Comparison of Therapies for Clinically Localized Prostate Cancer DRAFT

Transcription

1 Draft Comparative Effectiveness Review Number XX (Provided by AHRQ) Comparison of Therapies for Clinically Localized Prostate Cancer Prepared for: Agency for Healthcare Research and Quality U.S. Department of Health and Human Services 540 Gaither Road Rockville, MD Contract No , Task Order #4, Work Assignment #1 Prepared by: Minnesota Evidence-based Practice Center Minneapolis, Minnesota Investigators Timothy J. Wilt, MD, MPH Tatyana Shamliyan, MD, MS Brent Taylor, PhD Roderick MacDonald, MS James Tacklind, BS Indulis Rutks, BS Kenneth Koeneman, MD Chin-Soo Cho, MD Robert L. Kane, MD This information is distributed solely for the purposes of pre-dissemination peer review under 19 applicable information quality guidelines. It has not been formally disseminated by the Agency 20 for Healthcare Research and Quality. It does not represent and should not be construed to 21 represent an Agency for Healthcare Research and Quality or Department of Health and Human 22 Services determination or policy

2 This document is in the public domain and may be used and reprinted without permission except those copyrighted materials noted, for which further reproduction is prohibited without the specific permission of copyright holders. None of the investigators has any affiliations or financial 51 involvement that conflicts with the material presented in 52 this report ii

3 This report is based on research conducted by the Minnesota Evidence-based Practice Center 79 (EPC) under contract to the Agency for Healthcare Research and Quality (AHRQ), Rockville, 80 MD (Contract No , Task Order #4, Work Assignment #1). The findings and 81 conclusions in this document are those of the authors, who are responsible for its contents; the 82 findings and conclusions do not necessarily represent the views of AHRQ. Therefore, no 83 statement in this report should be construed as an official position of AHRQ or of the U.S. 84 Department of Health and Human Services The information in this report is intended to help health care decisionmakers patients and 87 clinicians, health system leaders, and policymakers, among others make well-informed 88 decisions and thereby improve the quality of health care services. This report is not intended to 89 be a substitute for the application of clinical judgment. Anyone who makes decisions concerning 90 the provision of clinical care should consider this report in the same way as any medical 91 reference and in conjunction with all other pertinent information, i.e., in the context of available 92 resources and circumstances presented by individual patients This report may be used, in whole or in part, as the basis for development of clinical practice 95 guidelines and other quality enhancement tools, or as a basis for reimbursement and coverage 96 policies. AHRQ or U.S. Department of Health and Human Services endorsements of such 97 derivative products may not be stated or implied iii

4 Preface The Agency for Healthcare Research and Quality (AHRQ) conducts the Effective Health Care Program as part of its mission to organize knowledge and make it available to inform decisions about health care. As part of the Medicare Prescription Drug, Improvement, and Modernization Act of 2003, Congress directed AHRQ to conduct and support research on the comparative outcomes, clinical effectiveness, and appropriateness of pharmaceuticals, devices, and health care services to meet the needs of Medicare, Medicaid, and the State Children s Health Insurance Program (SCHIP). AHRQ has an established network of Evidence-based Practice Centers (EPCs) that produce Evidence Reports/Technology Assessments to assist public- and private-sector organizations in their efforts to improve the quality of health care. The EPCs now lend their expertise to the Effective Health Care Program by conducting Comparative Effectiveness Reviews (CERs) of medications, devices, and other relevant interventions, including strategies for how these items and services can best be organized, managed, and delivered. Systematic reviews are the building blocks underlying evidence-based practice; they focus attention on the strength and limits of evidence from research studies about the effectiveness and safety of a clinical intervention. In the context of developing recommendations for practice, systematic reviews are useful because they define the strengths and limits of the evidence, clarifying whether assertions about the value of the intervention are based on strong evidence from clinical studies. For more information about systematic reviews, see AHRQ expects that CERs will be helpful to health plans, providers, purchasers, government programs, and the health care system as a whole. In addition, AHRQ is committed to presenting information in different formats so that consumers who make decisions about their own and their family s health can benefit from the evidence. Transparency and stakeholder input are essential to the Effective Health Care Program. Please visit the Web site ( to see draft research questions and reports or to join an list to learn about new program products and opportunities for input. Comparative Effectiveness Reviews will be updated regularly. iv

5 Acknowledgments We would like to thank David Jacobs, PhD, or his contribution to conceptualization and methodology of meta-analysis; Alisha Baines and Dianne Smith for programming support; the librarians Jim Beattie, MLIS, Lisa McGuire, MLIS, Judy Stanke, MA, and Delbert Reed, PhD, for their contributions to the literature search; and Marilyn Eells for editing and formatting the report. We would also like to acknowledge Jeremy Holtzman, MD, MS, for his successful efforts in preparing and submitting the original task order proposal. Technical Expert Panel Michael Barry, MD, Massachusetts General Hospital, Boston, Massachusetts D. Andrew Loblaw, MD, Toronto Sunnybrook Regional Cancer Centre, Toronto, Ontario Ian M. Thompson, MD, University of Texas Health Science Center, San Antonio, Texas EPC Program Director Timothy J. Wilt, MD, MPH Professor of Medicine Minneapolis VA Center for Chronic Disease Outcomes Research One Veterans Drive (111-0) Minneapolis, MN Phone: Fax: AHRQ Contacts William Lawrence, MD, MS Task Order Officer Center for Outcomes and Evidence Agency for Healthcare Research and Quality 540 Gaither Road Rockville, MD Phone: Fax: v

6 Contents Executive Summary... 1 Introduction Overview Background Clinically Localized Prostate Cancer Key Question 1. What are the comparative risks, benefits, short- and long-term conditions of therapies for clinically localized prostate cancer? Literature Search and Review Strategy Analysis Treatment Definitions Results Randomized Controlled Trials Demographic and baseline characteristics Efficacy and Adverse Outcomes Watchful waiting versus radical prostatectomy Radical prostatectomy versus external beam radiation therapy Radical prostatectomy combined with neoadjuvant hormonal therapy External beam radiation Hormonal therapy combined with external beam radiation Brachytherapy Adjuvant external beam radiation combined with brachytherapy Adjuvant bicalutamide therapy Vaccine versus nilutamide Androgen deprivation therapy Overall, Disease-Specific Survival, and Biochemical No Evidence of Disease Based on the AUA Clinical Guidelines Database and Other Nonrandomized Evidence Adverse Events Cryosurgery Laparascopic and Robotic Assisted Radical Prostatectomy Laparoscopic vs. Open Retropubic Radical Prostatectomy Transperitoneal vs. Extraperitoneal Laparoscopic Radical Prostatectomy Robotic Assisted Laparoscopic Radical Prostatectomy What is the Impact of Treatments on Overall and Disease Specific Quality of Life? Urinary dysfunction and bother Bowel dysfunction and bother Sexual dysfunction and bother Other outcomes Quality of Life Outcomes in Randomized Controlled Trials Summary vi

7 Key Question 2. How do specific patient characteristics, e.g. age, race/ethnicity, presence or absence of comorbid illness, preferences (e.g. tradeoff of treatmentrelated adverse events vs. potential for disease progression) affect the outcomes of these therapies, overall and differentially? Treatment Decisions According to Patient Factors Treatments for Localized Prostate Cancer by Race Results from observational studies Radical prostatectomy Electron beam radiation therapy Watchful waiting Androgen deprivation Summary Treatments for Localized Prostate Cancer by Age Results from RCTs Results from observational studies Evidence of treatment effect modification by age for survival Evidence of treatment effect modification by age for sexual function Summary Key Question 3. How do provider/hospital characteristics affect outcomes overall and differentially (e.g. geographic region and volume)? Background Methods for the Review Strategy for the Literature Search Inclusion Criteria for the Studies Exclusion Criteria for the Studies Assessment of the Methodological Quality of the Studies Data Extraction Data Synthesis Results Description of Studies Identified by the Search Strategy Association between provider specialty and prostate cancer screening and diagnosis Association between provider specialty and prostate cancer management Association between physician characteristics and patient outcomes How Does Geographic Region Affect Outcomes? Provider availability in geographic regions of the United States Screening and diagnosis of prostate cancer in U.S. regions Treatment options for localized prostate cancer in U.S. regions External Beam Radiation Therapy Brachytherapy Androgen Deprivation Therapy Radical Prostatectomy Mortality from prostate cancer in U.S. regions Summary for regional variations How Does Hospital and Provider Volume Affect Outcomes? Association between hospital volume and patient outcomes vii

8 Surgery related mortality Surgery related morbidity Surgery related quality measures (cancer control, urinary complications, and operative quality) Length of stay and readmission to hospital Association between hospital status and patient outcomes Association between surgeon volume and patient outcomes Summary of association between hospital and provider volume with patient outcomes Key Question 4. How do tumor characteristics, e.g. Gleason score, tumor volume, screen vs. clinically detected tumors, PSA levels, affect the outcomes of these therapies, overall and differentially? Key Question 5. What are the gaps in our knowledge that would allow patients to better understand the comparative risks, benefits, and outcomes of these treatment options for clinically localized prostate cancer, including for those with and without screen detected disease? Summary and Discussion References Abbreviations Tables Table 1. Treatment options for clinically localized prostate cancer Table 2. Prostate cancer treatments used from AUA database Table 3. Description of randomized studies of treatments for localized prostate cancer Table 4. Overall mortality or survival for randomized controlled trials Table 5. Main results. SPCG Table 6. Disease specific mortality or survival for randomized controlled trials Table 7. Biochemical progress/reoccurrence or bned for randomized controlled trials Table 8. Incidence of distant metastatic disease for randomized controlled trials Table 9. Adverse events and toxicity for randomized controlled trials Table 10. PCOS Comparison of 5-year responders to urinary, bowel, and sexual questions according to treatment Table 11. Patient outcomes after cryosurgical treatment of localized prostate cancer Table 12. Outcomes after laparoscopic extraperitoneal, transperitoneal, robot assisted, and radical retropubic prostatectomy (from systematic review by Tooher et al) Table 13. PCOS: Distribution of patient responses at 24 month followup by treatment Table 14. PCOS. Percent comparison of 24 month survey older responders on urinary, bowel, and sexual items Table 15. PCOS: Percentage overall effects of prostate cancer and treatment Table 16. QOL studies of treatments for localized prostate cancer in nonrcts Table 17. QOL studies of treatments for localized prostate cancer in RCTs Table 18. The association between hospital volume and mortality results from individual studies Table 19. Summary of evidence viii

9 Figures Figure 1. Overall survival at time points by treatment Figure 2. Disease specific survival at time points by treatment Figure 3. Biochemical no evidence of disease (bned) at time periods by treatment (all definitions) Figure 4. Bladder complications at time points by treatment Figure 5. Bowel complications at time points by treatment Figure 6. ED complications at time points by treatment Figure 7. Complications by treatment Figure 8. Comparative studies of LRP vs. RRP; functional and oncologic data (from the systematic review by Rassweiler et al) Figure 9. Comparative studies of LRP vs. RRP; operative data (from the systematic review by Rassweiler et al) Figure 10. Comparative studies of transperitoneal vs. extraperitoneal radical prostatectomy; operative data (from the systematic review by Rassweiler et al) Figure 11. Cumulative incidence of death from prostate cancer in the two study groups overall (Panel A) and according to age (Panel B) Figure 12. Conceptual Model Figure 13. Percentage of responses by family physicians and general internists to the question: Do you recommend the following for prostate cancer screening for patients 50 years old and older? Figure 14. Percentage of responses by urologists and primary care physicians to the question whether radical prostatectomy or external beam radiation probably or definitely offers survival benefits for patients with clinically localized prostate cancer Figure 15. Which therapy offers the best survival to patients with clinically localized prostate cancer? Treatment recommendations by urologists and radiation oncologists Figure 16. Percentage of responses by urologists and radiation oncologists to the question whether they believed that three main potentially curative prostate cancer therapies are overused or underused in the United States Figure 17. Regional variations in incidence of prostate cancer (CDC ) and distribution of urologists and radiation oncologists in U.S. regions Figure 18. Proportion of patients with localized prostate cancer treated with external beam therapy, brachytherapy, primary androgen deprivation therapy, radiation, and watchful waiting (%) in U.S. regions (pooled analysis) Figure 19. Utilization of radical prostatectomy, all cause and prostate cancer mortality in patients with prostate cancer treated with radical prostatectomy in U.S. regions (pooled analysis) Figure 20. Length of stay in hospital after radical prostatectomy in U.S. regions (pooled analysis) Figure 21. Mortality from prostate cancer (per 100,000 male populations) in U.S. regions (CDC data ) Figure 22. Mortality from prostate cancer (per 100,000 male population) among different races in U.S. regions (CDC data, ) ix

10 Figure 23. Regional variations in PSA testing and prostate cancer age adjusted mortality in the U.S Figure 24. Regional variations in prostate cancer age adjusted mortality and distribution of urologists and radiation oncologists in the United States (pooled analysis) Figure 25. Difference in surgery related death rate corresponding to an increase by 10 radical prostatectomies performed in hospital Figure 26. Difference in surgery related complication rates corresponding to an increase by 10 radical prostatectomies performed in hospital Figure 27. Difference in surgery related complications rate corresponding to an increase by 10 radical prostatectomies performed in hospital (pooled analysis) Figure 28. Difference in surgery related complications in categories of hospital volume (pooled analysis) Figure 29. Difference in length of stay and readmission rate corresponding to an increase by 10 procedures in annual hospital volume (results from individual studies) Figure 30. Difference in rates of surgery related urinary complications and long term incontinence corresponding to an increase by 1 radical prostatectomy performed by a surgeon and in categories of surgeon volumes above and below the median (the results from individual studies Figure 31. Difference in length of stay after radical prostatectomy by increase in surgeon annual volume (pooled analysis) Figure 32. Survival and cumulative mortality from prostate cancer causes up to 20 years after diagnosis, stratified by age at diagnosis and Gleason score Figure 33. Overall survival at time points by treatment and PSA level Figure 34. Disease specific survival at time points by treatment and PSA level Figure 35. Biochemical No Evidence of Disease (bned) at time points by treatment and PSA level Figure 36. Overall survival at time points by treatment and Gleason score Figure 37. Disease specific survival at time points by treatment and Gleason score Figure 38. Biochemical No Evidence of Disease (bned) at time points by treatment and Gleason score Appendixes Appendix A. Search Strategy Appendix B. List of Excluded Studies Appendix C. Evidence Tables Appendix D. Peer Reviewers Not included in the Draft Report Appendix E. Sample Abstraction Forms Appendix F. Definitions of Outcomes x

11 394 Comparison of Therapies for Clinically Localized 395 Prostate Cancer 396 Executive Summary The Effective Health Care Program was initiated in 2005 to provide valid evidence about the 399 comparative effectiveness of different medical interventions. The object is to help consumers, 400 health care providers, and others in making informed choices among treatment alternatives. 401 Through its Comparative Effectiveness Reviews, the program supports systematic appraisals of 402 existing scientific evidence regarding treatments for high-priority health conditions. It also 403 promotes and generates new scientific evidence by identifying gaps in existing scientific 404 evidence and supporting new research. The program puts special emphasis on translating 405 findings into a variety of useful formats for different stakeholders, including consumers The full report and this summary are available at Background Prostate cancer is the most common nondermatologic cancer in men. In 2006 an estimated 234,460 men were diagnosed with, and 27,350 deaths attributed to, prostate cancer in the United States. Approximately 90 percent of men have disease considered confined to the prostate gland (clinically localized disease). Prostate cancer incidence has increased coinciding with introduction of the Prostate Specific Antigen (PSA) blood test. Disease specific mortality rates have declined and an estimated 1.8 million men are living in the United States who have a diagnosis of prostate cancer. Prostate cancer is primarily a disease of elderly men. About 80 percent of cases are diagnosed in men over age 65, and 70 percent of men who die of prostate cancer do so after age 75. Autopsy studies indicate that the prevalence of subclinical prostate cancer is high at all ages: 30 percent for men ages years and more than 75 percent for men older than 85 years. The lifetime risk of being detected with prostate cancer in the United States has nearly doubled to 20 percent, while the risk of dying of prostate cancer has remained at approximately 3 percent. The primary goal of treatment for prostate cancer is to prevent prostate cancer death and disability while minimizing intervention related complications. The most commonly-employed treatments for clinically localized prostate cancer include watchful waiting (active surveillance) surgery to remove the entire prostate gland and surrounding seminal vesicles and ampulla of vas deferens (radical prostatectomy), external beam radiotherapy (EBRT) and interstitial radiotherapy (brachytherapy), freezing the prostate (cryotherapy), and androgen deprivation (ADT). The preferred treatment is not known, and wide variations among choices and recommendations exist. Options are associated with short and longer term risks including treatment-related death, urinary, bowel, and sexual dysfunction, as well as potential for death and disability due to disease spread. Complication risks apply to all treatment modalities, though frequency and severity may vary. Factors incorporated into the patient decision process include 1

12 cancer eradication, adverse effects, physician recommendations, convenience, and costs. This report summarizes the available evidence comparing the relative effectiveness and safety of treatment options for clinically localized prostate cancer. The report addresses the following questions: Key Question 1. What are the comparative risks, benefits, short- and long-term outcomes of therapies for clinically localized prostate cancer? Evidence regarding comparative treatment effectiveness is limited because: 1) few randomized trials compared the relative effectiveness of major treatment options; 2) wide variation existed in reporting and definitions of outcomes; and 3) there was little reporting of outcomes according to major patient and tumor characteristics. The search strategy identified 17 studies and one pooled analysis of three trials. The majority of randomized controlled trials (RCTs) reported biochemical progression or recurrence as the outcome. The definition of biochemical progression varied but generally was defined as an increase in PSA following treatment with or without local or distant (metastases) progression. Watchful waiting to radical prostatectomy (2 RCTs) Two RCTs compared watchful waiting (WW) to radical prostatectomy (RP). The Scandinavian Prostate Cancer Group trial No. 4, found significantly lower incidences of all-cause deaths, disease-specific deaths, and distant metastases for subjects treated with RP compared to subjects assigned WW after a median followup of 8.2 years. There were 106 (30 percent) all-cause deaths in the WW group compared to 83 (24 percent) deaths in the RP group, (relative risk (RR) = 0.74 [95 percent confidence interval (CI) 0.56 to 0.99]. Deaths attributable to prostate cancer occurred in 10 percent of the RP group and 15 percent in the WW group. Incidence of distant metastases was lower in the RP group compared to WW (14 percent vs. 23 percent, p=0.004). Surgery was associated with greater urinary (49 percent vs. 21 percent) and sexual dysfunction (80 percent vs. 45 percent) compared to WW. The Veterans Administration Cooperative Urological Research Group found no significant differences in overall survival between treatment groups after a median followup of 23 years, though small sample size limited study power. Radical prostatectomy versus radiation therapy (1 RCT) One RCT (n=106) conducted prior to the availability of PSA testing compared RP to external beam radiation therapy (EBR). RP was more effective in preventing treatment failure defined as acid phosphatase elevation on two consecutive followup visits or appearance of bony or parenchymal disease with or without concomitant acid phosphatase elevation. Treatment failure at 5 years of followup occurred in 39 percent for EBR compared to 14 percent in RP (p=0.037). Radical prostatectomy combined with neoadjuvant androgen deprivation therapy (5 RCTs). One small RCT comparing RP alone with RP combined with neoadjuvant ADT found no overall or disease-specific survival benefit with the addition of neoadjuvant ADT after a median of 6 years. The addition of neoadjuvant ADT did not prevent biochemical progression compared to RP alone in any of the four trials. The trial comparing 3 months to 8 months neoadjuvant ADT with RP, reported greater numbers of newly reported AEs in the 8 month group compared to the 3 month group (4.5 versus 2.9, p <0.0001), and higher incidences of hot flashes (87 percent versus 72 percent, p <0.0001). 2

13 External beam radiation (4 RCTs) No EBR treatment regimens were statistically superior to others in reducing death or disease-specific death in any trials. Four RCTs reported biochemical progression. In the trial comparing long arm (66 Gy in 33 fractions) to short arm (52.5 Gy in 20 fractions), the probability of biochemical or clinical progression at 5 years was lower in the long arm group (50 percent versus 56 percent respectively). There was no difference in PSA relapse events between conventional dose EBR (64 Gy in 32 fractions) and hypofractionated EBR (55 Gy in 20 fractions). Brachytherapy (Iridium implant) combined with EBR was superior to EBR alone in reducing biochemical or clinical progression over a median followup of 8.2 years. For clinical stage T2 patients, biochemical or clinical failure events occurred in 26 percent in the combined brachytherapy/ebr group compared to 56.3 percent for the EBR alone group (Hazard Ratio [HR] = 0.37 [95 percent CI 0.16 to 0.85]). High-dose EBR (79.2 Gy that included 3D conformal proton 50.4 Gy with 28.8 Gy proton boost) was more effective than conventional dose (70 Gy that included 19.8 Gy proton boost) in controlling biochemical failure. The percentages of men free from biochemical failure at 5 years was 80 percent [95 percent CI 74.7 to 86.1] in the high dose and 61 percent [95 percent CI 54.6 to 68.3] in the conventional dose group. Effectiveness was evident in low risk (PSA <10ng/ml; stage T2a tumors; or Gleason 6) and higher risk disease. Three RCTs reported on adverse events. Acute combined gastrointestinal (GI) and gentiourinary (GU) toxicity was lower in the long arm (7.0 percent) compared to the short arm (11.4 percent). Late toxicity was similar (3.2 percent in each arm). Both conventional and hypofractionated EBR resulted in increases from baseline for all GI symptoms. There were no significant differences between treatment groups with the exception of rectal bleeding at 2 years after therapy, which had a higher prevalence in the hypofractionated group (42 percent versus 27 percent for conventional group, p <0.05). In the trial comparing high dose to conventional dose regimens, subjects with acute GI or GU symptoms of at least moderate severity were similar, 2 percent versus 1 percent, respectively. For late symptoms, the percents were 1 percent and 2 percent. Androgen Deprivation Therapy combined with External Beam Radiation (3 RCTs) One RCT found conformal radiation therapy combined with 6 months of ADT reduced all-cause mortality and disease specific mortality compared with conformal EBR alone after a median followup of 4.5 years. The HR of EBR alone versus ADT+EBR therapy was 2.07 [95 percent CI 1.02 to 4.20]. Subjects randomized to ADT+EBR had lower PSA failure events compared to EBR alone (21 vs. 46 events; HR = 2.86 [95 percent CI 1.69 to 4.86] after a median followup of 4.5 years. There were significant increases in two events among subjects receiving ADT+EBR compared to EBR alone. Grade 1 and 2 gynecomastia occurred in 18 percent in the ADT+EBR group versus 3 percent for the EBR group (p=0.002). More men in the AST+EBR group were impotent after treatment compared to men treated with EBR alone, 26 versus 21 (p=0.02). There were no other significant differences in adverse events between treatments. One RCT found 6 months of ADT+EBR did not significantly reduce disease-specific mortality compared with conformal EBR alone in and T2b and T2c subjects after a median followup of 5.9 years. Six months combination therapy significantly reduced disease failure (defined as clinical failure at any site, biochemical failure, or death from any cause) compared to EBR alone in subjects with T2c disease (40 vs. 66 events, respectively; HR = 0.47 [95 percent CI 0.32 to 0.69]) but not for T2b subjects (34 vs. 48 events, respectively; HR = 0.68 [95 percent CI 0.44 to 1.06]). 3

14 Brachytherapy (1 RCT) One small trial comparing 125 I to 103 Pd brachytherapy found similar biochemical control at 3 years. There was a trend toward more radiation proctitis, defined as persistent bleeding, in the 125 I subjects. Actuarial estimates were 13 percent for the 125 I group and 8 percent for the 103 Pd group. Adjuvant external beam radiation combined with brachytherapy (1 RCT) One trial compared different doses of supplemental EBR, 20 Gy (n=83) versus 44 Gy (n=76), combined with brachytherapy ( 103 Pd). 1 There were no significant differences in the number of biochemical failure events and freedom from biochemical progression at 3 years. Adjuvant bicalutamide therapy (3 RCTs). At the median followup of 5.4 years there was no difference in total number of deaths between the bicalutamide and placebo groups for subjects receiving RP or radiation therapy. Among the WW subjects, there were significantly more deaths in the bicalutamide group versus placebo. The addition of bicalutamide to standard care did not reduce progression at 5.4 years. To supplement the findings from RCTs and summarize the existing literature on treatment for clinically localized prostate cancer we used the American Urological Association s (AUA) Clinical Guideline Panel for Treatment of Clinically Localized Prostate Cancer database. This work relied on data extracted from 436 articles published between 1991 and April Over 80 percent were case series and only 6 percent were controlled trials. The variability in reporting of results, lack of controls, and likelihood that the database contains results from multiple publications using identical or nearly identical populations limits data interpretation. Overall and disease specific mortality were infrequently reported. When reported, there was extremely wide variation within and between treatments, making overall estimates of outcomes difficult. More than 200 definitions of biochemical no evidence of disease (bned) were reported with results demonstrating extremely wide and overlapping ranges of outcomes at 5 and 10 years within and between treatments. Adverse events were reported, but the definition and severity varied widely. It was not possible to provide precise estimates regarding comparative effectiveness or particular adverse events for each treatment option. Urinary dysfunction appeared to be more common in patients treated with RP and bowel dysfunction than in patients treated with EBR. Sexual dysfunction was common following all treatments. Impotence rates ranged from less than 5 to approximately 60 percent in the few studies reporting on patients undergoing nerve sparing RP. Additional estimates for U.S. population-based adverse events at 5 years following treatment were obtained from a large survey of Medicare eligible men who had undergone treatment for localized prostate cancer. Urinary dysfunction defined as no control or frequent leaking of urine occurred in 14 percent of men undergoing RP and 5 percent undergoing EBRT (adjusted Odds Ratio [OR] = 4.4; 95 percent CI = 2.2, 8.6). The odds of wearing a pad to stay dry was nine fold higher with RP (29 percent) compared to EBRT (4 percent) (adjusted OR = 9.4; (95 percent CI = 4.7, 18.9). Bowel dysfunction was lower in men receiving RP than EBRT, though the only significant difference was related to bowel urgency (18 percent versus 33 percent; adjusted OR 0.56; 95 percent CI = 0.36, 0.87). Erection insufficient for intercourse occurred in approximately 4

15 three-quarters of men regardless of treatment. However, when adjusting for baseline factors, the odds of erectile dysfunction were greater with RP (OR = 2.5; 95 percent CI = 1.6, 3.8). No randomized trials evaluated cryosurgery, and the majority of reports included patients with T3-T4 stages. The sample size in the 15 included studies ranged from 54 to 1,467 patients followed for 3-69 months. Mean baseline PSA levels ranged from 6.5 to 26 ng/ml. The proportion of subjects with poorly differentiated tumors (Gleason 7 or more) varied from 14 percent to more than half of the total sample. Overall or prostate cancer specific survival was not reported. Progression free survival in patients with T1-2 stages ranged from percent. Adverse events were often not reported but, when described, included bladder outlet obstruction (3-21 percent), tissue sloughing (4-15 percent) and impotence ( percent). Outcomes may be biased by patient and provider characteristics. Three systematic reviews estimated the effectiveness and adverse effects of laparoscopic and robotic assisted prostatectomy from 21 nonrandomized trials and case series. Most originated from centers outside of the United States and the median followup was 8 months. Laparoscopic RP had longer operative but lower blood loss and improved wound healing vs. open retropubic radical prostatectomy. Reintervention rates were similar. Eight studies (all nonrandomized reports) were included to evaluate laparoscopic and robotic assisted RP. Total complications, continence rates, positive surgical margins, and operative time were similar to open RP. Median length of hospital stay (1.2 vs. 2.7 days) and median length of catheterization (7 vs. 13 days) were shorter after robotic assisted RP compared with open RP. Health status and quality of life studies published in English-language studies from 2000 to 2006 were included if they met the following criteria: localized prostate cancer; QOL outcomes measured by a standardized survey instrument; study duration of at least 1 year; and randomized controlled trials, or prospective longitudinal survey studies with 100 or more patients per treatment arm. This screening resulted in the inclusion of eight studies. In a large U.S. population based survey of men treated for clinically localized prostate cancer, bother due to dripping or leaking of urine was more than six fold greater in RP treated patients than in EBRT after adjusting for baseline factors. Bother due to bowel dysfunction (4 versus 5 percent) or sexual dysfunction (47 vs. 42 percent) was similar between RP and EBRT. In a subgroup of men ages 70 and over, bother due to urine, bowel, or sexual dysfunction was 5.1, 2.4, and 2.8 fold higher respectively for aggressive (RP/EBRT) vs. conservative (WW/ADT) therapy. Satisfaction with treatment was quite high with less than 5 percent reporting dissatisfaction, unhappy, or terrible about their treatment, though the highest percent was among those treated with RP. Treatment satisfaction was highly correlated with bowel, bladder, and erectile function; general health status; belief that the respondent was free of prostate cancer; and whether cancer treatments did not limit activity or relationships. More than 90 percent said they would make the same treatment decision again regardless of treatment received. Conclusions. There is very limited high-quality data regarding the relative effectiveness and safety of the primary treatment options for clinically localized prostate cancer, especially related to PSA detected disease. One large randomized trial found that radical prostatectomy reduced overall and disease specific mortality at 10 years by 5 percent and risk of metastatic disease in men by 8 percent with primarily nonpsa detected disease. Other randomized trials suggest that 5

16 the addition of ADT to EBRT (but not RP) in high risk patients may reduce the risk of disease progression. Urinary and sexual dysfunction was greater in men treated with radical prostatectomy. Additional randomized controlled trials are needed. All treatment options result in adverse effects (primarily urinary, bowel, and sexual) though the severity and frequency may vary between treatments. Patient satisfaction with therapy is high and associated with several clinically relevant outcome measures. Data from nonrandomized trials are inadequate to reliably assess comparative effectiveness and adverse effects. Key Question 2. How do specific patient characteristics, e.g. age, race/ethnicity, presence or absence of comorbid illness, preferences (e.g. tradeoff of treatmentrelated adverse events vs. potential for disease progression) affect the outcomes of these therapies, overall and differentially? Patient characteristics, particularly age, have an important role in predicting the mortality rate of a patient with localized prostate cancer and the likelihood of urinary, bowel, and sexual dysfunction. However, little is known regarding how patient characteristics modify the effect of treatment. While the list of patient characteristics for which we would like to be able to address this question is quite large (e.g. age, race/ethnicity, comorbid conditions, life expectancy, preferences, etc.), data on the affect of patient characteristics on treatment outcomes is quite rare. Therefore, only patient age and race/ethnicity were assessed. One population-based analysis of treatments for prostate cancer among Medicare beneficiaries indicated that radical treatment of any form increased by 15 percent for men ages 65 to 69 from but decreased in men 75 and older. While RP declined by 6 percent for men ages 70 to 74 and 25 percent for men 75 years or older, brachytherapy use increased especially in men 70 years or older. RP was the most frequently used therapy in men ages 65 to 69, but EBRT was most common in men 75 years and older, followed by brachytherapy. All RCTs were evaluated for any comparative effectiveness results according to patient (age, race, comorbidities) or tumor characteristics (PSA, tumor stage, histologic grade, tumor risk strata). Additionally, studies from the AUA database that had outcomes stratified according to age or race were extracted looking for comparative effectiveness between treatments according to these factors, (rather then absolute effectiveness of an individual treatment). No randomized controlled trials reported head-to-head comparisons of treatment outcomes stratified by race/ethnicity, and most did not provide baseline racial characteristics. While differences may exist in the incidence and morbidity of prostate cancer across racial or ethnic groups, there is little evidence of substantial differences in the effects of treatment by racial or ethnic groups. Reports of modest treatment differences in some nonrandomized reports have not been consistently reported in well-powered studies. Many studies reported poorer outcomes in older patients. However, there was little evidence of a differential effect of treatments based on age. While differences exist in the incidence and morbidity of prostate cancer based on patient age, and there are differences in the treatments offered to men at different age ranges, few studies directly compared the treatment effects of different therapies across age groups. Most RCTs did not have age exclusion criteria. The mean/median age ranged from a low of 63 years for trials of RP to 72 years for trials of EBR. 6

