Intensity-Modulated Radiation Therapy Leads to Survival Benefit Only in Patients with High-Risk Prostate Cancer: a Population-Based Study

Transcription

1 Annals of Oncology Advance Access published February 22, Intensity-Modulated Radiation Therapy Leads to Survival Benefit Only in Patients with High-Risk Prostate Cancer: a Population-Based Study G. Gandaglia 1,2, P.I. Karakiewicz 1,3, A. Briganti 1, Q. Trinh 4, J. Schiffmann 1, Z. Tian 1, S.P. Kim 5, P.L. Nguyen 6, M. Graefen 7, F. Montorsi 2, M. Sun 1,*, F. Abdollah 1,2,8,* 1 Cancer Prognostics and Health Outcomes Unit, University of Montreal Health Center, Montreal, Canada 2 Department of Urology, Urological Research Institute, Vita Salute San Raffaele University, San Raffaele Scientific Institute, Milan, Italy 3 Department of Urology, University of Montreal Health Center, Montreal, Canada 4 Department of Surgery, Division of Urology, Brigham and Women s Hospital/Dana-Farber Cancer Institute, Harvard Medical School, Boston, USA 5 Department of Urology, Yale University, New Haven, USA 6 Department of Radiation Oncology, Dana-Farber Cancer Institute, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, USA 7 Martiniclinic, University Medical Center Hamburg-Eppendorf, Hamburg, Germany 8 Vattikuti Urology Institute & VUI Center for Outcomes Research Analytics and Evaluation; Detroit, MI, USA Corresponding author: Firas Abdollah, MD, Vattikuti Urology Institute & VUI Center for Outcomes Research Analytics and Evaluation, Henry Ford Hospital; 2799 W Grand Blvd, Detroit, MI , Tel: ext: 35335; ; Fax: , firas.abdollah@gmail.com * Both senior authors contributed equally to the manuscript Key message: In patients with low/intermediate-risk PCa, IMRT was not associated with a PCa-specific survival improvement. Conversely, this approach leads to a significant survival advantage in patients with high-risk disease. The highest survival benefit was observed among younger and healthier patients. Keywords: Prostate Cancer, Intensity Modulated Radiation Therapy, Survival, Cancerspecific mortality, Competing-risks The Author Published by Oxford University Press on behalf of the European Society for Medical Oncology. All rights reserved. For permissions, please journals.permissions@oup.com.

2 2 ABSTRACT Background: During the last years, there has been a rapid adoption of intensitymodulated radiation therapy (IMRT) in patients with prostate cancer (PCa), despite the lack of randomized trials evaluating its effectiveness. The aim of our study was to evaluate the survival benefit associated with IMRT in patients with PCa. Patients and Methods: Overall, 42,483 patients with PCa treated with IMRT or initial observation between 2001 and 2007 within the Surveillance, Epidemiology, and End Results (SEER)-Medicare were evaluated. Patients in both treatment arms were matched using propensity-score methodology. After propensity-score matching, 19,064 patients remained in our analyses. 8-year cancer-specific mortality (CSM) rates were estimated, and the number needed to treat (NNT) was calculated. Competing-risks regression analyses tested the relationship between treatment type and CSM. Results: Overall, the 8-year CSM rates were 3.4 and 4.1% for patients treated with IMRT vs. initial observation, respectively (P<0.001). The corresponding 8-year NNT was 142. In patients with low-/intermediate-risk disease, IMRT was not associated with lower CSM rates compared to observation (P=0.7). In patients with high-risk disease, the 8-year CSM rates for IMRT vs. observation were 5.8 vs. 10.5%, respectively (P<0.001). The corresponding NNT was 21. When high-risk patients were stratified according to age (<73 vs. 73), and CCI ( 1 vs. >1) the 8-year CSM rates for IMRT vs. observation were 4.3 vs. 9.4% and 6.9 vs. 11.9% and 5.3 vs. 11.4% and 6.1 vs. 10.1%, respectively (all P<0.001). The corresponding NNTs were 19, 21, 16, and 25, respectively. In multivariate analyses the protective effect of IMRT was more evident in high-risk patients with younger age and lower comorbidities. Conclusions: IMRT leads to a survival advantage only in patients with high-risk disease. Conversely, patients with low/intermediate-risk disease did not benefit from IMRT at 8-year follow-up.

