Proposed prognostic scoring system evaluating risk factors for biochemical recurrence of prostate cancer after salvage radiation therapy

Transcription

1 Proposed prognostic scoring system evaluating risk factors for biochemical recurrence of prostate cancer after salvage radiation therapy Richard J. Lee, Katherine S. Tzou, Michael G. Heckman*, Corey J. Hobbs, Bhupendra Rawal*, Nancy N. Diehl*, Jennifer L. Peterson, Nitesh N. Paryani, Stephen J. Ko, Larry C. Daugherty, Laura A. Vallow, William Wong, Steven Schild, Thomas M. Pisansky and Steven J. Buskirk Department of Radiation Oncology, *Division of Biomedical Statistics and Informatics Jacksonville, FL, Department of Radiation Oncology Scottsdale, AZ, and Department of Radiation Oncology,Mayo Clinic,Rochester,MN,USA Objective To update a previously proposed prognostic scoring system that predicts risk of biochemical recurrence (BCR) after salvage radiation therapy (SRT) for recurrent prostate cancer when using additional patients and a PSA value of 0.2 ng/ml and rising as the definition of BCR. Patients and Methods We included 577 patients who received SRT for a rising PSA after radical prostatectomy in this retrospective cohort study. Clinical, pathological, and SRT characteristics were evaluated for association with BCR using relative risks (RRs) from multivariable Cox regression models. Results With a median follow-up of 5.5 years after SRT, 354 patients (61%) experienced BCR. At 5 years after SRT, 40% of patients were free of BCR. Independent associations with BCR were identified for the PSA level before SRT (RR [doubling]: 1.25, P < 0.001), pathological tumour stage (RR [T3a vs T2] 1.21, P = 0.19; RR [T3b/T4 vs T2] 2.09, P < 0.001; overall P < 0.001), Gleason score (RR [7 vs <7] 1.63, P < 0.001; RR [8 10 vs <7] 2.28, P < 0.001; overall P < 0.001), and surgical margin status (RR [positive vs negative] 0.71, P = 0.003). We combined these four variables to create a prognostic scoring system that predicted BCR risk with a c-index of Scores ranged from 0 to 7, and 5-year freedom from BCR for different levels of the score was as follows: Score = 0 1: 66%, Score = 2: 46%, Score = 3: 28%, Score = 4: 19%, and Score = 5 7: 15%. Conclusion We developed a scoring system that provides an estimation of the risk of BCR after SRT. These findings will be useful for patients and physicians in decision making for radiation therapy in the salvage setting. Keywords salvage radiation therapy, prostate cancer, prognostic factor, biochemical recurrence Introduction Radical prostatectomy (RP) is an effective treatment for localised prostate cancer and improves overall survival compared with observation, with the largest benefit in men aged <65 years and in those with intermediate-risk disease [1]. However, about one-third of men undergoing RP will experience recurrent disease that may lead to additional therapy including salvage radiation therapy (SRT). In 2013, the American Society for Radiation Oncology (ASTRO) and the AUA proposed guidelines for postoperative management [2] that also endorsed the AUA consensus definition of biochemical recurrence (BCR) as a serum PSA level of 0.2 ng/ml with a subsequent value >0.2 ng/ml [3]. Certain factors are associated with a greater risk of disease relapse after RP. These include higher preoperative PSA level [4,5], higher Gleason score [4 7], a positive surgical margin [4,5], tumour extension through the prostatic capsule [6,7], seminal vesicle invasion [4,7], and pelvic lymph node involvement [4,7]. Additional factors including shorter PSA doubling time, shorter time interval between RP and BCR, and elevated PSA level before SRT, are also associated with outcomes after SRT [8 10]. Using some or all of these risk factors, there are predictive tools available to physicians and patients for guidance in their decision making to tailor the most appropriate treatment. In the present study, we evaluated risk factors for BCR after SRT, using the AUA Panel definition of BCR, of a PSA level BJU Int 2016; 118: wileyonlinelibrary.com BJU International 2015 BJU International doi: /bju Published by John Wiley & Sons Ltd.

