesized that removing all circulating androgens-by blocking adrenal androgens in addition to inhibiting testicular androgen production-might be beneficial to patients. Combination treatment, in the form of surgical or medical castration plus administration of an anti-androgen [for example, flutamide, nilutamide, or cyproterone acetate (CPA)] is called "maximal androgen blockade" (MAB).
The use of MAB was first introduced in the early 1980s 5 . Since then, a large number of randomized controlled trials have been conducted to evaluate the efficacy of MAB as compared with castration alone. The trials yielded inconsistent results. Most failed to provide convincing evidence of improved survival with MAB; however, a few of the larger trials detected survival benefits with combined treatment [6][7][8] . Low statistical power, study immaturity, compliance to treatment, and imbalances in prognostic indicators between study arms of individual trials were implicated as potential sources of discrepancy [9][10][11][12][13] .
Recent attempts to determine the treatment efficacy of MAB have involved meta-analyses of the trials 14 . To determine whether MAB is associated with a survival advantage, the present review systematically examines the results of the meta-analyses comparing MAB with castration alone in patients with metastatic prostate cancer.

MATERIALS AND METHODS
The present systematic review was originally completed in the context of developing a clinical practice guideline for Cancer Care Ontario's Program in Evidence-Based Care (PEBC), using the methodology of the Practice Guidelines Development Cycle 15 . The literature was searched by one member of PEBC's Genitourinary Cancer Disease Site Group. Evidence was reviewed and selected by two members, and disagreements pertaining to eligibility were handled through consensus involving the five members of the writing group. Two reviewers assessed eligible reports for important aspects of methodologic quality as expressed in the Quorom statement 16 (Appendix A).

Literature Search Strategy
We conducted a systematic search of MEDLINE (1980 through July 2004), EMBASE (1980 through 2004 wk 27), CANCERLIT (1980through October 2002, and the Cochrane Library (2004, Issue 2) databases. In each database, subject headings were combined with disease-specific, treatment-specific, and design-specific search terms (Appendix B). The reference lists of all articles found, including reviews and articles held in personal files, were reviewed for additional citations. The search was restricted to reports published in the English language.

Eligibility Criteria
Published reports or abstracts of meta-analyses comparing MAB (orchiectomy or LHRH agonist plus administration of an anti-androgen) with castration alone (orchiectomy or LHRH agonist) in previously untreated men with metastatic prostate cancer (D1 or D2, N+/ M0 or M1) were eligible for inclusion. Papers were required to report overall mortality or disease progression-related outcomes, or both. Adverse effects and QOL were also outcomes of interest.

IPD Meta-analyses
Bertagna et al. 25 published the first IPD meta-analysis in 1994. That analysis was limited to seven doubleblind, placebo-controlled trials of MAB with nilutamide (1056 patients). An update published in abstract form by Debruyne et al. 26 provided extended follow-up data on survival and disease progression.
In 1995, the Prostate Cancer Trialists' Collaborative Group (PCTCG) published an IPD meta-analysis that included 22 MAB trials (5710 patients) 17 . All randomized trials that compared castration alone to MAB, both published and unpublished, were sought for inclusion. The main limitations of the report include the absence of explicitly defined eligibility criteria, a description of the methods used to identify and select trials, an appraisal of trial quality and its influence on the pooled results, and an indication of whether subgroup analyses were planned a priori (Appendix A). The report is also limited by the fact that overall mortality was the only outcome analyzed; other important endpoints, including toxicity and QOL were not examined. The meta-analysis was updated in 2000 18 to include a total of twenty-seven trials: twelve used flutamide, eight used nilutamide, and seven used CPA as the anti-androgen. In combining data on 8725 patients, this updated report represents the most extensive quantitative analysis of MAB trials conducted to date.