17 Only one RCT provided subgroup analysis according to age. Results suggest that survival benefits of radical prostatectomy when compared with watchful waiting may be limited to men less than 65 years of age. Practice patterns from observational studies show radical prostatectomy is the most common treatment option in younger men with localized prostate cancer. However, in men less than 70 years of age, radiation therapy and watchful waiting become more commonly used. These differences in practice patterns appear to be based more on differences in patient and provider preferences (that are related to age via lifestyle and life expectancy) than particular age independent treatment benefits and side effects. Conclusions. There is little evidence of substantial differences in the effects of treatment by racial or ethnic groups. One randomized trial found that the benefits of RP vs. WW were limited to men less than 65 years of age. Treatment recommendations vary according to patient age/life expectancy. Key Question 3. How do provider/hospital characteristics affect outcomes overall and differentially (e.g. geographic region and volume)? Provider/hospital characteristics including procedure volume, physician specialty, and geographic region were found to affect outcomes. While treatment selection and differential outcomes from treatment for patients with established prostate cancer are the most relevant issues for this review, understanding how provider/hospital characteristics relate to differences in treatment effectiveness requires information regarding how physician experience and screening practices might cause baseline differences in the population of men diagnosed with localized prostate cancer. The apparent effects of provider/hospital characteristics on outcomes can, therefore, be due to both the relative differences in effectiveness by provider/hospital characteristics as well as the relationship between provider/hospital characteristics on factors such as case mix. Screening practices varied by physician specialty, and these differences in screening practices may lead in part to differences in tumor/patient case mix by specialty. Clinicians were more likely to recommend procedures they performed regardless of tumor grades and PSA levels. Regional variation existed in physician availability, ratio of urologists and radiation oncologists per 100,000 adult citizens, screening practice, incidence, mortality, and treatment selection. However the direction of regional variation was not always consistent. Several studies reported geographic variation at the state or U.S. Census region level. However, overall there were many different methods used to report geographic variation, so pooling of results was difficult, and when results were pooled the geographic regions used were quite large. Surgeon volume was not associated with surgery related mortality and positive surgical margins. However, the relative risk of surgery related complications adjusted for patient age, race, and comorbidity and hospital type and location was lower in patients treated by higher volume surgeons. The rate of late urinary complications and incontinence was lower in surgeons that performed more than 10 radical prostatectomies per year. The length of hospital stay was shorter in patients operated by surgeons who performed more surgeries per year. Cost was not associated with surgeon volume. 7

18 Hospital characteristics such as procedure volume, teaching status, and profit status have important effects on outcomes. Hospital volume and teaching status were associated with patient outcomes. Despite different definitions of high and low hospital volumes in individual studies, pooled analysis showed that surgery related mortality and late urinary complications were lower and length of stay was shorter in the hospitals that performed more radical prostatectomies per year. Hospital readmission rates were lower in hospitals with greater volume. Teaching hospitals had a lower rate of surgery related complications and higher scores of operative quality. Several studies found differences in treatment and outcome based on whether the patient was seen in an HMO or fee-for-service organization and whether the patient was a Medicare beneficiary. Variability in the use of androgen deprivation therapy was more attributable to individual differences among urologists than tumor or patient characteristics. Conclusions. Patient outcomes varied in different locations and were associated with provider and hospital volume independent of patient and disease characteristics. Differences in structure of care including clinician availability and hospitals urologic technology may affect the association as well as selective referral patterns when providers with better results attract more patients. Variability in patient outcomes averaged over states and larger regions diminish possible differences in rural/urban areas with higher mortality in nonmetropolitan counties. Screening practices (and therefore the characteristics of patients diagnosed and tumors detected) and treatment choices varied across regions of the United States and did not correlate with clinician availability. Key Question 4. How do tumor characteristics, e.g. Gleason score, tumor volume, screen vs. clinically detected tumors, affect the outcomes of these therapies, overall and differentially? Prior to the advent of widespread PSA testing, most prostate cancers were detected clinically based on abnormalities on the digital rectal examination (DRE) (i.e. induration, nodularity) or incidentally from tissue obtained at surgery for treatment of symptoms due to benign prostatic obstruction. Among men with low PSA values (<4.0 ng/ml) the positive predictive value of an abnormal DRE is poor resulting in frequent prostate biopsies that do not detect cancer. Tumors that are detected in men with low PSA values are usually small volume, found serendipitously and of questionable clinical significance. The vast majority of prostate cancers detected in the United States are asymptomatic, clinically localized, and found on routine PSA testing. The incidence of prostate cancer has risen since the advent of widespread PSA testing, and has been associated with a near doubling of the lifetime risk of prostate cancer from 10 to 18 percent. PSA testing detects more tumors, at an earlier stage, with smaller volume within each stage and at an earlier period in a man s life then clinically detected tumors. The clinical significance, natural history, and comparative effectiveness of treatments in PSA detected cancer is not known. The primary measure of tumor aggressiveness is the Gleason histologic score, though efforts are underway to identify more reliable prognostic factors. A classification currently recommended incorporates PSA levels, histologic score, and tumor volume to identify low, intermediate, and high risk tumors based on their likelihood of progressing with no treatment as well as recurring (or failing to be eradicated) following early intervention. Very little data existed on the comparative effectiveness of treatment options based on this classification. Therefore, we focused on baseline PSA levels and Gleason histologic grade. 8

19 The natural history of men with PSA detected tumors is not known because few men remain untreated for a long followup period. One report assessed 20-year outcomes from a cohort of 767 U.S. men with prostate cancer detected prior to PSA testing and treated with WW. Men with low grade prostate cancers had a minimal risk of dying from prostate cancer (Gleason score 2-4, 7 percent died due to prostate cancer). Men with high grade prostate cancers had a high probability of dying from their disease within 10 years of diagnosis regardless of their age at diagnosis (Gleason score of 8-10, 53 percent died due to prostate cancer). Because PSA increases the time of detection by 5-15 years, it is likely that men with PSA detected tumors will have better 20 year disease specific survival than this cohort. Most RCTs did not exclude participants based on PSA levels or tumor histology and few provided comparative analysis according to these factors. One randomized trial concluded that disease specific mortality at 10 years due to RP compared to WW differed according to age but not baseline PSA level or Gleason score. Most RCTs providing outcomes reported biochemical progression rather than overall or disease specific mortality or development of metastatic disease. Men with Gleason scores 8-10 were more likely to have evidence of biochemical reoccurrence than men with Gleason scores 2-6, regardless of whether treatment was radical prostatectomy alone or combined with NHT. High dose EBRT was more effective in controlling biochemical failure (three successive increases in PSA level) than conventional dose therapy in both low risk disease (PSA <10 ng/ml; stage T2a tumors; or Gleason 6) and higher risk disease. However, when the higher risk subjects were further divided into intermediate risk and high risk groups, the benefit of high dose therapy remained for the intermediate risk (81 percent vs percent, p=0.02) but not for the high risk patients (p=0.80). Based on very limited nonrandomized trial data, disease specific survival appeared similar for subjects treated with EBR compared to RP in men with baseline PSA >10 ng/ml. Men with Gleason scores 8-10 appeared more likely to have biochemical reoccurrence than men with Gleason scores 2-6 regardless of type of treatment. Conclusions. The majority of prostate cancers currently detected in the United States are asymptomatic, clinically localized, and detected by routine PSA testing. PSA testing detects more tumors, at an early stage, with smaller volume within each stage, and at an earlier period in a man s life. Risk classifications incorporating Gleason histologic score, PSA levels, and tumor stage have been demonstrated to be predictive of tumor progression or recurrence regardless of treatment. There is very little information about the relative outcomes according to these parameters. One randomized trial found that the relative effectiveness of RP compared to WW differed according to age but not baseline PSA levels or Gleason histologic score. Key Question 5. What are the gaps in our knowledge that would allow patients to better understand the comparative risks, benefits, and treatment options for clinically localized prostate cancer? Uncertainty about the relative effectiveness and adverse effects of the main treatment options for clinically localized prostate cancer is the major gap in knowledge. This is primarily related to the paucity of randomized trials and the excess reliance on nonrandomized data comparing the most common treatment options: watchful waiting; radical prostatectomy, external beam radiotherapy, 9

20 brachytherapy, and androgen deprivation therapy. Initiation and completion of long-term, adequately powered randomized trials are critically needed. These trials should standardize reporting of key clinically relevant outcomes, including overall, disease specific and metastaticfree survival; biochemical no evidence of disease; adverse effects; and disease specific quality of life/health status. Ideally, relative effectiveness and adverse effects would be stratified according to tumor (PSA, stage, histologic grade) and patient (age, race, comorbidity) characteristics. There are currently several ongoing trials, but their results will not be available for several years. Patients, clinicians, researchers, and funders need to work together to ensure that initiation and completion are successful. The identification of biomarkers to provide reliable estimates about the aggressiveness of prostate cancer and the relative effectiveness of treatments is needed. This would reduce unnecessary interventions while focusing early treatment on patients most likely to benefit. A new generation of patient educational materials is required to provide balanced information about the risks and benefits of treatment options and assist in patient decisionmaking and incorporation of patient centric values (e.g. tumor eradication, impact of adverse effects, anxiety, costs, convenience, etc). There is beginning evidence that structure and process measures are associated with quality of prostate cancer care. Research across nationally representative databases using methods of risk adjustment is needed to clarify geographical differences in patient outcomes. Identification of factors associated with improved outcomes and development of systemwide methods to implement or improve them are needed. Conclusions. Randomized controlled trials comparing the relative effectiveness and adverse effects of the primary treatment options are required and urgently needed. Until then, patient decision aides that provide balanced information based on less reliable data and incorporate patient centric values are needed to assist in the treatment decision process. Systemwide methods to improve structure and process measures of prostate cancer care are needed. 10

21 Introduction Overview Prostate cancer is the most common nondermatologic cancer in men. In 2006 an estimated 234,460 men will be diagnosed with, and 27,350 deaths attributed to, prostate cancer in the United States. Approximately 90 percent of men have disease considered confined to the prostate gland (clinically localized disease). Prostate cancer incidence has increased coinciding with introduction of the PSA blood test. Disease specific mortality rates have declined and an estimated 1.8 million men are living in the United States who have a diagnosis of prostate cancer. 2 The primary goal of treatment for prostate cancer is to prevent prostate cancer death and disability while minimizing intervention related complications. The most commonly employed treatments for clinically localized prostate cancer include active surveillance (watchful waiting [WW]), surgery to remove the entire prostate gland and surrounding seminal vesicles and ampulla of vas deferens (radical prostatectomy), external beam radiation therapy (EBRT) and interstitial radiotherapy (brachytherapy), freezing the prostate (cryotherapy) and androgen deprivation therapy (ADT) (Table 1). The preferred treatment is not known, and wide variation among choices and recommendations exists. Options are associated with short- and long-term risks, including treatment related death, urinary, bowel, and sexual dysfunction, as well as the potential for death and disability due to disease spread. Complication risks apply to all treatment modalities, though frequency and severity may vary. Factors incorporated into patient decisionmaking include cancer eradication, adverse effects, physician recommendations, convenience, and costs. 3 This report was conducted for the Agency for Healthcare Research and Quality (AHRQ) under Section 1013 of the Medicare Modernization Act to address the following questions: 1) What are the comparative risks, benefits, short- and long-term outcomes of the following therapies for clinically localized prostate cancer? a. Radical prostatectomy, including perineal and retropubic approaches, and open vs. laparascopic vs. no lymphadenectomy b. External beam radiation (EBR) therapy, including standard therapy, and therapies designed to decrease exposure to normal tissues such as 3-D conformal radiation therapy and Intensity Modulated Radiation Therapy (IMRT) c. Interstitial brachytherapy d. Cryosurgery e. Expectant management ( watchful waiting ) a. Hormonal therapy as primary therapy, adjuvant or neoadjuvant to other therapies 2) How do specific patient characteristics, e.g. age, race/ethnicity, presence or absence of comorbid illness, preferences (e.g. tradeoff of treatment-related adverse events vs. potential for disease progression) affect the outcomes of these therapies, overall and differentially? 3) How do provider/hospital characteristics affect outcomes overall and differentially (e.g. geographic region and volume)? 4) How do tumor characteristics, e.g. Gleason score, tumor volume, screen vs. clinically detected tumors, PSA levels, affect the outcomes of these therapies, overall and differentially? 11

22 ) What are the gaps in our knowledge that would allow patients to better understand the comparative risks, benefits, and outcomes of these treatment options for clinically localized prostate cancer, including for those with and without screen detected disease? Background Prostate cancer is primarily a disease of elderly men. About 80 percent of cases are diagnosed in men over age 65, and 70 percent of men who die of prostate cancer do so after age Autopsy studies indicate that the prevalence of subclinical prostate cancer is high at all ages: 30 percent for men ages years and more than 75 percent for men older than 85 years. 4 With widespread PSA testing, the lifetime risk of being detected with prostate cancer in the United States has nearly doubled to 20 percent, while the risk of dying of prostate cancer has remained at approximately 3 percent. Incidence of clinically detected disease varies widely by ethnic group and country. The highest rates occur in men of black ethnicity living in the United States and the lowest among men living in China. Geographic variation is likely due to racial, dietary, and environmental factors as well as differences in the intensity of cancer detection efforts. 5 Prostate cancer is more common in men with a first degree relative who has had prostate cancer and mortality may be associated with obesity. 6 High intake of dairy products and calcium have been claimed to increase prostate cancer risk, though any risk is small. 7 It is not known whether testosterone replacement to restore low serum levels increases prostate cancer risk. 8 Bothersome lower urinary tract symptoms due to benign prostatic obstruction common in aging men do not increase prostate cancer risk but result in increased PSA levels. 9 Few proven strategies exist to prevent prostate cancer. Observational studies suggest that diets high in soy, selenium, vitamin E, fruits, and lycopenes (tomato based products) are associated with lower prostate cancer incidence and mortality. However, no association with nutritional interventions and cancer mortality or disease free survival was found in patients with cancer or preinvasive lesions. 10 One randomized trial demonstrated that the 5 alpha reductase inhibitor, finasteride, reduced the risk of incident prostate cancer at 7 years by about six men in The percentage of detected cancers considered high-grade and thus potentially of greater risk for causing morbidity and mortality was higher in the finasteride than the placebo group. Strategies for early detection of prostate cancer include the DRE and PSA testing. The DRE has not been proven to improve morbidity or mortality. Sensitivity, specificity and interexaminer agreement with findings are poor. The DRE requires considerable experience to achieve the tactile sensitivity for detection of early tumors. About 3-6 percent of DREs can be expected to be abnormal and suspicious for cancer (i.e. induration, nodularity) with risk of cancer based substantially on other factors (such as PSA levels). More than half of subjects with 13 DRE detected cancer will have disease that has spread beyond the gland at diagnosis. PSA is a protein found in the blood of all men. Higher levels are noted with prostate cancer, though PSA is elevated due to prostate enlargement, infection, or inflammation. In the United States nearly three-quarters of men over age 50 have had at least one PSA test. PSA testing finds more cancers, shifts detection to tumors of lower stage, smaller volume, and at earlier time periods (stage, lead, and length shift) compared to DRE. Sensitivity and specificity of the PSA 12 12

23 test vary with test thresholds of abnormality as well as other factors such as family history, age, gland size, findings on DRE, and whether prior biopsies (negative) have been obtained. The greatest factor leading to a diagnosis of prostate cancer is aggressive testing. The lifetime risk of prostate cancer diagnosis for men in their 50s in the United States was approximately 10 percent prior to widespread PSA testing. This nearly doubled to 19 percent during with widespread PSA testing. With increasing regular and repeated PSA testing, lower PSA thresholds considered normal, and obtaining a greater numbers of core prostate specimens during biopsy, the lifetime risk of being diagnosed with prostate cancer is likely to exceed 20 percent. An individual's risk of both any prostate cancer and potentially aggressive cancers can be calculated using a risk assessment tool ( and may be useful for decisionmaking. 14 Prostate cancer can cause signs or symptoms due to local (hematuria, urinary obstruction), regional (edema), or metastatic progression (bone pain). However, the vast majority of newly diagnosed prostate cancers in the United States are asymptomatic and detected by elevated levels or rate change of PSA tests. Estimates for the lead time associated with PSA detected tumors range from 5-15 years. Many tumors detected by PSA testing are found serendipitously, may never cause signs or symptoms. Increased detection of localized disease has resulted in more frequent utilization of interventions that are potentially effective but have adverse effects, thus complicating treatment decisionmaking. This may be particularly problematic in men with a life expectancy <10-15 years due to age or comorbid conditions. For example, among men >75 years almost half have received PSA screening, including those in poor health. 15 The likelihood of detecting clinically insignificant disease in men over age 75, based on histopathologic criteria, has been estimated to be 56 percent. 16 Despite widespread testing, there is no conclusive evidence that screening improves morbidity or mortality. Prostate cancer screening is associated with adverse effects, including anxiety related to abnormal results, pain, infection, and bleeding due to diagnostic prostate biopsies and detection/treatment of prostate cancers unlikely to cause health problems While prostate cancer mortality rates have been declining in several countries and some age groups, it is not clear if this is due to increased PSA testing. In the United States, where PSA testing is common, prostate cancer incidence is approximately seven-fold higher than in England and Wales, where testing is not routinely practiced. An annual 4 percent decline in prostate cancer mortality from has occurred in the United States, similar to that reported in the United Kingdom. 20 Age standardized mortality rates in the United States, while lower than peaks observed in the early to mid 1990s, remain at levels reported from the 1970s-80s; eras prior to widespread testing and increased utilization of aggressive interventions. Clinically Localized Prostate Cancer Pretreatment assessment of whether prostate cancer is localized is determined by tumor stage based on clinical examination; primarily the DRE. Prostate cancer believed confined to the prostate gland (T1 or T2, NxM0 or Stage 1-2) is considered clinically localized, forms the foundation for treatment decisionmaking, and is the focus of this report. T1 tumors include those with a normal DRE (typically detected by abnormalities of PSA tests but also diagnosed on 13

24 histopathology from specimens obtained during surgical resection of the prostate for treatment of benign prostate conditions) or T2 (abnormal DRE but no evidence of disease spread beyond the confines of the prostate). Additional tests including x-rays, bone scans, computerized tomography (CT) or magnetic resonance imaging (MRI) are of limited use for these patients and not typically performed. Because of limited sensitivity of pretreatment evaluations, some men with clinically localized disease may have disease that has spread outside of the gland (i.e. pathologically nonlocalized). The risk of pathologically nonlocalized disease is associated with several pretreatment classification factors. Classification includes measures of tumor volume/extent determined by tumor stage, number of biopsy cores with cancer, and extent of cancer in the involved core(s). The primary measure of aggressiveness is the Gleason histologic score. Gleason scores range from Gleason 8-10 tumors are considered the most aggressive, Gleason 7 tumors somewhat less, and Gleason 6 tumors potentially indolent. Pretreatment histology is determined based on a pathologist s examination of several small cores of prostate tissue. Typically, six cores are obtained during a prostate biopsy (sextant biopsy that includes both lobes of the prostate). However, the number has increased over time to 12, 24, and even saturation techniques. This has led to an increasing amount of prostate gland sampled with enhancement in the likelihood of detecting even small volume disease. In addition to the histologic score, the number of biopsy cores that contain prostate cancer and the percent within each core containing tumor is recorded. Risk stratification strategies have incorporated PSA level, biopsy Gleason score, and clinical tumor category because these appear to be associated with risk of PSA failure and prostate cancer specific mortality. Readily available tables have been designed to help men and their doctors predict the definitive pathological stage (determined after surgery, when a pathologist examines the removed prostate for the presence of cancer) and are often used in treatment decisionmaking. 21 Because Gleason score, tumor volume, and PSA levels do not appear to be complete indicators of an individual tumor risk characteristic, efforts are underway to identify more reliable prognostic factors. One risk classification currently recommended is: Low Risk: PSA 10 ng/ml, Gleason score 6, and clinical stage T1c or T2a Intermediate Risk: PSA >10 20 ng/ml, or Gleason score 7, or clinical stage T2b High Risk: PSA >20 ng/ml or Gleason score 8-10 or clinical stage T2c The most common Gleason score is 6 or 7 disease. 22,23 Most men diagnosed with prostate cancer have a PSA between 4 to 10 ng/ml; increasingly between 2.5 and 4.0 ng/ml. Therefore, the average man currently diagnosed with prostate cancer and facing uncertainty about the comparative risks, benefits, and outcomes of treatment decisions is between years of age and has low-risk disease. However, changes in the application of the Gleason scoring has resulted in contemporary uropathologists assigning these grades more commonly than in the past when these tumors were more likely to receive a grade one or two scores lower. 22,23 A resultant improved survival relative to historical controls assigned similar scores has been reported. Additionally, as noted above, as thresholds to define PSA abnormalities are lowered and a greater number of prostate cores obtained at biopsy, an individual diagnosed with prostate cancer in the future is likely to have a lower PSA level, smaller tumor volume, and better long-term natural tumor history than currently. 14

25 Key Question 1: What are the comparative risks, benefits, short- and long-term outcomes of therapies for clinically localized prostate cancer? (Table 1) The main treatment options for clinically localized prostate cancer are identified in the full version of the key question and briefly described in Table 1. Literature Search and Review Strategy To address key questions 1, 2, 4, and 5 we relied on several sources of data. First randomized controlled trials published through mid-november 2006 were identified using the Cochrane Library and the Cochrane Review Group in Prostate Diseases specialized registry. Trials were included if the randomized treatment allocation was based on patients with clinically localized disease and reported clinical outcomes (though some had contamination with patients having T3 or T4 prostate cancer). Two randomized trials were excluded because the treatment assignments were based on pathologic staging, even though patients had clinically localized disease. 24,25 The trial by Thompson and colleagues evaluated radiotherapy adjuvant to radical prostatectomy for pathologically advanced prostate cancer (pt3n0m0) while the Messing trial assessed immediate androgen deprivation therapy compared to observation after radical prostatectomy and pelvic lymphadenectomy in men with node-positive prostate cancer (pt1-2,n+, M0). An additional RCT by Fransson 26 comparing external beam radiotherapy versus deferred therapy was only included in the quality of life data because no further description of deferred therapy or the study protocol were available, despite contacting the senior author. Because our search of RCTs yielded very few trials directly comparing the major treatment options, especially for PSA detected prostate cancer, we used results from a database primarily comprised of nonrandomized studies and previously extracted by our group. These data were used for development of the Guidelines for the Management of Clinically Localized Prostate Cancer: 2006 update (in press). Studies were identified by a series of four PubMed searches conducted by the AUA Guideline Panel between May 2001 and April This search captured articles published from 1991 through April Search terms included MESH major headings of prostate cancer and prostatic neoplasm and were limited to human subjects and English language. Articles were rejected if patients with disease higher than clinical T1 or T2 were included and the outcomes were not stratified by stage (the AUA Guideline Panel Document provides complete details about this search, inclusion criteria, and list of excluded articles). The search identified 13,888 citations that were retrieved and reviewed. Of these, 1,764 (13 percent) met initial inclusion criteria for extraction. Further review yielded 592 articles that were extracted with 436 meeting full inclusion criteria and fully extracted. Among the 436 extracted articles 352 (81 percent) were case series. Only 28 (6 percent) were controlled trials. Due to the initial findings by the AUA Treatment Guideline Panel indicating poor methodologic design quality and reporting of outcomes from nonrandomized trials, our Technical Expert Panel (TEP) members felt that an updated search or inclusion of nonrandomized studies published after April 2004 would be of limited use, biased in evaluating comparative effectiveness, and therefore unanimously recommended against such an update. An updated case series assessing long-term outcomes of men in the United States managed with active surveillance was included because little is known about the natural history of prostate cancer, especially stratified by patient s age and Gleason score. Articles of cryotherapy (12) and laparoscopic or robotic assisted prostatectomy (two systematic reviews of 14 studies) identified 15

26 through Medline, contact with Endocare (a manufacturer of cryotherapy devices) and published between April 2004 and September 2006 were included because little published literature on these emerging technologies was available for the AUA Guideline report. For health status and quality of life studies a literature search was conducted on Ovid MEDLINE, using the search terms prostatic neoplasms, quality of life, QOL, HRQOL, and health status. The search was limited to English language clinical trials published from 2000 to September The resulting list of 494 references was screened to exclude articles that did not meet the following inclusion criteria: localized prostate cancer; QOL outcomes measured by a standardized survey instrument; study duration of at least 1 year; and randomized controlled trials, or prospective, longitudinal survey studies with 100 or more patients per treatment arm. This screening resulted in the inclusion for data extraction of 11 references describing eight studies (Appendix C, Figure C1). CaPSURE (a national disease registry of more than 10,000 men with prostate cancer accrued at 31 sites across the United States) was excluded because the authors noted: the sites were not chosen at random and thus they cannot be assumed to represent a statistically valid sample of U.S. practice patterns only diagnostic and therapeutic interventions ordered or coordinated by participating urologists are recorded 27 Key questions 2 and 4 were addressed by reviewing all RCTs for any comparative effectiveness results according to patient (age, race, comorbidities) or tumor characteristics (PSA, tumor stage, histologic grade, tumor risk strata). Additionally, any study from the AUA database that had outcomes stratified according to age or race was extracted, again looking for comparative effectiveness between treatments according to these factors (rather than absolute effectiveness of an individual treatment). Methods and search strategy for KQ3 are described in that section. Analysis Due to differences in study designs, treatments tested, patient and tumor characteristics, and reporting of outcomes, we did not conduct pooled analysis. Summaries of effectiveness and adverse event outcomes with ranges according to treatment option, tumor characteristics, and group sample size are provided. Results are provided separately for randomized trials and nonrandomized studies. For the AUA database, reviewers divided patients into multiple groups for which the article provided data. For example, disease stage, PSA and Gleason scores, risk categories, and race were used to define groups. Within each group there were sometimes multiple subgroups. It was possible for subgroups to overlap. For the included graphs, each point represents an article/group combination. Some articles may have multiple points for any given time period or treatment. Due to the overlap between subgroups, the most inclusive groups available for each article were selected. When multiple subgroups overlapped, the total patients in the parent group along with the subgroup definitions were used to select which subgroups would be used in the graphs. Gleason score was used for some of the graphs, so when a more inclusive definition was not available and Gleason score was used to define subgroups for an article, we tried to use those subgroups rather than subgroups defined by tumor characteristic or PSA, for example. Our primary goal was to assess the comparative effectiveness and adverse effects of the major treatment options for men with clinically localized prostate cancer overall and according to clinically relevant patient and tumor characteristics including: age (< vs. 65 years), race (white, 16

27 black, Hispanic, other), tumor stage (T1c [PSA detected] vs. other), PSA levels ( 4.0; ; ; 20.0 ng/ml), and Gleason histologic scores (2-4, 5-6, 7, and 8-10). Treatment Definitions From the AUA database seven treatment categories with 19 predefined treatments, and the option of describing others that fit into each category, were identified. Four main categories were selected: Prostatectomy (P), External Beam Radiation (EBR), Brachytherapy (Brachy) and Watchful Waiting. For some graphs, Prostatectomy was broken down into Radical Prostatectomy (RP) and Nerve-Sparing Prostatectomy (NSP); and EBR was divided into EBR and Conformal EBR. If a second treatment, such as hormone therapy, was used, that group was excluded. Results The evidence regarding comparative effectiveness of treatment options for clinically localized prostate cancer is limited in the following ways: 1) very few randomized trials compared the relative effectiveness of major treatment options (Table 2 compares major primary treatment options and reports clinical outcomes for RCTs); 2) among these trials many did not enroll patients with PSA detected tumors, did not report survival outcomes, and/or were of insufficient size or duration to adequately address survival (Table 3 summarizes RCT treatment options and reported outcomes); 3) wide variation existed in reporting of treatment outcomes and great variation in definitions of outcomes including disease recurrence, biochemical progression, and measures of the primary outcome of bowel, bladder, and sexual dysfunction over time; and 4) reporting of outcomes according to major patient and tumor characteristics was limited, including race, age, functional status/life expectancy, and tumor risk strata. Many of these prognostic variables likely vary between and within studies. Therefore, caution must be taken when trying to assess the relative outcomes across treatment for an individual patient. We summarize results for individual treatment options and then describe in greater details findings from randomized controlled trials and additional data. Randomized Controlled Trials Demographic and baseline characteristic (Table 3). The search strategy identified 18 randomized studies 1,28-44 and one pooled analysis of three trials. 45 Descriptions of these studies are summarized in Table 3. Only three studies directly compared the primary treatment options (i.e. radical prostatectomy versus radiation versus watchful waiting) and none were conducted in patients primarily detected by PSA testing. Instead, most randomized trials evaluated variations of a particular treatment approach (e.g. different doses, isotopes, or duration of radiation therapy or addition of androgen deprivation therapy to RP or EBR). A total of 13,964 patients were enrolled to date (some trials had not yet completed randomization). Thirteen trials were conducted in North America (United States or Canada), 1,29,30,32,34-37,39-43 two in Europe, 28,33 one in Japan, 31 and two in Australia and/or New Zealand. 38,44 The three trials of a pooled analysis were conducted in North America, Europe, Israel, Australia, and Mexico. 45 Six studies enrolled subjects with advanced prostate cancer (tumor stage T3 or T4), comprising 24 percent of all subjects. 31,33,37,40,44,45 These subjects were excluded from the baseline demographic, Gleason, efficacy, and adverse events/toxicity analyses. Mean age of the subjects for eight studies reporting was 65.4 years (n=2,945) 1,28,29,34-36,42,43 In studies reporting median age, the ages ranged from 63.6 to 72.5 years. 32,38,39,41 Two 17

28 trials reported on ethnicity, and over 90 percent of subjects in both studies were white. 35,39 The majority of subjects were classified as having T2 tumor stage (75.5 percent versus 23.5 percent T1). 28,31-36,39,41 Only one trial enrolled more T1 subjects than T2. 39 Among the eight studies reporting on Gleason score based on the combined score at randomization, 65.1 percent had a score of 6 or less, 31 percent had a score of 7 or greater, and the score was unknown in 3.9 percent. 1,28,29,32,35,36,39,41 One trial enrolled only subjects with a Gleason score no greater than Six studies reported study eligibility based on level of serum prostate-specific antigen ranging from <15 ng/ml 39 to <40 or 50 ng/ml, 28,32,34,36,41 and most began enrollment prior to widespread use of PSA testing. Close to half of the total subjects were randomized to radical prostatectomy (n=6,550), 28-35,45 31 percent to radiation treatment (n=4,352), 36-41,44,45 20 percent to watchful waiting (n=2,729), 28,29,45 nearly 1 percent to brachytherapy (n=115) 42 and brachytherapy with adjuvant radiation treatment (n=165), 1 and less than 1 percent to either vaccine or nilutamide (n=21 each). 43 Almost 17 percent of subjects assigned radical prostatectomy and 13 percent of subjects assigned radiation treatment received adjuvant or neoadjuvant hormonal therapy. In the pooled three trial analysis, subjects were randomized to either adjuvant bicalutamide (n=4,052) or placebo (4,052) combined with standard care including radical prostatectomy (estimated n=4,445), radiation treatment (estimated n=1,379), or watchful waiting (estimated n=2,313). 45 Efficacy and Adverse Events Outcomes Survival outcomes, biochemical progression or recurrence, distance metastases and adverse events are summarized in the following tables: (overall mortality/survival Tables 4 and 5; disease specific survival Table 6; biochemical progression or reoccurrence Table 7 and Appendix C, Table C1; incidence of distant metastases Table 8; adverse events and toxicity Table 9 and Appendix C, Table C2.) Eight trials reported overall mortality/survival or provided actuarial estimates of overall survival, 28,29,32,36,39,41,43,45 and disease specific PC deaths. 28,32,36,38,39,41,43,44 The majority (n=15) of the RCTs reported biochemical progression or recurrence as an outcome ,36-45 Shown in Appendix C, Table C1, definitions of biochemical progression varied but generally were defined as an increase in PSA following treatment with or without local or distant (metastases) progression. Seven RCTs reported incidence of distant metastases 28,30,32-34,36,43 and seven reported on adverse events or toxicity. 35,36,38,39,41,43,46 1. Watchful waiting versus radical prostatectomy. Two RCTs compared WW to RP. 28,29 Few men had tumors detected by PSA testing. The most recent trial, the Scandinavian Prostate Cancer Group No.4 (SPCG-4), randomized 695 subjects with T1 or T2 localized PC who had a life expectancy of more than 10 years (Table 5). 28 Only 5 percent of enrollees had prostate cancer detected by PSA testing. After a median followup of 8.2 years, all-cause mortality was higher in the WW group compared with the RP group, 106 (30 percent) versus 83 (24 percent), with a relative risk (RR) of 0.74 [95 percent confidence interval (CI) 0.56 to 0.99; p=0.04]. After 5 and 10 years, the absolute risk reductions (ARR) in mortality were 2 percent [95 percent CI -2.2 to 6.2] and 5 percent [95 percent CI -2.8 to 13.0], respectively. There was also a significantly lower risk of disease specific death for subjects treated with RP compared to subjects assigned WW. 28 There were 30 deaths (9.6 percent) attributable to prostate cancer (PC) in the RP group and 50 18