2 Scoring system for BCR after SRT of 0.2 ng/ml and rising following the post-srt nadir [3]. Our findings confirm independent associations with BCR after SRT for pathological stage, Gleason score, and PSA level before SRT [11]. Additionally, a positive surgical margin was identified as an independent prognostic factor for decreased BCR. Using these four variables, we propose the SRT BCR Risk Score 2.0 as an update to our previously reported BCR prognostic scoring system [11]. Patients and Methods In all, 577 patients who received SRT between July 1987 and February 2012 for a rising PSA level after RP at the Mayo Clinic Florida (308 patients), Mayo Clinic Rochester (126), and Mayo Clinic Arizona (143) were included in this Institutional Review Board-approved retrospective cohort study. Information regarding PSA level before RP, PSA level before SRT, SRT dose, patient age, time interval between RP and SRT initiation, race, pathological tumour stage, surgical margin status, Gleason score, androgen-suppression therapy (AST) before SRT, date of RP, date of SRT, date of BCR, and date of last PSA level measurement was abstracted from the medical record of patients who provided consent. Information was unavailable for relatively few patients for PSA level before RP (43 patients), pathological tumour stage (six), surgical margin status (15), and Gleason score (43). BCR was defined as a PSA value of 0.2 ng/ml and rising folllowing the post-srt nadir. The date of BCR was the initial date of a PSA level 0.2 ng/ml without backdating. SRT was defined as RT administered for BCR after RP, or for RT administered >6 months after RP. Patients included in our review had more than one recorded PSA value after SRT. Patients who received AST before SRT were included (13% of all patients), but patients who received AST during or immediately after SRT were not. The PSA level before SRT in patients who received AST was defined as the PSA level immediately before initiation of neoadjuvant AST. Patients whose PSA level was detectable after RP ( 0.1 ng/ml) and were treated with RT to the prostate fossa at 6 months of RP were considered to be treated adjuvantly, and therefore were excluded. The SRT technique has been described previously [11]. All patients were treated with SRT using megavoltage (6 20 mv photons) linear accelerators. SRT was delivered to the prostate fossa consistent with contemporary guidelines [12], with the exception of variable inclusion of the entire seminal vesicle bed, which was at the discretion of the treating physician. Five patients also received treatment to the pelvic lymphatic regions ( Gy). The treatment volume was defined by skeletal landmarks with or without surgical clip localisation until the mid-1990s, when CT-based treatment planning was performed on all patients. Retrograde urethrography was performed in many patients. The median (range) dose to the prostate bed was 66.6 ( ) Gy. Freedom from BCR after the start of SRT was estimated using the Kaplan Meier method, censoring on the date of the last PSA measurement. Associations of patient clinical, pathological, and SRT characteristics with the endpoint of BCR after SRT were evaluated using Cox proportional hazards regression models. Relative risks (RRs) and 95% CIs were estimated. Single variable models adjusted only for Mayo Clinic site were used in an exploratory analysis, while multivariable analyses were adjusted for Mayo Clinic site and any additional variable with a P 0.05, using a forward selection approach. Based on the results of the multivariable Cox regression analysis, we created a risk score that classifies patients according to their likelihood of experiencing BCR after SRT, and we refer to this as the SRT BCR Risk Score 2.0. To assess the predictive value of the SRT BCR Risk Score 2.0, we estimated the c-index [13], where a value of 1.0 indicates that the SRT BCR Risk Score 2.0 perfectly separates patients with and without BCR, and a value of 0.5 indicates that the SRT BCR Risk Score contains prognostic information equal to that of chance alone. The SRT BCR Risk Score 2.0 was internally validated using two