Literature-based Meta-analyses
The number of trials included in the four literaturebased meta-analyses ranged from nine (1978 patients) to twenty (6745 patients; Table I). The largest analysis was conducted for the Agency for Health Care Policy Research (AHCPR) by Aronson et al. 20 . The review was well conducted, with trials systematically identified through a prospectively designed protocol that specified the objectives, literature search strategy, eligibility criteria, method of assessing trial quality, and subgroup and sensitivity analyses, including an assessment of publication bias (Appendix A). Twenty-seven MAB trials were identified, including twelve using flutamide, eight using nilutamide, and seven using CPA. Overall mortality was the only outcome for which data were statistically pooled, but data on disease progression, QOL, and adverse effects were also summarized.
The three other literature-based meta-analyses were restricted in scope, analyzing trials that compared castration with MAB using nonsteroidal antiandrogens (NSAA) [22][23][24] . The largest of those analyses, carried out by Schmitt et al. 22 for the Cochrane Prostatic Diseases and Urologic Cancers Group, included twenty trials and pooled data on overall mortality, progression-free survival (PFS), and cancer-specific survival. Bennett et al. 23 and Caubet et al. 24 both included nine trials and pooled data on overall mortality 23,24 and PFS 24 .  (Table II). Further analyses at different follow-up periods also showed no differ-ence in mortality and suggested an absolute 5-year survival difference of approximately 2% in favour of MAB. Subgroup analyses were performed by method of androgen suppression (orchiectomy vs. LHRH agonist), type of anti-androgen, patient age, stage of disease (metastases vs. no metastases), and non-prostate cancer mortality. With the exception of type of anti-androgen, no significant differences in treatment effect were observed within any of those subgroups. A small and statistically significant survival benefit was detected for MAB with flutamide (HR = 0.92; 95% CI: 0.86-0.98; p = 0.02), and a similar but nonsignificant result was observed for nilutamide. MAB with CPA was associated with a significantly worse survival outcome than castration alone (HR = 1.13; 95% CI: 1.01-1.25; p = 0.04). Treatment with MAB containing either of the NSAAs increased 5-year survival over castration alone by 3% (27.6% vs. 24.7%, p = 0.005).

Overall Survival IPD Meta-analyses
Debruyne et al. 26 reported a reduction in the odds of death in patients treated with nilutamide-containing MAB; MAB was associated with a 16% reduction in mortality as compared with castration alone [odds ratio (OR) = 0.84; 95% CI: 0.71-0.99; p = 0.038]. Table III summarizes the results for overall mortality from the four literature-based meta-analyses. Aronson et al. 20 detected no significant difference in overall mortality at 2 years, although at 5 years, overall mortality was significantly improved with MAB (HR = 0.87; 95% CI: 0.81-0.94). However, the 5-year estimate was based on half the trials (10 trials, 66% of patients) that contributed to the 2-year estimate. No differences in treatment effect were detected in any of the subgroup analyses performed (method of androgen suppression, stage of disease, type of anti-androgen, or trial quality).

Literature-based Meta-analyses
Schmitt et al. 22 reported no difference in mortality at 1 or 2 years between NSAA MAB and castration-only arms, but 5-year mortality was better with MAB (OR = 1.29; 95% CI: 1.11-1.50; p = 0.0009). The two other literature-based reports examining NSAA MAB detected significant reductions in the risk for mortality with MAB that ranged between 10% and 22% 23,24 .

Disease Progression
Pooled analyses of disease progression data were available from three of the six meta-analyses 22,24,26 ; however, those analyses are limited by the inclusion of a small proportion of MAB trials. Each of those reports combined data from trials of MAB using NSAA. Debruyne et al. 26 reported that, among seven trials, the odds of progression were reduced by 17% by MAB with nilutamide (OR = 0.83; 95% CI: 0.70-0.98; p = 0.031). Schmitt et al. 22 pooled published PFS data at 1 (seven trials), 2 (five trials), and 5 years (two trials); the odds of progression were significantly  24 reported a 23%-26% reduction in the risk for progression with MAB depending on the type of metaanalytic method used [relative risk (RR) = 0.74, p < 0.001 among seven trials; RR = 0.77, p < 0.001 among seven trials].

QOL and Adverse Effects
Aronson et al. 20 wrote the only report that reviewed QOL and adverse effects. Among the 27 randomized controlled trials that those authors reviewed, only one formally assessed QOL. The authors summarized two reports 61,62 of the large National Cancer Institute (NCI) INT-0105 Southwest Oncology Group/Eastern Cooperative Oncology Group trial 29 , which compared orchiectomy plus flutamide with orchiectomy plus placebo. Measures of QOL included three treatmentrelated symptoms (diarrhea, gas pain, and body image), physical functioning, and emotional functioning; all were assessed at 1, 3, and 6 months after the start of treatment. Patients treated with MAB experienced significantly more diarrhea at 3 months (p < 0.001) and worse emotional functioning at 3 and 6 months (p < 0.003) than did patients treated with castration alone. Nonsignificant trends toward poor scores on measures of physical functioning, fatigue, abdominal gas, overall pain, and body image were also observed with MAB. Major limitations were present in the reporting of adverse effects in the MAB trials. Because of those limitations, data on adverse effects were supplemented with similar data from package inserts that accompany therapeutic agents marketed in the United States and from phase II studies. Tables IV and V summarize the adverse effects data contained in the Aronson et al. report 20 . Compared with castration alone, treatment with MAB that included NSAA appeared to produce more gastrointestinal-related problems and more frequent withdrawal from treatment. When MAB contained CPA, treatment was associated with more complications related to endocrine function, but a withdrawal pattern similar to that in patients receiving monotherapy was demonstrated.