3 3 INTRODUCTION A continuous shift toward the adoption of novel techniques has been witnessed in the treatment of prostate cancer (PCa) over the past few years.[1] In this context, intensitymodulated radiotherapy (IMRT) rapidly took place over standard conformal radiation therapy, and now it represents the most commonly used treatment options for clinically localized PCa.[1, 2] Lower rates of treatment-related toxicity and increased patient expectation might have significantly contributed to the rapid dissemination of this novel approach, despite a concomitant substantial increase in expenditures for health-care providers.[3] Of note, these observations were carried out in an era where the benefit of active management for clinically localized PCa has been extensively questioned, particularly in the context of low-risk disease.[4-6] Consequently, the increasing utilization of advanced treatment technologies even in men at low-risk of cancer-specific mortality (CSM) and at high risk of other-cause mortality (OCM) raises concerns regarding the substantial risk of overtreatment in these patients.[1] In the lack of prospective randomized-controlled trials comparing initial observation and IMRT, we sought to assess the survival benefit associated with IMRT in a large populationbased cohort representing the United States. Particularly, we aimed at evaluating the risk of dying form PCa in the overall population, and according to tumor risk, age, and comorbidity status. MATERIALS AND METHODS Population source The current study relied on the most recent version of the SEER-Medicare insurance program linked database. This database is 98% complete for case ascertainment. The

4 4 SEER registries cover approximately 28% of the United States population with Medicare administrative data. Medicare insurance includes approximately 97% of Americans aged 65 years. Study population Overall 141,961 patients with histologically confirmed non-metastatic PCa (International Classification of Disease for Oncology [ICD-O] site code 61.9, histologic code 8140) aged 65 years or older diagnosed between 2001 and 2007 were identified. Using the Medicare Provider Analysis and Review outpatient, and carrier files, we identified patients who were primarily treated with IMRT (Healthcare Common Procedure Coding System [HCPS] codes 77418, G0174, and 0073T) ± ADT within the first 6 months from diagnosis.[1] Patients included in the observation group did not receive any treatment within 6 months from diagnosis. For the aim of our study, we focused exclusively on patients treated with IMRT as initial treatment or observation. This resulted in 68,805 assessable patients. Additional exclusions criteria consisted of men aged >80 years old (N=18,407), those with unknown clinical tumor stage (N=3,867), unknown Gleason score (N=4,695), unknown socioeconomic status (N=3,326), unknown cause of death (N=173), and follow-up shorter than 6 months (N=722), resulting in 42,483 assessable patients. Covariates For each patient, age at diagnosis, year of diagnosis, race, population density, marital status, 2000 census tract percent with 4-year college education, 2000 census tract annual median income, region, Gleason score, clinical stage, were assigned.[7] The Charlson comorbidity index (CCI) was derived from the Medicare claims one year prior to PCa diagnosis using a previously validated algorithm.[8] The use of salvage treatments after 6

5 5 months from diagnosis (i.e., radiotherapy, radical prostatectomy, and/or ADT) was recorded, consistent with previous methods.[9] Patients were divided according to their PCa risk group: low/intermediate (biopsy Gleason score 7 and clinical stage T2b) and high-risk (biopsy Gleason score 8-10 or clinical stage T2c).[7, 10] Outcomes The cause of death was defined using the SEER cause of death code. Patients who died from PCa (ICD or ICD-10 C619) were classified as CSM, while patients who succumbed to all other causes were classified as other-cause mortality (OCM).[11] The duration of survival was defined as the time interval from PCa diagnosis to the date of death. Statistical analyses Our statistical analyses consisted of three main steps. First, due to inherent differences between patients undergoing IMRT and observation, adjustment was performed using a 0.5 to 0.5 propensity-score matching. This methodology is used in observational studies to select control subjects who are matched with treated subjects on the controlled background covariates, which, if uncontrolled for, might lead to biased estimates of treatment effects. When matching is performed, the covariates of the control and treatment groups are balanced in order to reduce possible biases to a minimum.[12] Propensity scores were computed by modeling a logistic regression with the dependent variable as the odds of receiving IMRT, and the independent variable as age, race, marital status, population density, CCI, education, income, year of diagnosis, clinical stage, and Gleason score. Subsequently, covariate balance between the matched groups was examined.[13] Second, matched patients were stratified according to tumor risk groups in low/intermediate- vs. high-risk disease. Cumulative incidence CSM and OCM rates were