different methods. First, the aforementioned c-index was estimated using bootstrapping with resamples of the SRT BCR Risk Score 2.0, where the final c-index reported herein is the median c-index from these resamples. Second, we assessed calibration by comparing the predicted proportion of patients free of BCR at 5 years after the start of BCR from Cox regression analysis including SRT BCR Risk Score 2.0 as a covariate with the actual Kaplan Meier proportion free of BCR at 5 years after the start of SRT. The PSA level before RP, PSA level before SRT, and the time interval from RP to SRT initiation were considered on the logarithm scale in all Cox regression analyses owing to their skewed distributions. To account for the number of statistical tests that were performed in our evaluation of risk factors for BCR and control the family-wise error rate at 5%, we used a Bonferroni correction, after which P were considered as statistically significant. All statistical tests were two-sided. All statistical analysis was performed using SAS (version 9.2; SAS Institute, Inc., Cary, NC, USA) and R Statistical Software (version ; R Foundation for Statistical Computing, Vienna, Austria). Results A summary of the clinical, pathological, and SRT characteristics is given in Table 1. The median (range) PSA level before RP was 7.7 ( ) ng/ml and before SRT was 0.6 ( ) ng/ml. Most patients had either pathological tumour stage T2 (38.5%) or T3a (42.2%). BJU International 2015 BJU International 237

3 Lee et al. Table 1 Summary of patients clinical, pathological, and SRT characteristics. Variable Value Number of patients 577 Median (range) PSA level before RP*, ng/ml 7.7 ( ) PSA level before SRT, ng/ml 0.6 ( ) SRT dose, Gy 66.6 ( ) Age at SRT initiation, years 68.1 ( ) Time from RP to SRT initiation, months 24.6 ( ) N (%) Year of start of SRT (18.4) (28.1) (17.3) (36.2) Race* Caucasian 544 (96.6) African-American 17 (3.0) Other 2 (0.4) Pathological tumour stage* T2 220 (38.5) T3a 241 (42.2) T3b 108 (18.9) T4 2 (0.4) Surgical margin* Positive 299 (53.2) Negative 263 (46.8) Gleason score* (30.7) (50.2) (19.1) AST before SRT Yes 74 (12.8) No 503 (87.2) *Information was not available for the following variables: PSA level before RP (43 patients), race (14), pathological tumour stage (six), surgical margin (15), and Gleason score (43). Gleason scores were between 3 and 6 in 164 patients (30.7%), 7 in 268 patients (50.2%), and 8 10 in 102 patients (19.1%). With a median (range) follow-up of 5.5 years ( ) years, 354 patients (61%) experienced BCR. Kaplan Meier estimates of the proportion of patients free of BCR after the start of SRT at 1, 3, and 5 years, were 71% (95% CI 67 75%), 52% (95% CI 48 56%), and 40% (95% CI 36 45%), respectively (Fig. S1). Associations of patient clinical, pathological, and SRT characteristics with BCR after SRT are given in Table 2. In multivariable analysis, adjusting for Mayo Clinic site and all variables with a P 0.05 in forward selection (PSA level before SRT, pathological tumour stage, Gleason score, and surgical margin), independent and significant associations with BCR were identified for the PSA level before SRT [RR (per each doubling) 1.25, P < 0.001), pathological tumour stage [RR (T3a vs T2) 1.21, P = 0.19; RR (T3b/T4 vs T2) 2.09, P < 0.001; overall P < 0.001), Gleason score [RR (7 vs 3 6) 1.63, P < 0.001; RR (8 10 vs 3 6) 2.28, P < 0.001; overall P < 0.001), and surgical margin [RR (positive vs negative) 0.71, P = 0.003) (Figs S2 S5). Of note, the lack of significant association between surgical margin and BCR in single variable analysis appears to be the result of confounding with pathological tumour stage, where positive surgical margins were more common in patients with more advanced pathological tumour stage (P < 0.001) causing an artificially higher risk of BCR in patients with positive surgical margins until accounted for in multivariable analysis. We further explored the association between the PSA level before SRT and BCR by considering a three-level categorical variable (<0.5, , >1.0 ng/ml). In multivariable analysis compared with patients with a PSA level before SRT