29 deaths (14.9 percent) in the WW group. The RR at 10 years was 0.56 [95 percent CI 0.36 to 0.88, p=0.01] with an ARR of 5.3 percent [95 percent CI -0.3 to 11.0]. Incidence of distant metastases was lower in the RP group compared to WW (14.4 percent vs percent, p=0.004). 28 The cumulative incidences at 5 and 10 years were 8.1 percent and 15.2 percent for the RP group and 9.8 percent and 25.4 percent for the WW group. At 10 years the ARR was 10.2 percent [95 percent CI 3.1 to 17.2] and the RR was 0.60 [95 percent CI 0.42 to 0.86]. The Veterans Administration Cooperative Urological Research Group (VACURG) study randomized 142 subjects with stage I or II localized PC recruited from Veterans Administration Hospitals between 1967 and After a median followup of 23 years the median overall survival was 10.6 years for the RP group and 8 years for the WW group. Results were not statistically significantly different, but this study was underpowered. Two ongoing trials are evaluating primary treatment options in men with primarily PSA detected clinically localized prostate cancer. The U.S. based VA/NCI/AHRQ funded CSP#407: Prostate cancer Intervention Versus Observation Trial (PIVOT) is comparing RP versus WW in 731 men 47 and completed recruitment. Results are due after The Prostate Testing and Cancer Treatment study, based in the United Kingdom, is comparing surgery (radical prostatectomy), radiotherapy (radical conformal) and active monitoring (monitoring with regular check-ups. A combined U.K., U.S., and Canadian trial in its pilot phase will compare active surveillance versus immediate treatment (patient's choice). 2. Radical prostatectomy versus external beam radiation therapy. One small (N=106) RCT evaluated effectiveness of RP compared to EBRT in subjects with clinical stage A2 or B (T1/T2) (non-psa detected disease) prostate cancer and normal serum prostatic acid phosphatase levels. 30 RP was more effective in controlling disease progression defined as acid phosphatase elevation on two consecutive followups or by appearance of bony or parenchymal disease with or without concomitant acid phosphatase elevation. Treatment failure at 5 years was an estimated 39 percent for the RT group compared to 14 percent in the RP group (p=0.037). Two events (positive bone scans) occurred in the RP group compared to 14 (11 positive bone scans, one pulmonary, lymph node, parenchymal metastases each) in the RT group Radical prostatectomy combined with neoadjuvant hormonal therapy. One small RCT compared RP alone (n=101) versus RP combined with neoadjuvant hormonal therapy (NHT) (n=112). 32 NHT consisted of 300 mg of cyproterone acetate daily for 3 months prior to surgery. After a median followup of 6 years, there was no benefit with the addition of NHT. Overall survival at 5 years was 88.4 percent [95 percent CI 80.6 to 96.3] and 93.9 percent [95 percent CI 88.6 to 99.1] for the NHT and RP alone groups, respectively (p=0.38). There were five total deaths in the RP group and eight in the NHT group. NHT did not reduce disease specific deaths compared to RP alone (1 vs. 0). 32 Four RCTs reported biochemical progression outcomes All defined progression based on PSA rises, although two trials included local recurrence, distant metastases, 31 or death due to prostate cancer. 32.NHT did not prevent biochemical progression or recurrence, distant metastases or death due to prostate cancer more effectively than RP alone. 19

30 A Japanese study reported 11 (16 percent) clinical relapse events in the NHT combined with RP group versus 9 (14.3 percent) in the RP alone group for stage A2 and B subjects. 31 Only one event was reported for stage A2 subjects. 31 At a median followup of 6 years, 34 percent and 38 percent of RP subjects and NHT subjects had biochemical recurrence (HR=0.98 NHT vs. RP alone, [95 percent CI 0.61 to 1.56], p=0.92). 32 A Gleason score of 8 to 10 at biopsy was a significant predictor of recurrence (HR=2.82 score 8 to 10 vs. 2 to 6, [95 percent CI 1.52 to 5.22], p=0.001), regardless of type of treatment. One small trial evaluated biochemical recurrence at 5 years followup, defined as a PSA value >0.4 ng/ml, based on biopsy and surgical specimen Gleason scores. 34 For Gleason score of 8 to 10, 8/14 of RP subjects had biochemical failure compared to 13/15 of NHT combined with RP subjects (p=0.173) Approximately 65 percent in the NHT group and 68 percent in the RP group had evidence of bned (p=0.663). Three RCTs reported distant progression The addition of NHT did not reduce the risk of developing distant metastases. A trial comparing 3 months to 8 months of NHT focused on adverse events of treatment rather then effectiveness. There were no fatal AEs and no difference between the groups in the causality and severity of AEs. 35 Within the 8 month group there were significantly greater numbers of newly reported AEs compared to the 3 month group (4.5 versus 2.9, p<0.0001), defined as the first occurrence of an event regardless of the ongoing status, and higher incidences of hot flashes (87 percent versus 72 percent, p<0.0001). 4. External beam radiation. As noted above, only one small trial compared EBRT to RP. Despite the findings that RP was superior to EBRT in preventing disease progression, the study was small, conducted prior to PSA testing and prior to refinements in both surgical and radiation therapy. Therefore, the results may not be applicable to current practice. 30 The majority of RCTs have evaluated different doses/duration of EBRT or use in combination with adjuvant ADT. None have directly evaluated EBRT with WW. Recent modifications to EBRT include high dose conformal EBRT which uses three dimensional radiotherapy planning systems and methods to match radiation treatment to prostate and tumor volumes as well as IMRT that uses multiple beams of EBRT to deliver radiation to a small area while attempting to avoid healthy tissue. These modifications have not been directly compared with other primary options. Variations in EBRT regimens have not demonstrated that any provide differences in overall or disease specific survival. Most RCTs are of insufficient size or duration to adequately assess survival or metastases and focus on adverse events or biochemical outcomes. Compared to conventional radiotherapy, high-dose conformal EBRT decreased the rate of PSA failure without increasing acute or late serious urinary or rectal complications. 39,48 Two RCTs compared two dose fractionation schemes, a long arm (66 Gy in 33 fractions over 45 days) and a short arm (52.5 Gy in 20 fractions over 28 days), and reported overall survival. 36,39 The first study, a multicenter Canadian trial, randomized 936 men with early-stage PC. At the median followup of 5.7 years, neither approach was superior in improving overall survival. Overall survival at 5 years was estimated at 85.2 percent for the long arm and 87.6 percent in the short arm. A multicenter American trial compared conventional dose EBR (64 Gy in 32 fractions) to hypofractionated EBR (55 Gy in 20 fractions) in 393 men with stage T1b through T2p PC. 37 There was no difference in overall survival between groups at 5 years (97 percent conventional 20

31 dose versus 96 percent high dose, p=0.8). No EBR treatment regimens were statistically superior in reducing prostate cancer specific deaths in the three trials reporting. 36,38,39 Incidences of reported disease specific deaths were low, ranging from 0 to 2 percent. In the study comparing long arm EBR to short arm, there were three (<1 percent) prostate cancer deaths in the long arm group and none in the short arm group. 36 One (2 percent) prostate cancer death was reported in the conventional arm compared to zero in the hypofractionated arm. 38 The conventional dose group had two deaths due to prostate cancer, but there were none in the high dose group. 39 Four RCTs reported biochemical progression Two trials used a composite definition of progression, including death due to prostate cancer and clinical failure 36,37 and all used increases in serum PSA. In the Lukka trial, the probability of biochemical or clinical progression at 5 years favored the long arm, 53 percent versus 60 percent for the short arm, yielding an ARR of -7 percent [95 percent to -1.4]. 36 There were 263 (56.4 percent) and 236 (50.2 percent) events for the short and long arm groups, respectively. There was no difference in PSA relapse events between conventional EBR and hypofractionated arm. 38 Brachytherapy (Iridium implant) combined with EBR was superior to EBR alone in reducing biochemical or clinical progression over a median followup of 8.2 years. For clinical stage T2 patients, biochemical or clinical failure events occurred in 26 percent in the combined brachytherapy/ebr group compared to 56.3 percent for the EBR alone group (HR=0.37 [95 percent CI 0.16 to 0.85]). High dose EBR (79.2 Gy that included 3D conformal proton 50.4 Gy with 28.8 Gy proton boost) was more effective in preventing biochemical failure than conventional dose (70 Gy that included 19.8 Gy proton boost). 39 Biochemical failure at 5 years occurred in 80.4 percent [95 percent CI 74.7 to 86.1] in the high dose group and 61.4 percent [95 percent CI 54.6 to 68.3] in the conventional-dose group (p<0.001). Superior effectiveness was reported in both low risk disease (n=227, PSA <10 ng/ml; stage T2a tumors; or Gleason 6) and high risk disease. In the low risk subgroup, the percentages were 80.5 percent for the high dose group and 60.1 percent in the conventional dose group (p<0.001). For the high risk subjects, the percentages were 79.5 percent and 63.4 percent (p=0.03) for the respective groups. However, when the higher risk subjects were further divided into intermediate risk and high risk groups, the benefit of high dose therapy remained for the intermediate risk (81 percent vs percent, p=0.02) but not for the 33 high risk patients (p=0.80). One trial found slightly more distant failure events in the short arm (ten events, 2 percent) compared to the long arm (four events, 1 percent) at the median followup of 5.4 years. 36 Three RCTs reported on toxicity/adverse events associated with EBR. 36,38,39 The trial by Lukka found acute ( 5 months) combined GI and GU toxicity lower in the long arm (7.0 percent) compared to the short arm (11.4 percent), a difference of -4.4 percent [95 percent CI -8.1 to - 0.6]. 36 Late toxicity was similar in both arms (3.2 percent each). Both conventional and hypofractionated EBR resulted in increases from baseline for all GI symptoms and for five symptoms characterizing GU symptoms 1 month after completion of therapy. 38 For GI symptoms, increases in four of the six symptoms (rectal pain, mucus discharge, urgency of defecation, and rectal bleeding) remained 2 years after EBR compared to baseline. There were no differences between treatment groups with the exception of rectal bleeding at 2 years after therapy, which had a higher prevalence in the hypofractionated group (42 percent versus 27 percent for conventional group, p<0.05). Fewer subjects had urinary frequency equal to or more 21

32 than every 3 or 4 hours compared to baseline (70 percent versus 81 percent, p<0.05). Only 25 of the 120 subjects completed the sexual function questionnaire (European Organization for Research and Treatment of Cancer). Nine (36 percent) were impotent at baseline. One month after treatment the number of subjects reporting erectile dysfunction increased to 13 (52 percent). Two years after EBR treatment erectile dysfunction was reported by 9 of 17 subjects. In the trial comparing high dose to conventional dose regimens, subjects with acute severe GI or GU symptoms (RTOG 3) were similar, 2 percent versus 1 percent. 39 For late severe GI or GU symptoms (RTOG 3), the percents were 1 percent and 2 percent for high dose and conventional dose groups. For acute GI symptoms, 57 percent of high dose subjects experienced grade 2 GI morbidity compared to 41 percent of conventional dose subjects (p=0.004). The difference remained significant for late grade 2 GI morbidity, although proportions decreased (17 percent high dose versus 8 percent conventional dose, p=0.005). 5. Hormonal therapy combined with external beam radiation. The use of short duration (3-8 months) hormonal (ADT)) neoadjuvant to external beam radiation (combined ADT + EBRT) has been evaluated in several RCTs. The results generally indicate that combined therapy is more effective than EBRT alone in selected patients, though the studies are relatively small, of short duration, and with few overall or prostate cancer specific deaths. One RCT (n= 206) compared conformal radiation therapy combined with 6 months of ADT with EBRT alone in localized prostate cancer with PSA levels of <10 ng/ml. 41 ADT consisted of an LHRH agonist (leuprolide acetate) or goserelin and a nonsteroidal anti-androgen (flutamide). Combined EBRT +ADT reduced all-cause mortality versus EBRT alone: 12 deaths versus 23 deaths for EBRT alone. The hazard ratio (HR, EBRT alone versus ADT+EBR therapy) was 2.07 [95 percent CI 1.02 to 4.20, p=0.04]. Overall survival at 5 years was 88 percent [95 percent CI 80 to 95] for the combined group compared to 78 percent [95 percent CI 68 to 88] for EBRT alone. The addition of ADT also reduced disease specific mortality compared to EBRT alone (zero versus six deaths (5.8 percent, p=0.02). 41 A RCT conducted in Australia and New Zealand reported fewer prostate cancer deaths with 6 months of ADT added to EBR versus EBR alone (8 deaths vs. 17 deaths, respectively), though the confidence intervals were wide and the results not statistically different in men with T2 disease after a median followup of 5.9 years. 44 The HR for T2b subjects was 0.22 [95 percent CI 0.03 to 1.88] and 0.57 [95 percent CI 0.22 to 1.44] for T2C subjects. Two RCTs reported biochemical progression outcomes based on rising PSA levels. 40,41 One evaluated different durations of ADT (3 months vs. 8 months) combined EBR. 40 The overall median followup, which included subjects with stage T3 disease, was 3.7 years. For the low risk subjects (n=92, PSA <10 ng/ml; stage T1c to T2a tumors; Gleason 6), the actuarial estimate of freedom from biochemical failure was 61 percent for the 3 month group compared to 72 percent for the 8 month group. In the D Amico trial, subjects randomized to combined therapy had lower PSA failure events compared to subjects randomized to EBR alone (21 vs. 46 events, HR=2.86 [95 percent CI 1.69 to 4.86], p<0.001) after a median followup of 4.5 years. 41 Survival without salvage ADT was also higher in the combination group versus the EBRT alone group (p=0.002). Denham found combination therapy reduced clinical failure at any site, biochemical failure, or death from any cause, in subjects with T2c disease but not for T2b subjects. 44 There were 66 events in the EBR alone group compared to 40 in the ADT+EBR group in T2c subgroup, with a 22

33 HR of 0.47 [95 percent CI 0.32 to 0.69] favoring the ADT+EBR group. In the T2b subgroup, there were 48 and 34 events with a HR of 0.68 [95 percent CI 0.44 to 1.06]. One study found significant increases in two events among subjects receiving combined therapy compared to EBRT alone. 41 There were 18 (18.4 percent) incidences of grade 1 and 2 gynecomastia in the ADT+EBRT group versus 3 (2.9 percent) incidences for the EBR group (p=0.002). More men in the ADT+EBRT group who were potent at baseline became impotent after treatment compared to men treated with EBR alone, 26 versus 21 (p=0.02). There were no other significant differences in toxicity between the treatment groups. 6. Brachytherapy. Brachytherapy delivers radiation with small radioactive pellets implanted into the prostate gland under general or spinal anesthesia. These needles deliver the pellets, which can be left either permanently or temporarily (high dose), and give off radiation at a low dose over several weeks or months. Brachytherapy is used as a primary therapy in combination with EBRT boost or with ADT. Brachytherapy is increasingly used for selected men with low to moderate risk prostate cancers despite no survival data from randomized trials. A recent review of case series data concluded that efficacy and some long-term adverse effects were comparable to surgery or EBRT in well-selected patients. 49 Preliminary results of RCT comparing different isotopes or adjuvant therapies 42,46 and other underpowered studies have been published 37,50 but preclude conclusions regarding the relative efficacy versus other treatments, as well as conclusion regarding optimal forms of brachytherapy. Wallner (N=115), compared 125 I (144 Gy) to 103 Pd (125 Gy). They found similar biochemical control for both treatments at 3 years. 42 Actuarial estimate of freedom from biochemical progression, defined as PSA 0.5 ng/ml at last followup, was 89 percent for the 125 I group versus 91 percent for 103 Pd group (p=0.76). A trend toward more radiation proctitis, defined as persistent bleeding, was found in the 125 I subjects (p=0.21). Actuarial estimates were 13 percent for the 125 I group and 8 percent for the 103 Pd group. 7. Adjuvant external beam radiation combined with brachytherapy. Use of EBRT adjuvant to brachytherapy has been proposed to be more effective than brachytherapy alone, particularly for patients considered at higher risk for disease recurrence. One trial compared different doses of supplemental EBRT, 20 Gy (n=83) versus 44 Gy (n=76), combined with brachytherapy ( 103 Pd). 1 There were no significant differences between EBRT groups in the number of biochemical failure events and the actuarial estimates of freedom from biochemical progression at 3 years. The estimated freedom from biochemical failure was 83 percent in the 20 Gy group versus 88 percent in the 44 Gy group (p=0.64). The estimated percents of freedom from biochemical failure in patients with a pre-treatment PSA <10 ng/ml (n=112) were 84 percent and 94 percent for the 20 and 44 Gy groups, respectively (p=0.16). For the 47 subjects with a pretreatment PSA >10 ng/ml, the percents were 82 percent for the 20 Gy group and 72 percent for the 44 Gy group (p=0.38). 8. Adjuvant bicalutamide therapy. The bicalutamide Early Prostate Cancer Program was a pooled analysis of three international RCTs assessing the effectiveness of adjuvant bicalutamide combined with standard care (RP, EBRT, or WW) compared to placebo and standard care. 45 The trials enrolled subjects with both clinically localized (two-thirds of all subjects, n=5,426) and 23

34 locally advanced prostate cancer. The majority of the subjects received RP (55 percent) followed by WW (28.5 percent) and EBRT (17 percent). At the median followup period of 5.4 years, there was no difference in total number of deaths between the bicalutamide and placebo groups for subjects receiving RP or radiation therapy (3,799). There were 187 (9.8 percent) and 182 (9.6 percent) deaths for the respective groups with a HR of 1.01 [95 percent CI 0.82 to 1.23, p=0.97]. Among the WW subjects with clinically localized disease (n=1,627), there were significantly more deaths in the bicalutamide group (196, 25.2 percent) versus placebo (174, 20.5 percent) with a HR of 1.23 [95 percent CI 1.00 to 1.50, p=0.05]. Progression was defined as death from any cause or objective progression confirmed by bone scan, computerized tomography/ultrasound/mri, or histological evidence of distant metastases. 45 Among subjects with localized disease (stage T1/T2), the addition of bicalutamide to standard care did not significantly reduce objective progression at the median followup period of 5.4 years. Among subjects who received RP (n=2,734), progression events in the bicalutamide group was 8.4 percent versus 8.8 percent for placebo (HR=0.93 [95 percent CI 0.72 to 1.20], p=0.57). Progression events for radiation therapy subjects (n=1,065) were 21.2 percent and 24.3 percent for the bicalutamide and placebo groups respectively (HR=0.80 [95 percent CI 0.62 to 1.03], p=0.09). 9. Vaccine versus nilutamide. One small RCT (N=42) compared a vaccine designed to enhance T-cell responses and anti-tumor activity to the antiandrogen, nilutamide, in men with nonmetastatic hormone refractory PC. 43 Overall followup times were not reported. There were three deaths in vaccine group compared to seven in the nilutamide group. There were four reported prostate cancer deaths in the nilutamide group, including two deaths among subjects who had vaccine added. 43 Among the vaccine subjects, there was one prostate cancer death. Treatment failure was a composite outcome, defined as PSA progression, development of secondary malignancies or toxicity, and were either removed from study or crossed over to the other arm as determined by study protocol. 43 Median time to treatment failure was 9.9 months for the vaccine group compared to 7.6 months for the nilutamide arm. There were twice as many progressive disease events (metastases on bone scans) for subjects initially treated with vaccine (14 total, five events after crossover to nilutamide) than the subjects initially treated with nilutamide (seven total, one event after crossover to vaccine). 43 Three subjects in the nilutamide arm (14.3 percent) were removed from the study due to grade 3 toxicities 43 and 38 percent in the vaccine arm experienced pain at the injection site. Both arms reported grade 2 and 3 toxicities. Dyspnea, fatigue, and hot flashes were reported for nilutamide patients. Toxicities in the vaccine group included arthralgia, fatigue, dyspnea, and cardiac ischemia (3.4 percent). The vaccine regimen also included injections of adesleukin (IL-2). Grade 2 and 3 toxicities associated with IL-2 included fever, arthralgias, hyperglycemia (20.7 percent grade 2, 6.9 percent grade 3), lymphopenia (13.8 percent grade 2, 6.9 percent grade 3), dehydration/anorexia, and diarrhea. 10. Androgen deprivation therapy (ADT): Use of continuous or intermittent long-term ADT as primary therapy has increased. A previous AHRQ evidence report 51 examined randomized trials of different methods of ADT for advanced prostate cancer. Survival after treatment with a 24

35 Luteinizing Hormone-Releasing Hormone (LHRH) agonist was equivalent to survival after orchiectomy. The available LHRH agonists were equally effective and no LHRH agonist was superior to others when adverse effects are considered. There was a trend toward lower survival with use of a nonsteroidal antiandrogen compared to orchiectomy or LHRH agonists HR=1.13; 95 percent CI 0.92 to 1.39). Individual patient level meta-analysis suggested an improvement in survival of about 2 percent at 5 years (median survival benefit of 2-3 months) of combined 52 androgen blockade compared to monotherapy. Primary ADT can last for 20 years or more in men with localized disease, but no randomized trials have compared the relative effectiveness of ADT in localized disease. In addition to treatment costs, adverse effects of ADT include erectile dysfunction, loss of libido, breast tenderness, hot flashes, depression and mood changes, memory difficulties, fatigue, muscle and bone loss, and fractures. 53 Long-term costs, sequelae, and/or use of additional medications to mitigate these adverse effects, such as androgen deprivation syndrome and osteoporosis (e.g. bisphosponates for bone loss, anxiolytics, and antidepressants) are issues of greater importance compared to treatment of advanced disease Overall, Disease-Specific Survival, and Biochemical No Evidence of Disease Based on the AUA Clinical Guidelines Database and Other Nonrandomized Evidence Data from the AUA Clinical Guidelines Database were used to assess overall and disease specific survival and bned at 5, 10, 15, and 20 years according to treatment options and size of reported patient group. We assessed this among cohorts regardless of risk strata and then separately according to Gleason score when available. Findings are limited, as noted previously, because studies frequently did not report certain outcomes, may have provided multiple publications of identical or nearly identical cohorts but did not clearly differentiate these reports, used various definitions, used different followup times, and/or did not provide standard classification of patient/tumor risk characteristics. The vast majority of data comes from uncontrolled case series. For overall, disease specific survival and bned, there were very wide variations in outcome estimates resulting in considerable overlap within and between treatments (e.g. at 10 years overall survival for any of the therapies ranged from approximately 15 percent to percent; disease specific survival ranged from approximately 40 percent to nearly 100 percent). Variation in outcomes within and between treatments could be related to provider, patient (age, race, comorbidities) and/or tumor (stage, PSA, histologic grade) factors. Treatment related outcomes according to provider and patient and tumor factors are described in Key Questions 2, 3, and 4. Given the limitations of the results and the quality of the studies, it is not possible to accurately estimate the relative effectiveness of options beyond that available from the few randomized trials. Figures describe the range of outcomes reported. (Figures 1-3). Overall and disease specific survival at 10 years and beyond was most commonly reported in patients treated with EBRT and rarely reported with brachytherapy. bned was much more commonly reported than overall or disease specific survival. There were more than 200 definitions of bned. Our figures included all definitions of bned and likely account for some variability in percent bned within and between treatments. While bned has not been clearly demonstrated to correlate with survival, additional treatments are often based on followup PSA levels. 25

36 Adverse Events Adverse effects due to treatments based on the few reported randomized trials have been noted above. Several studies have used national data bases to assess 30-day mortality following radical prostatectomy. Based on a 20 percent random sample from of male Medicare beneficiaries, Lu-Yao and colleagues found that approximately 1 percent of men between the ages of died within 30 days of RP. The risk of mortality and morbidity increased for older men and exceeded 4 percent for men ages 80 or greater. 54 A more recent analysis of Medicare recipients ages 65 years or older, indicated that from the 30-day mortality following RP was approximately 0.5 percent. 55 Major treatment related morbidity was common in these older men with cardiopulmonary complications occurring in 4-10 percent and need for surgical repairs in 0.5 to 1 percent. Thirtyday readmissions per 1,000 operations declined from about per 1,000 in the late 1980s to about 5 per 1,000 in the mid 1990s. 55 Similar results were found using a national sample of male veterans receiving RP at VA medical centers. 51 (A more detailed analysis of provider and hospital factors on mortality and morbidity is described in Key Question 3.) The Prostate Cancer Outcomes Study (PCOS) 56 was begun in 1994 to prospectively collect individual level data from a population-based cohort of men with newly diagnosed prostate carcinoma. The PCOS is based on an existing tumor registry system, the National Cancer Institute s Surveillance, Epidemiology, and End Results (SEER) program that provides information on cancer incidence and survival for the United States. PCOS assessed the effects of cancer treatments, including RP, EBRT, and ADT on health-related quality of life outcomes. With regard to adverse effects, PCOS focused on bladder, bowel, and sexual function. PCOS was initiated prior to widespread PSA testing. Some baseline characteristics and findings may differ from patients currently diagnosed with prostate cancer. Survey results indicate that sexual dysfunction was commonly associated with all treatments, urinary incontinence was more common after RP, and bowel dysfunction more common after EBRT. Sexual dysfunction was the most common adverse effect related to prostate cancer treatments. Approximately half the individuals receiving either RP or EBRT had no or little interest in, as well as no sexual activity. Three-quarters of men had erections that were insufficient for intercourse. Inability to achieve an erection was also commonly reported by men treated with ADT (86 percent) though one-third of men treated with WW reported inability to achieve any erections. At 24 month followup, urinary leakage occurring at least daily was three to five times more commonly reported in men treated with RP than with other options; reported in 7, 11, 12, and 35 percent of men who were treated with WW, ADT, EBRT/brachytherapy, and RP respectively. Five years after diagnosis, 14.4 percent of PCOS subjects who underwent RP versus 4.9 percent who were treated with some form of EBRT reported that they had no control or frequently leaked urine OR=4.4, 95 percent CI= ). Twenty-nine percent versus 4 percent of subjects reported that they wore pads to stay dry. (Table 10). 26

37 Bowel dysfunction was more commonly noted after EBRT. At five years significant differences between the two treatments after adjustment for baseline factors and treatment propensity included bowel urgency (33.4 percent vs percent) and painful hemorrhoids (15.7 percent vs percent). Daily bowel urgency was reported by about 3 percent of individuals treated with ADT or radiation therapy but occurred in less than 1 percent of men receiving either WW or RP. Both types of primary ADT (orchiectomy or LHRH agonist) 57 had a large adverse impact on sexual interest, activity, and ability to maintain an erection, though there were no significant differences between options. About 30 percent of individuals reported that they had no sexual interest before treatment. This increased to 64 percent and 58 percent at 5-year followup of orchiectomy or LHRH agonist. About 69 percent of men who were potent before treatment were impotent after, regardless of treatment. Only 10 and 13 percent of subjects treated with orchiectomy or LHRH respectively were able to maintain an erection sufficient for sexual intercourse. Breast swelling after treatment was reported by 24.9 percent in LHRH patients compared with 9.7 percent in orchiectomy patients. Hot flashes were similar in both treatment groups (56.5 percent versus 67.9 percent). PCOS results are consistent with findings from the randomized trials evaluating RP, WW, ADT, and EBRT. Shahinian and colleagues used SEER Medicare data to evaluate the risk of androgen deprivation syndrome in a cohort of 50,613 men receiving ADT for incident prostate cancer. 53 Of men surviving at least 5 years after diagnosis, 31.3 percent of those receiving ADT developed at least one depressive, cognitive, or constitutional diagnosis compared with 23.7 percent who did not receive ADT. However, the risk differences were substantially reduced when adjusting for age, comorbid conditions, and more advanced prostate cancer. The risk of fracture after ADT appears to be increased. Shahinian used the SEER Medicare linked database to assess fracture risk in 50,163 men who had a diagnosis of prostate cancer from Of men surviving at least 5 years after diagnosis, 19.4 percent of those who received ADT had a fracture, compared with 12.6 percent of those not receiving ADT. After controlling for patient and tumor characteristics, there was a statistically significant relation between the numbers of doses of gonadotropin-releasing hormone received during the 12 months after diagnosis and the subsequent risk of fracture. A previous AHRQ report 51 examined randomized trials of different methods of androgen suppression in the treatment of advanced prostatic cancer. No LHRH agonist was found to be superior to others when adverse effects were considered. Adverse effects leading to withdrawal from therapy and drug costs were greater with combination therapy (LHRH agonist or orchiectomy plus antiandrogen) than with monotherapy. The AUA Guideline Panel had 24 predefined complications. These included bladder complications (seven), bowel (six), erectile dysfunction (one), deep venous thromboses and others (nine). Authors infrequently used the same definition for a given complication, often did not report outcomes during the same time period, varied in whether they reported on all subjects, only those with or without dysfunction at baseline, and how the outcome was assessed. For example, we identified 112 different definitions of incontinence, erectile dysfunction (79), 27

38 bladder (203), bowel (87), and 336 definitions of other complications. The vast majority of definitions were only used once. A report for adverse events was included if 1) it provided one of the predefined complications or 2) additional definitions of bowel, bladder, erectile dysfunction were used in at least three reports, and 3) the percent of subjects with complications was provided (or the ability to calculate this). At any time point the number of reports providing definitions and the number of reports that we could use were bowel (57/5); bladder (79/19), and ED (44/13) (Appendix C, Table C3). A series of figures (Figures 4-7) illustrate the major complications according to treatment, time period and group sample size. Results were not assessed according to baseline patient or tumor characteristics. Based on the AUA database as well as surveys or administrative datasets of men treated for prostate cancer (PCOS), 56 Medicare, and VA) 54,59 described above, we make the following general conclusions. All treatments can cause bladder, bowel, and sexual dysfunction. Frequency and severity of these adverse events may vary by treatment, length of followup, reporting method, definition of adverse event, patient baseline characteristics, and provider/facility factors (Question 3). Bladder complications including hematuria, incontinence, cystitis, and urethral stricture were more commonly reported in patients treated with surgery and persisted beyond 24 months of treatment. Bladder neck contracture occurred in 5-20 percent of subjects treated with RP. Incontinence of any severity was the most frequently assessed bladder complication, though it was rarely reported. Incontinence rates were reported in: brachytherapy (2-32 percent); RP (5-35 percent); and EBRT (2-6 percent). Urethral stricture and hematuria were more frequent with EBRT. Bowel complications including diarrhea, fecal incontinence, and rectal bleeding were rarely reported in studies evaluating patients undergoing RP. When reported, they occurred less commonly than in men treated with radiation therapy (15-30 percent), either EBRT or conformal EBRT. Except for rectal injury, bowel complications were present beyond 6 months followup. Erectile dysfunction/impotence was common with all treatments ranging from 5 to 95 percent. NSP has been utilized in selected patients in attempts to maintain erectile function. Four patient groups treated with nerve sparing RP were assessed. Impotence rates ranged from less than 5 percent to as high as 60 percent. Cryosurgery Cryosurgery induces cell death by two main mechanisms: direct cellular toxicity from disruption of the cellular membrane by ice ball crystals and vascular compromise from thrombosis and ischemia. The degree of cell destruction is dependent on rapid freezing, the lowest temperature achieved, and slow thawing. Cryotherapy was first used in the treatment of prostate cancer in Newer generation cryosurgery uses pressurized gas-driven probes to both freeze and actively thaw. Transrectal ultrasound guidance assists in probe placement and real time monitoring while urethral warmers have reduced urethral sloughing. However, the requirement to both rapidly freeze the prostate while protecting surrounding structures may affect therapeutic efficacy and/or limit the type of patients/cancers that are candidates for this treatment. Use of cryotherapy has not reached levels comparable to other treatment options. We searched research publications using the MeSH terms "Prostatic Neoplasms" and "Cryosurgery" limited to clinical trial and randomized controlled trials. None of the studies used 28

39 randomization or control group. The majority of the studies included patients with T3-T4 stages of cancer (Appendix C, Table C4). An overview of the studies that reported patient outcomes after cryosurgery as a primary treatment option is presented in Table 11. The sample size ranged from to 1, patients followed for months. Patients with advanced cancer constituted 7.4 percent 63 to 57 percent of the total samples. 76 Mean baseline PSA levels ranged from to ng/ml. The proportion of subjects with poorly differentiated tumors (Gleason 7 or more) varied from 14 percent 78 to more than half of the total sample. 61,65,67,73,76 Progression free survival in men with T1-2 stages was 39 percent percent. 61 Positive biopsy after cryosurgery was detected in 11 percent percent. 63 Progression free and positive biopsy rates varied by tumor characteristics and length of followup. Prevalence of urethrorectal fistula, epididymitis, and sepsis was low in the majority of the studies. Tissue sloughing was observed in 4 percent percent, 76 urethral stricture in 1 percent percent, 60 bladder obstruction in 3 percent 61,62,79-29 percent, 64 and perineal pain in 1 percent percent. 64 Urinary tract infection was diagnosed in 2 percent 62,79-33 percent 60 of patients and incontinence in 2 percent percent. 64 The majority of patients reported impotence (40 percent percent). 80 Quality of life was assessed with FACT-G scale (160 maximum possible scores) 74,75 in men followed for months. Physical well-being was estimated as 26.0 ± 2.9, social/family wellbeing as 23.5 ± 4.6, functional well-being as 24.3 ± 4.0, and emotional well-being as 17.9 ± 2.9 in subscales with 28 maximum possible scores. 47 percent of patients were able to have sexual intercourse at 3 years. Scores did not improve over the time of observation. Authors compared outcomes after cryosurgery with other treatments 68,77 and concluded that effectiveness and safety are comparable. However, one phase II clinical trial was stopped due to poor outcomes. 60 Improved techniques including direct transperineal cryoneedles and percutaneous approach monitored by real-time transrectal ultrasound may reduce complications. 72,82 There is no direct comparative effectiveness evidence of cryosurgery for localized prostate cancer. Studies have not assessed long-term outcomes, including overall and disease specific survival. Outcomes may be biased by patient and provider characteristics. Laparoscopic and Robotic Assisted Radical Prostatectomy Laparoscopic and robotic assisted laparoscopic radical prostatectomy (LRP and RLRP) have risen in popularity since being introduced in 1998 as a minimally invasive surgical method to remove the prostate. Video-assisted endoscopic surgery may result in fewer complications, especially intraoperative blood loss, and quicker recovery time than conventional open radical prostatectomy. However, LRP and RLRP appear to cost more, may not be applicable to all patients (e.g. those with large prostate glands), and require a learning curve for proficiency as well as purchase of laparoscopic and robotic surgical systems. Because they have only been used since 1998, long term-outcomes, including overall and disease specific mortality, are not available. Three systematic reviews estimated the effectiveness and adverse effects of LRP and RLRP from 21 nonrandomized clinical trials and case series (Appendix C, Tables C5-C7). These involved 2,301 and 1,757 patients respectively. Most reports originate from centers outside of the United States with followup ranging from immediate postoperative period to almost 6 years 29