6 6 generated for different groups and compared with the Gray test.[14] When active treatment leads to a statistically significant risk reduction, the number needed to treat (NNT) was calculated. The latter was defined as the number of patients who must be treated with IMRT to prevent 1 CSM event.[15] Finally, competing-risks multivariate regression analyses were performed to test the effect of treatment on CSM and OCM, after accounting for confounders. We repeated all the analyses among patients with low/intermediate- and high-risk after stratifying patients according to median age (<73 vs. 73 years), and CCI ( 1 vs. 2). All statistical tests were performed using the R statistical package (v3.0.0). All tests were 2-sided with a significance level set at P<0.05. Sensitivity analyses In separated sensitivity analyses we repeated our analyses including patients receiving IMRT ± ADT within 12 months from diagnosis. Patients included in the observation group did not receive any treatment within 12 months from diagnosis. This resulted in a population of 42,319 patients (19,759 [42.7%] and 22,380 [57.3%] patients included in the observation and IMRT groups, respectively). After propensity-score matching, 19,516 patients remained. RESULTS Overall, 42,483 patients with PCa were included in the study (Table 1). Average age at diagnosis was 72.5 (median: 73; interquartile range: 69-76). Respectively, 23,261 (54.8%) and 19,222 (45.2%) patients were treated with observation and IMRT. Among men included in the IMRT group, 5,013 (52.6%) and 2,514 (26.4%) individuals received ADT concomitantly and after initial treatment, respectively. Table 1 shows baseline characteristics of patients included in the study. When patients were stratified according to

7 7 initial treatment (observation vs. IMRT), statistically significant differences were recorded with respect to year of diagnosis, race, population density, marital status, income, education, CCI, clinical stage, and biopsy Gleason score (all P<0.001). Following propensity-score matching, 9,532 (50%) patients treated with observation and 9,532 (50%) patients treated with IMRT remained. The mean standardized differences of all patient characteristics between the two groups were <10%, which indicate a high degree of similarity in the distribution of both populations. All subsequent analyses were based on the propensity-matched cohort. The median follow-up (mean) was 50 months (median: 53). Overall, the 8-year CSM rates were 3.4 vs. 4.1% for patients treated with IMRT vs. observation, respectively (Fig. 1a; P<0.001). The absolute risk reduction was 0.7% and the corresponding 8-year NNT was 142. In multivariate analyses, patients undergoing initial observation had 1.8-fold higher probability of dying from PCa compared to their counterpart treated with IMRT (Table 2a; P<0.001). When focusing on patients with low/intermediate-risk disease, the 8-year CSM rates were 2.5 for IMRT vs. 1.7% for observation (Fig. 1b; P=0.7). After stratifying patients according to age and CCI, no statistically significant differences were observed in 8-year CSM rates between patients treated with IMRT vs. observation (Table 2b; all P 0.5). In multivariate analyses, IMRT was not associated with lower risk of dying from PCa, even after stratifying patients according to age and CCI (all P 0.3). When focusing on patients with high-risk disease, the 8-year CSM rates were 5.8 for IMRT vs. 10.4% for observation (Fig 1c; P<0.001). The absolute risk reduction was 4.8% (8-year NNT: 21). After stratifying patients according to age and CCI, for individuals treated with