of <0.5 ng/ml, risk of BCR was higher for patients with a PSA level before SRT of ng/ml (RR 1.35, 95% CI , P = 0.024) and also for patients with a PSA level before SRT of >1.0 ng/ml (RR 2.04, 95% CI , P < 0.001). Based on the results of our multivariable regression analysis, we created the SRT BCR Risk Score 2.0. In these calculations, we used the four aforementioned variables that were significantly associated with BCR in multivariable analysis. We assigned each level of these variables an individual score based on both the relative risk estimate and the pair-wise comparison P value in relation to the reference category. The reference category of each variable was assigned a score of 0, levels with a RR <1.25 and/or a P > 0.05 compared with the reference group were assigned a score of 0, levels with a RR of were assigned a score of 1, and levels with a RR > 1.75 were assigned a score of 2. These individual scores were summed across the four variables for a given patient to create the SRT BCR Risk score 2.0, which has a possible range of 0 7. A higher risk score indicates an increased risk of BCR (Table 3). Patients with missing information for any one of the four variables involved in the SRT BCR Risk Score 2.0 were excluded from its calculation. A reference category of positive surgical margins was used in order to obtain a RR estimate >1 for that variable. A detailed summary of the ability of the SRT BCR Risk Score 2.0 to predict risk of BCR after SRT is provided in Table 4. The 5-year freedom from BCR was 66% for patients with a risk score of 0 1 (138 patients), 46% for patients with a risk score of 2 (145), 28% for patients with a risk score of 3 (116), 19% for patients with a risk score of 4 (75), and 15% for patients with a risk score of 5, 6, or 7 (47) (Fig. 1). Compared with patients with a SRT BCR Risk Score 2.0 of 0 1, those with risk scores of 2, 3, 4, and 5 7 were 1.9-, 3.1-, 3.5-, and 4.9-times more likely to experience BCR, and these differences were all highly significant (all P 0.001, Table 4). The SRT BCR Risk Score 2.0 predicted risk of BCR after SRT with a c-index of 0.66 in internal validation, and agreement was good between the predicted and observed proportion of patients free of BCR at 5 years after the start of SRT (Fig. 2). 238 BJU International 2015 BJU International

4 Scoring system for BCR after SRT Table 2 Associations of clinical, pathological, and SRT characteristics with BCR after SRT. Variable Single variable analysis Multivariable analysis RR (95% CI) P RR (95% CI) P PSA level before RP (doubling) 1.11 (1.00, 1.23) (0.85, 1.05) 0.32 PSA level before SRT (doubling) 1.22 (1.12, 1.33) < (1.14, 1.37) <0.001 SRT dose (5 Gy increase) 0.80 (0.65, 0.98) (0.68, 1.08) 0.20 Age at SRT initiation (10 year increase) 0.99 (0.84, 1.16) (0.77, 1.08) 0.28 Time from RP to SRT initiation (doubling) 0.93 (0.87, 1.00) (0.89, 1.04) 0.33 Year of start of SRT (5 year increase) 0.98 (0.89, 1.09) (0.98, 1.27) 0.11 Race (non-caucasian) 0.56 (0.28, 1.13) (0.27, 1.10) Pathological tumour stage Test of overall difference: P < Test of overall difference: P < T (reference) N/A 1.00 (reference) N/A T3a 1.21 (0.93, 1.58) (0.91, 1.62) 0.19 T3b/T (1.78, 3.19) < (1.51, 2.89) <0.001 Surgical margin (positive) 0.85 (0.69, 1.05) (0.56, 0.89) Gleason score (reference) N/A 1.00 (reference) N/A (1.17, 2.00) (1.24, 2.14) < (1.85, 3.46) < (1.65, 3.16) <0.001 AST before SRT 1.08 (0.80, 1.48) (0.58, 1.14) 0.23 RRs, 95% CIs, and P values result from Cox proportional hazards regression models. RRs correspond to presence of the given characteristic (categorical variables) or the increase given in parenthesis (continuous variables). For pathological tumour stage and Gleason score, results of tests of overall difference are given as well as pair-wise comparisons vs the reference category. Single variable models were adjusted for Mayo Clinic site only. Multivariable models were adjusted for Mayo Clinic site and any additional variable that was associated with BCR with a P 0.05 using a forward selection approach, which included PSA level before SRT, pathological tumour stage, Gleason score, and