DISCUSSION
Six meta-analyses form the evidence base of the present review 17,18,20,[22][23][24][25] . Evidence from those analyses suggests that patient outcomes depend on the type of anti-androgen used with MAB. The PCTCG meta-analysis 18 showed that MAB was not associated with a statistically significant improvement in overall survival. However, when outcomes were analyzed by type of anti-androgen, varying treatment efficacies among the agents were evident. Small but statistically significant survival benefits in the range of 3% at 5 years were detected among trials that used MAB with an NSAA (as compared with castration alone). Compared with castration alone, MAB with CPA (a steroidal anti-androgen) was associated with an approximate 3% increased risk of death.
Variability in the magnitude of outcome among meta-analyses may arise from a number of factors, including the number and size of the trials contributing to the pooled estimate, the type of anti-androgens being evaluated, and the use of published summary data or IPD for the analyses. The four literature-based meta-analyses 20,22-24 included fewer trials (and fewer patients) than did the PCTCG meta-analysis, but the resulting pooled estimates were of greater magnitude (in favour of MAB) than were those generated using IPD 18 . In meta-analyses based on published data, publication bias is more likely to exaggerate treatment effects 63 . Only one of the four literature-based meta-analyses assessed the influence of publication status on the overall pooled result 20 . With IPD, many of the problems associated with published data that introduce bias are eliminated by the ability to incorporate all trial data (published and unpublished), to check the integrity of patient randomization, and to perform proper time-to-event analyses (as compared with fixed time point) by intent-to-treat 64,65 . Further, because of greater patient numbers, IPD often provides greater statistical power to properly perform subgroup analyses 65 . The methodologic weaknesses of the PCTCG have been identified, but the advantages of using IPD currently make the PCTCG meta-analysis the most reliable evidence comparing MAB with castration alone.
To decide whether MAB should be the preferred treatment for patients, the small survival benefit and the additional adverse effects of combined treatment must be balanced. The clinical significance of a statistically significant 3% improvement in survival with NSAA MAB is questionable, especially when the toxicity of MAB is considered. Data on adverse effects and QOL are limited, but they suggest increased toxicity and a concomitant decline in QOL in MABtreated patients. In addition, data on disease progression provide further evidence that MAB does not provide superior treatment efficacy over castration alone 20 .
Based on the evidence reviewed, MAB should not be routinely offered to patients with metastatic prostate cancer. Monotherapy, consisting of orchiectomy or the administration of an LHRH agonist, should be recommended as standard treatment.
It is important to distinguish between MAB as longterm treatment and short-term use of MAB in the prevention of testosterone flare. In patients treated with REVIEW: MAXIMAL ANDROGEN BLOCKADE FOR METASTATIC PROSTATE CANCER medical castration, initial treatment with an LHRH agonist is accompanied by a surge in serum testosterone during the first week or weeks of therapy, followed by a decline. That surge may exacerbate existing metastatic disease 66,67,68 , therefore short-term use of an anti-androgen is indicated to prevent or block the flare phenomenon 68 . Administration of an anti-androgen is reasonable for a period of 2-4 weeks when treatment with an LHRH agonist is initiated.
Because of the small survival improvement observed with MAB, some clinicians may still choose MAB over monotherapy for individual patients. If MAB is administered with this intent, MAB containing a NSAA is suggested. Given its higher mortality, MAB with CPA should be avoided as compared with castration alone 18 .
The present review did not identify any metaanalyses that included trials evaluating MAB with the newer anti-androgen bicalutamide. Bicalutamide-  Progressive prostate cancer is usually detected through a rise in prostate-specific antigen (PSA), which usually predates clinical or radiologic evidence of metastases. Most patients included in MAB trials had documented metastases (stage D2), and whether results from those trials are generalizable to patients with a rising PSA without evidence of metastatic disease is unknown. Only a handful of trials have analyzed outcomes by extent of metastatic involvement 4,6,11,29,34 . Most of those have not shown a benefit of MAB in patients with minimal disease, although the subgroup analyses included small numbers of patients. Only 12% of patients (approximately 1000) in the PCTCG meta-analysis 18 had documented non-metastatic prostate cancer. An analysis of those patients showed slightly worse survival with MAB, although the difference did not reach statistical significance. Prospective randomized trials to investigate the efficacy of MAB in that subgroup of patients are warranted.

CONCLUSIONS
The small survival benefit conferred by MAB with NSAA is of questionable clinical significance given the added toxicity and concomitant decline in QOL observed in patients treated with MAB. Therefore, combined treatment with flutamide or nilutamide should not be routinely offered to patients with metastatic prostate cancer (beyond the purpose of blocking testosterone flare). Monotherapy consisting of orchiectomy or the administration of a LHRH agonist is recommended as standard treatment.