8 8 IMRT vs. observation the 8-year CSM rates were 4.3 vs. 9.4%for men aged <73 years, 6.9 vs. 11.5% for men aged 73 years, 5.3 vs. 11.4% for men with CCI 1, and 6.1 vs. 10.1% for men with CCI 2 (Table 2b; all P<0.002). The absolute risk reduction was respectively 5.1, 4.6, 6.1 and 4.0% (8-year NNT: 19, 21, 16, and 25, respectively). In multivariate analyses, observation was associated with a 2.4-fold higher risk of dying from PCa compared to IMRT in high-risk patients, after accounting for confounders (P<0.001). This held true even after stratifying patients according to age and CCI (all P 0.01). Sensitivity analyses In separate sensitivity analyses including patients who received IMRT ± ADT within 12 months from diagnosis, the 8-year CSM rates were 3.3 vs. 4.6% for patients treated with IMRT vs. observation, respectively (Supplemental Table 1; P<0.001). The absolute risk reduction was 1.3% and the corresponding 8-year NNT was 77. In multivariate analyses, observation was associated with a 2.4-fold higher risk of dying from PCa compared to IMRT, after accounting for confounders (Supplemental Table 2; P<0.001). When focusing on patients with low/intermediate-risk disease, the 8-year CSM rates were 3.1 for IMRT vs. 2.4% for observation (Supplemental Table 1; P<0.001). The absolute risk reduction was 0.7% (8-year NNT: 143). In multivariate analyses, observation was associated with a 1.9- fold higher risk of dying from PCa compared to IMRT in low/intermediate-risk patients, after accounting for confounders (Supplemental Table 2; P<0.001). When focusing on patients with high-risk disease, the 8-year CSM rates were 7.4 for IMRT vs. 11.6% for observation (Supplemental Table 1; P<0.001). The absolute risk reduction was 4.2% (8- year NNT: 24). In multivariate analyses, observation was associated with a 2.5-fold higher risk of dying from PCa compared to IMRT in high-risk patients, after accounting for confounders (Supplemental Table 2; P<0.001).

9 9 DISCUSSION IMRT represents a novel treatment modality for patients with localized PCa, allowing the delivery of higher doses of radiation with improved precision and lower treatment-related toxicities.[2, 16] The use of this novel technique considerably increased during the last decade, where roughly 100% of radiotherapy treatments are performed with this approach.[2] Nonetheless, solid data from randomized controlled trials supporting the efficacy of IMRT are still lacking. Furthermore, it has been estimated that the adoption of IMRT instead of traditional conformal radiotherapy might translate into excess spending of approximately $285 million per year in the United States alone.[3] Consequently, there is an impending need for comparative effectiveness studies evaluating the costs and benefits of advanced treatment technologies among PCa patients.[1, 3] Although previous retrospective observational studies supported the oncological safety of IMRT compared to conformal radiation therapy,[2, 17] none of them assessed the survival benefit associated with this novel approach compared to initial observation. This is crucial, particularly in an era where randomized trials questioned the effectiveness of active treatment in the improvement of overall survival in PCa patients.[6] Identifying individuals who might benefit the most from IMRT would help reduce the proportion of men receiving unnecessary treatment, thereby minimizing substantial personal and societal costs.[3, 18] To address this issue, we estimated the survival benefit of IMRT relative to initial observation on PCa-specific mortality after accounting for the risk of dying from other causes. First, we selected a large cohort of patients receiving IMRT or observation. Due to the prolonged natural history of the disease, we excluded patients diagnosed after 2007 to maintain adequate follow-up time. Additionally, in order to reduce treatment selection biases to the minimum, the two treatment groups were matched using propensity-score.