surgical margin. Table 3 SRT BCR Risk Score 2.0 description. Variable Discussion Individual score PSA level before SRT, ng/ml < >1.0 2 Pathological tumour stage T2 0 T3a 0 T3b/T4 2 Surgical margin Positive 0 Negative 1 Gleason score Individual scores were summed to create the SRT BCR Risk Score 2.0, which has a possible range of between 0 and 7. Patients with missing information for any one of the four variables involved in the SRT BCR Risk Score 2.0 were excluded (56 patients). The distribution of the SRT BCR Risk Score 2.0 was as follows: Score = 0 (24 patients), Score = 1 (114), Score = 2 (145), Score = 3 (116), Score = 4 (75), Score = 5 (34), Score = 6 (11), and Score = 7 (two). There are various predictive tools available to guide physicians and patients in making treatment decisions after disease recurrence after RP. In 2006, we developed a prognostic scoring system for patients who developed BCR and received SRT after RP. In this system, the PSA level before SRT, tumour stage, and the Gleason score provided a prognostic framework for physicians to provide patient guidance when counselling treatment options in the event of BCR after RP [11]. At that time, biochemical relapse after RT was defined as a PSA level after SRT of 0.4 ng/ml and rising [14]. The current AUA/ASTRO guidelines define BCR after SRT as a PSA level 0.2 ng/ml and rising [2]. Therefore, the aim of our present study was to update our previous analysis using this definition of BCR in a larger study population. This updated scoring algorithm is the second largest retrospective analysis published to date on this topic. With a c-index of 0.66, our present SRT BCR Risk Score 2.0 predicts risk of BCR slightly better than our aforementioned previous scoring system, which has a c-index of 0.65 when applied to these data, and comparably to the nomogram proposed in by Stephenson et al. [10], which had a c-index of 0.69 in the original study of patients and c-indices of 0.65 and 0.71 in two subsequent smaller validation studies [15,16]. After RP, a detectable PSA level of 0.2 ng/ml has been associated with a median time to distant metastasis from prostate cancer of 7 8 years, with death following metastasis 5 years later [17,18]. There has been debate about the absolute biochemical threshold that best correlates with eventual disease relapse (mainly in the PSA level range of ng/ml). Historically, a biochemical failure threshold of a PSA level of 0.4 ng/ml was found to be more significantly related to eventual disease progression [14]. In our original scoring proposal, we used this 0.4 ng/ml PSA level threshold in defining BCR. Currently, the AUA/ASTRO BJU International 2015 BJU International 239

5 Lee et al. Table 4 Risk of BCR according to SRT BCR Risk Score 2.0. SRT BCR Risk Score 2.0 N Proportion free of BCR, % (95% CI) Association with BCR 1 year 3 years 5 years RR (95% CI) P (84 95) 80 (74 88) 66 (57 77) 1.00 (reference) N/A (73 86) 56 (48 65) 46 (38 56) 1.89 (1.32, 2.72) (49 68) 37 (29 47) 28 (20 38) 3.11 ( ) < (50 73) 34 (24 47) 19 (12 33) 3.49 ( ) < (29 57) 26 (16 42) 15 (8 30) 4.88 ( ) <0.001 RRs, 95% CIs, and P values results from a single variable Cox proportional hazards regression model. Patients with missing information for any one of the four variables involved in the SRT BCR Risk Score 2.0 were excluded (56 patients). The SRT BCR Risk Score 2.0 predicted risk of BCR with a c-index of Fig. 1 Proportion of patients free of BCR after SRT according to SRT BCR Risk Score 2.0. No. at risk Proportion free of BCR (%) Time from the start of salvage radiation therapy, years Fig. 2 Calibration plot for the proportion of patients free of BCR at 5 years after the start of SRT. Patients were grouped according to their SRT BCR Risk Score 2.0 (0 1, 2, 3, 4, or 5 7). The dashed line indicates the ideal reference line where predicted 5-year BCR-free proportions are identical to the actual Kaplan Meier estimated 5-year BCR-free proportions. Vertical lines represent the 95% CI for the given Kaplan Meier estimated actual proportion of patients BCR-free at 5 years after the start of SRT. Actual proportion free of BCR at 5 years