40 (median about 8 months). Important differences in patient and tumor characteristics as well as variable duration of followup make accurate estimates of effectiveness problematic. The authors compared outcomes after several laparoscopic techniques including transperitoneal prostatectomy with initial retrovesical dissection of the seminal vesicles, transperitoneal ascending prostatectomy, extraperitoneal descending technique, extraperitoneal ascending technique, robotic assisted laparoscopic prostatectomy, as well as standard open retropubic radical prostatectomy. Pooling was not appropriate due to differences in study design. Laparoscopic vs. Open Retropubic Radical Prostatectomy (RRP) Overall survival was reported in two studies (n=1,058) (Figure 8) with improved survival in both studies for patients treated with LRP. PSA relapse was assessed in three studies (n=941). There were no statistical differences between treatments with risk estimates ranging from 28 percent worse to 90 percent greater risk of PSA relapse with LRP. Six studies compared positive surgical margins after treatments and did not find significant differences. Patient reported continence (proportion of pad free patients) were similar after two treatments in three studies and better after laparoscopic approach in two studies. 86,87 Long-term potency is not known after LRP. Few studies reported erectile function. All ten comparative studies showed longer operative time for laparoscopic ( minutes) compared to open RRP ( minutes). The majority of the studies demonstrated a lower blood loss after laparoscopic vs. open RRP (189 1,100 ml vs ,550 ml respectively) and transfusion rate with laparoscopy (Figure 9). Bleeding, urine extravasation, wound healing, and thrombo-embolic events were better after laparoscopic surgery. Re-intervention rates were comparable between LRP and retropubic radical prostatectomy (RRP). Recurrence free survival of 84 percent and 99 percent was reported in two studies, though results are limited by study duration and number of patients enrolled. Transperitoneal vs. Extraperitoneal Laparoscopic Radical Prostatectomy The authors reported that both techniques demonstrated comparable outcomes (Figure 10). Sample sizes were fairly small and confidence intervals wide, thus precluding the detection of clinically important differences. Extraperitoneal LRP had shorter learning curve and operating times, lower risk of bleeding, and permitted the elimination of the initial retrovesical dissection of the seminal vesicles. The majority of studies compared outcomes with historical controls. The transperitoneal approach reduced the risk of lymphocele formation. No differences were found in overall morbidity, complications, continence, and positive surgical margins. Robotic Assisted Laparoscopic Radical Prostatectomy Evidence of comparative effectiveness between robotic assisted laparoscopic radical prostatectomy, retropubic, and transperitoneal laparoscopic radical prostatectomy are primarily limited to short-term outcomes from nonrandomized trials (eight studies) (Table 12). Total complications and continence rates, positive surgical margins, and operative time were comparable to RRP. Blood loss was less (median 153 ml) after robotic assisted RP compared to RRP (median 910 ml). Transfusion rate demonstrated the same tendency with lower median (0) after robotic assisted vs. open approach (median 38 percent). Recurrence-free survival was 92 percent and 95 percent for robot assisted RP vs. 85 percent and 95 percent for RRP in two studies, though there were relatively few patients and followup duration was short. Length of stay after robotic assisted was less than half of that for RRP (median 1.2 and 2.7 days 30

41 respectively). The length of catheterization was shorter after robotic assisted RP compared with RRP (median 7 vs. 13 days respectively). Comparative effectiveness of LRP on quality of life has not been established. European experience reported that quality of life scores improved in 7.8 percent and remained the same in 37.4 percent of the first 500 patients who underwent laparoscopic RP. Authors reported scores to analyze the global results of RP grading biochemical progression (0 4), incontinence (0 2), and impotence (0 1) with comparable quality of life after LRP and RRP. Estimates did not include morbidity and adjustment for baseline patient functional characteristics. What is the Impact of Treatments on Overall and Disease Specific Quality of Life? Our review included nationally representative prospective studies of men with clinically localized prostate cancer using standardized quality of life instruments. More recently developed therapies such as brachytherapy, cryotherapy, laparoscopic or robotic prostatectomy, and conformal radiotherapy were not specifically assessed in PCOS but are addressed based on results from other studies. We describe quality of life data as reported in randomized studies. All men up to age 90 years at the time of diagnosis were eligible for entry into PCOS. We focused on a cohort of over 2,000 men with clinically localized prostate cancer who provided survey responses at least 24 months post diagnosis. 56,88 The study cohort had an average age of 66 years (range 39-88). Fifty-seven percent of men undergoing RP were <65 in contrast to 13 percent, 15 percent, and 23 percent of those receiving no treatment, ADT, or EBRT respectively. Men were primarily of white race (72 percent) with approximately 13 percent Black and 13 percent Hispanic. Treatment received did not vary substantially by race (71-77 percent were white across treatment categories). Baseline health status varied according to treatment received with fewer patients undergoing androgen deprivation (35 percent) or no treatment (43 percent) reporting excellent to very good baseline general health compared to 52 percent of those receiving RP and 45 percent of men undergoing radiation. Primary health status domains for prostate cancer treatments include urinary, bowel, and sexual questions. Additional measures include general health status, impact of cancer or its treatment on daily activities or relationships with spouse or friends; belief that one is free of cancer, satisfaction with treatment selected, and likelihood of making the same treatment decision again. PCOS assessed the prevalence and severity of factors at baseline and followup as well as their overall bother. Questions typically referred to health status/events over the 1 month prior to the completed questionnaire. Treatment decisions, outcomes, and bother may vary by patients baseline health status. In separate PCOS reports, baseline urinary, bowel, and sexual dysfunction and bother were greater in men who received EBRT than in men who received RP. 89,90 Urinary dysfunction and bother. The odds of being bothered due to dripping or leaking of urine (13.9 percent versus 3.0 percent) was more than six fold greater in RP treated patients than in EBRT after adjusting for baseline variables including age, race, clinical stage, and comorbidity index (Table 10). 56 Incontinence summary scores (Scale 0-100) varied by treatment received and baseline function. Men with normal baseline function declined from 100 to 60 at 6 months for RP and from 100 to 95 for EBRT but increased to 76 and 96 at 2 years. In comparison, patients with lower baseline urinary function scores had no decline at either 6 31

42 months or 5 years when treated with EBR and a modest decline at 5 years if treated with RP (79 at baseline compared with 72 at 5 years; data not shown). The main reasons patients reported bother included night time urination urgency, slow or difficult urination, and frequent urination, which all were reported by greater than 30 percent of individuals who reported bother but not different between groups. (Appendix C, Table C8) Bowel dysfunction and bother. Bowel dysfunction was more frequent in men receiving EBRT. Five percent versus 4.3 percent of men undergoing EBRT and RP respectively were bothered by frequent bowel movement, pain, or urgency. Bowel summary scores changed little from baseline during 5 years of followup for either treatment regardless of baseline function. Sexual dysfunction and bother. Sexual dysfunction and bother related to sexual dysfunction was the most common adverse health status effect related to RP or EBRT. The impact on sexual function at 5 years did not differ by these two treatments. The two most frequent reasons for sexual bother were erectile difficulties and inability to satisfy spouse or partner. Sexual function summary scores decreased markedly within 6 months and remained much lower than baseline throughout 5 years of followup for both RP and EBRT regardless of baseline function. For individuals with normal baseline sexual function, sexual summary scores declined from 91 at baseline for both RP and EBT treated patients to 37 and 67 respectively at 6 months and 47 and 50 at 5 years (data not shown). The percentage of individuals in PCOS treated with ADT who stated that they had a big/moderate overall problem with sexual function increased from 22 percent to 26 percent (4 percent increase) in the orchiectomy group and from 33 percent to 38 percent (5 percent increase) in the LHRH group. Nearly one-quarter of men who reported no problems before treatment reported they had some problem with sexual function after treatment. Other outcomes. Over three-quarters of patients treated with RP and one-half of those treated with EBRT or brachytherapy believed that they were free of prostate cancer compared to 16 percent versus 9 percent of those receiving ADT or no treatment respectively (Table 13). General perceptions of prostate cancer health slightly favored orchiectomy compared to LHRHa. For example, the percentage of individuals reporting physical discomfort or worry due to prostate cancer as well as rating their overall health as fair or poor was greater with LHRHa. However, more (47 percent vs. 40 percent) LHRHa patients believed they were free from cancer. Satisfaction with treatment and willingness to choose the same treatment again was similarly high in both groups (Table 13). Scores on the SF-36 general health status scale or any of its domains did not differ between treatments. As noted earlier, a previous AHRQ EPC report addressed ADT for patients with advanced prostate cancer. In these studies, the mean duration of treatment and patient survival was less than 5 years. Therefore, adverse impact on quality of life or other adverse effects due to prolonged treatment were not adequately addressed. Satisfaction with treatment, general health status, overall impact of cancer, or treatment on daily activities was reported by Hoffman and colleagues at 24 months of followup (Table 13). 88 Less than 5 percent of patients reported that they were dissatisfied, unhappy, or felt terrible about their treatment, with the highest percent (4.9 percent) occurring in those who underwent radical prostatectomy. Patients treated with RP more frequently reported that cancer or treatment affected the relationship with their spouse or friends. Financial problems due to cancer or treatment were highest in patients treated with ADT followed by patients treated with RP. 32

43 Treatment satisfaction was highly correlated with bowel, bladder, and erectile function; general health status; belief that the respondent was free of prostate cancer; and whether cancer treatments did not limit activity or affect relationships with spouses or friends (Appendix C, Tables C9 and C10). Between 91 percent and 95 percent said they would definitely or probably make the same treatment decision again, with the highest percent reported from patients treated with primary ADT and the lowest with RP. Additional analysis of PCOS data assessed prostate cancer specific health status and bother among men ages 70 years an older (Table 14). 88 Prevalence, severity, and health impact of urinary, bowel, and sexual dysfunction in these older men was similar to the entire cohort. Separate results for men under the age of 65 versus those over 65 were not provided. Men who underwent aggressive therapy defined as RP/EBRT or brachytherapy were more bothered by dripping or leaking of urine, and bowel or sexual problems than men treated with conservative therapy. When adjusted for treatment propensity score, baseline function, age, race, education, and comorbidity score, the extent of bother due to urine, bowel, or sexual dysfunction was 5.1; 2.4, and 2.8 fold higher respectively for older men treated with aggressive rather than conservative therapy. Despite these findings, men treated with aggressive therapy more frequently reported that they were delighted or very pleased with their treatment (68.1 percent vs percent) than those treated conservatively (Table 15). 88 There were no differences in physical discomfort, health worry, limitation in daily activities, overall bother, or decisions on whether they would undergo the same treatment again if given the chance. Other longitudinal cohort studies assessed quality of life in men treated for localized prostate cancer using validated disease specific health status measures (Table 16) Differences between treatment options were few and of small magnitude. Sample sizes ranged from subjects and followup lasted months. Lee evaluated patients treated with brachytherapy, EBRT, and RP. Compared to baseline, none of the treatments altered scores on the Functional Assessment of Cancer Therapy-Prostate (FACT-P) overall scores, FACT-General scores, physical well being, or functional well-being scores. International prostate symptom scores (IPSS), a validated symptom scale score typically used to evaluate the presence and severity of benign lower urinary tract symptoms, demonstrated slight worsening with brachytherapy and improvement with both EBRT and RP. Comparative evaluations between treatments for each of the health related quality of life (HRQOL) scores demonstrated statistically significant differences of small magnitude. Schapira assessed bother with urinary, sexual, and bowel function using the UCLA Prostate Cancer Instrument (PCI) in 122 patients treated with RP, EBRT, or expectant management. 92 At 12 months there were no significant differences between treatments in any of the domains. Soderdahl also used the UCLA PCI instruments and determined that patients treated with brachytherapy were significantly less bothered by sexual and urinary problems than those treated with either open or laparoscopic RP. 93 Brachy treated patients had marginally more bother with bowel function. Fulmer reported no differences in urinary symptoms between patients treated with RP and those receiving hormonobrachy therapy with or without EBR. Hormonobrachy therapy patients had less sexual function bother than RP patients. Galbraith evaluated 185 men 33

44 treated with WW, RP, and various forms of radiation therapy. 94 After 18 months there were no significant differences in health related quality of life, physical functioning, or general health. Quality of Life Outcomes in Randomized Controlled Trials Two RCTs reported on quality of life in patients treated with early intervention compared with deferred treatment or WW. Fransson and colleagues (Table 17) evaluated conventional or conformal EBRT versus deferred treatment (not further stated) in 166 enrollees. 26 Questionnaire data were available in 108 subjects with a median time from randomization of 41 months in the radiation therapy group and 30 months in the deferred therapy group. There were no differences in global health status. Subjects treated with EBR therapy were more likely to note limitation in daily activity due to prostate cancer, incontinence, and limitation in daily activity due to intestinal or urinary problems. Steineck reported on 4-year health status results for 379 surviving subjects enrolled in the SPCG- 4 study comparing RP to WW and who completed the survey. 95 Sexual dysfunction, urinary leakage, and distress were greater with RP. Bowel and urinary obstructive symptoms were greater with WW. The relative risk of sexual dysfunction as measured by desire, penile stiffness, intercourse, orgasm, and distress from compromised sexuality was higher in subjects randomized to RP than in those randomized to WW (RR for specific domains). Urinary tract dysfunction was markedly higher for domains of urinary leakage with 18 percent versus 2 percent reporting moderate or severe leakage and 29 percent versus 9 percent saying they had moderate or great distress. Distress from obstructed voiding of moderate to great degree was similar between treatments. Overall distress from all urinary symptoms was reported in 27 percent of those receiving RP and 18 percent in those receiving WW (RR 1.5 [ ]). Bowel function was worse in patients treated with WW with distress from all bowel symptoms occurring in 6 percent of subjects treated with WW compared with 3 percent of those receiving RP. Physical and psychological function did not differ between treatments. Forty to fifty percent reported low or moderate physical well being and subjective quality of life. Moderate or high worry was reported by 39 and 45 percent of individuals receiving RP and WW respectively. Summary Assessing the comparative effectiveness and adverse effects of primary treatment options for clinically localized prostate cancer is limited by the lack of high-quality data regarding the relative effectiveness and safety of the primary treatment options, especially related to PSA detected disease. One large randomized trial found that RP reduced overall and disease specific mortality at 10 years by 5 percent and risk of metastatic disease in men by 8 percent with primarily non PSA detected disease. Other randomized trials suggest that the addition of ADT to EBRT (but not RP) in high risk patients may reduce the risk of disease progression. All treatment options result in adverse effects and impact on disease specific health status (primarily urinary, bowel, and sexual) though the severity and frequency may vary between treatments. Patient satisfaction with therapy is high and associated with several clinically relevant outcome measures. Data from nonrandomized trials are inadequate to reliable assess comparative effectiveness and adverse effects. Methodologic limitations include lack of randomization, wide variety in reporting of patient and tumor characteristics, and outcome measures. 34

45 Table 1. Treatment options for clinically localized prostate cancer Does not remove prostate gland and may not eradicate cancer; impotence, incontinence, scrotal edema, pelvic pain; sloughed urethral tissue; prostatic abscess; urethrorectal fistula. No long- Treatment Option Treatment Description Potential Benefits Potential Risks Radical retropubic or perineal prostatectomy (RP) Complete surgical removal of prostate gland with seminal vesicles, ampulla of vas and sometimes pelvic lymph nodes. Sometimes done laparascopically or with robotic assistance and attempt to preserve nerves for erectile function. May eliminate cancer; generally well tolerated. 1 RCT showed improved overall, prostate cancer survival and metastasis versus surveillance. External-beam radiation Multiple doses of radiation from an May eliminate cancer; (EBRT) external source applied over several generally well tolerated, weeks. Dose and physical and avoids operative risk characteristics of beam may vary. Conformal radiotherapy uses 3 dimensional planning systems to maximize dose to prostate cancer and attempt to spare normal tissue. Brachytherapy Radioactive implants placed under May eliminate cancer; anesthesia using radiologic guidance. generally well tolerated; Lower dose/permanent implants avoids operative risk; typically used. External beam boost single outpatient session radiotherapy and/or androgen deprivation sometimes recommended. patients with prior TURP. Cryoablation Destruction of cells through rapid May eliminate cancer; freezing and thawing using transrectal generally well tolerated; guided placement of probes and avoids operative risk; injection of freezing/thawing gases. single outpatient session term outcomes from national sample. Androgen deprivation Oral or injection medications or surgical Avoids risks of RP and therapy removal of testicles to lower or block EBRT. Usually lowers circulating androgens. PSA levels and may slow cancer progression. Active surveillance Active plan to postpone intervention. No immediate side (watchful waiting) May involve monitoring with DRE/PSA effects or complications; and repeat prostate biopsy with further low initial cost; most men therapy (either curative or palliative ) do not need therapy and based on patient preference, symptoms survive at least 10 years. and/or clinical findings PSA = prostate-specific antigen; TURP = transurethral resection of the prostate Hospitalization for major surgery; operative-related death, perioperative cardiovascular complications and bleeding. May not eradicate cancer. Long-term urinary incontinence, urethral stricture, bladder neck contracture, erectile dysfunction. Does not remove prostate gland and may not eradicate cancer; 6-8 weeks of outpatient therapy; treatment related death, incontinence, proctitis, cystitis, impotence, urethral stricture, bladder neck contracture, bleeding. Not indicated in men with inflammatory bowel disease because of risk of bowel injury. Does not remove prostate gland and may not eradicate cancer. May not be effective for larger prostate glands or more aggressive tumors; urinary retention, incontinence, impotence, cystitis/urethritis, proctitis; long-term outcomes from representative national sample not reported. Not indicated in Gynecomastia, impotence, diarrhea, osteoporosis, lost libido, hot flashes, androgen deprivation syndrome (i.e. depression, memory difficulties, fatigue) Cancer could advance, become incurable, and cause death; patients quality of life could be painfully restricted before he dies; additional treatments may be necessary, not effective, and have side effects. Patients may be too anxious or worried to monitor cancer without treatment. 35

46 Table 2. Randomized controlled trials comparing major primary treatment options and reporting any clinical outcome External Beam Androgen Adjuvant or Radical Watchful Radiation Deprivation Neoadjuvant Prostatectomy Waiting Therapy Therapy Therapy Brachytherapy Radical Prostatectomy OS; DSS; DM; bned; DM OS; DSS; bned; AE; QoL DM; AE Watchful Waiting OS; DSS; DM; AE; QoL External Beam bned; DM OS; DSS; OS; DSS; Radiation Therapy bned; DM; AE bned; AE Androgen Deprivation Therapy Adjuvant or OS; DSS; OS; DSS; bned Neoadjuvant Androgen bned; DM; AE bned; AE Deprivation Therapy Brachytherapy bned bned; AE Cryrotherapy OS = Overall survival DSS= Disease-specific survival bned = Biochemical no evidence of disease DM = Distant metastasis AE = Adverse events/toxicity QoL = Quality of life Cryrotherapy 36

47 Table 3. Description of randomized studies of treatments for localized prostate cancer Study (ref) Study Characteristics Interventions Watchful waiting versus radical prostatectomy Bill-Axelson, Radical prostatectomy (RP) (n=347) Method of allocation: Adequate, telephone outside clinic 2. Watchful waiting (WW) (n=348) Intention to treat: yes Iversen, / Graversen, 1. RP + oral placebo (n=74) Oral placebo/expectant Method of allocation: Unclear management (n=68) Radical prostatectomy verses radiation therapy Paulson, 1982, RP (n=47) Method of allocation: Unclear 2. Radiation Therapy (n=59) totaling 4,500 to 5,000 rad Intention to treat: no Radical prostatectomy with or without neoadjuvant therapy Homma, RP followed by leuprolide + chlormadione for 3 months Method of allocation: Unclear, followed by leuprolide alone dynamic balancing method (n=86). used so stage and histiologic differentiation were evenly 2. Neoadjuvant androgen distributed between groups. deprivation group: leuprolide + chlormadione for 3 months Blinding: The pathologist was followed by RP and adjuvant informed of the patient s endocrine therapy (leuprolide treatment group (endocrine alone) (n=90). therapy may interfere with accuracy of judgment of histiologic differentiation) Median Followup Years Description of Subjects; Inclusion Criteria Swedish, Finnish men and Icelandic men, mean age 65 years. Mean PSA (ng/ml): RP 13.5; WW Tumor stage: T1b 11.9%; T1c 11.7%; T2 76.1%; unknown 0.3%. Gleason score: %; %; %; %. Subjects were eligible if <75 years of age; had newly diagnosed untreated localized PC, confirmed with histiologic or cytologic exam, with a tumor stage T0 to T2 (tumor had to be well to moderately differentiated WHO definition); had to be healthy enough to undergo RP; and had a life expectancy >10 years; bone scan had to show no abnormalities; and PSA had to be less than 50 ng/ml. 23 (19-27) 142 American men with early carcinoma Unclear, analysis up to 5 years 106 American men with clinical stage A2 or B (T1/T2) PC. Analyses excludes 9 subjects. Exclusion criteria: Subjects with occult focal carcinoma or patients with stage C disease Japanese men (48 excluded from interim analyses). Tumor stage: A 2 12%; B 1 25%; B 1 38%. Age, PSA, and histologic differentiation contaminated with C stage subjects. Subjects were eligible if they had histologically confirmed untreated clinical stage A 2, B, or C PC; serum testosterone concentration of 1.0 ng/ml; age 80 years; and absence of any contraindication to RP or the teat drugs. 37

48 Table 3. Description of randomized studies of treatments for localized prostate cancer (continued) Study (ref) Study Characteristics Klotz, 2003, ,97 Method of allocation: Adequate, computer-generated randomization schedule by strata (according to clinical stage, baseline PSA, and Gleason sum) Open label Intention to treat: yes Schulman, Method of allocation: unclear, stratified according to clinical stage and pathological grade of tumor Soloway, Lupron Depot Prostate Cancer Study Group Method of allocation: Unclear Gleave, ; Toxicity only Method of allocation: Unclear Open label Intention to treat: no, interim analysis 1. RP (n=101) Interventions 2. Neoadjuvant androgen ablation + RP (n=112): cyproterone acetate 300g daily 3 months prior to RP 1. RP (n=210) 2. Neoadjuvant androgen ablation + RP (n=192): goserelin acetate monthly and flutamide 250mg t.i.d. 3 months prior to RP 1. RP (n=154) 2. Neoadjuvant androgen ablation + RP (n=149): leuprolide acetate 7.5mg per month and flutamide 250mg t.i.d 3 months prior to RP 1. Neoadjuvant androgen ablation + RP (n=273): leuprolide acetate 7.5mg per month and flutamide 250mg t.i.d 3 months prior to RP. 2. Neoadjuvant androgen ablation + RP (n=274): leuprolide acetate 7.5mg per month and flutamide 250mg t.i.d 8 months prior to RP Median Description of Subjects; Inclusion Criteria Followup Years Canadian men, median age PSA (ng/ml): <10 54%, %, > %. Tumor stage: T1b/c 8.9%; T2a 33.3%; T2b 18.8%, T2c 34.3%. Gleason score sum: %, %, %. Subjects were eligible if they had histologically confirmed untreated clinically localized PC (stages T1/T2); negative bone scan; enzymatic prostatic acid phosphatase less than twice normal (<1.8 units/l); and PSA <50ng/mL European men locally confined PC. Tumor stage: T2 54.3%. Subjects were eligible if they histologically confirmed T2/T3NxM0 PC with a PSA < 100 ng/ml American men (21 excluded from trial), mean age 65 years, with tumor stage T2b PC. Mean PSA (ng/ml) Mean Gleason score 6. Mean prostate volume 36.1cc. White race 68%. 3 and 8 months Subjects were eligible if they were <75 years of age, had a PSA <50 ng/ml, and had a normal bone scan 547 Canadian men, mean age 62.6 years, with clinically confirmed PC. PSA (ng/ml): % ; %; >20 9.5%. Tumor stage: T1b 2.6%; T1c 29.1%; T2a 28.9%; T2b 32.9%, T2c 6.6%. Gleason score: %, %. White race 93%. Subjects were eligible if they required RP for previously untreated, histologically confirmed clinical stage T1b to T2 PC. Exclusion criteria: prior RT or hormonal therapy, concomitant use of medications with antiandrogen activity, prior history of cancer (except basal cell carcinoma of the skin), or severe renal or hepatic impairment. 38

49 Table 3. Description of randomized studies of treatments for localized prostate cancer (continued) Study (ref) Study Characteristics External beam radiation (EBR) Lukka, Method of allocation: Adequate, central computer-generated randomization schedule by strata. Intention to treat: yes Sathya, Method of allocation: Adequate, centralized using sealed opaque envelopes. Intention to treat: no Yeoh, Method of allocation: Adequate, computer-generated randomization. Intention to treat: no, interim analysis Interventions 1. Long arm EBR (n=470): 66 Gy in 33 fractions 2. Short arm EBR (n=466): 52.5 Gy in 20 fractions 1. Iridium implant + EBR (n=51) 2. EBR alone (n=53) 1. Conventional EBR (n=61): 64 Gy in 32 fractions over 6.5 weeks. 2. Hypofractionated EBR group (n=59): 55 Gy in 20 fractions in 4 weeks. Median Followup Years Description of Subjects; Inclusion Criteria Canadian men with early stage PC (T1 or T2), mean age 70 (range 53-84). Mean PSA (ng/ml): Tumor stage: T1a <1%; T1b 2%; T1c 25%; T2a 27%; T2b 27%; T2c 18%. Gleason score: 2-4 8%; 5 14%; 6 38%; 7 31%; 8-9 9%. Subjects were eligible if the had early stage PC (T1 or T2). Exclusion criteria: PSA >40 ng/ml; previous therapy for PC; previous hormone therapy; prior or active malignancy (except nonmelanoma skin cancer, colon cancer, or thyroid cancer treated 5 years before trial and presumed cured); previous pelvic radiotherapy; inflammatory bowel disease; a serious nonmalignant disease that would preclude radiotherapy or surgery biopsy; psychiatric or addictive disorder Canadian men (138 randomized, 34 subjects excluded post-randomization due to positive lymph nodes), mean age 66 (range 49-74). Tumor stage: T2 61%. Age, PSA, and Gleason score contaminated with T3 stage subjects. Subjects were eligible if they had histologically proven PC with clinical stage T2 or T3, N0, M0 and had to be fit to undergo pelvic lymphadenectomy as a staging procedure. Exclusion criteria: Prior history of pelvic radiotherapy or RP, androgen ablation, or TURP or evidence of metastatic disease using computed tomography (CT) scan and bone scan Australian men, median age 63.6 years (range 44.3 to 82.8), with localized, early stage (T1/T2N0M0) PC 39

50 Table 3. Description of randomized studies of treatments for localized prostate cancer (continued) Study (ref) Study Characteristics Zietman, Method of allocation: Unclear, centralized, stratified to ensure balanced groups Intention to treat: Partially (one subject from high dose group refused treatment and was not included in analyses) Interventions 1. Conventional dose group (n=197): 70 Gy (3D conformal proton 50.4 Gy and proton boost 19.8 Gy) 2. High dose group (n=196): 79.2 Gy (3D conformal proton 50.4 Gy and proton boost 28.8 Gy) Hormonal therapy combined with radiation therapy Denham, Radiotherapy (RT) 66 Gy and no androgen deprivation Method of allocation: Adequate, (n=276) centralized using the minimization technique. 2. RT and 3 months neoadjuvant and adjuvant Intention to treat: yes androgen deprivation (goserelin acetate monthly and flutamide 250 mg t.i.d.) starting two months prior to RT (n=270) 3. RT and 6 months neoadjuvant and adjuvant androgen deprivation (goserelin acetate monthly and flutamide 250 mg t.i.d.) starting five months prior to RT (n=272) Median Description of Subjects; Inclusion Criteria Followup Years American men, median age 67 years (range 45-91). Median PSA (ng/ml): 6.3. Tumor stage (AJC 1992): T1b <1%; T1c 61.2%; T2a 23.7%; T2b 14.8%. Gleason score: %; %; %. Risk group: Low (PSA < 10 ng/ml, T2a stage, Gleason 6) 57.9%; Race: white 90.3%; black 4.3%; Hispanic 2.8%; other 2.6%. Subjects were eligible if they had clinically localized PC (stage T1b through T2b, PSA <15 ng/ml and no evidence of metastatic disease as assessed by both whole-body bone scan (with PSA level >10 ng/ml, tumor stage T2b, or Gleason score 7) and abdominopelvic computed tomography scan. There was no exclusion from entry based on basis of tumor histology (Gleason score) Australian and New Zealand men, median age 68 (range 41-87). Tumor stage (TNM 1992): T2b 26%; T2c 34%. Age, PSA, and Gleason score contaminated with T3/4 stage subjects. Subjects were eligible if did not have substantial comorbidity (of a severity that would limit survival to 5 years in the absence of PC) or previous malignant disease; stage T2b-T4 PC without evidence of lymph-node involvement, bony metastases, or metastases at other sites. 40

51 Table 3. Description of randomized studies of treatments for localized prostate cancer (continued) Study (ref) Study Characteristics Crook, Method of allocation: Adequate, central generation of a list that was transferred into sequentially numbered sealed envelopes for on-site randomization at each center. Intention to treat: no D Amico, Method of allocation: Adequate, centralized, permuted blocks algorithm Intention to treat: yes Brachytherapy Wallner, / Herstein, Method of allocation: Adequate, centralized, computerized random number generator Interventions 1. Neoadjuvant hormonal therapy 3 months prior to RT (n=177): goserelin acetate monthly and flutamide 250 mg t.i.d. 2 weeks before the initiation of goserelin injections and continued throughout the course of neoadjuvant therapy 2. Neoadjuvant hormonal therapy 8 months prior to RT (n=184) 1. Conformal RT (CRT) Group (n=104): 70 Gy 2. CRT + androgen suppression therapy (ADT) group (n=102): 70 Gy - ADT consisted of leuprolide acetate 7.5 mg or 22.5 mg IM every 3 months (n=88) or goserelin 3.6 or 10.8 mg sq (n=10) combined with flutamide 250 mg tid I 144 Gy (n=57) Pd 125 Gy (n=58) Intention to treat: yes Adjuvant EBR combined with brachytherapy Wallner, Pd 125 Gy + EBR (20 Gy) (n=85). Method of allocation: Adequate, centralized, computerized Pd 125 Gy + EBR (44 random number generator Gy) (n=80). Median Description of Subjects; Inclusion Criteria Followup Years (17 excluded from analyses) Canadian men, median age 72 (range 50-85) with clinically localized PC. Tumor stage (TNM 1997): T1c/T2a 52.6%; T2b/T2c 34.1%. Age, PSA, and Gleason score contaminated with T3 stage subjects. Subjects were eligible if they had a histologic diagnosis of adenocarcinoma of the prostate with all Gleason scores and PSA levels, and clinical stages from T1c to T4), normal baseline hepatic and renal function, and estimated life expectancy >5 years American men, mean age 72.5 years. Mean PSA (ng/ml): 11. Tumor stage AJC 1992): T1b 1.9%; T1c 46.1%; T2a 22.3%; T2b 22.8%. Gleason score: mean 7; 5 or %; %; %; %. Prostate volume (ml) 39. Subjects were eligible if patient had PSA 10 ng/ml (40 ng/ml maximum) or a Gleason score 7 (range 5-10). Low-risk patients ineligible unless there was radiographic evidence using MRI of extracapsular extension or seminal invasion. Exclusion criteria: prior history of malignancy (except nonmelanoma skin cancer) or any history of hormone use. 3 (Wallner) 126 (of 492 of a planned total of 600) American men with 1997 American Joint Commission on Cancer (AJC) clinical stage T1c-T2a, Gleason 2-6 (mean 5.9), PSA 4-10 ng/ml(mean 6.9) PC. 115 subjects included in the analyses. Herstein (2004) assesses long-term radiation related morbidities in 314 men (of a planned total of 600) American men with 1997 American Joint Commission on Cancer (AJC) clinical stage T1c-T2a, Gleason 7-10 (mean 7.0) and/or PSA ng/ml(mean 6.9) PC. 159 subjects included in the analyses. 41