10 10 Finally, since the risk of CSM and of OCM significantly varies according to the tumor risk classification, age, and comorbidity status, we stratified our analyses according to these important variables. The results of our study are several-fold. First, we observed that IMRT was associated with an overall CSM reduction of 0.7% at 8-year follow-up. This implies that overall 142 men should be treated with IMRT in order to prevent one single detrimental outcome at 8 years. This was confirmed in multivariable analyses, where patients receiving initial observation had roughly 1.8-fold higher risk of dying from PCa compared to their counterpart receiving active treatment. Second, we observed that IMRT was not associated with lower CSM rates in men with low- and intermediate-risk PCa 8 years after diagnosis. This held true even in younger patients without comorbidities. Conversely, in patients with high-risk disease, IMRT was associated with an absolute CSM-risk reduction of roughly 5%, corresponding to a NNT of 21. As expected, the highest survival benefit was observed in younger and healthier patients. Specifically, individuals with a CCI less or equal to 1 and aged less than 73 years had an absolute risk reduction of 5.1 and 6.1%, respectively. From a practical perspective, this translated in 19 and 16 patients respectively to be treated in order to prevent a single PCa-related death. These results were confirmed in multivariable regression analyses, where high-risk patients treated with initial observation had a 2.5-fold higher risk of dying from PCa compared to their counterpart receiving IMRT. Additionally, after accounting for confounders we confirmed that the protective effect of IMRT on CSM is more evident in patients with younger age at diagnosis and lower comorbidities. On the other hand, older and sicker patients with high-risk disease might also benefit from active treatment. However, the number of men to be treated to prevent one death related to the tumor was

11 11 substantial higher in these categories. These results were confirmed when evaluating patients receiving IMRT vs. observation within 12 months from surgery. Indeed, although IMRT leads to a statistically significant benefit in terms of CSM both in patients with low/intermediate- and high-risk disease, the absolute risk reduction in the former group was only 0.7%, with a corresponding NNT of 142 patients. On the other hand, the absolute risk reduction was substantially higher when considering patients with high-risk disease (4.2%), where only 24 patients needed to be treated in order to prevent a cancer-related death at 8-year follow-up. Additionally, we should highlight that the inclusion of individuals treated within 12 months from diagnosis might results in a heterogeneous patient populations, where men who experienced upgrading and/or upstaging might be erroneously included in the low/intermediate-risk group. As a consequence, although IMRT within 12 months from diagnosis led to a statistically significant survival benefit as compared to initial observation in individuals included in the lower risk group, these results should be interpreted with caution. Based on the current findings, several clinical implications are noteworthy. First, we show that IMRT is not effective in reducing 8-year CSM rates in a clinically significant manner among patients with low- and intermediate-risk PCa. These findings corroborate Abdollah et al.[15] report, which focused exclusively on PCa patients with localized disease included in the SEER-Medicare database. Likewise, a recent prospective randomized trial failed to demonstrate a survival benefit at 12-year follow-up for men with low-risk disease receiving radical prostatectomy vs. observation.[6] This raises concerns regarding the substantial risk of overtreatment in these patients.[18] Indeed, an increasing body of evidence shows that the use of advanced treatment technologies, such as robotic radical prostatectomy or IMRT, is continuously growing, particularly in men at lower risk of dying from the tumor itself.[1] This phenomenon is mainly driven by aggressive marketing strategies and

12 12 patients expectative regarding reduced treatment-related toxicity and better oncologic outcomes. In this context, it has been recently shown that the adoption of this novel technology is associated with an additional expenditure of nearly $11,000 per patient compared to conformal radiotherapy.[3] If we translate our findings to the general United States population, we can estimate that IMRT results in an excess spending of approximately $250 million per year to treat low-intermediate risk patients, who do not experience any clinically significant survival benefit from IMRT.[3] Moreover, it should be noted that IMRT is not free from short-term and long-term complications.[2, 16, 17, 19] These findings are worrisome and highlight the imperative need to reduce overtreatment of IMRT in populations that do not seem to benefit from the treatment. As such, we should aim to appropriately select patients who may benefit from IMRT. This step would result in sparing patients and society with unnecessary treatment-related burdens.[1] Second, our results show a substantial survival benefit of IMRT in patients with high-risk disease. This is in line with what previously reported by Abdollah et al.[15]. Particularly, the authors showed that 18 patients with high-risk disease should be treated to prevent one adverse outcome at 10-year follow-up. That being said, important considerations need to be factored in the interpretation of their results: 1) the lack of discrimination between IMRT and conformal radiation therapy and 2) their report focused exclusively on patients with clinically localized PCa (ct1 and ct2).[15] The current study was able to overcome the aforementioned points. Particularly, we confirmed the effectiveness of IMRT in high-risk patients, even when including men with locally advanced disease. Despite several strengths, our study is not devoid of limitations. First, lack of randomization between IMRT and initial observation might in part limit the validity of our results. Indeed, patients included in the observation arm of our study might represent older