after SRT, % Predicted proportion free of BCR at 5 years after SRT, % defines BCR as a threshold PSA level of 0.2 ng/ml and rising, which we had previously identified as a level of failure [19]. In the present study, we used this threshold to define BCR and confirmed associations for our three previously identified independent risk factors (PSA level before SRT, tumour stage, Gleason score) and also identified an additional significant independent association with BCR for surgical margins. Patients with locally advanced prostate cancers that have pathological features of extraprostatic extension (T3a) or seminal vesicle invasion (T3b) are at increased risk of BCR. In patients who have not received adjuvant therapy immediately after RP, SRT should be considered when PSA levels after RP are rising [2]. In three randomised clinical trials of adjuvant RT, about one-third to one-half of the patients assigned to postoperative observation arm eventually received SRT [20 22]. Our present study confirms that patients with seminal vesicle involvement are at a higher risk for BCR after SRT. However, extraprostatic extension alone (T3a) is not a significant factor for BCR. Our present findings 240 BJU International 2015 BJU International

6 Scoring system for BCR after SRT also confirmed that elevated Gleason score is an independent risk factor for a higher risk of biochemical failure. In the Southwest Oncology Group (SWOG) 8794 adjuvant trial, a rising PSA level after RP was found to be attributable to the persistence of cancer in the prostate fossa [23]. A single-institution retrospective analysis revealed a reduction in metastases and prostate cancer-related mortality when SRT was used in patients with elevated PSA levels after RP [24]. These observations suggest that the local control of residual prostate cancer in the prostate fossa reduces a late wave of secondary metastasis. The European Organisation for the Research and Treatment of Cancer (EORTC) trial found that adjuvant RT immediately after RP improved biochemical progression-free survival and local control in patients with positive surgical margins or pt3 prostate cancer [20]. Additional review of the pathology revealed that patients with a positive surgical margin benefited from immediate RT in contrast to patients with a negative margin [25]. In our present multivariable analysis, patients with positive surgical margins who underwent SRT had a decreased risk of BCR. Of note, PSA doubling time has been associated with risk of BCR after SRT by several groups, and was included in the aforementioned nomogram by Stephenson et al. [8,10]. We did not evaluate doubling time as a potential risk factor for BCR because calculation of doubling time is not possible for some patients who are candidates for SRT, and our goal was to develop a prediction tool that can be applied to all SRT candidates as long as standard clinicopathological data are collected. Specifically, because it is recommended that doubling time should only be calculated for patients that satisfy certain requirements for the number, frequency, and magnitude of PSA level measurements before SRT, where PSA levels are not influenced by AST [26], doubling time information will not be available for some men who need to make a decision about whether or not to undergo SRT. Therefore, although PSA doubling time has been shown to be a valuable prognostic factor when available, our present SRT BCR Risk Score 2.0 provides physicians with a tool that should be able to be used for all SRT candidates, rather than only the subset for whom doubling time can be calculated. With that said, both our SRT BCR Risk Score 2.0 and the Stephenson nomogram (which as mentioned does use PSA doubling time information) appear to provide valuable prognostic information, and therefore we recommend that physicians use both tools when attempting to determine risk of BCR after SRT for a given patient. The question is raised for both clinicians and patients, does SRT add value? The retrospective nature of our prognostic scoring system limits our ability to provide a conclusion to this question. The determination