52 Table 3. Description of randomized studies of treatments for localized prostate cancer (continued) 1950 Study (ref) Study Characteristics Interventions Median Followup Years Description of Subjects; Inclusion Criteria Intention to treat: no Adjuvant bicalutamide versus placebo; both treatment arms combined with standard care (radical prostatectomy/radiation therapy or watchful waiting) Wirth, Analysis of 3 RCTs 1. Bicalutamide 150mg daily (n=4,052) 5.4 8,113 multinational men with localized or locally advanced PC Trial 23: 3,292 North American men, mean age 64 (38-85). Tumor stage: T1/T2 2. Placebo (n=4,061) 73%. Vaccine versus nilutamide Arlen, Phase II trial conducted at NCI Method of allocation: Unclear Intention to treat: yes Trial 23: Standard care received: RP 80%; RT 20% Trial 24: Standard care received: RP 46%; RT 18%; WW 36% Trial 25: Standard care received: RP 13%; RT 5%; WW 81% 1. Vaccine consisting of recombinant vaccinia viruses containing PSA and B7.1 costimulatory genes (prime vaccinations) and avipox PSA (as boosters) (n=21). Patients also received granulocytemacrophage colony stimulating factor and interleukin-2 as part of their vaccination schedule. 12 subjects received nilutamide at time of PSA progression. 2. Nilutamide 300mg qd x 1 month, then 150 mg qd (n=21). 8 subjects received vaccine at time of PSA progression. Trial 24: 3,603 European, South African, Israeli, Australian, and Mexican men, mean age 69 (48-93). Tumor stage: T1/T2 65%. Trial 25: 1,218 Scandinavian men, mean age 69 (48-87). Tumor stage: T1/T2 60%. Age, PSA, and Gleason score contaminated with T3 stage subjects. Subjects were eligible if patient was 18 (75 years of age upper limit for trial 25) with clinically or pathologically diagnosed T1-T4 PC with no distant metastases Unclear 42 American men with hormone refractory PC, mean age 68 years (range 51-87). Mean PSA: vaccine 35.1; nilutamide Gleason score: mean 7; score %; %; %; unknown 7.1%. Current testosterone decreasing Rx: goserelin acetate 21.4%; leuprolide acetate 54.8%; orchiectomy 23.8%. Prior antiandrogens: %; %; %. Subjects were eligible if castrate levels of serum testosterone <50 ng/dl. Subjects with failed prior antiandrogen Rx were required to have 2 consecutive increasing serum PSA levels a week apart, measure 6 weeks after bicalutamide withdrawal or 4 weeks after flutamide withdrawal. Subjects need Zubrod performance status 0 or 1. Have adequate hematological, hepatic and renal function. No evidence of an immunocompromised condition, no diagnosis of altered immune function, no prior radiotherapy to more than 50% of nodal groups. Exclusion criteria: Egg allergy, skin disorder, history of seizures, serious intercurrent illnesses, close contact with immunocompromised individuals, contact with individuals with skin disorders or children <5 years old, prior nilutamide therapy. 42

53 Table 4. Overall mortality or survival for randomized controlled trials Study Treatment Group Control Group Analyses; p-values Outcomes Watchful waiting versus radical prostatectomy Bill-Axelson, Radical prostatectomy (n=347) Watchful waiting (n=348) Absolute Risk Reduction (ARR) [95% CI] Relative Risk (RR) [95% CI] Total number of deaths Cumulative incidence of death Median followup: 8.2 years % [5.4 to 11.2]* at 5 years 27.0% [21.9 to 33.1] at 10 years % [ 7.1 to 13.5] at 5 years 32.0% [26.9 to 38.2] at 10 years p=0.04 all deaths ARR: 2.0 [-2.2 to 6.2] at 5 years ARR: 5.0 [-2.8 to 13.0] at 10 years RR: 0.74 [0.56 to 0.99] at 10 years Iversen, Radical prostatectomy plus placebo Watchful waiting and placebo (n=68) p value (n=74) Total number of deaths Median survival Median followup: 23 years (19 to 27) years 63 8 years Not significant Radical prostatectomy with or without neoadjuvant therapy Klotz, Radical prostatectomy (n=101) Neoadjuvant androgen ablation + RP p value (n=112) Total number of deaths Overall survival at 5 years Median followup: 6 years (0.6 to 9.8) 88.4% [80.6 to 96.3] 93.9% [88.6 to 99.1] External beam radiation Trials Lukka, Long arm (66 Gy in 33 fractions) EBR Short arm (52.5 Gy in 20 fractions) Hazard Ratio [95% CI] (n=470) EBR (n=466) Total number of deaths [0.63 to 1.15] Overall survival at 5 years Median followup: 5.7 years (4.5 to 8.3) 85.2% 87.6% Zietman, Conventional dose (70 Gy) group High dose (79.2 Gy) group p Value (n=197) (n=196) Total number of deaths Overall survival at 5 years Median followup: 5.5 years (1.2 to 8.2) 10 97% 8 96% 0.80 Hormonal therapy combined with radiation therapy D Amico, Conformal radiation therapy (70 Gy) Group (n=103) Conformal radiation therapy and androgen suppression therapy group p value Hazard ratio [95% CI] Total number of deaths Overall survival at 5 years Median followup: 4.5 years 23 78% [68 to 88] (n=98) 12 88% [80 to 95] [1.02 to 4.20] 43

54 Table 4. Overall mortality or survival for randomized controlled trials (continued) Study Outcomes Treatment Group Control Group Analyses; p-values Bicalutamide versus placebo; both treatment arms combined with standard care (adjuvant radical prostatectomy/radiation therapy or watchful waiting) Wirth, Localized disease population (T1/T2) Bicalutamide and adjuvant therapy (estimated n=1,908) Placebo and adjuvant therapy (estimated n=1,891) p value Hazard ratio [95% CI] Total number of deaths Median followup: 5.4 years 187 (9.8%) 182 (9.6%) [0.82 to 1.23] Bicalutamide and watchful waiting Placebo and adjuvant therapy p value (estimated n=777) (estimated n=850) Total number of deaths Median followup: 5.4 years 196 (25.2%) 174 (20.5%) [1.00 to 1.50] Vaccine versus nilutamide Arlen, Vaccine Group (n=21) Nilutamide Group (n=21) Total number of deaths 3** 7** * 95% Confidence intervals ** Includes crossover deaths 44

55 Table 5. Main results. SPCG Main results SPCG-4 Over the next 10 years, what percent of men with new, clinically diagnosed prostate cancer will experience each of the following if they undergo: Outcomes Radical prostatectomy Watchful waiting Benefits due to treatment Dying from prostate cancer 9.6% 14.9% Developing metastatic disease 15.2% 25.4% Developing local progression 19.2% 44.3% Harms due to treatment at 4 years Impotence 80% 45% Incontinence 49% 21% Weak urinary stream 28% 44% Overall benefits vs. harms Quality of life after 4 years No difference Dying for any reason at 10 years 27.0% 32.0% Number men needed to treat with RP to prevent 1 death at 10 years 19 45

56 Table 6. Disease specific mortality or survival for randomized controlled trials Study Treatment Group Control Group Analyses; p-values Outcomes Watchful waiting versus radical prostatectomy Bill-Axelson, Radical prostatectomy (n=347) Watchful waiting (n=348) p value Absolute risk reduction (ARR) [95% CI] Relative risk (RR) [ 95% CI] Total number of PC deaths Cumulative Incidence of death Median followup: 8.2 years % [1.2 to 4.6]* at 5 years 9.6% [6.5 to 14.2] at 10 years % [2.6 to 7.1] at 5 years 14.9% [11.2 to 19.8] at 10 years 0.01 ARR: 2.0 [-0.6 to 4.7] at 5 years ARR: 5.3 [-0.3 to 11.0] at 10 years RR: 0.56 [0.36 to 0.88] at 10 years Radical prostatectomy with or without neoadjuvant therapy Klotz, Radical prostatectomy (n=101) Neoadjuvant androgen ablation + RP (n=112) Total number of PC deaths 0 1 Median followup: 6 years (0.6 to 9.8) External beam radiation Lukka, Long arm (66 Gy in 33 fractions) EBR (n=470) Short arm (52.5 Gy in 20 fractions) EBR (n=466) Total number of PC deaths 3 (<1%) 0 Median followup: 5.7 years (4.5 to 8.3) Yeoh, Conventional (64 Gy) EBR (n=61) Hypofractionated (55 Gy) EBR group (n=59) Total number of PC deaths 1 (1.6%) 0 Median follow-up: 3.6 years (1.9 to 5.2) Zietman, Conventional dose (70 Gy) group High dose (79.2 Gy) group (n=196) (n=197) Total number of PC deaths Median followup: 5.5 years (1.2 to 8.2) 2 0 Hormonal therapy combined with radiation therapy Denham, Radiotherapy (66 Gy) Group (n=164, Radiotherapy (66 Gy) Group and 6 Hazard ratio [95% CI] T2 only) months androgen suppression therapy group (n=162, T2 only) Total number of PC deaths: T2b Total number of PC deaths: T2c [0.03 to 1.88] 0.57 [0.22 to 1.44] Median followup: 5.9 years (0.1 to 8.5) D Amico, Conformal radiation therapy (70 Gy) Conformal radiation therapy and p value Group (n=103) androgen suppression therapy group (n=98) Total number of PC deaths Median followup: 4.5 years

57 Table 6. Disease specific mortality or survival for randomized controlled trials (continued) Study Treatment Group Control Group Analyses; p-values Outcomes Vaccine versus nilutamide Arlen, Vaccine Group (n=21) Nilutamide Group (n=21) Total number of Deaths 1** 4** * 95% Confidence intervals ** Includes crossover deaths 47

58 Table 7. Biochemical progression/reoccurrence or bned for randomized controlled trials Study Outcomes Treatment Group Control Group Analyses; p-values Radical prostatectomy versus radiation therapy Paulson, Radical prostatectomy (n=47) Radiation therapy (n=59) p value K-M estimate of failure at 5 years 14% (estimated from graph) 39% (estimated from graph) Radical prostatectomy with or without neoadjuvant therapy Homma, Neoadjuvant androgen deprivation No neoadjuvant androgen deprivation p value group (n=69, stage A and B) group (n=63, stage A and B) Clinical relapse events at 5 years 11 9 <0.05 relapse vs. no relapse 0/9 stage A 11/60 stage B 1/11 stage A 8/51 stage B Klotz, Radical prostatectomy (n=101) Neoadjuvant androgen ablation + RP Hazard ratio [95% CI*] (n=112) Biochemical reoccurrence events at 6 years Gleason score at biopsy: (33.7%) 42 (37.5%) 0.98 [0.61 to 1.56] [0.72 to 2.31] 2.82 [1.52 to 5.22] No evidence of biochemical disease (bned) at 5 years Estimated bned survival at 7 years: Gleason: 2-6 (n=142) 7 (n=36) 8-10 (n=22) Median followup: 6 years (0.6 to 9.8) Schulman, % [58.5 to 77.8] % [50.4 to 70.0] Radical prostatectomy (n=115, T2 only) Neoadjuvant androgen ablation + RP (n=105, T2 only) PSA progression >1 ng/ml events at 4 26/114 (22.8%) 18/102 (17.6%) years T2 and PSA <20ng/ml T2 and PSA <20ng/ml 16/94 (17.0%) 13/84 (15.4%) Soloway, Radical prostatectomy (n=154) Neoadjuvant androgen ablation + RP (n=149) Biochemical reoccurrence at 5 years Gleason, biopsy: 2-6 (n=159) 22/80 (27.1%) 18/79 (22.6%) 7 (n=56) 10/25 (41.3%) 15/31 (47.7%) 8-10 (n=29) 8/14 (57.4%) 13/15 (89.6%) Gleason, specimen: 2-6 (n=105) 7 (n=96) 8-10 (n=38) bned at 5 years 18/66 (27.1%) 17/48 (38.9%) 4/12 (33.7%) 67.6% /39 (26.0%) 15/48 (30.5%) 17/26 (64.6%) 64.8% p value p Value p value

59 Table 7. Biochemical progression/reoccurrence or bned for randomized controlled trials (continued) Study Outcomes External beam radiation Lukka, Biochemical or clinical failure (BCF) events K-M estimate of BCF at 5 years Median followup: 5.7 yrs (4.5 to 8.3) Treatment Group Control Group Analyses; p-values Short arm (52.5 Gy in 20 fractions) EBR (n=466) 263 (56.4%) 59.95% Long arm (66 Gy in 33 fractions) EBR (n=470) 236 (50.2%) Absolute risk reduction (ARR) [95% CI] Hazard ratio (HR) [95% CI] 52.95% ARR: -7.0 [-12.6 to -1.4] HR: 1.18 [0.99 to 1.41] Yeoh, Conventional EBR n=61) Hypofractionated EBR group (n=59) p value PSA relapse events 9 8 Not significant K-M estimate of biochemical relapsefree survival at 4 years 85.5% 86.2% Not significant Sathya, Iridium implant + EBR (n=31, T2 only) EBR (n=32, T2 only) Hazard ratio [95% CI] BCF events [ 0.16 to 0.85] K-M estimate of BCF at 5 years Median followup: 8.2 years Zietman, Freedom from biochemical failure at 5 years Median followup: 5.5 years (1.2 to 8.2) 25.8% Hormonal therapy combined with radiation therapy Denham, Radiotherapy (66 Gy) Group (n=164, T2 only) Failure events: T2b Failure events: T2c Median followup: 5.9 years (0.1 to 8.5) Crook, K-M estimate of freedom from biochemical failure at 5 years Median followup: 3.7 years (10 months-7 years) D Amico, PSA failure events 57.3% Conventional dose (70 Gy) group High dose (79.2 Gy) group (n=196) p value (n=197) 61.4% [54.6 to 68.3] 80.4% [74.7 to 86.1] < Radiotherapy (66 Gy) Group and 6 months androgen suppression therapy group (n=162, T2 only) month group 8 month group (n=51, Low risk T1c-T2a; PSA <10 (n=41, Low risk T1c-T2a; PSA<10 ng/ml; Gleason 6) ng/ml; Gleason 6) 61% (estimated from graph) 72% (estimated from graph) Conformal radiation therapy (70 Gy) group (n=103) 46 Conformal radiation therapy and androgen suppression therapy group (n=98) 21 Hazard ratio [95% CI] 0.68 [0.44 to 1.06] 0.47 [0.32 to 0.69] p value Hazard ratio [95% CI] < [1.69 to 4.86] Survival free of salvage ADT events

60 Table 7. Biochemical progression/reoccurrence or bned for randomized controlled trials (continued) Study Outcomes KM estimated salvage therapy free survival at 5 years Median followup: 4.5 years Brachytherapy Wallner, Biochemical failure events KM estimated biochemical freedom from failure at 3 years Treatment Group Control Group Analyses; p-values 57% [46 to 69] 82% [73 to 90] 2.30 [1.36 to 3.89] 125 I (144 Gy) (n=57) 6 89% Adjuvant EBR combined with brachytherapy Wallner, Pd + EBR (20 Gy) (n=83) Biochemical failure events 12 KM estimated biochemical freedom 83% from failure at 3 years Subjects with pretreatment PSA <10ng/mL (n=112) KM estimated biochemical freedom from failure at 3 years Subjects with pretreatment PSA >10ng/mL (n=47) KM estimated biochemical freedom from failure at 3 years 84% 82% 103 Pd (125 Gy) (n=58) p value 5 91% Pd + EBR (44 Gy) (n=76) 9 88% Bicalutamide versus placebo; both treatment arms combined with standard care (adjuvant radical prostatectomy/radiation therapy or watchful waiting) Wirth, Localized disease population (T1/T2) Bicalutamide and radical prostatectomy (estimated n=1365) Placebo and radical prostatectomy (estimated n=1,369) p value Hazard ratio (HR) [95% CI] Event-to-time ratio (ETR) [95% CI] Progression events Median followup: 5.4 years Progression events Median followup: 5.4 years 94% 72% p value 115 (8.4%) 121 (8.8%) 0.57 HR: 0.93 [0.72 to 1.20] ETR: 1.06 [0.87 to 1.28] Bicalutamide and radiation therapy (estimated n=538) Placebo and radiation therapy (estimated n=527) 114 (21.2%) 128 (24.3%) 0.09 HR: 0.80 [0.62 to 1.03] ETR: 1.16 [0.98 to 1.37]

61 Table 7. Biochemical progression/reoccurrence or bned for randomized controlled trials (continued) Study Treatment Group Control Group Analyses; p-values Outcomes Vaccine versus nilutamide Arlen, Vaccine Group (n=21) Nilutamide Group (n=21) p value Median time to treatment failure, in months Crossover: 12 subjects had added nilutamide 13.9 after crossover, 25.9 from initiation of therapy Crossover: 8 subjects had added vaccine 5.2 after crossover, 15.5 from initiation of therapy 1967 * Confidence intervals 51

62 Table 8. Incidence of distant metastatic disease for randomized controlled trials Study Outcomes Treatment Group Control Group Analyses; p-values Watchful waiting versus radical prostatectomy Bill-Axelson, Radical prostatectomy (n=347) Watchful waiting (n=348) p value Absolute Risk Reduction (ARR) [95% CI] Relative Risk (RR) [95% CI] Total number of events Cumulative incidence of metastases Median followup: 8.2 years % [5.7 to 11.6]* at 5 years 15.2% [11.4 to 20.3] at 10 years % [7.1 to 13.5] at 5 years 25.4% [20.4 to 31.5] at 10 years ARR: 1.7 [-2.5 to 6.0] at 5 years ARR: 10.2 [3.1 to 17.2] at 10 years RR: 0.60 [0.42 to 0.86] at 10 years Radical prostatectomy versus radiation therapy Paulson, Radical prostatectomy (n=47) Radiation therapy (n=59) Total number of events at followup (time unclear) 2 (positive bone scan) 14 (positive bone scan (11); pulmonary metastases (1); lymph node metastases (1); parenchymal metastases (1) Radical prostatectomy with or without neoadjuvant therapy Klotz, Radical prostatectomy (n=101) Neoadjuvant androgen ablation + RP p value (n=112) Total number of events Median followup: 6 years (0.6 to 9.8) Schulman, Radical prostatectomy (n=115, T2 Neoadjuvant androgen ablation + RP p value only) (n=105, T2 only) Total number of events at 4 years 6/114 (5%) 6/102 (6%) 0.84 Soloway, Radical prostatectomy (estimated Neoadjuvant androgen ablation + RP p value n=135) (estimated n=133) Events at 5 years urethral margin involvement seminal vesicle involvement positive lymph nodes 23 (17%) 30 (22%) 9 (6%) 8 (6%) 20 (15%) 8 (6%) <0.01 Not significant Not significant External beam radiation Lukka, Long arm (66 Gy in 33 fractions) EBR (n=470) Short arm (52.5 Gy in 20 fractions) EBR (n=466) Distant failure events 4 (1%) 10 (2%) Median followup: 5.7 years (4.5 to 8.3) Vaccine versus nilutamide Arlen, Vaccine Group (n=21) Nilutamide Group (n=21) Progressive disease (metastasis on scans) 14 (5 events after crossover addition of nilutamide) 7 * 95% Confidence intervals (1 event after crossover addition of nilutamide) 52

63 Table 9. Adverse events and toxicity for randomized controlled trials Study Outcomes Treatment Group Control Group Analyses; p-values Radical prostatectomy with or without neoadjuvant therapy Gleave, months neoadjuvant androgen 8 months neoadjuvant androgen ablation + p value ablation + radical prostatectomy (n=273) radical prostatectomy (n=274) Newly reported adverse events Hot flushes Fatal adverse events (AEs) Severity of AEs Causality of AEs Increased liver enzymes Diarrhea % None Not reported Not reported Not reported Not reported % None Not reported Not reported Not reported Not reported < < External beam radiation Lukka, Long arm (66 Gy in 33 fractions) EBR Short arm (52.5 Gy in 20 fractions) EBR % Difference [95% CI]* (n=470) (n=466) Number of subjects with Acute NCIC** Grade 3/4 toxicity, 5 months Gastrointestinal (GI) system Genitourinary (GU) system GI or GU Number of subjects with Late NCIC** Grade 3/4 toxicity, >5 months GI GU GI or GU 12 (2.6%) 23 (4.9%) 33 (7.0%) 6 (1.3%) 9 (1.9%) 15 (3.2%) 19 (4.1%) 40 (8.6%) 53 (11.4%) 6 (1.3%) 9 (1.9%) 15 (3.2%) -1.5 [-4.0 to 0.8] -3.7 [-7.0 to -0.5] -4.4 [-8.1 to -0.6] 0.0 [-1.7 to 1.6] 0.0 [-1.9 to 1.9] 0.0 [-2.4 to -2.3] Yeoh, Conventional (64 Gy) EBR (n=61) Hypofractionated (55 Gy) EBR group (n=59) p value compared to Series 1 GI Symptoms, patients scoring 1 before RT Abnormal frequency of bowel movements Diarrhea Pain on using bowels Mucous discharge from bowel Urgency of defecation Rectal bleeding 1 month after RT Abnormal frequency of bowel movements Diarrhea Pain on using bowels Mucous discharge from bowel Urgency of defecation Rectal bleeding 27 (44%) 13 (21%) 2 (3%) 4 (5%) 22 (38%) 5 (9%) 43 (69%) 27 (44%) 23 (37%) 17 (27%) 29 (47%) 8 (13%) 31 (53%) 15 (26%) 4 (7%) 5 (9%) 16 (28%) 4 (7%) 43 (75%) 21 (37%) 24 (41%) 22 (39%) 36 (59%) 13 (23%) 53

64 Table 9. Adverse events and toxicity for randomized controlled trials (continued) Study Outcomes Treatment Group Control Group Analyses; p-values 2 years after RT Abnormal frequency of bowel movements Diarrhea Pain on using bowels Mucous discharge from bowel Urgency of defecation Rectal bleeding 32 (58%) 18 (33%) 9 (16%) 17 (31%) 24 (44%) 15 (27%) 30 (59%) 18 (35%) 6 (12%) 10 (20%) 26 (51%) 21 (42%) 64 Gy <0.05; 55 Gy <0.05 GU Symptoms, patients scoring 1 before RT Abnormal urinary frequency by day Abnormal urinary frequency by night Hematuria Urgency of urination Dysuria 1 month after RT Abnormal urinary frequency by day Abnormal urinary frequency by night Hematuria Urgency of urination Dysuria 2 years after RT Abnormal urinary frequency by day Abnormal urinary frequency by night Hematuria Urgency of urination Dysuria 52 (84%) 25 (40%) 2 (3%) 23 (37%) 8 (13%) 56 (92%) 39 (64%) 3 (5%) 31 (51%) 17 (28%) 38 (69%) 23 (42%) 2 (4%) 20 (36%) 5 (9%) 45 (80%) 27 (47%) 3 (5%) 26 (46%) 11(19%) 52 (92%) 41 (72%) 3 (6%) 36 (63%) 21 (37%) 36 (71%) 24 (47%) 2 (4%) 24 (47%) 3 (6%) Zietman, Conventional dose (70.2 Gy) group High dose (79.2 Gy) group (n=195) p value Acute symptoms, RTOG scale 0-4: GU, Grade 1 GU, Grade 2 GU, Grade 3 GU, Grade 4 GI, Grade 1 GI, Grade 2 GI, Grade 3 GI, Grade 4 Late symptoms, RTOG scale 0-4: GU, Grade 1 GU, Grade 2 GU, Grade 3 GU, Grade 4 (n=196) 79 (40%) 82 (42%) 2 (1%) 0 62 (31%) 81 (41%) 2 (1%) 0 85 (43%) 35 (18%) 3 (2%) 0 69 (35%) 95 (49%) 2 (1%) 1 (1%) 48 (25%) 112 (57%) (43%) 39 (20%) 1 (1%) 0 Not significant

65 Table 9. Adverse events and toxicity for randomized controlled trials (continued) GI, Grade 1 GI, Grade 2 GI, Grade 3 GI, Grade 4 Study Outcomes Treatment Group Control Group Analyses; p-values 71(36%) 15 (8%) 1 (1%) 0 Actuarial risk of a GU event of 15% Grade 2 at 3 years Actuarial risk of a GU event of 19% Grade 2 at 5 years Hormonal therapy combined with radiation therapy D Amico, Conformal radiation therapy (70 Gy) Toxicity, number of events: Urinary incontinence (UI), complete: Grade 1 Grade 2 Grade 3 UI, stress: Grade 1 Grade 2 Grade 3 Hematuria: Grade 1 Grade 2 Grade 3 Diarrhea: Grade 1 Grade 2 Grade 3 Rectal bleeding: Grade 1 Grade 2 Grade 3 Impotence (men potent at baseline): Grade 1 Grade 2 Grade 3 Gynecomastia: Grade 1 Grade 2 Grade 3 Liver dysfunction group (n=103) : Grade 3 (1); Grade 4 (1) 84 (43%) 33 (17%) 1 (1%) 0 13% 18% Conformal radiation therapy and androgen suppression therapy group (n=98) p value Not significant for all events unless noted for Grades 1 and 2 combined 55

66 Table 9. Adverse events and toxicity for randomized controlled trials (continued) Study Outcomes Treatment Group Control Group Analyses; p-values Brachytherapy Herstein, / Wallner, I (144 Gy) (n=159) Pd (125 Gy) (n=155) p value Actuarial risk of radiation proctitis Estimated 13% Estimated 8% 0.21 (persistent bleeding) at 5 years 29 events total (n=314) for both groups, trending more toward 125 I Vaccine versus nilutamide Arlen, Vaccine Group (n=21) Nilutamide Group (n=21) Subjects removed from trial due to 0 3 toxicity Injection site reaction 39.7 Percent patients with toxicity Grade 2 toxicity included: arthralgia Grade 2 toxicity included: dyspnea (13.8%); fatigue (10.3%); dyspnea (15.2%); fatigue (15.2%); hot flashes (6.9%) (15.2%) Grade 3 AEs included: cardiac ischemia Grade 3 toxicity included: dyspnea (3%); (3.4%) fatigue (3%) Interleukin-2 Grade 2 toxicity included: fatigue (48.3%); fever (13.8%); arthralgias (6.9%); hyperglycemia (20.7%); lymphopenia (13.8%); dehydration/anorexia (10.3%); diarrhea (10.3%) Grade 3 toxicity included: fatigue (10.3%); fever (6.9%); hyperglycemia (6.9%); lymphopenia (6.9%); dehydration/anorexia (3.4%) * CI = Confidence intervals ** NCIC = National Cancer Institute of Canada 56

67 Figure 1. Overall survival at time points by treatment % Overall Survival 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% Brachy EBR C-EBR RP WW Brachy EBR C-EBR RP WW Brachy EBR RP WW EBR RP WW 5 Years 10 Years 15 Years 20 Years Brachy = Brachytherapy; EBR = External Beam Radiotherapy; C-EBR = Conformal External Beam Radiation, RP = Radical Prostatectomy; WW = Watchful Waiting Point size indicates N, <50 (smallest), (next smallest) (next largest) and >300 (largest) 57

68 1980 % Disease Specific Survival Figure 2. Disease specific survival at time points by treatment 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% Brachy EBR C-EBR RP WW EBR C-EBR RP WW EBR RP WW EBR RP WW 5 Years 10 Years 15 Years 20 Years Brachy = Brachytherapy; EBR = External Beam Radiotherapy; C-EBR = Conformal External Beam Radiotherapy; RP = Radical Prostatectomy; WW = Watchful Waiting Point size indicates N, <50 (smallest), (next smallest) (next largest) and >300 (largest) 58

69 1981 Figure 3. Biochemical no evidence of disease (bned) at time points by treatment (all definitions) % with bne 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% Brachy EBR C-EBR RP Brachy EBR C-EBR RP Years 10 Years Brachy = Brachytherapy; EBR = External Beam Radiotherapy; C-EBT = Conformal External Beam Radiotherapy; RP = Radical Prostatectomy Point size indicates N, <50 (smallest), (next smallest) (next largest) and >300 (largest) 59

70 Table 10. PCOS: Comparison of 5-year responders to urinary, bowel, and sexual questions according to treatment* 56 Domain RP (n=901) EBR (n=286) Odds ratio (95% CI) Urinary No control or frequent leaks vs. total control or occasional leaks 14.4 (15.3) 4.9 (4.1) 4.4 ( ) Leaks 2 times per day 15.6 (16.1) 4.1 (3.6) 5.3 ( ) Wears any pads to stay dry 28.6 (28.6) 4.2 (4.2) 9.4 ( ) Frequent urination more than half the time 10.6 (10.1) 8.9 (9.3) 1.1 ( ) Bothered by dripping or leaking urine 13.9 (14.3) 3.0 (2.6) 6.5 ( ) Bowel Diarrhea 23.3 (23.9) 28.8 (26.7) 0.84 ( ) Painful bowel movement 10.4 (11.5) 12.2 (9.4) 1.31 ( ) Bowel urgency 17.7 (19.3) 33.4 (28.5) 0.56 ( ) Wetness in rectal area 13.8 (14.8) 20.6 (18.3) 0.75 ( ) Painful hemorrhoids 11.0 (10.2) 15.7 (19.6) 0.43 ( ) Bothered by frequent bowel movement to pain or urgency 4.3 (4.8) 5.0 (4.0) 1.23 ( ) Sexual No/little interest in sexual activity 46.5 (48.9) 55.2 (47.4) 1.1 ( ) No sexual activity 48.9 (50.7) 51.3 (43.9) 1.4 ( ) Erection insufficient for intercourse 76.9 (79.3) 73.1 (63.5) 2.5 ( ) Bothered by sexual dysfunction 47.4 (46.7) 42.0 (44.6) 1.1 ( ) * External beam radiation is the referent group. Adjusted percentages are from separate logistic regression models, each adjusted for treatment propensity score, age, baseline function, race, comorbidity, and educational level. All estimates were weighted to total eligible cases. RP = radical prostatectomy, EBR = external beam radiation. Values in columns are unadjusted percentages, in parentheses, adjusted percentages. Percentages and odds ratios for yes versus no/none. For bother items, percentages refer to patients reporting a large or moderate problem versus a small or no problem. For the 5 bowel functions, percentages refer to patients reporting having the problem every day or some days versus rarely or never. 60

71 Figure 4. Bladder complications at time points by treatment % with complication 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% EBR Radical Prostatectomy Brachytherapy EBR Radical Prostatectomy Hematuria Brachytherapy Incontinence Cystitis Urethral Stricture 6 months 7-24 months >24 months Point size indicates N, <50 (smallest), (next smallest) (next largest) and >300 (largest) EBR 61

72 Figure 5. Bowel complications at time points by treatment % with complication 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% EBR 6 months Radical Prostatectomy Rectal Bleeding Diarrhea Rectal Injury Fecal Incontinence Point size indicates N, <50 (smallest), (next smallest) (next largest) and >300 (largest) EBR 7-24 months Radical Prostatectomy 62

73 Figure 6. ED complications at time points by treatment % with complication 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% EBR RP NSP WW Brachy EBR RP Brachy EBR C-EBR 6 months 7-24 months >24 months Point size indicates N, <50 (smallest), (next smallest) (next largest) and >300 (largest) EBR Radical Prostatectomy (RP) Nerve-Sparing Prostatectomy (NSP) Watchful Waiting (WW) Brachytherapy (Brachy) Conformal REB (C-EBR) 63

74 2005 % with complication Figure 7. Complications by treatment 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% Bladder Neck Contracture Brachytherapy Radical Prostatectomy EBR Conformal EBR Nerve-Sparing Prostatectomy Watchful Waiting Hematuria Incontinence Urethral Stricture Point size indicates N, <50 (smallest), (next smallest) (next largest) and >300 (largest) Diarrhea Rectal Bleeding Impotence 64

75 Table 11. Patient outcomes after cryosurgical treatment of localized prostate cancer Author Long Cohen Sosa Coogan Wieder Cox Bahn Shinohara Shinohara Bahn Koppie Donnelly Ellis Aus Wake Year Followup, months NR NR Number of patients * % with T > NR NR NR Mean PSA ng/ml NR NR NR % Gleason (>8) 5(>8) NR NR NR NR Outcomes in patients with localized PC Progression free NR NR NR 92.6 NR NR survival, % % positive biopsy NR NR NR NR NR 1.4 NR NR 25 Mean PSA ng/ml NR NR NR NR NR NR NR NR NR 1.15 % Complications Urethrorectal fistula NR <0.1 NR NR Tissue sloughing NR 23 NR NR NR NR Urethral stricture NR NR NR NR NR NR BOO NR NR NR NR NR NR 21.2 Incontinence NR 4.3 NR NR Impotence 88 NR NR NR NR 94.9 NR NR NR Perineal pain NR 3 NR NR NR NR NR NR 9.6 Urinary tract infection NR NR 3 NR NR NR NR NR 33.3 NR Epididymitis NR NR NR 3 NR NR NR NR NR NR NR Sepsis < NR 3 NR NR NR NR NR NR NR Urethral obstruction NR NR 4 NR NR NR NR NR NR NR Total complications 17 NR NR NR NR NR NR NR NR NR NR NR NR Number of operations 17 NR NR NR NR 43 NR NR NR NR NR NR NR NR NR Any treatment (n) 21 NR NR NR NR NR NR NR NR NR NR NR NR NR NR * Probably same patients NR = not reported BOO = Bladder outlet obstruction 65