13 13 and sicker patients compared to those receiving active treatment. In order to minimize this potential limitation, we performed propensity-score matching.[20] Additionally, we relied on competing-risks methodology to account for the risk of dying from OCM. Second, our study is limited by the relatively short follow-up. For this reason, we were able to compare the oncologic outcomes of IMRT and initial observation only at 8-year follow-up. In this context, we can not exclude that a longer follow-up might have resulted in a significant benefit also for patients with low/intermediate-risk PCa. However, given the highly overlapping survival curves observed in our study, it is unlikely that longer follow-up will translate in better survival in favor of IMRT. On the other hand, it might result in enhanced survival benefit for patients with high-risk disease. Additionally, prostate-specific antigen was not available for the majority of patients included in our study. Thus, we stratified patients according to tumor stage and grade, which represent reliable proxies of the extent of the disease.[10, 15] However, the nature of our database prevented us to sub-stratify patients included in the low/intermediate-risk group. Consequently, we can not exclude that patients with more aggressive disease characteristics included in this group (i.e., biopsy Gleason score 4+3) might benefit from IMRT in terms of CSM. Further studies are needed to better address this issue and help identifying patients who might benefit the more from active treatments in the context of intermediate-risk disease. Finally, SEER- Medicare data did not include information about oral medications. That said, treatment with androgen monotherapy is not approved for PCa in the United States. FUNDINGS: None. CONFLICTS OF INTEREST: The authors have declared no conflicts of interest.

14 14 REFERENCES: 1. Jacobs BL, Zhang Y, Schroeck FR et al. Use of advanced treatment technologies among men at low risk of dying from prostate cancer. JAMA 2013; 309: Sheets NC, Goldin GH, Meyer AM et al. Intensity-modulated radiation therapy, proton therapy, or conformal radiation therapy and morbidity and disease control in localized prostate cancer. JAMA 2012; 307: Nguyen PL, Gu X, Lipsitz SR et al. Cost implications of the rapid adoption of newer technologies for treating prostate cancer. J Clin Oncol 2011; 29: Hayes JH, Ollendorf DA, Pearson SD et al. Active surveillance compared with initial treatment for men with low-risk prostate cancer: a decision analysis. JAMA 2010; 304: Hayes JH, Ollendorf DA, Pearson SD et al. Observation versus initial treatment for men with localized, low-risk prostate cancer: a cost-effectiveness analysis. Ann Intern Med 2013; 158: Wilt TJ, Brawer MK, Jones KM et al. Radical prostatectomy versus observation for localized prostate cancer. N Engl J Med 2012; 367: Wong YN, Mitra N, Hudes G et al. Survival associated with treatment vs observation of localized prostate cancer in elderly men. JAMA 2006; 296: Klabunde CN, Potosky AL, Legler JM, Warren JL. Development of a comorbidity index using physician claims data. J Clin Epidemiol 2000; 53: Hu JC, Gu X, Lipsitz SR et al. Comparative effectiveness of minimally invasive vs open radical prostatectomy. JAMA 2009; 302: Abdollah F, Sun M, Thuret R et al. A competing-risks analysis of survival after alternative treatment modalities for prostate cancer patients: Eur Urol 2011; 59:

15 Hu CY, Xing Y, Cormier JN, Chang GJ. Assessing the utility of cancer-registryprocessed cause of death in calculating cancer-specific survival. Cancer D'Agostino RB, Jr. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med 1998; 17: Ho DE, Imai K, King G, Stuart E. MatchIt: nonparametric preprocessing for parametric causal inference. J Stat Softw 2011; Gray R. A class of K-sample tests for comparing the cumulative incidence of a competing risk. Ann Stat 1988; 16: Abdollah F, Sun M, Schmitges J et al. Competing-risks mortality after radiotherapy vs. observation for localized prostate cancer: a population-based study. Int J Radiat Oncol Biol Phys 2012; 84: Jacobs BL, Zhang Y, Skolarus TA et al. Comparative Effectiveness of External- Beam Radiation Approaches for Prostate Cancer. Eur Urol Spratt DE, Pei X, Yamada J et al. Long-term survival and toxicity in patients treated with high-dose intensity modulated radiation therapy for localized prostate cancer. Int J Radiat Oncol Biol Phys 2013; 85: Esserman LJ, Thompson IM, Jr., Reid B. Overdiagnosis and overtreatment in cancer: an opportunity for improvement. JAMA 2013; 310: Vora SA, Wong WW, Schild SE et al. Outcome and toxicity for patients treated with intensity modulated radiation therapy for localized prostate cancer. J Urol 2013; 190: Austin PC. The relative ability of different propensity score methods to balance measured covariates between treated and untreated subjects in observational studies. Med Decis Making 2009; 29:

16 16 FIGURE LEGEND Figure 1. Cumulative incidence plot depicting cancer-specific mortality (CSM) rates stratified according to intensity-modulated radiotherapy (IMRT) vs. initial observation (A) in the overall population; (B) in patients with low/intermediate-risk; and (C) in patients with high-risk prostate cancer (PCa).

17 17

18 18

19 Table 1. Descriptive statistics of 42,483 patients treated with observation vs. Intensity-modulated radiation therapy (IMRT) for prostate cancer (PCa) between 2001 and 2007 within the Surveillance Epidemiology and End Results (SEER) database stratified according to the type of treatment. Age at diagnosis Mean (Median) IQR Year of diagnosis Mean (Median) IQR Race (%) White African-American Other Income <34,837 34,838-46,648 46,649-62,339 >62,339 Education (%) < >39.2 Marital status (%) Married Unmarried Population density (%) Metropolitan Non Metropolitan CCI (%) 1 2 Gleason score (%) Before propensity-score matching Total (n=42483) Observation IMRT (n=23261, (n=19222, 54.8%) 45.2%) 72.5 (73) (2005) (78.3) 5982 (14.1) 3253 (7.3) (25.6) (24.8) (24.9) (24.7) (26.2) (25.0) (25.1) (23.7) (67.2) (32.8) (84.5) 6585 (15.5) (69.5) (30.5) 72.5 (73) (2004) (76.5) 3617 (15.5) 1852 (8.0) 6483 (27.9) 5896 (25.3) 5781 (24.9) 5101 (21.9) 6560 (28.2) 5977 (25.7) 5658 (24.3) 5066 (21.8) (63.1) 8591 (36.9) (83.6) 3806 (16.4) (71.5) 6622 (28.5) 72.6 (73) (2005) (80.4) 2365 (12.3) 1401 (7.3) 4402 (22.9) 4643 (24.2) 4785 (24.9) 5392 (28.1) 4577 (23.8) 4633 (24.1) 5021 (26.1) 4991 (26.0) (67.2) 5352 (27.8) (85.5) 2779 (14.5) (67.1) 6322 (32.9) Mean standardized difference Total (n=19064) (73) (2005) (80.0) 2434 (12.8) 1374 (7.2) (23.2) 4561 (23.9) 4875 (25.6) 5211 (27.3) (23.8) 2684 (24.6) 4967 (26.1) 4873 (25.6) (71.7) 5402 (28.3) (85.8) 2704 (14.2) (66.1) 6466 (33.9) After propensity-score matching Observation IMRT (n=9532, 50%) (n=9532, 50%) 72.5 (73) (2005) (79.9) 1226 (12.9) 693 (7.3) 2196 (23.0) 2283 (24.0) 2534 (26.6) 2519 (26.4) 2264 (23.8) 2368 (24.8) 2478 (26.0) 2422 (25.4) 6817 (71.5) 2715 (28.5) 8192 (85.9) 1340 (14.1) 6247 (65.5) 3285 (54.5) 72.6 (73) (2005) (80.2) 1208 (12.7) 681 (7.1) 2221 (23.3) 2278 (23.9) 2341 (24.6) 2692 (28.2) 2276 (23.9) 2316 (24.3) 2489 (26.1) 2451 (25.7) 6845 (71.8) 2687 (28.2) 8168 (85.7) 1364 (14.3) 6351 (66.6) 3181 (33.4) Mean standardized difference