of whether men with a poor score will ultimately benefit from SRT compared with no treatment will require a randomised trial. Small retrospective studies have shown improved response rates to SRT, but a large randomised study would be difficult to perform randomising men between SRT and observation alone. In the meantime, we must extrapolate from previously conducted randomised trials (EORTC 22911, SWOG 8794, and ARO 96-02) that do reveal a benefit with adjuvant therapy, one of which revealed a survival advantage with treatment in the adjuvant setting. As in any retrospective study, there are limitations to our present analysis. Because this scoring system was created from a retrospective analysis of patients treated during a 25-year period, patient selection and treatment bias may have influenced the outcome. Additionally, the primarily Caucasian population may limit the ability to generalise our findings. Finally, a median follow-up duration of 5.5 years may limit our conclusions for long-term benefits to SRT. In conclusion, in the present study we have updated our previous scoring algorithm, developing a risk score that classifies patients according to the likelihood of BCR after SRT using the most recent definition of BCR after RP. Multivariable analysis has confirmed pathological stage T3a, Gleason score 7, low PSA levels before SRT and positive surgical margin as independent prognostic factors for decreased BCR. We recommend that this scoring system be validated in an additional set of patients before we would fully advocate its routine use in a clinical setting. In the meantime, this scoring system is a tool available for physicians and patients to use when weighing the risks vs potential benefits of SRT, and when deciding whether to add AST to SRT in patients at higher risk for BCR after SRT alone. Funding Source There was no funding source for this study. Conflicts of Interest The authors have no conflicts of interest to disclose. References 1 Bill-Axelson A, Holmberg L, Garmo H et al. Radical prostatectomy or watchful waiting in early prostate cancer. N Engl J Med 2014; 370: Valicenti RK, Thompson I Jr, Albertsen P et al. Adjuvant and salvage radiation therapy after prostatectomy: American Society for Radiation Oncology/American Urological Association guidelines. Int J Radiat Oncol Biol Phys 2013; 86: Cookson MS, Aus G, Burnett AL et al. Variation in the definition of biochemical recurrence in patients treated for localized prostate cancer: the American Urological Association Prostate Guidelines for Localized Prostate Cancer Update Panel report and recommendations for a standard in the reporting of surgical outcomes. J Urol 2007; 177: Thompson RH, Blute ML, Slezak JM, Bergstralh EJ, Leibovich BC. Is the GPSM scoring algorithm for patients with prostate cancer valid in the contemporary era? J Urol 2007; 178: BJU International 2015 BJU International 241

7 Lee et al. 5 Mullins JK, Feng Z, Trock BJ, Epstein JI, Walsh PC, Loeb S. The impact of anatomical radical retropubic prostatectomy on cancer control: the 30- year anniversary. J Urol 2012; 188: Han M, Partin AW, Zahurak M, Piantadosi S, Epstein JI, Walsh PC. Biochemical (prostate specific antigen) recurrence probability following radical prostatectomy for clinically localized prostate cancer. J Urol 2003; 169: Roehl KA, Han M, Ramos CG, Antenor JA, Catalona WJ. Cancer progression and survival rates following anatomical radical retropubic prostatectomy in 3,478 consecutive patients: long-term results. J Urol 2004; 172: Jackson WC, Johnson SB, Li D et al. A prostate-specific antigen doubling time of <6 months is prognostic for metastasis and prostate cancerspecific death for patients receiving salvage radiation therapy post radical prostatectomy. Radiat Oncol 2013; 8: Johnson S, Jackson W, Li D et al. The interval to biochemical failure is prognostic for metastasis, prostate cancer-specific mortality, and overall mortality after salvage radiation therapy for prostate cancer. Int J Radiat Oncol Biol Phys 2013; 86: Stephenson AJ, Scardino PT, Kattan MW et al. Predicting the outcome of salvage radiation therapy for recurrent