76 Figure 8. Comparative studies of LRP vs. RRP; functional and oncologic data (from the systematic review by 83 Rassweiler et al) 2014 Relative risk of outcomes Author (Size) (95% CI) Incontinence (1year) Rassweiler (657) 0.81 (0.42, 1.54) Anastasidis (300) 0.79 (0.45, 1.41) Roumeguere (162) 1.25 (0.55, 2.81) Artibani (121) 2.43 (1.07, 5.55) Keller (150) 0.14 (0.05, 0.38) Overall mortality Salomon (401) 0.57 (0.35, 0.91) Rassweiler (657) 0.40 (0.11, 1.55) PSA-relapse Rassweiler (657) 0.72 (0.43, 1.22) Roumeguere (162) 1.27 (0.40, 4.06) Artibani (121) 1.90 (0.65, 5.51) Positive margins Salomon (401) 0.86 (0.57, 1.29) Rassweiler (657) 0.66 (0.43, 1.02) Roumeguere (162) 0.52 (0.27, 1.01) Artibani (121) 1.36 (0.60, 3.09) Remzi (121) , 2.98) Keller (150) 0.76 (0.27, 2.12) Potency bilat ns. Anastasidis (300) 1.44 (0.84, 2.46) Roumeguere (162) 1.57 ( 0.83, 2.96) Artibani (121) 0.78 (0.22, 2.76) LRP is safer Relative risk of outcomes RRP is safer 66

77 Figure 9. Comparative studies of LRP vs. RRP; operative data (from the systematic review by Rassweiler et 83 al) Relative risk of outcomes Author (size) (95% CI) Complications Salomon (401) 0.71 (0.45, 1.13) Rassweiler (657) 0.68 (0.41, 1.14) Bhayani (60) 1.02 (0.29, 3.64) Roumeguere (162) 0.41 (0.18, 0.95) Artibani (121) 2.35 (1.01, 5.46) Brown (120) 1.49 (0.62, 3.58) Fornara (64) 0.31 (0.03, 3.20) Reintervention Rassweiler (657) 0.60 (0.26, 1.40) Bhayani (60) 0.45 (0.07, 2.79) Roumeguere (162) 0.43 (0.10, 1.78) Artibani (121) 1.05 (0.17, 6.55) Brown (120) 1.00 (0.13, 7.41) Remzi (121) 1.06 (0.20, 5.60) Transfusion Guilloneau (220) 0.25 (0.12, 0.51) Salomon (401) 0.08 (0.04, 0.20) Rassweiler (657) 0.34 (0.23, 0.51) Artibani (121) 3.31 (1.55, 7.05) Brown (120) 0.02 (0.00, 0.12) Fornara (64) 0.31 (0.06, 1.75) Keller (150) 4.57 (0.92, 22.73) Favors LRP Relative risk of outcomes Favors RRP 67

78 Figure 10. Comparative studies of transperitoneal vs. extraperitoneal radical prostatectomy; operative data 83 (from the systematic review by Rassweiler et al) Relative risk of outcomes Author (size) 95% CI) Complication Cathelineau (200) 1.53 (0.77, 3.08) Erdogru (106) 2.54 (0.88, 7.29) Brown (156) 0.90 (0.27, 2.95) Remzi (80) 0.43 (0.16, 1.19) Positive margins Hoznek (40) 0.53 (0.11, 2.60) Cathelineau (200) 1.87 (0.93, 3.76) Erdogru (106) 0.89 (0.35, 2.26) Brown (156) 1.19 (0.47, 3.00) Remzi (80) 1.42 (0.49, 4.08) Reintervention Remzi (80) 3.34 (0.33, 33.54) Transfusion Hoznek (40) 1.59 (0.24, 10.70) Cathelineau (200) 1.35 (0.29, 6.18) Ruiz (330) 0.21 (0.04, 1.00) Erdogru (106) 0.83 (0.28, 2.50) Favors LRP Relative risk of outcomes Favors RRP 68

79 Table 12. Outcomes after laparoscopic extraperitoneal, transperitoneal, robot assisted, and radical 84 retropubic prostatectomy (from systematic review by Tooher et al) Outcome (Number of Studies) Effect Size % Conversions to open procedure (12) 0 14 (median 2) % Total complications Transperitoneal LRP vs. RRP (6) 0 25 (median 17) vs (median 19) Extraperitoneal endoscopic RP vs. RRP (3) (median 14) vs (median 20) Robot assisted RP vs. RRP (2) vs Blood loss (ml) Transperitoneal LRP vs. RRP (5) 317 1,100 (median of means 800) vs. 1,325 1,550 (median of means 1,400) Extraperitoneal endoscopic RP vs. RRP (2) 795 vs. 829, 522 vs. 1,514 Robot assisted RP vs. RRP (3) (median of means 153), 418 1,330 (median of means 910) % Transfusions Transperitoneal LRP vs. RRP (5) 0 18 (median 2) vs (median 26) Extraperitoneal endoscopic RP vs. RRP (1) 34 vs. 64 Robot assisted RP vs. RRP (3) 0 28 (median 0) vs (median 38) Operative time (minutes) Transperitoneal LRP or extraperitoneal endoscopic RP vs. RRP (10) (median of means 288) vs (median of means 168) Robot assisted RP vs. RRP (4) (median of means 231) vs (median of means 214) Length of stay (days) Transperitoneal LRP vs. RRP (8) 2 12 (median of means 5) vs (median of means 7) Extraperitoneal endoscopic RP vs. RRP (1) 7.2 vs Robot assisted RP vs. RRP (3) (median of means 1.2) vs (median of means 2.7) Catheterization (days) Transperitoneal LRP vs. RRP (5) (median of means 7) vs (median of means 15) Extraperitoneal endoscopic RP vs. RRP (2) (median of means 8) vs (median of means 14) Robot assisted RP vs. RRP (3) (median of means 7) vs (median of means 13) % Positive margins Transperitoneal LRP vs. RRP (8) (median 23) vs (median 29) Extraperitoneal endoscopic RP vs. RRP (3) (median 26) vs (median 24) Robot assisted RP vs. RRP (4) 6 26 (median 16) vs (median 20) % Recurrence-free survival Transperitoneal LRP vs. RRP (2) vs Extraperitoneal endoscopic RP vs. RRP (2) vs Robot assisted RP vs. RRP (2) vs Continence (% pad free) Transperitoneal LRP vs. RRP (3) (median 80) vs (median 89) Extraperitoneal endoscopic RP vs. RRP (3) (median 81) vs (median 83.9) Robot assisted RP vs. RRP (3) (median 78) vs (median 75) % Potency Transperitoneal LRP vs. RRP (1) 41 vs. 30 Extraperitoneal endoscopic RP vs. RRP (2) vs Robot assisted RP vs. RRP (2) vs * PSA 0.2 m/l or greater on more than 1 occasion LRP= laparoscopic radical prostatectomy RRP= retropubic radical prostatectomy 69

80 Table 13. PCOS: Distribution of patient responses at 24 month followup by treatment* 88 Characteristic WW** (n = 230) % AD** (n = 179) % EBR/Brachy** (n = 583) % RP** (n = 1373) % p value Satisfied with treatment 0.00 Delighted Pleased Mostly satisfied Mixed Dissatisfied/unhappy/feel terrible Would make same decision again 0.02 Definitely yes Probably yes Definitely/probably not Free of PC 0.00 No Don t know Yes Bowel urgency 0.00 Rarely or not at all Some days Almost everyday Urinary leakage 0.00 Not at all Once per week or less Daily or more often Erectile dysfunction 0.00 None or only a little Some or a lot No erections at all Cancer or treatment limits activities 0.00 None A little Some/a lot Cancer or treatment caused financial problems 0.00 None A little Some/a lot Cancer or treatment affects relationships with spouse/friends 0.00 None A little Some/a lot General health 0.00 Excellent Very good Good Fair or poor * Weighted to reflect all patients in the study. ** WW = watchful waiting; AD = androgen deprivation; EBR/Brachy = external beam radiation/brachytherapy; RP = radical prostatectomy 70

81 Table 14. PCOS: Percent comparison of 24-month survey older responders on urinary, bowel, and sexual 98 items Domain and Survey Items *Conservative AD/WW** (n=290) Aggressive RP/EBR/ Brachy** (n = 175) Odds Ratio (95% CI) p Value Urinary No control of frequent leakage ( ) 0.17 Leaks more than once daily ( ) 0.01 Frequent urination more than half the time ( ) 0.46 Bothered by dripping or leaking urine ( ) 0.02 Bowel Frequency some/almost all days ( ) 0.04 Bowel urgency some/almost all days ( ) 0.19 Painful bowel movements some/almost all days ( ) 0.24 Bothered by bowel function problems ( ) 0.12 Sexual No/little interest in sexual activity ( ) 0.75 No sexual activity ( ) 0.63 Erections not firm enough for intercourse ( ) 0.16 No erections/a lot of difficult keeping erections ( ) 0.63 Bothered by sexual function problems ( ) 0.01 Percents and odds ratios adjusted for treatment propensity score, baseline function, age, race, education, and comorbidity score * Reference group ** AD/WW = androgen deprivation/watchful waiting; RP/EBR/Brachy = radical prostatectomy/external beam radiation/brachytherapy 71

82 Table 15. PCOS: Percentage overall effects of prostate cancer and treatment 98 Outcome *Conservative AD/WW** (n = 290) Aggressive RP/EBR/ Brachy** (n = 175) Odds Ratio (95% CI) Treatment satisfaction Delighted/very pleased ( ) Make same treatment decision again if given chance Definitely yes ( ) Physical discomfort related to cancer/treatment A lot/some ( ) Health worries related to cancer/treatment A lot/some ( ) Limited daily activities related to cancer/treatment A lot/some ( ) Overall bother related to cancer/treatment A lot/some ( ) Percents and odds ratios adjusted for treatment propensity score, age, race, education, and comorbidity score * Reference group ** AD/WW = androgen deprivation/watchful waiting; RP/EBR/Brachy = radical prostatectomy/external beam radiation/brachytherapy 72

83 Table 16. QOL studies of treatments for localized prostate cancer in nonrcts Study Study Type/Patients/Interventions Quality of Life Outcomes Lee, Prospective cohort study HRQOL scores Measure/ Mean (SD) - Mean (SD) Mean (SD) N = 98 T1/T2 patients enrolled treatment baseline 1 month 12 months Interstitial brachytherapy (IB): n = 44 [I-125, 144 Gy; 11 patients received androgen deprivation therapy before IB, to reduce size of prostate gland] Electron beam radiotherapy (EBRT): n = 23 [median dose = 70.2 Gy; 2 patients received androgen deprivation therapy before EBRT] Radical prostatectomy (RP): n = 23 [3 patients received androgen deprivation therapy before RP] Age [median (range)] IB = 67.1 (49 79) EBRT = 68.8 (51 79) RP = 61.0 (42 68) Response rate All patients = 91% (90 patients) p-value overall within treatment group p-value B vs. 12 months FACT-P IB (17.0) (21.7) (14.2) EBRT (12.1) (21.0) (15.6) RP (14.7) (18.3) (14.9) FACT-G IB (11.8) 92.5 (14.7) (9.1) EBRT 99.9 (7.9) 96.1 (12.5) (10.1) RP 99.8 (9.6) 88.9 (13.2) (11.3) PWB IB 25.9 (2.8) 21.6 (4.7) 25.3 (2.6) EBRT 25.2 (2.2) 22.6 (4.7) 25.1 (4.1) RP 26.3 (2.3) 20.9 (5.0) 26.3 (2.5) FWB IB 23.3 (4.4) 18.9 (6.4) 24.1 (3.7) EBRT 22.9 (3.2) 21.7 (5.0) 23.2 (4.8) RP 23.6 (4.3) 16.5 (5.1) 23.3 (4.1) IPSS IB 8.3 (6.0) 20.8 (7.7) 10.4 (7.3) EBRT 11.9 (6.4) 13.8 (7.5) 8.5 (5.4) RP 12.5 (9.3) 17.2 (10.3) 5.5 (3.7) FACT-P = Functional Assessment of Cancer Therapy Prostate; FACT-G = FACT general; PWB = physical well-being; FWB = functional well-being; IPSS = International Prostate Symptom Score 73

84 2063 Table 16. QOL studies of treatments for localized prostate cancer in nonrcts (continued) Study Study Type/Patients/Interventions Quality of Life Outcomes Schapira, Prospective observational study UCLA-PCI scores Mean score - Domain/treatment N = 122 T1/T2 patients baseline Radical prostatectomy (RP): n = 42 [during the course of the study, 6 patients received hormonal therapy and 2 had radiation therapy] Radiation therapy (RT): n = 51 [during the course of the study, 12 patients received hormonal therapy] Expectant management (EM): n = 29 [during the course of the study, 7 patients received hormonal therapy and 4 had radiation therapy] Age [median (range)] RP = 64 (58-68) RT = 73 (68-75) EM = 73 (70-78) Mean score 3 months Mean score 12 months Urinary bother RP RT EM p-values (across treatment groups) NS NS NS Sexual bother RP RT EM p-values (across treatment groups) NS 0.02* NS Bowel bother RP RT EM p-values (across treatment groups) NS 0.05** NS * RP vs. RT and RP vs. EM ** RP vs. RT Incontinence and impotence rates (based on UCLA-PCI responses) Side-effect/treatment Baseline n (%) 3 months n (%) 12 months n (%) Incontinence RP 1 (2) 31 (76) 16 (44) RT 2 (4) 4 (9) 3 (8) EM 0 (0) 0 (0) 1 (4) Impotence RP 14 (35) 42 (100) 33 (89) RT 30 (61) 33 (77) 30 (75) EM 16 (55) 17 (63) 17 (68) SF-36 scores Domain/treatment Mean score - Mean score Mean score - baseline 3 months 12 months General health perceptions RP RT EM p-values (across treatment groups) NS NS NS 74

85 2064 Table 16. QOL studies of treatments for localized prostate cancer in nonrcts (continued) Study Study Type/Patients/Interventions Quality of Life Outcomes Soderdahl, Prospective longitudinal comparative study UCLA-PCI scores Mean Mean 1 Mean 6 Mean 12 N = 452 T1c/T2 patients recruited [334 responded Domain score - month months months * to baseline questionnaire; 250 completed all baseline surveys] Sexual bother ORP Open radical prostatectomy (ORP): n = 86 Laparoscopic radical prostatectomy (LRP): n = 93 Brachytherapy (Brachy): n = 71 [Pd-103, 115 Gy mean dose] Age [median]: ORP = 59 LRP = 61 Brachy = 68 P-values LRP Brachy P <0.001 [brachy were significantly less bothered by sexual problems than surgery] Urine bother ORP LRP Brachy P <0.001 [RP more bother than brachy] Bowel bother ORP LRP Brachy P = 0.07 [brachy had marginally lower scores (more bother) than surgery] Scale range = SF-36 scores After each treatment, there was an initial drop-off of general domain scores, which returned to baseline over time. Profiles of changes were not different between groups for these scales Physical functioning (P = 0.26); Limitations secondary to emotional problems (P = 0.57) Vitality (P = 0.46); Mental health (P = 0.35); Bodily pain (P = 0.46); General health (P = 0.92); Health transition (P = 0.55). Scale change profile was marginally significantly different between groups for: Limitations attributable to physical problems (P = 0.052) Social functioning (P = 0.06) At 12 months, there were no significant differences among groups. 75

86 2065 Table 16. QOL studies of treatments for localized prostate cancer in nonrcts (continued) Study Study Type/Patients/Interventions Quality of Life Outcomes Potency (sexual function) Mean Mean Mean Mean Treatment score - score - 1 score - 6 score - 12 P-values baseline month months months ORP Bilateral NS Unilateral NS Non NS No statistical difference by ORP LRP Bilateral NS Unilateral Non NS P < [baseline sexual function was worse in non-ns LRP] 76

87 2066 Table 16. QOL studies of treatments for localized prostate cancer in nonrcts (continued) Study Study Type/Patients/Interventions Quality of Life Outcomes Fulmer, Prospective cohort study Voiding and sexual function scores Baseline - 3 months - fold Domain/ N = 127 T1/T2 patients mean score change treatment (SE) (95% CI)* Hormonobrachytherapy with external beam radiotherapy (HBTC): n = 40 [high-risk patients] [received Gy of EBRT followed by brachytherapy + antiandrogen + LHRH agonist] Hormonobrachytherapy (HBT): n = 45 [low-risk patients] [received brachytherapy + antiandrogen + flutamide] Radical retropubic prostatectomy (RP): n = 42 Age [median] HBTC = 68 HBT = 66 RP = 59 Median followup (time since surgery) 8.2 months (range months) 6 months - fold change (95% CI)* 12 months - fold change (95% CI)* AUASS RP 7.17 (0.98) HBTC 7.57 (1.06) 1.66 ( ) 1.55 ( ) 1.36 ( ) HBT 6.69 (0.93) 1.40 ( ) 1.38 ( ) 1.32 ( ) IAUA RP 3.87 (0.42) HBTC 4.63 (0.61) 1.26 ( ) 1.26 ( ) 1.27 ( ) HBT 4.10 (0.48) 1.02 ( ) 1.04 ( ) 1.10 ( ) OAUA RP 3.32 (0.62) HBTC 3.25 (0.70) 2.48 ( ) 2.18 ( ) 1.68 ( ) HBT 2.64 (0.55) 2.28 ( ) 2.17 ( ) 1.95 ( ) VB RP 3.58 (0.55) HBTC 4.03 (0.58) 0.76 ( ) 0.85 ( ) 1.08 ( ) HBT 3.30 (0.48) 0.71 ( ) 0.78 ( ) 0.96 ( ) SF RP 4.24 (0.73) HBTC (n = 36) 9.75 (1.21) ( ) HBT (n = 43) 7.69 (0.96) ( ) SFB RP 3.42 (0.60) HBTC (n = 36) 6.36 (0.76) ( ) HBT (n = 43) 4.90 (0.70) ( ) AUASS = AUA symptom score; IAUA = AUA Irritative subscale; OAUA = AUA Obstructive subscale; VB = Voiding Bother; SF = Sexual function; SFB = Sexual Function Bother score *Relative change or fold change of given BT group compared to RP (BT/RP). Numbers lower than one indicate lower scores in BT than in RP. 77

88 Table 16. QOL studies of treatments for localized prostate cancer in nonrcts (continued) Study Study Type/Patients/Interventions Quality of Life Outcomes Galbraith, Longitudinal survey study Health-related quality of life score Treatment Baseline 6 months 18 months N = 185 localized prostate cancer patients Watchful waiting Surgery Watchful waiting (WW): n = 30 Conventional radiation Mixed-beam radiation Surgery (S): n = 59 Proton-beam radiation p value Conventional radiation (CR): n = 25 Higher score = higher QOL (range = 0-100) Proton beam therapy (PB): n = 24 Mixed-beam radiation (MB): n = 47 Age [mean] WW = 73 S = 65 CR = 71 PB = 68 MB = 69 Physical functioning Treatment Baseline 6 months 18 months Watchful waiting Surgery Conventional radiation Mixed-beam radiation Proton-beam radiation p value <0.05 (PB vs. WW) Higher score = higher functioning (range = 0-100) General health Treatment Baseline 6 months 18 months Watchful waiting Surgery Conventional radiation Mixed-beam radiation Proton-beam radiation p value <0.001 (MB vs. WW) <.0.01 (S vs. WW) Higher score = better health (range = 0-100) 78

89 Table 17. QOL studies of treatments for localized prostate cancer in RCTs Study Study Type/Patients/Interventions Fransson, Randomized controlled trial N = 166 T1/T2 patients enrolled [30 patients died by date of questionnaire submission; 11patients were excluded due to followup <6 months or progressed disease with progressive treatment before questionnaire] Radiation therapy (RT): n = 59 Deferred therapy (DT): n = 49 Age [mean (range)] RT = 71.3 ( ) DT = 72.8 ( ) Median followup (from randomization date to time of questionnaire) RT = 40.6 months DT = 30.4 months [p = 0.055] Response rate RT = 90% DT = 85% Quality of Life Outcomes EORTC QLQ-C30 (+3) Questionnaire RT mean DT mean Domain p value scores scores Global health/qol Scores were linearly transformed to a 0 to 100 scale according to EORTC recommendations Increasing score = higher level of functioning QUFW94 Questionnaire Variable RT mean score DT mean score (95% CI) (95% CI) p value General function Limitation to daily activity caused by prostate carcinoma 1.8 ( ) 0.7 ( ) Urinary symptoms Incontinence 1.5 ( ) 0.6 ( ) Limitation in daily activity caused by urinary problems 1.1 ( ) 0.9 ( ) 0.06 Urinary problems in general 1.8 ( ) 1.2 ( ) Intestinal symptoms Limitation in daily activity caused by intestinal problems 1.3 ( ) 0.3 ( ) Planning of daily activity caused by intestinal problems 1.9 ( ) 1.0 ( ) QUFW94 scores range from 0 (no problem/very good function) to 10 (many problems/very bad function) 79

90 Table 17. QOL studies of treatments for localized prostate cancer in RCTs (continued) Study Study Type/Patients/Interventions Steineck, Randomized controlled trial: SPCG-4 N = 376 T0/T1/T2 patients enrolled Age [mean (range)]: RP = 64.1 ( ) WW = 64.8 ( ) Median followup (from randomization date to time of questionnaire) RP = 50.0 months WW = 48.0 months Sexual Dysfunction Function - outcome Quality of Life Outcomes RP [n / N responding (%)] WW [n / N responding (%)] Unadjusted relative risk (95% CI) Desire Importance of sexual function no or little importance 79 / 159 (50) 64 / 154 (42) 1.2 ( ) Penile stiffness Erectile function - seldom or never sufficient for intercourse 129 / 161 (80) 71 / 158 (45) 1.8 ( ) Distress from erectile dysfunction moderate or great distress 90 / 155 (58) 65 / 152 (43) 1.4 ( ) Distress from erectile dysfunction great distress 46 / 155 (30) 26 / 152 (17) 1.7 ( ) Intercourse Distress from decreased frequency moderate or great distress 94 / 160 (59) 66 / 153 (43) 1.4 ( ) Orgasm Distress from decreased frequency moderate or great distress 88 / 158 (56) 65 / 152 (43) 1.3 ( ) Distress from compromised sexuality Distress (if sexual function has declined) moderate or great distress 87 / 156 (56) 62 / 154 (40) 1.4 ( ) For each question, some men did not respond. A number above unity indicates better function in the watchful waiting group. Urinary Tract Dysfunction Function - outcome RP [n / N responding (%)] WW [n / N responding (%)] Unadjusted relative risk (95% CI) Global features Distress from obstructed voiding moderate or great distress 34 / 164 (21) 34 / 157 (22) 1.0 ( ) Urinary leakage Subjective estimation of the degree of leakage moderate or severe leakage 30 / 163 (18) 3 / 152 (2) 9.3 ( ) 80

91 Table 17. QOL studies of treatments for localized prostate cancer in RCTs (continued) Study Study Type/Patients/Interventions Quality of Life Outcomes Distress from urinary leakage moderate or great distress 47 / 164 (29) 15 / 158 (9) 3.0 ( ) Regular dependence on some form of protective aid - yes 71 / 165 (43) 16 / 154 (10) 4.1 ( ) Regular dependence on diaper or urine bag - yes 23 / 165 (14) 1 / 154 (1) 21.5 ( ) Overall distress from all urinary symptoms moderate or great distress 44 / 163 (27) 28 / 157 (18) 1.5 ( ) For each question, some men did not respond. A number above unity indicates better function in the watchful waiting group. Psychological symptoms RP [n / N WW [n / N Unadjusted Function - responding responding relative risk outcome (%)] (%)] (95% CI) Physical function Decreased general physical capacity the lowest five of seven possible categories 89 / 164 (54) 89 / 157 (57) 1.0 ( ) Low or moderate physical well-being the lowest five of seven possible categories 68 / 164 (41) 78 / 157 (50) 0.8 ( ) Psychological function Worry (moderate or high) - the highest five of seven possible categories 64 / 164 (39) 71 / 157 (45) 0.9 ( ) Low or moderate psychological wellbeing the lowest five of seven possible categories 57 / 164 (35) 57 / 158 (36) 1.0 ( ) Low or moderate subjective quality of life the lowest five of seven possible categories 64 / 159 (40) 68 / 151 (45) 0.9 ( ) For each question, some men did not respond. CI denotes confidence interval. 81

92 Table 17. QOL studies of treatments for localized prostate cancer in RCTs (continued) 2071 Study Study Type/Patients/Interventions Bowel function Function - outcome Quality of Life Outcomes RP [n / N responding (%)] WW [n / N responding (%)] Unadjusted relative risk (95% CI) Fecal leakage once a week or more 1 / 164 (1) 9 / 157 (6) - Distress from fecal leakage moderate to great 3 / 164 (2) 7 / 155 (5) - Distress from all bowel symptoms 5 / 159 (3) 10 / 156 (6) - 82

93 Key Question 2: How do specific patient characteristics, e.g. age, race/ethnicity, presence or absence of comorbid illness, preferences (e.g. tradeoff of treatment-related adverse events vs. potential for disease progression) affect the outcomes of these therapies, overall and differentially? Treatment Decisions According to Patient Factors Factors that influence patient treatment preferences are poorly understood. A systematic review of patient decisionmaking for localized prostate cancer treatment indicated that many factors are incorporated into the decision process, including cancer eradication, adverse effects, physician recommendations, convenience, and costs. 3 The various weights that patients attributed to these factors in decisionmaking and/or satisfaction with treatment outcomes is difficult to determine. Based on indepth semistructured interviews of 20 men with newly-diagnosed clinically localized prostate cancer, one study concluded that treatment preferences were not based on careful assessment of numerical risks for various clinical outcomes. Instead, feelings of fear and uncertainty contributed to a desire for rapid treatment; preferences were influenced by misconceptions, especially about RP and anecdotes about the experiences of others with cancer. 100 Few patients sought second opinions. At 6-8 months of post treatment followup, justification for treatment choices was based on similar anecdotal influences and misconceptions that were present during their initial treatment deliberations. Results described in Question 1 (Quality of Life) indicate that frequency, severity, and bother associated with adverse events varied by baseline general and condition specific functional status and whether the adverse condition existed at baseline. Despite patients having more adverse effects and condition specific bother with early intervention, they generally were as satisfied with treatment and likely to choose this therapy again if they had another chance. In question 3 we describe that treatment recommendations vary according to physician specialty. Treatments for Localized Prostate Cancer by Race While some studies identified differences in the rate of prostate cancer outcomes by racial or ethnical groups, the purpose of this section is to highlight the evidence that race or ethnicity might modify the effect of treatments on outcomes. Since no randomized trials investigated the role of patient race or ethnicity on the efficacy and adverse events associated with localized prostate cancer and its treatment, we are left with only observational studies. Confounding from observational studies is a concern since observed differences in health status across and within racial or ethnic groups is likely due to a complex interaction of numerous factors, most of which are unmeasured and therefore impossible to control for statistically. Results from observational studies. Few studies reported treatment by racial or ethnic groups and most of these studies were limited by the small numbers of non-white participants. 88, Also several of the larger studies included participants who received one of several types of treatments; therefore, the number of participants receiving any one type of treatment was sometimes small even before stratifying by race and ethnicity. Radical prostatectomy. Several studies have failed to find evidence that race or ethnicity has an impact on treatment efficacy 104,106 or adverse effects 107 related to RP. Powell et al found no evidence of a difference 83

94 in biochemical recurrence following a RP between African American and white men with organconfined prostate cancer. 106 In multivariable analyses of 278 men (African American=100 and Caucasian=178) with organ-confined disease, after adjusting for Gleason score and PSA, race was not a statistically significant predictor (p=0.41). In a study of 693 men (44 percent African American and the rest Caucasian) treated at the same cancer institute of whom 391 were treated with RP, there was no significant evidence that African American men had a differential biochemical disease free survival compared to Caucasian men. 104 In multivariate analyses of disease free survival that included stage, PSA, Gleason score, and procedure (combined EBRT and RP) the Cox s proportional hazards ratio for race (African American/Caucasian) was not significant (1.22, 95 percent CI 0.87 to 1.72). Finally, in a cohort study of 802 men of whom 385 received RP (white=285 and African American=92), there was no statistically significant difference in potency between white or African American men. 107 A few large studies have published treatment efficacy, 101 treatment satisfaction, 88 and adverse effect 108 differences in subgroup analyses by race or ethnicity, including the Prostate Cancer Outcomes Study (PCOS) 88,108 and a pooled analysis of nine U.S. military medical centers. 101 In a study of 3,162 men (African American=626 and white=2,299) with localized prostate cancer who were treated with RP, African American men had a somewhat greater risk of biochemical recurrence than white men HR=1.22 (95 percent CI 1.03 to 1.45, p=0.021), even after controlling for stage, Gleason margin status, and seminal vesicle involvement. 101 This magnitude of effect is similar to that reported in a smaller study. 104 While the smaller study was not statistically significant, it was significant in the larger study. Within the PCOS study, Stanford et al. found that sexual function following RP varied by race with 38 percent of African American men reporting firm erections at >18 months versus 26 percent of Hispanic and 21 percent of white men (p=0.001). 108 At 60 months after diagnosis, African American men reported better recovery of sexual and urinary function following RP, despite reporting feeling higher level problems. 109 Also, from the PCOS, Hispanic men were somewhat less satisfied with RP than either African American or white men (p=0.05). 88 These divergent findings between level of functional loss and perceived problem suggest that racial/ethnic differences in the perception and reporting of functional problems may exist. Electron beam radiation therapy (EBRT). In a cohort study of 467 men with localized prostate cancer between the ages of 46 and 82 (of whom a quarter were African American and three quarters were white) who were treated with definite radiotherapy, race was not a significant factor in biochemical relapse free survival. 105 In another study of 893 men treated with conformal radiotherapy within one large department, while African Americans presented with more advanced disease, within similar risk strata African American men had statistically similar five-year bned. 103 In a third study of 693 men (44 percent African American and the rest Caucasian) of whom 302 men were treated with conformal radiation, African American men did not have a differential biochemical disease free survival compared to Caucasian men. 104 As noted above in the section on RP, while the effect of race in this study in multivariate analyses was not statistically significant when EBRT and RP participants were combined, it was similar in magnitude (HR=1.22) to the statistically significant finding from another study. 101 In a cohort study of 802 men, of whom 305 received EBRT, there was some statistically significant evidence (p=0.035) that white men had a greater decrease in potency following EBRT. 107 The reasons for this difference were not clear, and since followup for potency was months, it is not 84

95 possible to know if this difference would persist with longer followup. The PCOS did have followup reported up to 60 months post diagnosis and found no statistically significant evidence of a difference in treatment satisfaction 88 or functional outcomes 109 by race or ethnicity for men who received EBRT. The only exception was a borderline (p=0.05) better bowel function in African American men. Watchful waiting. Only three studies stratified results by race and included participants who 88,102,107 were either enrolled in WW or had no treatment (a combined total of 607 participants). None of these studies found statistically significant differences between racial groups. With such a small number of participants over three studies focused on different outcomes, there is not adequate information to tell whether there are meaningful racial differences with WW. In the first study of 313 men who chose WW (white=209, African American=76 and Asian or Hispanic=19), there was no significant race effect on the likelihood to have a secondary treatment. In Cox proportional hazards models adjusted for clinical stage, PSA doubling time, age, PSA at diagnosis, Gleason score, number of comorbidities, and family history of disease, white men had a 1.13 times greater rate of secondary treatments (95 percent CI of 0.73 to 1.76, p=0.586) compared to African American men. 102 In a second cohort study of 802 men who received either RP, EBRT, or WW (n=64), with a reported outcome of erectile function, there was no statistically significant difference in potency between white or African American men treated with WW. 107 Finally, a report from the PCOS study of 230 men who reported receiving no treatment for their prostate cancer found no statistically significant evidence that treatment satisfaction varied by race at 24 months post diagnosis. 88 Androgen deprivation. Only the PCOS included participants treated with androgen deprivation and stratified by racial or ethnic groups. 88 Among the 179 men treated with androgen deprivation, Hispanic men reported lower treatment satisfaction at 24 months post diagnosis (30 percent) than either white (72 percent) or black (57 percent) men (p=0.0014). Summary While there may be differences in the incidence and morbidity of prostate cancer across racial or ethnic groups, there is little evidence of substantial differences in the effects of treatment by racial or ethnic groups. Modest treatment differences in some studies have not been consistently reported in well-powered studies. Future research is needed to better explore potential racial/ethnic differences in the perception and reporting of sexual function and incontinence. There are no randomized controlled trials of such effects. Treatments for Localized Prostate Cancer by Age U.S. population based trends in radical prostatectomy, brachytherapy and external beam radiotherapy for older men were assessed using Medicare data from RP was less frequently used in men older than 70 years than in the past. However, because brachytherapy increased, the total population-based treatment rates changed little over time. Use of any of these interventions increased 15 percent for men ages 65 to 69 but decreased for older men. From RP remained the most common intervention for men years and years of age but decreased by 6 percent and 34 percent respectively during this time period. Among men 75 years and older EBRT was the most common intervention. Brachytherapy was used twice as often as RP, which declined 50 percent during this period. 85