20 (83.9) 6829 (16.1) (88.1) 2762 (11.9) (78.8) 4067 (21.2) (80.3) 3765 (19.7) Clinical stage (%) T2b 1.2 T2c Risk group (%) Low-intermediate risk (81.2) (85.6) (76.0) NA (77.9) High-risk 7968 (18.8) 3357 (14.4) 4611 (24.0) 4207 (22.1) IMRT: Intensity-modulated radiation therapy; IQR: interquartile range; CCI: Charlson comorbidity index 7796 (81.8) 1736 (18.2) 7614 (79.9) 1918 (20.1) 7503 (78.7) 2029 (21.3) 7243 (76.0) 2289 (24.0) 7.5 NA

21 Table 2a. Multivariable competing-risks regression analysis predicting cancer-specific mortality (CSM) and other-cause mortality (OCM) in 42,483 patients with prostate cancer (PCa) treated with observation vs. intensity-modulated radiation therapy (IMRT) within the Surveillance Epidemiology and End Results (SEER)-Medicare database between 2001 and Cancer-specific mortality Other-cause mortality IMRT vs. observation P-value IMRT vs. observation P-value HR (95% CI) HR (95% CI) Overall* 0.55 ( ) < ( ) <0.001 Risk group* Low-intermediate High 0.88 ( ) 0.42 ( ) 0.5 < ( ) 0.77 ( ) < Low/intermediate-risk, <73 years** Low/intermediate-risk, 73 years** Low/intermediate-risk, CCI 1*** Low/intermediate-risk, CCI 2*** 0.76 ( ) 0.93 ( ) 0.85 ( ) 0.88 ( ) ( ) 0.74 ( ) 0.94 ( ) 0.68 ( ) 0.1 < <0.001 High-risk, <73 years** High-risk, 73 years** High-risk, CCI 1*** High-risk, CCI 2*** 0.31 ( ) 0.54 ( ) 0.39 ( ) 0.47 ( ) <0.001 <0.001 < ( ) 0.77 ( ) 0.72 ( ) 0.84 ( ) IMRT: intensity-modulated radiation therapy; HR: Hazard ratio; CI: confidence interval; CCI: Charlson Comorbidity Index *Model adjusted for: age at diagnosis, year of diagnosis, region, race, marital status, population density, Gleason score, clinical stage, Charlson comorbidity index, and the use of salvage therapies. **Model adjusted for: year of diagnosis, region, race, marital status, population density, Charlson comorbidity index, Gleason score, clinical stage, and the use of salvage therapies. ***Model adjusted for: age at diagnosis, year of diagnosis, region, race, marital status, population density, Gleason score, clinical stage, and the use of salvage therapies

22 Table 2b. Cancer-specific mortality (CSM) rates in 42,483 patients treated with observation vs. intensity-modulated radiation therapy (IMRT) within the Surveillance Epidemiology and End Results (SEER)-Medicare database between 2001 and Low/intermediate-risk, <73 years Low/intermediate-risk, 73 years Low/intermediate-risk, CCI 1 Low/intermediate-risk, CCI 2 High-risk, <73 years High-risk, 73 years High-risk, CCI 1 High-risk, CCI 2 Cancer-specific mortality rates (95% CI) IMRT Observation P-value NNT 1.6 ( ) 0.5 NA 1.9 ( ) 0.8 NA 1.3 ( ) 0.8 NA 2.5 ( ) 0.9 NA 2.2 ( ) 2.9 ( ) 2.4 ( ) 2.7 ( ) 4.3 ( ) 6.9 ( ) 5.3 ( ) 6.1 ( ) 9.4 ( ) 11.5 ( ) 11.4 ( ) 10.1 ( ) < <