prostate cancer after radical prostatectomy. J Clin Oncol 2007; 25: Buskirk SJ, Pisansky TM, Schild SE et al. Salvage radiotherapy for isolated prostate specific antigen increase after radical prostatectomy: evaluation of prognostic factors and creation of a prognostic scoring system. J Urol 2006; 176: Michalski JM, Lawton C, El Naqa I et al. Development of RTOG consensus guidelines for the definition of the clinical target volume for postoperative conformal radiation therapy for prostate cancer. Int J Radiat Oncol Biol Phys 2010; 76: Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 1996; 15: Amling CL, Bergstralh EJ, Blute ML, Slezak JM, Zincke H. Defining prostate specific antigen progression after radical prostatectomy: what is the most appropriate cut point? J Urol 2001; 165: Moreira DM, Jayachandran J, Presti JC Jr et al. Validation of a nomogram to predict disease progression following salvage radiotherapy after radical prostatectomy: results from the SEARCH database. BJU Int 2009; 104: Johnson S, Jackson W, Speers C, Feng F, Hamstra D. A comprehensive assessment of the prognostic utility of the Stephenson nomogram for salvage radiation therapy postprostatectomy. Pract Radiat Oncol 2014; 4: Pound CR, Partin AW, Eisenberger MA, Chan DW, Pearson JD, Walsh PC. Natural history of progression after PSA elevation following radical prostatectomy. JAMA 1999; 281: Freedland SJ, Humphreys EB, Mangold LA et al. Risk of prostate cancerspecific mortality following biochemical recurrence after radical prostatectomy. JAMA 2005; 294: Schild SE, Wong WW, Novicki DE, Ferrigni RG, Swanson SK. Detection of residual prostate cancer after radical prostatectomy with the Abbott IMx PSA assay. Urology 1996; 47: Bolla M, van Poppel H, Tombal B et al. Postoperative radiotherapy after radical prostatectomy for high-risk prostate cancer: long-term results of a randomised controlled trial (EORTC trial 22911). Lancet 2012; 380: Wiegel T, Bartkowiak D, Bottke D et al. Adjuvant radiotherapy versus wait-and-see after radical prostatectomy: 10-year follow-up of the ARO 96-02/AUO AP 09/95 trial. Eur Urol 2014; 66: Thompson IM, Tangen CM, Paradelo J et al. Adjuvant radiotherapy for pathological T3N0M0 prostate cancer significantly reduces risk of metastases and improves survival: long-term followup of a randomized clinical trial. J Urol 2009; 181: Swanson GP, Hussey MA, Tangen CM et al. Predominant treatment failure in postprostatectomy patients is local: analysis of patterns of treatment failure in SWOG J Clin Oncol 2007; 25: Trock BJ, Han M, Freedland SJ et al. Prostate cancer-specific survival following salvage radiotherapy vs observation in men with biochemical recurrence after radical prostatectomy. JAMA 2008; 299: Van der Kwast TH, Bolla M, Van Poppel H et al. Identification of patients with prostate cancer who benefit from immediate postoperative radiotherapy: EORTC J Clin Oncol 2007; 25: Arlen PM, Bianco F, Dahut WL et al. Prostate Specific Antigen Working Group guidelines on prostate specific antigen doubling time. J Urol 2008; 179: Correspondence: Katherine S. Tzou, Department of Radiation Oncology, Mayo Clinic, 4500 San Pablo Rd, Jacksonville, FL 32224, USA. Tzou.Katherine@mayo.edu Abbreviations: AST, androgen-suppression therapy; ASTRO, American Society for Radiation Oncology; BCR, biochemical recurrence; EORTC, European Organisation for the Research and Treatment of Cancer; RP, radical prostatectomy; RR, relative risk; (S)RT, salvage radiation therapy; SWOG, Southwest Oncology Group. Supporting Information Additional Supporting Information may be found in the online version of this article: Fig. S1 Proportion of patients free of BCR after SRT. Dashed lines represent 95% CIs. Fig. S2 Proportion of patients free of BCR after SRT according to PSA level before SRT. Fig. S3 Proportion of patients free of BCR after SRT according to pathological tumour stage. Fig. S4 Proportion of patients free of BCR after SRT according to Gleason score. Fig. S5 Proportion of patients free of BCR after SRT according to combination of surgical margins and pathological stages. 242 BJU International 2015 BJU International