96 Additional emphasis of this section is whether the effects of treatment depend on the patients age. More specifically, we assessed whether age modifies effect of treatments on outcomes. Therefore, the important question is whether there is evidence that either Treatment A s benefit (reduced death) or Treatment A s side effect (increased impotence) is different relative to Treatment B, depending on the age of the patient. RCTs including multiple treatments and a large number of patients with diverse range of ages would be the ideal situations to address this issue. However, this type of evidence is lacking; therefore, we summarize the information that is available and highlight some of the gaps in knowledge. Life expectancy may be a more relevant characteristic to use when deciding on treatment options than age. However, age will be used primarily as a proxy for life expectancy in this review. While individual treatment consideration should include attention to patient life expectancy, which includes the consideration of competing comorbid conditions, life expectancy is not easily obtained and is rarely a characteristic reported in journal articles. Many articles regarding treatments for localized prostate cancer have reported results by age; 28,89,103,105,107,108, however, few studies have included multiple treatments and reported whether age was an effect modifier for the treatment effects. 28,108,112,121,122,127 Fewer still reported differences in survival or biochemical disease free survival. 28,112,122 One study reported on longterm overall and disease specific survival by different age and Gleason histologic strata in men treated with WW who were diagnosed prior to the PSA error. 132 Regardless of age, prostate cancer specific mortality after 20 years of followup was low in men with well differentiated tumors (Gleason 2-4). However, for men with poorly differentiated prostate cancer (Gleason 8-10), death from prostate cancer comprised the majority of deaths within 5-10 years, even those >75 years at the time of diagnosis. As noted in Question 1, the prevalence, severity, and bother related to bladder, bowel, and sexual dysfunction did not appear to differ in men >70 years compared to the entire cohort of PCOS participants. Results from RCTs. There are very limited data from RCTs regarding the role of patient age on the efficacy of treatments for localized prostate cancer. We identified only one study that reported whether or not they found evidence their intervention groups differed based on the age of the participant. 28 In the long-term SPCR Study Number 4 comparing RP (n=347) to WW (n=348), men ranged in age from with some sites only including men <70 years and the remaining sites including men years if they were considered otherwise healthy. In the overall cohort, the WW group was more likely to die from any cause or from prostate cancer. However, the WW group was less likely to experience urinary leakage or erectile dysfunction. 95 In a subgroup analysis of men <65 years compared to men 65 years, the difference in prostate cancer mortality between RP and WW appeared to be primarily in younger men (Figure 11). Differences in adverse events were not reported by age strata. While these results support the idea that RP has benefits greater than WW, particularly in men under 65, the clinical decision is clouded somewhat by the fact that prostate cancer in this study was not primarily PSA identified, as is currently the case in the United States. 86

97 Results from observational studies. Only a few studies that have included multiple treatments and reported whether the effects of the different treatment options differ by the age of the patient. Most of these studies have been small, so the power to detect statistically significant differences is limited. Evidence of treatment effect modification by age for survival. Two studies reported age stratified results comparing the survival or biochemical relapse free survival of patients treated with either surgery or radiation therapy. 112,122 In a combined study of 354 men treated with RP and 253 men treated with EBRT, there was no evidence that biochemical relapse free survival was different for the two treatments. 122 This study tested whether age (<65 vs vs. 70) was an effect modifier of treatment effect for biochemical relapse free survival and found no statistically significant evidence that the treatment effects differed in the three age strata (univariate adjusted p=0.77 and multivariable adjusted p=0.59). In a second larger study including 44 institutions and 7,316 men, patients treated with RP were much more likely (p <0.0001) to be <70 years old (86 percent) compared to men treated with radiation therapy (42 percent). 112 This study found evidence, consistent they claimed, with practice patterns during the time period, that men who were <70 and treated with RT tended to be less healthy than men <70 who were treated with RP. Following PSA failure, men initially treated with radiation therapy had a significantly higher rate of non-prostate cancer related death compared to men who were treated with RT (p=0.03); however, this effect was only seen in men <70 years old (p=0.007) and not in men 70 (p=0.58). Evidence of treatment effect modification by age for sexual function. Three observational studies have reported results for multiple treatments on sexual function stratified by age group. The largest study included 802 men with 52 percent receiving RP, 40 percent receiving EBRT, and 8 percent in the WW group. 107 Across all three groups, baseline pre-treatment potency was lower in the older men (60-70 and >70) compared to the younger men (<60). In general, the EBRT group had somewhat lower baseline potency. Following treatment, the post-treatment potency was substantially reduced in both active treatment groups, and the post-treatment potency levels were similar between RP and EBRT groups. There was no evidence that the effects of the treatments on potency varied by age. The other two studies compared patients with nerve sparing versus patients with partial or non-nerve sparing radical prostatectomy. 121,127 Due to small numbers in the treatment by age strata and a lack of adjustment for baseline differences between the groups, it is not possible to draw robust conclusions from either of these two studies. Summary. In spite of the lack of evidence for an interaction between age and treatment effects, treatment options have different risks and benefits and these different risks and benefits may be more or less desirable depending on a patient s current age, life expectancy, and lifestyle. While there are definitely differences in the incidence and morbidity of prostate cancer based on patient age, and there are differences in the treatments offered to men at different age ranges, there are few studies that directly compare the treatment effects of different therapies across age groups. Evidence from a subgroup analysis of one randomized trial suggests that survival benefits of radical prostatectomy when compared with WW may be limited to men <65 years of age. Practice patterns show radical prostatectomy is the most common treatment option in younger men with localized prostate cancer; however, in older men (>70) radiation therapy and WW become more commonly used treatment options. These differences in practice patterns appear to 87

98 be based more on preferences of the patients and providers that are related to age via lifestyle and life expectancy than particular age independent treatment benefits and side-effects. Figure 11. Cumulative incidence of death from prostate cancer in the two study groups overall (Panel A) and according to age (Panel B). (Source: New England Journal of Medicine, used with permission [Bill-Axelson A, Holmberg A, Ruutu M, et al. Radical prostatectomy versus watchful waiting in early prostate cancer. N Engl J Med (19): ] 28 Copyright 2005 Massachusetts Medical Society 88

99 Key Question 3. How do provider/hospital characteristics affect outcomes overall and differentially (e.g. geographic region and volume)? Background Evidence suggests that provider characteristics, including higher volume, 133 affiliation with academic center, 134,135 and profit status 135,136 are associated with improved quality of care and better outcomes. The association can be partially explained by patient selection, aging and comorbidities, and differences in process of care. 137 Public dissemination of the knowledge that high volume health services provide better patient care for many surgical procedures and conditions resulted in selective referral to such providers. While more patients were treated in high volume hospitals from , operative mortality decreased by percent for cardiovascular diseases, but not for many cancers. 138 Earlier studies reported significant reduction in mortality after transurethral prostatectomy performed in high volume hospitals and surgeons. 139,140 One study found substantial differences in published definitions of volume categories and its effects on surgical mortality and complications after urological cancer procedures. 141 Volume thresholds and patient distributions in low and high volume hospitals are defined for several cardiovascular and oncology operations, but not for prostate cancer 142 The effect size of provider volumes on clinical outcomes in patients with localized prostate cancer is not well established. Because prostate cancer is the second most expensive cancer organ site for Medicare with approximately $8 billion annual expenditure, 143 improved understanding of the role of provider/hospital characteristics is important. Specialty and geographical location of providers influence diagnostic strategies and the management of localized prostate cancer. 54,144,145 Variability in the management of localized prostate cancer is often based on physician opinions and specialty. 144,145 Diagnosis of localized disease is based primarily on a screening of asymptomatic patients. Therefore, differences in screening practices lead to length bias in the stage of tumors detected and referral onward to more likely recommend intervention. Physician recommendations play an important role in patient decisions on treatment preferences. 146 A systematic review of treatment choices for localized prostate cancer concluded that variations in treatment decisions are attributable to differences in physician recommendations more than on patient and tumor characteristics. 3 Recent studies showed that patient and physicians treatment preferences reflect perceived personal factors more than evidence-based recommendations. 3,100 We assessed how provider characteristics impact care for patients with localized prostate cancer. Methods for the Review The Conceptual Model (Figure 12) outlines the hypothesized relationships between the exposures (bold), outcomes (italic bold), and effect modifiers (underlined) variables. Strategy for the Literature Search The following databases were searched to identify reports of the human studies published in English from 1992 to August 2006: The National Library of Medicine via PubMed ; Cochrane Library; CDC Website; Catalog of U.S. Government Publications (U.S. GPO); LexisNexis Government Periodicals Index; Digital Dissertations; and the Agency for Health Care Research and Quality. The MeSH terms, key words, and its combinations are presented in Appendix A. 89

100 Inclusion Criteria for the Studies The following studies were included: administrative reports that measured outcomes in different locations, administrative surveys that measured physician distribution in regions of the United States, and epidemiologic studies that evaluated the association between provider characteristics and patient outcomes and had a control group. Inclusion criteria for the meta-analysis were as follows: studies reporting outcome rates by surgeon and hospital volume categories or relative risk of outcomes between groups with different surgeon and hospital volumes, and studies with reported outcomes rates in different locations in the United States. Exclusion Criteria for the Studies Studies were excluded if they were not published in English, target population was outpatients or patients in long-term care facilities, there was no information regarding provider characteristics, or were administrative reports and single hospital studies with no control comparisons that did not test an associative hypothesis. Assessment of the Methodological Quality of the Studies Assessment of study quality was based on Systems to Rate the Strength of Scientific Evidence scored from 0 (poorest) to 5 (highest). 147 Summated scores were used to establish study quality. The level of evidence was estimated using U.S. Preventive Services Task Force criteria. 148 Data Extraction Study evaluation and data extraction were performed by two researchers. Standard errors, regression coefficients, and 95 percent CI were calculated from reported means, standard deviations, and sample size. 149 Eligibility information and outcomes were collected. Decisions of study eligibility were made with no relation to authors and institutions. 150 Data Synthesis The impact of the provider/hospital characteristics on clinical outcomes was estimated analyzing published evidence of the associations. Since no randomized trials were identified, observational studies were used to calculate the associations between outcomes and provider/hospital characteristics; both crude estimates and estimates adjusted for confounding factors. Relative risks of the outcomes among different providers/hospitals were analyzed. The results of individual studies were summarized with relation to sample size and 95 percent CI. Weighted by the sample size (number of patients and hospitals) odds ratios and 95 percent CI were calculated with fixed and random effects models. 151 Meta-regression models analyzed possible interactions with the year of data collection, databases to measure outcomes, and adjustment for confounding factors. 152,153 The calculations were performed using STATA 154 and SAS 9.2 packages, Proc Mixed. 155 Consistency in the results was tested comparing the direction and strength of association in models with provider variables as continuous (overall trend) and categorical, in studies reporting outcomes rates and adjusted relative risk, and with goodness of fit analysis. Chi squared test were obtained to assess a heterogeneity in study results

101 Results Description of Studies Identified by the Search Strategy Of the 850 potentially relevant references identified (literature search strategy is presented in Appendix A; excluded studies are listed in Appendix B), 91 percent were excluded (Appendix C, Figure C2). We identified 75 eligible original studies; five reports presented a geographic distribution of physicians in the United States, 23 investigated regional variations of screening and treatment of prostate cancer, 18 articles analyzed the differences of physician characteristics on the diagnosis and management of prostate cancer, 15 articles analyzed the impact of physician experience on learning curves of treatment procedures, six tested the association between hospital volume and patient outcomes, and eight evaluated the role of surgeon volume and patient outcomes. All eligible studies were original epidemiologic investigations to test associative hypotheses 147 between provider characteristics and patient outcomes with IIA-III levels of evidence. Study quality varied from percent of the maximum score possible, in the studies that analyzed 147,156,157 the difference in outcomes in regions of the United States to percent of maximum possible in surveys of clinicians to estimated differences in practice patterns, individual preferences, and opinions. The quality scores of the studies that examined the differences in outcomes in regions of the United States averaged 82.2 percent, with 77 percent for the sampling of subjects and 44 percent for the adjustment for confounding factors. Retrospective cohorts, that investigated the association between provider volumes on patient outcomes, averaged 65 percent of the maximum possible quality score (Appendix C, Table C11), with the highest quality in studies that measured morbidity and urinary complications after RP. The quality of the studies that assessed provider volumes was not associated with sponsorship, country, or data sources. The average applicability of the studies was 70 percent, compared with criteria of external 156 validity of nationally representative cohorts. Few authors analyzed the differences between selected patients and target population; as a result the mean quality score for adequacy of sampling was 58 percent. The majority of the authors selected patients from existing databases (Appendix C, Tables C12 and C13) including The National Center for Health Statistics, 158,159 Cancer Surveillance System, and the National Center for Health Statistics and the North American Association of Central Cancer Registries (SEER), the Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSURE), 168,169 and Medicare health claims national database. 55, , Several single hospital and multi hospital studies selected patients in clinics to analyze medical records (Appendix C, Tables C14-C22). One study obtained the Quebec Healthcare Plan database to identify eligible subjects, 182 one study selected participants within the State Cancer control map, 183 and one study was conducted in the Veteran s Affairs Medical system. 59 Few authors reported random sampling of subjects. 135,165,166 We compared time periods when patient events occurred and databases the authors obtained to select participants (Appendix C, Tables C23-C25) to avoid including the same patients more than once in the analysis. Studies of geographical variations adjusted for patient age, 54,55,59,135, ,168,169,172,174,184 race or ethnicity, 54,135,159,161,162,169,172,174 comorbidity, 55,135,171,174 cancer stage and grade 135,159,160,164 (Appendix C, Tables C12-C13). Studies that investigated provider volumes of RP adjusted for patient age and comorbidity, ,164,165,167,170,178 race, 135,136,167,170,176 cancer stage and 91

102 grade, 166,167,176 provider location and teaching status, 134,135,164,170,180,182 and clustering of patients and providers. 139,140,167,185 Authors stated that the target population included patients with localized prostate cancer, 162 reported the number of participants with localized cancer, adjusted for cancer stage and grade, 135,159,161,162,166,167,176 or assumed that all patients treated primarily with RP would have localized disease. 182 Pooled adjusted rates and relative risk of patient outcomes estimated the association with provider location and volumes independent of cancer stage. Interaction models examined the effect modification by adjustment for cancer stage. Association between provider specialty and prostate cancer screening and diagnosis. Different screening practices can result in variation in the incidence and stages of detected prostate cancer across various patient characteristic categories and consequently affect treatment selection and outcomes. Several physician surveys found differences in prostate cancer screening and referral 144,145, (Appendix C, Table C26) according to provider characteristics. Members of the Academy of Family Physicians (113 family physicians and 238 general internists) were asked which test they would recommend for prostate cancer screening for patients 50 years old and older. 186 Physician preferences were different for all screening practices (Figure 13): family physicians more often recommended digital rectal examination (87 percent vs. 69 percent of positive responses) and PSA testing (67 percent vs. 40 percent of positive responses, p<0.05) and would screen patients in all ages. General internists prefer to refer men with elevated PSA to urologists rather than repeat PSA tests at 4-6 weeks. Board certified physicians in three states (231 urologists and 205 family physicians) responded on their screening preferences in older and asymptomatic patients (Appendix C, Figure C3). 187 Urologists believe that PSA is the best screening test to detect prostate cancer but would generally not screen males older than 70 years and asymptomatic patients less than age 50. Another survey conducted in a random sample of the American Medical Association Registry of Physicians (444 primary care physicians and 394 urologists) 144 showed that more than half of clinicians perform PSA testing as a part of routine health care in patients years of age with higher prevalence of using PSA among urologists compared with primary care physicians. Radiation oncologists more often than urologists recommended that primary care physicians include PSA testing as a part of the routine examination as reported in a survey of 504 urologists and 559 radiation oncologists randomly selected from the American Medical Association Registry of Physicians. 145 Radiation oncologists recommended that primary care clinicians include PSA for males years of age (77 percent of positive responses vs. 51 percent among urologists, p<0.05) and older than 80 years (43 percent of positive responses vs. 16 percent among urologists, p<0.05) (Appendix C, Figure C4). Association between provider specialty and prostate cancer management. Treatment recommendations and beliefs in effectiveness were associated with clinician specialty, training, experience, and gender. Nearly all urologists responded that RP provided a survival benefit for patients with localized prostate cancer and life expectancy more than 10 years (Figure 14). 144 Primary care physicians believed EBRT offered a survival benefit for patients independent of life expectancy. Radiation oncologists responded that radiation therapy offers better survival for patients with localized disease and baseline life expectancy of more than 10 years (Figure 15). In contrast, urologists responded that RP is better than radiation (17 percent of positive responses among urologists compared with 2 percent among radiation oncologists, p<0.001). 145 Urologists and radiation oncologists differed regarding survival benefit of treatment options for patients 92

103 with localized prostate cancer. For instance, 93 percent of urologists believe that RP is preferred compared with EBRT (vs. 20 percent of radiation oncologists). Moreover, 82 percent of radiation oncologists reported that RO is overused as a potentially curative treatment, while brachytherapy and EBRT are underused (Figure 16). Urologists who performed more than ten RPs in residency recommended this treatment more often (53 percent vs. 21 percent) compared to urologists who performed less than ten RPs in residency. 189 Older physicians recommend noncurative approaches, including WW and hormone therapy, more often than younger colleagues (adjusted OR 1.1; 95 percent CI ). 190 Female general practitioners referred elderly patients with prostate cancer to a specialists more often than male general practitioners (OR 2.3, p=0.03) after adjustment for patient and physician characteristics. 190 Several surveys found little consensus about which treatment provides the best benefits for patients with localized prostate cancer (Appendix C, Table C27). 188, Clinical oncologists favored radical radiotherapy for localized prostate cancer (52 percent of positive responses) and for poorly differentiated local tumor (77 percent of positive responses). 190 Urologists and radiation oncologists in Florida recommended EBRT and early androgen deprivation for patients with <10 years of life expectancy more often compared to their colleagues in the Northeast, Midwest, and Western regions of the United States. 188 Recommended diagnostic and treatment procedures differed from evidence-based guidelines and suggested overuse of computerized tomography, pelvic magnetic resonance imaging, radionuclide bone imaging, and high prevalence of hormone therapy in young patients. 191,192 One large retrospective cohort 193 reported that the proportion of variation in use of androgen deprivation therapy attributable to urologists (22.6 percent) was more than tumor (9.7 percent) or patient characteristics (4.3 percent). The significant influence of urologists personal opinions on use of ADT increased over time from 16 percent in 1992 to 25.6 percent in For patients receiving EBRT who had T3 tumors or who had T2 tumors with high grade histology, approximately 12 percent of urologists had significantly lower prescribing rates and 5 percent had higher prescribing rates than the mean rate of 71 percent of patients receiving ADT (range = percent). For patients not in this group ( uncertain-benefit group ) the mean prescribing rate was 36 percent with a range from 5 percent to 92 percent. One-quarter of urologists had a rate of ADT use that was significantly different from the mean. Association between physician characteristics and patient outcomes. Several studies (two non randomized interventions, 194,195 three prospective cohorts, and two surveys, 199,200 examined the association between physician characteristics with patient outcomes (Appendix C, Table C28). Implementation of evidence based physician education programs reduced lengths of stay and total hospital charges in patients undergoing RP. 194,195 Patients receiving care in a health maintenance organization (HMO) were treated more often with radiation therapy (OR 2.99, 95 percent CI ) and had lower mortality (OR 0.3, 95 percent CI ) compared with those treated in fee-for-service settings independent of patient characteristics. 201 Medicare beneficiaries were treated with RP more often compared with HMO members and had lower mortality (OR 0.8, 95 percent CI ). 198 Studies that examined the learning curves of treatment procedures in patients with prostate cancer included a single surgeon experience and did not test the association with provider 93

104 characteristics; rather they reported the improvement in outcomes in early experience (Appendix C Table C29). Some evidence suggests that the operative time for laparoscopic RPs was less among senior compared with junior surgeons. 208,209 Experienced surgeons had less crude, but not adjusted rates, of positive margins after laparoscopic RP. 209 Rate of positive margins after performed procedures reduced from percent in one surgeon (150 previous RPs) but increased from percent in another (600 previous RPs). 205 Relative risk of acute and prolonged acute urinary retention was less after 400 brachytherapy procedures by percent. 210 In summary, evidence from observational studies suggests substantial differences in physicians screening and treatment recommendations; that is partly related to clinician specialty, age and experience, gender, and clinical settings. How Does Geographic Region Affect Outcomes? Provider availability in geographic regions of the United States. Distribution of urologists and radiation oncologists at a state level was obtained from surveys conducted by the American Medical Association from (Appendix C, Tables C30-C32 and Figures C5-C6) The ratio of physicians per 100,000 adult citizens in each U.S. Census region in was largest in the Middle Atlantic and lowest in the West North (Appendix C, Figures C7 and C8). Screening and diagnosis of prostate cancer in U.S. regions. Differences in PSA testing were derived from the Behavioral Risk Factor Surveillance System Survey Data in U.S. regions from (Appendix C, Figure C9 and Appendix C, Tables C29 and C30). 217 Participants were asked whether they ever or within the past 2 years had PSA testing. Prevalence of PSA testing was higher in the South and lower in North East regions (Appendix C Figure C10). PSA testing prevalence did not correlate with a distribution of urologists and radiation oncologists (p = 0.17 and 0.36 respectively). Incidence of prostate cancer in U.S. regions varied with the data source. Three large nationally representative cohorts examined the incidence of prostate cancer in U.S. regions 160,184,218 as well as the U.S. Cancer Statistics data from (Appendix C, Table C11). Results from the U.S. Cancer Statistics (Appendix C, Figure C11) and from individual studies compared with the CDC data are presented in Appendix C, Table C35. Incidence differed among regions with the highest in Middle Atlantic and lowest in Mountain regions (Appendix C, Figure C12). Incidence of prostate cancer in the Middle Atlantic was significantly higher (24.3±12 per 100,000 males) compared with the national average (Appendix C, Table C36). Increased incidence was greater among Hispanics (36.1±13.9 per 100,000 males (Appendix C, Table C37). The incidence was significantly lower in African-American males, residents of Mountain and West North regions, compared with the national average among Blacks by 43.4±13.1 and by 33.1±12.9 per 100,000 males, respectively. The year when cancer was diagnosed did not modify the associations between incidence and location. Two cohorts 160,218 examined incidence of localized prostate cancer in U.S. regions (Appendix C, Table C38), one included African-American men only. Incidence of localized disease was the highest in Middle Atlantic and East North and lowest in East South (heterogeneity not significant), but the differences were not statistically significant. Patient race and the time when the cancer was diagnosed did not modify the association between incidence and location. 94

105 Incidence of prostate cancer was not correlated with the number of urologists or radiation oncologists in U.S. regions (Figure 17) with or without stratification for race. Regional variation in incidence of prostate cancer was not correlated with regional differences in the prevalence of PSA testing (Appendix C Tables C39-C41); however, there was little variation in testing between regions. In contrast with ecologic analysis, individual studies reported that prostate cancer incidence reflected screening prevalence, even though difference in incidence diminished over time. 173 Treatment options for localized prostate cancer in U.S. regions. The prevalence of treatment options for patients with localized prostate cancer varied substantially among U.S. regions (Figure 18 and Appendix C, Table C.20). The prevalence of therapies was not associated with the number of urologists and radiation oncologists despite a borderline significant negative tendency for lower use of RP in regions with higher numbers of radiation oncologists (correlation coefficient -0.7, p=0.05)(appendix C, Table C.43). External Beam Radiation Therapy Four studies examined the percentage of patients with prostate cancer treated with external beam radiation as the first treatment option (Appendix C, Table C13). 55,158,167,168 The probability of receiving EBRT as a primary treatment was the lowest in Mountain and the highest in New England (Appendix C, Figure C13) (p for heterogeneity <0.05). The largest differences in prevalence of EBRT were observed between North East and West (11 percent, 95 percent CI percent) and Midwest (-7.8 percent, 95 percent CI -6 to -9 percent) (Appendix C, Table C44). Two studies 160,184 examined the prevalence of EBRT as a primary treatment (Appendix C, Figure C14) (p for heterogeneity not significant) with small nonsignificant differences in U.S. regions. Three studies evaluated the prevalence of WW 160,168,169 (Appendix C, Figure C15) reporting the higher prevalence in West, Mountain, and Pacific (Appendix C Table C45), heterogeneity not significant). Brachytherapy. Three studies examined the prevalence of brachytherapy in U.S. regions (Appendix C, Figure C16). 168,169,171 The prevalence of brachytherapy was less than 11 percent in all regions with the lowest 4 percent in Middle Atlantic (p for heterogeneity <0.05) and the highest in West South (mean difference 6.6 percent, 95 percent CI percent) (Appendix C, Table C43). Androgen Deprivation Therapy Three studies examined the prevalence of primary androgen deprivation in U.S. regions. 163,168,169 Middle Atlantic had the lowest prevalence of ADT while West South the highest (Appendix C, Figure C17) (p for heterogeneity <0.05) by 10 percent (95 percent CI percent) (Appendix C, Table C47). One study 168 reported the relative risk of utilization of primary ADT in U.S. regions compared with West (Appendix C, Figure C18). The RR was highest in Pacific and in Mountain (relative risk 1.25, 95 percent CI 1.20, 1.30) and lowest in North East. Radical Prostatectomy. The majority of studies evaluated the probability of receiving RP ,168,169,171, Results differed substantially. Prevalence was highest in Mountain (36 percent) 95

106 and lowest in Middle Atlantic (22.7 percent) (Appendix C, Figure C19). The probability of RP was 12.7 percent greater in Mountain compared with New England (95 percent CI percent). Four studies reported age adjusted rate of RP in U.S. regions (Appendix C, Table C48). 54,59,172,184 Rates were lower than the national average in North East by 56 per 100,000 males (95 percent CI 55-75) and in New England by 58 per 100,000 males (95 percent CI 39-78) (Appendix C, Table C49). Mountain and West regions had higher rates of RP by 38 and 33.5 per 100,000 males respectively. Three large studies evaluated the utilization rate of RPs in U.S. regions, 54,55,59 (Appendix C, Figure C20, and Appendix C, Table C48), two in nationally representative Medicare samples, 54,174 and one in the Department of Veterans Affairs Patient Treatment File and Outpatient Clinic File. 213 All studies reported a consistent (heterogeneity not significant) decrease in utilization of RP by 35 percent (95 percent CI ) in North East and increase by 38 percent (RR 1.38, 95 percent CI ) in West compared with the U.S. average (Figure 19). Despite the difference in treatment utilization, all cause and cancer specific mortality 59,161 were the same across regions (Figure 19). Limited evidence 135 suggests that hospital complications after RP varied in U.S. regions with a 42 percent reduction in West compared with North East (RR 0.58, 95 percent CI ) independent of patient age, race, comorbidities, and hospital type. The rates of anastomatic stricture after RP did not differ in the studied locations. Few studies examined length of hospital stay after RP in U.S. regions (Appendix C, Tables C51). 135,220 Length of hospital stay did not differ in the regions compared to the national average (Figure 20). However, it varied from the lowest in Pacific (3.8 days, 95 percent CI ) to the highest in Middle Atlantic (5.7 days, 95 percent CI ). The cost of RP in various regions did not differ from the U.S. average (Appendix C, Table C52-C53) with the lowest ($14,103, 95 percent CI $4,707-$23,498) in East South and the highest in Middle Atlantic ($20,915, 95 percent CI $11,519-$30,311). The substantial differences in cost of $6,075 (95 percent CI 14,285-10,721) were also observed between Middle Atlantic and Mountain (Appendix C, Figure C21). Mortality from prostate cancer in U.S. regions. Four large cohort studies 158,160,183,218 examined age adjusted mortality in the U.S. regions in addition to the U.S. Cancer Statistics data from Results from the individual studies are presented in Appendix C, Table C54). Mortality was highest in East South and lowest in the Pacific region (Figure 21). Age adjusted mortality in U.S. regions did not differ compared with the national average (Appendix C, Table C55). The highest age adjusted mortality was observed among African American males in South Atlantic (5.55±2.67 per 100,000 males above the U.S. average, p=0.04) and in East South (6.39±2.79 per 100,000 males above the U.S. average, p=0.02) and the lowest was among Hispanic males in West South, and New England (Appendix C, Table C56 and Figure C22). African American males in New England had lower by 12.1±2.9 per 100,000 males age adjusted mortality compared with the national average among Blacks (Figure 22). The year of death did not modify the association between mortality and location. 96

107 Adjustment for years of schooling and for the proportion employed in agriculture did not change geographical differences in mortality in either Whites or Blacks. 158 Prostate cancer death rates were higher in non metro than metro areas by 12 percent in Blacks and 4 percent in Whites. 218 Mortality in Blacks increased over time from 1973 to 1998 in Connecticut and Iowa and decreased in New Mexico. 160 The observed trend in mortality among Blacks was not attributable to regional differences in PSA testing and treatment utilization. Another study reported that more intensive screening for prostate cancer was not associated with lower mortality. 173 The PSA testing rate was five times (RR 5.39, 95 percent CI 4.76; 6.11) in Seattle vs. Connecticut with no difference in mortality. However, in ecological analysis, regional variation in age-adjusted mortality was positively correlated with the prevalence of PSA testing within the last 2 years in males older than 40 years (correlation coefficient 0.43, p=0.002) and with the prevalence of ever having a PSA test (correlation coefficient 0.44, p=0.001) (Figure 23 and Appendix C, Tables C39-C41). The states with higher PSA testing (ranges percent) had higher age-adjusted mortality (26-54 per 100,000 males) (Appendix C Figure C23). Ecologic correlations ignored the regional differences in patient baseline risk including PSA levels, Gleason score, and tumor stage as possible explanatory factors for mortality variations. Age adjusted mortality was not correlated with numbers of urologists and radiation oncologists (Figure 25) when patients with all races were combined and among African American males, but mortality rates were lower among Whites in regions with higher numbers of urologists (correlation coefficient -0.2, p=0.01) and radiation oncologists (correlation coefficient -0.2, p=0.02) (Appendix C, Table C34). Increased PSA screening and cancer detection may result in attribution bias when patients diagnosed with prostate cancer died from other diseases but had cancer as underlying cause of death. 221 No studies examined regional difference in misclassification of prostate cancer mortality, so we were unable to determine the extent that increased cancer screening or diagnoses contributed to increased differences in cancer specific mortality. Summary for regional variation. Differences in structure (number of physicians involved in prostate cancer care) and process variables (screening and treatment practices) in U.S. regions were not correlated in ecologic analysis. Incidence and mortality varied in regions but with no significant differences compared with the national average. Significant geographic differences in incidence and mortality vs. the U.S. average were observed in African American males with the highest mortality in East South and South Atlantic regions. Physician availability negatively correlated with mortality in Whites but not in Blacks. Limited evidence suggests variations in morbidity related to RP and in cost of this procedure. Pooling analysis at state and regional levels may diminish differences in access to and quality of care in smaller urban/rural areas. 54,184 Patient characteristics, including ethnicity and socioeconomic status, were not associated with treatment choices in the study conducted in the SEER database. 162 However, the Cancer of the Prostate Strategic Urologic Research Endeavor database reported that patients with higher annual incomes and with fee-for-service patients more likely received RP than other treatments. 169 Managed care was not associated with mortality of Medicare beneficiaries with prostate cancer 198 African American patients had 168 percent higher mortality compared with Whites in 97

108 the private sector but not in Veteran Affairs within one region. 222 Uncertain effectiveness of treatment options may contribute to differences in patient outcomes. 59 Future research should address geographic differences in process variables (distribution of hospital technology, and quality of care) and in patient characteristics (distribution of socioeconomic status and access to care). How Does Hospital and Provider Volume Affect Outcomes? Association between hospital volume and patient outcomes. Several epidemiologic investigations examined the associations between hospital volume of RP and patient outcomes. 134,135, ,170,223 We evaluated studies that examined hospital volumes of perineal and retropubic RP. Authors defined volume as an annual average of procedures 165,167 or the total number of procedures during the time of the study performed in the each hospital. 134,135,170 Volumes were measured from linked SEER and the Medicare hospital claims database (Appendix C, Tables C14-C17). Authors compared volume measurements from different databases (State Discharge Registry and Medicare database) and concluded that both approaches yield the same results. 167 The distribution of hospital volumes is presented in Appendix C, Table C57. Surgery related mortality. Four retrospective cohorts examined the association between hospital volumes of RP and surgery related mortality (Appendix C, Table C14). 134,165,167,170 Authors defined mortality related to RP as in-hospital death 165 or postoperative death within days after surgery. 134,167,170 We combined these two measurements. Authors reported death rate in different categories of hospital volume. We computed death rate corresponding to an increase by ten procedures performed annually in hospitals for a pooling analysis (Figure 25). One study, 170 of three 165,167 that reported random changes in mortality, showed a significant reduction in death rate that resulted in a random pooled estimate (p for heterogeneity = 0.017). One study 166 did not find a significant association between hospital volume and all cause and cancer specific mortality 10 years after surgery. Three studies reported adjusted relative risk of surgery related mortality (Appendix C, Figure C24). 134,165,167 The relative risk of death in hospitals that performed vs. more than 55 RPs per year was 1.71 (95 percent CI ). 165 Considering that around 70 percent of patients were treated in hospitals with a volume of less than 50 procedures per year, 165,167 more than 41 percent of deaths in such hospitals might be attributable to a low volume independent of patient characteristics (Table 18). Among all patients treated with RP, percent of deaths might have been avoided if the patients had been treated in hospitals with higher volume. Another large study 170 showed a 42 percent higher relative increase (RR 1.42, 95 percent CI ) of surgery-related mortality in hospitals that performed less than 27 vs. more than 36 procedures per year. In terms of absolute risk reduction, 5.3 deaths per 1,000 RP patients might be avoided if patients had been treated in hospitals that performed more than 36 procedures per year. One earlier study 134 did not find an association with mortality. Our pooled analysis showed a relative reduction of 13 percent (RR 0.87, 95 percent CI ) corresponding to ten additional RPs performed annually in hospitals (p for heterogeneity 0.11). The time when events occurred, the database, and the sampling strategy did not modify the association between hospital volume and surgery related mortality. 98

109 The relative risk of surgery related mortality in categories of quartiles of hospital volume was estimated (Appendix C, Figure C25). Relative risk of death was almost twice that in hospitals performing less than 22 (1 st quartile) (RR 1.97, 95 percent CI ) and 64 percent higher in hospitals with operations per year (2 nd quartile)(rr 1.64, 95 percent CI ) compared with hospitals that had done more than 85 surgeries per year (4 th quartile) (p for heterogeneity = 0.08). The mean hospital volume appears to be a reasonable cut point to identify a threshold in volume effect (p for heterogeneity = 0.24). The relative risk of death related to a surgery was 0.62 times less in hospitals that performed more than 43 RPs per year (RR 0.62, 95 percent CI ). Appendix C, Figure C26, presents the number of avoided deaths per 1,000 hospitalized patients in hospitals with volume above an average mean level and number of excessive deaths in hospitals with lower volumes. Surgery related morbidity. Four cohorts 135,167,170,176 examined the association between hospital volume and surgery related morbidity including cardiac, respiratory, and vascular complications, bleeding, renal failure, shock, and need for re-operation (Appendix C, Table C14). Three of four studies reported a significant reduction in complication rates among higher categories of hospital volume. We calculated pooled rates corresponding to ten additional procedures in three studies (Figure 26). Overall, every ten RPs performed annually in hospitals was associated with absolute reduction in complications by 1.1 percent (95 percent CI ). The association was significant with a reduction in complication rate by 7.3 percent per natural logarithm of ten surgeries per year (95 percent CI 4-10, p <0.001). At the time when events occurred, the database, and the sampling strategy did not modify the association between hospital volume and surgery related morbidity. We found an absolute reduction in complication rate of 2.8 percent among hospitals that performed surgeries per year (2 nd quartile) vs. fewer than 22 (1 st quartile) (95 percent CI percent). Hospitals with more than 85 procedures per year (4 th quartile) vs. fewer than 22 (1 st quartile) had 7.3 percent reduction in complication rates (95 percent CI ) (p for heterogeneity = 0.07). The absolute differences in complication rate of 9.7 percent (95 percent CI ) were observed among hospitals above vs. below the mean of volume (43 procedures per year) (p for heterogeneity = 0.23). Three studies evaluated relative risk of morbidity after RP. 135,170,176 Hu et al 135 compared complication rates in hospitals with greater vs. less than 60 procedures per year: 85 percent of procedures were performed in hospitals with lower volume. The authors reported a nonsignificant 16 percent reduction in complications (RR 0.84, 95 percent CI ) in hospitals with higher volume. Yao et al 170 defined low volume as less than nine procedures per year and reported a 1.43 fold increase in the adjusted complication risk (RR 1.43, 95 percent CI ). The reduction in relative risk was consistent across the categories of hospital volume in this study with 8 percent relative reduction corresponding to additional ten procedures per year in hospital (RR 0.92, 95 percent CI ). However, pooled analysis of all studies did not detect a significant association between hospital volume and relative risk of complications. Surgery related quality measures (cancer control, urinary complications, and operative quality). One cohort study 166 included 5,837 patients with prostate cancer followed for 10 years after RP performed in 348 hospitals. The authors examined the association between hospital 99

110 volume and use of adjuvant therapy started more than 6 months after surgery adjusted for tumor stage and grade and patient comorbidity. Patients operated in hospitals with low volume (<16 procedures per year) were treated 1.25 times more often with adjuvant therapy (95 percent CI ) compared with those operated in hospitals that performed more than 85 surgeries per year. The association was consistent across volume categories with a significant reduction in risk of adjuvant therapy by 2 percent (RR 0.98, 95 percent CI ) per ten additional surgeries per year (Appendix C, Figure C27). Patients operated in hospitals with less than 22 procedures per year had a percent increase in use of adjuvant therapy compared with those treated in hospitals that performed and more than 85 RPs per year, respectively (Appendix C, Figure C27). Two cohort studies 135,167 examined the association between urinary and incontinence complications and hospital volume with different definitions of low and high volumes. We calculated changes in frequency of diagnosed events (in hospitals claims) and symptoms (in medical charts) corresponding to additional ten RPs performed annually in a hospitals (Appendix C, Figure C28). Rates of any urinary complications but not incontinence were lower by percent (95 percent CI -1.12;-0.36) in patients sampled from SEER database 167 and by (95 percent CI -3.57,-0.09) in Medicare beneficiaries. 135 A pooled analysis was conducted to estimate the association between hospital volume as a continuous variable and rates of surgery related complications and detected a decrease by.85 percent (-1.53, -0.17) in diagnosed events of late urinary complications (Figure 27) (p for heterogeneity = 0.02). A small but significant increase of 0.16 percent (95 percent CI 0.01; 0.30) in the rates of long-term incontinence corresponded to an additional ten RPs per year. Patients treated in hospitals with volume above vs. below the mean (43 procedures) (Figure 28) had lower rates of urinary complications by 5.3 percent (95 percent CI -9.3;-1.3). A further increase in hospital volume (4 th quartile, more than 85 procedures per year) was not associated with the larger benefit reduction of 5.3 percent (95 percent CI 0.6; 10) in rate of late urinary complications compared with hospitals that performed procedures (2 nd quartile) annually. One study examined operative quality indicators in 133 hospitals from New York state and reported an increase in quality scores of 2.7 (95 percent CI 0.9-4) for every additional RP performed annually. 223 Hospitals that performed procedures per year vs. less than 23 had higher operative quality by a score of 62 (95 percent CI p=0.002). We could not find studies that examined positive surgical margins in relation to hospital volume. Length of stay and readmission to hospital. Hospital volume was associated with reduced length of stay and readmission. Four studies 135,165,170,178 reported length of stay and readmission rate in relation to hospital volume; three 135,165,170 tested the associative hypothesis (Appendix C, Table C17). The authors obtained the Nationwide Inpatient Sample 165 and the Medicare claims databases 135,170 to analyze hospital volume, length of stay and readmission rate, and adjusted for patient 135,165,170 and hospital 135,170 characteristics to estimate the effect of volume. We calculated the differences in outcomes corresponding to an increase by ten procedures in annual hospital volume (Figure 29). All studies reported a small reduction in length of stay by an increase in volume. Pooled analysis with four studies detected a decrease in length of stay by 0.32 days (95 percent CI ) corresponding to additional ten procedures and by 1.7 days (95 percent CI 100

111 ) corresponding to an increase in natural logarithm of ten RPs performed annually in hospitals (Appendix C, Figure C29). Hospital volume was categorized to find a threshold in volume effects (Appendix C. Figure C30). Hospitals in the highest volume quartile (more than 85 procedure per year) had lower lengths of stay by 0.9 days (95 percent CI ) compared with those that performed RPs annually and by 1.5 days (95 percent CI ) compared with those in the lowest quartile (<22 surgeries per year). The decrease in length of stay was 0.9 days (95 percent CI ) in hospital above vs. below the mean (43 procedures per year). Crude readmission rates did not differ in relation to hospital volume (Appendix C, Figures C29- C30). However, adjusted for patient age, race, and comorbidity, surgeon specialty, and hospital teaching status, relative risk of readmission was 1.3 times higher among patients operated in hospitals that performed an average of nine surgeries per year compared with patients operated in high (more than 36 procedures per year) volume clinics. 170 The relative risk was 1.16 times higher in patients operated in hospitals with an average of 14 RPs per year. We estimated that an increase in hospital volume by ten surgeries per year was associated with a decrease in relative risk of readmission by 10 percent (RR 0.9, 95 percent CI ). Association between hospital status and patient outcomes. Teaching status was defined by an affiliation with an academic center 182 or membership in the Council of Teaching hospitals 136 (Appendix C, Table C18). The ownership of the hospitals was defined as for profit or not for profit institutions, government, or public hospitals. 135,136,181 Higher rates of surgery related mortality 182 and increased relative risk of death were reported in nonteaching hospitals. 136 A relative increase in surgery related mortality by 18 percent was shown in for profit institutions compared with teaching not-for-profit institutions (RR 1.18, 95 percent CI ).The majority of the authors combined hospitals with different status in the investigations. We included all studies that reported rates 134,136,165,167,170,174,181 and relative risk 134,136,165,170 of surgery related mortality in pooled analysis and did not find significant associations with hospital status. However, teaching hospitals had lower rates of surgery related complications (p for heterogeneity not significant) by 17.6 percent (95 percent CI ) 167,170,176,181 and higher scores of operative quality (mean 141 scores, 95 percent CI ). 223 Length of stay was higher in private institutions compared with academic centers. 177 Public nonteaching hospitals had lower length of stay by 1.5 days (95 percent CI ) compared with academic centers 135,136,165,170,177,178,181 (p for heterogeneity <0.01). Hospital charges 165,181 were not associated with teaching status being higher in for-profit clinics. 181 Association between surgeon volume and patient outcomes. Authors defined surgeon volume as an average of RPs performed annually by a surgeon 167,176 or the total number of procedures during the time of the study. 164,175,178,180,182 Two studies 167,182 examined the association with surgery related mortality and did not find differences in death rates in relation to surgeon volumes (Appendix C, Table C19). Five studies examined the association between surgeon volumes and complications 135,164,167,176,178 including cardiac, respiratory, or vascular events, the need for reoperation bleeding, renal failure, and shock in different volume categories (Appendix C, Table C20). We calculated the difference 101

112 in complication rates corresponding to an increase by one surgery per year in individual studies (Appendix C, Figure C31), when possible, and pooled estimates including all studies (Figure 30). Patients operated by surgeons with higher volume had lower rates of complications (-0.19; 95 percent CI -0.07;-0.3 per one RP per year) (p for heterogeneity = 0.01). This decrease was larger by 2.8 percent (95 percent CI 0.5;5) per natural logarithm of surgeon volume. The relative risk of surgery related complications adjusted for patient age, race, and comorbidity and hospital type and location was 0.53 times lower in men treated by higher volume surgeons (>40 vs. <40 surgeries per year). 135 Patients needed blood transfusion 8.6 times more often when the operating surgeon performed fewer than 15 RPs per year. 176 Cohort studies that examined the association between surgeon volume and quality measures 135,164,167,175,180 reported rates of late urinary complications and long-term incontinence (events and symptoms), and positive surgical margins among different categories of surgeon volumes (Appendix C, Table C21). We calculated the difference in outcome rates corresponding to an increase of one procedure in surgeon annual volume (Appendix C, Figure C31) and found a reduction in urinary complications and symptoms of long-term incontinence. In pooled analysis (Figure 38) the rate of late urinary complications was lower by 0.24 percent (95 percent CI percent) and the rate of long-term incontinence was lower by 0.12 percent (95 percent CI ) corresponding to an increase of one RP per year in surgeon experience. The rate of long-term incontinence was less by 0.6 percent (95 percent CI ) in men operated by surgeons that performed more than ten RPs per year. Surgeon volume was not associated with positive surgical margins. Length of stay in hospitals after RP was assessed in four studies (Appendix C Table C22). 135, Pooled analysis showed reduction in length of stay by 0.97 days (95 percent CI -1.45,-0.48) corresponding to an increase in surgeon volume logarithm (p for heterogeneity = 0.13). The hospital stay for patients operated by the higher volume surgeons (4-9 RPs per year, 2 nd quartile) decreased by 2.2 days (95 percent CI day) compared with surgeons who performed less than three surgeries per year (1 st quartile) (Figure 31). Length of stay was shorter by 3.3 days (95 percent CI 0.5-6) in men operated by surgeons who performed more than 15 (4 th quartile) vs. fewer than three surgeries (1 st quartile) per year (p for heterogeneity = 0.5). Cost was not associated with surgeon volume. Summary of association between hospital and provider volume with patient outcomes. Observational studies suggest that hospital volume of RP was associated with a decrease in surgery related mortality independent of measured confounding factors. Limited evidence suggests a reduction in relative risk of readmission and rate of adjuvant therapy in association with increased hospital volume. The decrease in length of stay was significant in most reports, but the pooled estimate may not be valid due to heterogeneity in the results from individual studies. Hospital volume was associated with decreased rates of surgery related morbidity and complications. Patient referral patterns and clustering patients among hospitals can affect the association and cannot be estimated from the reports. Despite different definitions of high and low hospital volumes in individual studies, pooled analysis showed that facilities with above average numbers of RPs per year had better patient outcomes, including lower surgery-related mortality, late urinary complications, and length of stay. Surgeon volume was also inversely 102

113 associated with surgery-related late urinary complications, long-term incontinence, and length of stay. Whether patients who attended lower volume facilities would, in fact, on average have had better outcomes had they attended higher volume facilities cannot be absolutely confirmed from these observational studies, but the consistency of the results implies that further research into this issue, including possible randomized assignment of some patients to higher volume facilities, is warranted. 103

114 2965 Figure 12. Conceptual Model Location Time periods 2966 Screening and diagnosis of localized prostate cancer Providers availability, specialty, and preferences Patient Factors -race -stage of cancer -comorbidities -socio-economic status -access to care Outcomes -Mortality -Complications -Quality of life -Length of stay -Cost -Readmission rate Treatment options Hospital technology Hospital teaching status Hospital volume Physician specialty, practice patterns Surgeon volume 104

115 Figure 13. Percentage of responses by family physicians and general internists to the question: Do you 186 recommend the following for prostate cancer screening for patients 50 years old and older? % of positive responses Family physician * * DRE PSA No screening >60 years General internist DRE = digital rectal examination; PSA = prostate specific antigen * significant differences at 95% CI * * Age not a criteria 105

116 Figure 14. Percentage of responses by urologists and primary care physicians to the question whether radical prostatectomy or external beam radiation probably or definitely offers survival benefit for 144 patients with clinically localized prostate cancer Life expectancy >10 years Life expectancy <10 years % Urologists * * RP EBRT RP EBRT Primary care physicians * significant differences at 95% CI * 106

117 Figure 15. Which therapy offers the best survival to patients with clinically localized prostate cancer? 145 Treatment recommendations by urologists and radiation oncologists % Urologists 98* Life expectancy <10 years Life expectancy >10 years * 39 14* 22 Brachy EBRT RP Brachy EBRT RP Radiation oncologists * significant differences at 95% CI 66* 67* 107

118 Figure 16. Percentage of responses by urologists and radiation oncologists to the question whether they believed that three main potentially curative prostate cancer therapies are overused or underused in the 145 United States % Radiation oncologists Therapy overused Therapy underused * * * * significant differences at 95% CI * RP EBRT Brachy RP EBRT Brachy Urologists * * 108

119 Figure 17. Regional variations in incidence of prostate cancer (CDC ) and distribution of urologists and radiation oncologists* in U.S. regions physicians/100,000 adult population East North East South Middle Mountain Atlantic Urologists /100,000 adult population Incidence/100,000 male population New England Pacific South Atlantic West North West South * an average of absolute number of physicians who identified themselves as radiation oncologists in the USA states obtained from the surveys conducted by the American Medical Association in the ratio per 100,000 adult population was calculated with U.S. Census data Radiation oncologists/100,000 adult population Incidence/100,000 male population 109

120 Figure 18. Proportion of patients with localized prostate cancer treated with external beam therapy, brachytherapy, primary androgen deprivation therapy, radiation, and watchful waiting (%) in U.S. regions (pooled analysis) % of patients with localized PC treated with therapy East North East South Middle Atlantic Mountain New England Pacific South Atlantic West North West South Radical Prostatectomy External Beam Radiation Brachytherapy Primary Androgen Deprivation Watchful waiting 110

121 Figure 19. Utilization of radical prostatectomy, all cause and prostate cancer mortality in patients with prostate cancer treated with radical prostatectomy in U.S. regions (pooled analysis) Outcomes Relative risk of outcomes in regions vs. national average (95% CI) Mortality in patients treated with radical prostatectomy Midwest 1.04 (0.80, 1.35) Northeast 0.89 (0.66, 1.19) South 1.00 (0.75, 1.34) West 0.99 (0.78, 1.26) PC mortality in patients treated with radical prostatectomy Midwest 1.20 (0.74, 1.95) Northeast 0.86 (0.49, 1.50) South 0.78 (0.45, 1.36) West 1.04 (0.67, 1.61) Utilization of radical prostatectomy compared with the national average Midwest 1.08 (0.93, 1.25) Northeast 0.65 (0.56, 0.75) South 1.05 (0.91, 1.21) West 1.38 (1.19, 1.60) Relative risk of outcomes in regions vs. national average 111

122 Figure 20. Length of stay in hospital after radical prostatectomy in U.S. regions (pooled analysis) Region 0 7 Length of stay in days Length of stay in days (95% CI) East North 4.63 (2.61, 6.65) East South 4.81 (2.73, 6.89) Middle Atlantic 5.74 (3.65, 7.82) Mountain 3.76 (1.74, 5.78) New England 4.70 (2.64, 6.76) Pacific 3.77 (1.68, 5.85) South Atlantic 5.00 (2.99, 7.02) West North 4.77 (2.71, 6.83) West South 4.39 (2.33, 6.45) Midwest 4.65 (2.56, 6.74) Northeast 5.19 (3.09, 7.29) South 4.73 (2.65, 6.81) West 3.72 (1.63, 5.81) USA 4.62 (2.24, 7.01) 112

123 Figure 21. Mortality from prostate cancer (per 100,000 male populations) in U.S. regions (CDC data ) Region Mortality/100,000 male population (95% CI) East North (34.66, 41.36) East South (41.66, 50.09) Middle Atlantic (31.97, 40.40) Mountain (30.04, 36.15) New England (29.74, 36.70) Pacific (28.56, 36.59) South Atlantic (39.04, 44.60) West North (30.77, 38.26) West South (35.56, 43.48) USA (28.49, 44.81) 0 60 Mortality/100,000 male population 113

124 Figure 22. Mortality from prostate cancer (per 100,000 male population) among different races in U.S. regions (CDC data, ) Region Age-adjusted mortality/100,000 male population (95% CI) African American East North (62.45, 66.29) East South (71.23, 75.70) Middle Atlantic (59.75, 64.74) Mountain (60.10, 65.44) New England (52.56, 57.55) Pacific (55.46, 61.48) South Atlantic (71.09, 74.25) West North (61.70, 67.46) West South (65.21, 69.67) USA (62.11, 72.09) Hispanic East North (14.87, 20.63) Middle Atlantic (20.20, 25.96) Mountain (24.78, 29.24) New England (20.01, 27.07) Pacific (17.98, 25.04) South Atlantic (19.99, 25.76) West South ( 18.97, 26.03) USA (17.61, 27.59) Whites East North (26.70, 30.47) East South (26.41, 30.88) Middle Atlantic (24.44, 29.43) Mountain (27.81, 30.96) New England (27.04, 30.57) Pacific (27.68, 31.85) South Atlantic (24.89, 27.97) West North (26.84, 30.36) West South (24.82, 29.28) USA ( 22.26, 32.24) 0 75 Age-adjusted mortality/100,000 male population 114

125 Figure 23. Regional variations in PSA testing* and prostate cancer age adjusted mortality** in the U.S % PSA tested East North East South Middle Atlantic Mountain New England Pacific South Atlantic West North West South PSA test in men older than 40 years during last 2 years PSA test ever Mortality * an average responses in U.S. regions to the BRFSS questionnaire 217 having ever had PSA test and during last 2 years in males older than 40 years ** average of age adjusted prostate cancer mortality (U.S. Cancer statistic) deaths/ male population 115

126 Figure 24. Regional variations in prostate cancer age adjusted mortality* and distribution of urologists and radiation oncologists** in the United States (pooled analysis) Physicians/100,000 adult population East North East South Urologists /100,000 adult population Mortality/100,000 male population Middle Mountain New Pacific Atlantic England South Atlantic West North West South 40 Radiation oncologists/100,000 adult population * average of age adjusted prostate cancer mortality (the U.S. Cancer statistic) 219 ** an average of absolute number of physicians who identified themselves as radiation oncologists in the U.S. states obtained from surveys conducted by the American Medical Association from Deaths/100,000 male population 116

127 Figure 25. Difference in surgery related death rate corresponding to an increase by 10 radical prostatectomies performed in hospital Study Difference in surgery related death rate (95% CI) Yao (1999) (-0.13,-0.03) Ellison (2000) Begg (2002) Difference in surgery related death rate (-0.06, 0.05) (-0.02, 0.01) Overall (-0.07, 0.02) 117

128 Table 18. The association between hospital volume and mortality results from individual studies Reference Number of Attributable % in Number Attributable % in Volume Categories Crude Adjusted Excess or Patients Operated Needed Population with PC, (% Hospitals) Rate, % Volume Effect Avoided in Low Volume to Treat 95% CI Deaths/1,000 Hospitals, 95% CI Ellison, <25 (76) ( ) (17) ( ) >54 (7) (reference) Increase by 10 procedures per year 0.83 ( ) 0.5 Yao, Hospital Volume <38 (9/year) ( ) (14/year) ( ) (27/year) ( ) >141 (36/year) (reference) 118

129 Figure 26. Difference in surgery related complications rate corresponding to an increase by 10 radical prostatectomies performed in hospital Difference in complications rate Study (95% CI) Begg (2002) (-0.66,-0.62) Yao (1999) (-2.51,-0.75) Hu (2003) (-2.80, 0.86) Overall (-1.72,-0.26) Difference in complications rate 119

130 Figure 27. Difference in surgery related complications rate corresponding to an increase by 10 radical prostatectomies performed in hospital (pooled analysis) Increase by 10 procedures in annual hospital volume Difference in rate Difference in rate (95% CI) Late urinary complications (-1.53, -0.17) Symptoms of late urinary complications (-2.70, 0.38) Long term incontinence 0.16 (0.01, 0.30) Symptoms of long term incontinence (-0.44, 0.17) 120

131 Figure 28. Difference in surgery related complications in categories of hospital volume (pooled analysis) Annual hospital volume Difference in outcome rate (95% CI) Late urinary complications <22 vs (-5.45, 5.60) <22 vs. > (-0.77, 11.50) vs. > (0.62, 9.97) <43 vs. > (1.31, 9.34) Long-term incontinence <22 vs (-6.80, 6.40) <22 vs. > (-8.72, 6.52) vs. > (-7.50, 5.70) <43 vs. > (-2.56, 0.63) Difference in outcome rate 121

132 Figure 29. Difference in length of stay and readmission rate corresponding to an increase by 10 procedures in annual hospital volume (results from individual studies) Increase by 10 procedures in annual hospital volume Difference in outcome rate (95% CI) Length of stay (days) Ellison (2000) (-0.39, -0.39) Yao (1999) (-0.47, -0.08) Hu (2003) (-0.36, -0.04) Readmission Yao (1999) (-0.54, 0.06) Difference in outcome rate 122

133 Figure 30. Difference in rates of surgery related urinary complications and long term incontinence corresponding to an increase by 1 radical prostatectomy performed by a surgeon and in categories of surgeon volumes above and below the median (the results from individual studies) Outcomes Difference in outcome rate (95% CI) Incontinence symptoms Increase in annual surgeon volume by 1 procedure (-0.25, -0.00) Long-term incontinence >10 vs. <10 procedure per year (-0.84, -0.34) Urinary symptoms Increase in annual surgeon volume by 1 procedure (-0.48, -0.00) Difference in outcome rate 123

134 Figure 31. Difference in length of stay after radical prostatectomy by increase in surgeon annual volume (pooled analysis) Difference in length of stay (95% CI) Surgeon annual volume Increase by one procedure in surgeon* annual volume Difference in length of stay (-1.45, -0.48) 4-9 vs. <3 procedures per year (-4.43, 0.06) >15 vs.<3procedures per year (-6.12, -0.53) * in logarithmic scale 124

135 Key Question 4. How do tumor characteristics, e.g. Gleason score, tumor volume, screen vs. clinically detected tumors, PSA levels, affect the outcomes of these therapies, overall and differentially? Tumor characteristics are often utilized by patients and providers when choosing or recommending treatments because they are believed to affect outcomes. The primary measure of aggressiveness is the Gleason histologic score which ranges from Gleason 8-10 tumors are considered the most aggressive, Gleason 7 tumors somewhat less, and Gleason 6 tumors potentially indolent. Pretreatment Gleason scores are determined based on a pathologist s examination of several small cores of prostate tissue. Typically, six cores are obtained during a prostate biopsy (sextant biopsy that includes both lobes of the prostate). However, the number has increased over time to 12, 24, and even saturation techniques which increase the amount of the prostate gland sampled and enhance the likelihood of detecting even small volume disease. In addition to the histologic score, the number of biopsy cores that contain prostate cancer and the percent within each core containing tumor is recorded. Because the Gleason score and tumor volume do not appear to be ideal or complete indicators of an individual tumor risk characteristic, additional efforts are underway to identify more reliable prognostic factors for individual tumors. 224 Risk stratification strategies have incorporated PSA level, biopsy Gleason score, and clinical tumor category because these appear to be associated with risk of PSA failure and prostate cancer specific mortality. 21 A risk classification currently recommended is: Low Risk: PSA <10 ng/ml, Gleason score 6 and clinical stage T1c or T2a Intermediate Risk: PSA >10 20 ng/ml, or Gleason score 7, or clinical stage T2b High Risk: PSA >20 ng/ml or Gleason score 8-10 or clinical stage T2c Because little information exists on the comparative effectiveness of treatment options based on these low, intermediate, or high risk classifications, our analysis was confined to baseline PSA levels and Gleason histologic grade. The true natural history of prostate cancer is not well known because patients rarely remain untreated for the full duration of their disease. A recent report assessed 20-year outcomes among 767 men diagnosed in the United States with clinically localized prostate cancer between 1971 and 1984 (pre-psa era) and followed with WW and delayed palliative interventions. Sixty percent were diagnosed by transurethral resection of the prostate, 58 percent had no treatment within 6 months of diagnosis, and the remainder had some form of androgen deprivation therapy. The median age at diagnosis was 69 years, and all but 6 percent have died. Overall, 29 percent died of prostate cancer. Both overall survival and cumulative mortality from prostate cancer and other causes varied according to age at diagnosis and comorbidities as measured by Charlson Comorbidity score and Gleason score. Among men with no or only minor comorbidities, 26 percent, 15 percent, and 8 percent survived at least 15, 20, and 25 years respectively. In comparison, in men with significant comorbidities, only 11 percent, 6 percent, and 3 percent, respectively, survived at least 15, 20, and 25 years (Figure 32). 132 Men with low grade prostate cancers managed with WW had a minimal risk of dying from prostate cancer during 20 years of followup (Gleason score of 2-4, six deaths per 1,000 person years; 7 percent died due to prostate cancer). 132 Men with high grade prostate cancers had a high probability of dying from prostate cancer within 10 years of diagnosis regardless of their age at 125

136 diagnosis (Gleason score of 8-10, 121 deaths per 1,000 person years; 53 percent died due to prostate cancer). Death due to prostate cancer over this time period was 20 percent in men ages 55-59; 27 percent in men ages 65-69, and 30 percent in men ages 70-74, though the percent of high grade tumors was greater in older patients. Annual prostate cancer specific mortality rates were similar when assessed before and after 15 years of followup. Because PSA testing increases the time of detection by 5-15 years, it is likely that men with PSA detected tumors treated with WW will have a better 20 year disease specific survival than this cohort AUA database results were infrequently stratified by PSA or Gleason score. When results were stratified, studies often used varying followup times, making comparative effectiveness difficult. Results were not controlled for confounding variables including age, comorbid conditions, or histologic score. No studies assessed survival or bned in patients with baseline PSA levels 4.0. Outcomes were stratified according to baseline PSA <10 vs. 10 ng/ml and reported for two time periods: months (short term) and months (mid to long term). Based on very limited nonrandomized trial data, mid term disease specific survival appeared similar for subjects treated with EBR compared to RP in men with baseline PSA >10 ng/ml. Men with Gleason scores 8-10 appeared more likely to have biochemical reoccurrence than men with Gleason scores 2-6 regardless of type of treatment. It was not possible to clearly determine whether comparative effectiveness between treatments varies by Gleason scores (Figures 33-35). In contrast to the limited survival data, many patient groups treated with brachytherapy, EBR, or RP reported bned. While there was a wide range in outcomes within and across treatment modalities, there appears to be an inverse association of mid and longer term bned with baseline PSA categories across these three treatments. However, data were too variable to make conclusive statements, and it is not possible to determine if the relative effectiveness between brachytherapy, EBR, and RP varies according to these PSA categories. Overall and disease specific survival were infrequently stratified by Gleason scores (Figures 36-38).When provided, the full spectrum of Gleason scores for each treatment modality was rarely available. Within and between treatment comparisons were limited for reasons noted above. For long-term outcomes, limited data from nonrandomized trials suggest that both overall and disease specific survival are associated with Gleason score regardless of treatment. Because of the paucity of information, wide variation in outcomes, and the lack of controlling for potentially prognostic factors, it is not possible to determine if survival outcomes varied between treatments according to Gleason score. Results suggest that both mid and long-term bned were inversely associated with Gleason score (Figure 38). The wide range of outcomes and the lack of controlling for confounding variables preclude determining whether one treatment provides superior outcomes based on baseline Gleason score. As described in Key Question 1, some randomized trials reported outcomes according to baseline PSA levels, Gleason scores, or risk strata. Most reported biochemical progression rather than overall or disease specific mortality or development of metastatic disease. In preplanned multivariate analyses from the SPCG-4, randomized study reduction in disease specific mortality at 10 years due to RP compared to WW differed according to age but not baseline PSA level or Gleason score. Only five percent of subjects enrolled in SPCG-4 had PSA detected disease. The 126

137 number needed to treat and the duration of time needed to achieve a benefit is likely greater in PSA detected compared to nonpsa detected disease. 225 In another randomized trial, men with Gleason scores 8-10 were more likely to have evidence of biochemical reoccurrence than men with Gleason scores 2-6, and the results did not differ whether treatment was radical prostatectomy alone or combined with NHT. High dose EBRT was more effective in controlling biochemical failure (three successive increases in PSA level) than conventional dose therapy 39 in both low risk disease (n=227, PSA <10 ng/ml; stage T2a tumors; or Gleason 6) and high risk disease. In the low risk subgroup, the percentages were 80.5 percent for the high dose group and 60.1 percent in the conventional dose (p <0.001), a 51 percent risk reduction. For the high risk subjects, the percentages were 79.5 percent and 63.4 percent (44 percent risk reduction; p=0.03) for the respective groups. However, when the high risk subjects were further divided into intermediate risk and high risk groups, the benefit of high dose therapy remained for the intermediate risk (81 percent vs percent, p=0.02) but not for the 33 high risk patients (p=0.80). In summary, tumor characteristics such as Gleason score, tumor volume, screen detected tumors, and PSA levels (and rates of PSA change) affect the overall outcomes of therapies. However, the affect on relative effectiveness of therapies is not well established. In a single randomized trial comparing RP with WW, disease specific mortality at 10 years due to RP compared to WW differed according to age but not baseline PSA level or Gleason score. Only 5 percent of subjects enrolled in this trial had PSA detected disease. The relative effect of these and other tumor prognostic characteristics require evaluation through the conduct of adequately powered RCTs. 127

138 Figure 32. Survival and cumulative mortality from prostate cancer causes up to 20 years after diagnosis, stratified by age at diagnosis and Gleason score (Source: American Medical Association, 2005, used with permission [Albertsen PC, Hanley JA, Fine J. 20-year outcomes following conservative management of clinically localized prostate cancer. JAMA 2005;293: ] 132 Copyright 2005 American Medical Association