A comprehensive assessment of the prognostic utility of the Stephenson nomogram for salvage radiation therapy postprostatectomy

Practical Radiation Oncology (2014) xx, xxx–xxx

www.practicalradonc.org

Original Report

A comprehensive assessment of the prognostic utility of the Stephenson nomogram for salvage radiation therapy postprostatectomy Skyler Johnson BS 1 , William Jackson BS 1 , Corey Speers MD, PhD, Felix Feng MD, Daniel Hamstra MD, PhD ⁎ Department of Radiation Oncology, University of Michigan, Ann Arbor, Michigan Received 7 November 2013; revised 24 January 2014; accepted 11 February 2014

Abstract Purpose: To investigate the prognostic utility of the Stephenson nomogram for clinically relevant endpoints, freedom from metastasis (FFM), and prostate cancer–specific survival (PCSS) in patients treated with salvage external beam radiation therapy (SRT) following a rising prostatespecific antigen (PSA) after radical prostatectomy (RP). Methods and materials: From an institutional cohort of 575 patients treated with SRT between 1986 and 2010, the Stephenson nomogram variables were retrospectively collected and available for 179 patients. The prognostic impact of the Stephenson nomogram on 6-year freedom from biochemical failure (FFBF), FFM, and PCSS was assessed on univariate and multivariate analysis using Kaplan-Meier and Cox proportional hazards models. The prognostic utility of the Stephenson nomogram was compared with individual pretreatment, treatment, and clinical characteristics using concordance indices. Results: In the 179 patients with all available nomogram variables, median follow-up was 85.0 months (interquartile range [IQR], 53-113) and 6-year FFBF, FFM, and PCSS were 38% (95% confidence interval [CI], 30-46), 79% (95% CI, 73-85), and 96% (95% CI, 92-100), respectively. Univariate analysis, demonstrated that the Stephenson nomogram, as a continuous variable and as a risk stratified group, was prognostic of FFBF (both, P b .0001), FFM (both, P b .0001), and PCSS (both, P ≤ .0005). When analyzing individual Stephenson nomogram variables, multivariate analysis revealed that positive surgical margins (P = .02; hazard ratio [HR], 0.4; 95% CI, 0.2-0.8) and pre-RT PSA (P = .0001; HR, 1.6; 95% CI, 1.3-2.0) were prognostic for FFM, while pre-RT PSA (P = .03; HR, 1.2; 95% CI, 1.0-1.4) was the only prognostic variable for PCSS. Concordance indices revealed the Stephenson nomogram to have superior prognostic capability for biochemical failure (0.71), distant metastasis (0.75), and prostate cancer–specific mortality (0.75) when compared with individual variables (BF all ≤ 0.65, DM all ≤ 0.67, PCSM all ≤ 0.71).

Supplementary material for this article (http://dx.doi.org/10.1016/j.prro.2014.02.003) can be found at www.practicalradonc.org. Conflict of Interest: None. ⁎ Corresponding author. The University of Michigan Medical Center, 1500 East Medical Center Dr, Ann Arbor, MI 48109. E-mail address: [email protected] (D. Hamstra). 1 Co-first authorship. 1879-8500/$ – see front matter © 2014 American Society for Radiation Oncology. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.prro.2014.02.003

2

S. Johnson et al

Practical Radiation Oncology: Month 2014

Conclusions: For patients treated with SRT for a rising PSA postprostatectomy, the Stephenson nomogram is an appropriate prognostic tool for estimating the response to treatment; however, there remains a need for improvement in current and future nomograms. © 2014 American Society for Radiation Oncology. Published by Elsevier Inc. All rights reserved.

Introduction Following radical prostatectomy (RP) for clinically localized prostate cancer (PCa), approximately 30% of patients experience biochemical failure (BF). 1 Salvage external beam radiation therapy (SRT) with or without androgen deprivation therapy (ADT) is often utilized in this setting. In retrospective studies it has been suggested that SRT improves prostate cancer–specific survival (PCSS), 1 but a long-term durable response to treatment is seen in b 50% of patients. 2 Therefore, determination of which patients will most likely benefit from SRT has been of particular research interest, which has led to the development of prognostic nomograms to assist physicians and patients in making challenging treatment-related decisions. 3,4 The Stephenson nomogram is the current standard for prediction of outcome following salvage radiation therapy. 4 This nomogram has been shown to be prognostic for biochemical failure (BF) 4-6 but it has not been evaluated for other clinical endpoints. Many but not all men who have a PSA recurrence after salvage RT will subsequently develop metastasis or death from PCa. 2,7-9 Therefore, it was our goal to establish the utility of this nomogram as a prognostic tool for freedom from metastasis (FFM) and PCSS. Further, past validation studies of the Stephenson nomogram have demonstrated conflicting results in regard to which variables in the model are the best predictors of FFBF 5,6; therefore, we also wished to assess which variables were prognostic for FFM and PCSS.

Gleason score (2-6, 7, 8-10) (n = 563), seminal vesicle invasion (SVI) (n = 567), extracapsular extension (ECE) (n = 560), surgical margins (SM) (n = 552), lymph node invasion (LNI) (n = 336), persistently elevated post-RP PSA ng/mL (n = 301), pre-RT PSA (n = 553), PSA doubling-time (n = 277), neoadjuvant ADT (n = 575), and RT dose gray (Gy) (n = 574). Persistently elevated post-RP PSA was defined as serum PSA N 0.1 ng/mL, which was within the limits of detectability from the reporting laboratory. 10 PSADT was calculated using a minimum of 3 PSA values at least 2 months apart postoperatively using previously described methods. 11,12 Given many patients were referred for salvage radiation treatment at our institution after BF following RP at outside institutions; we did not have pathologic data available for all patients. Of the 575 patients, all 11 variables that make up the Stephenson nomogram were only available for 179 (31%) patients. To establish differences between patients with and without all 11 variables available, we compared demographic, pretreatment and treatment related characteristics (supplementary Table 1; available online only at www.practicalradonc.org). Additionally, to account for unknown confounders, 6-year outcomes for FFBF, FFM, and PCSS were compared between those with and without all 11 variables and there was no difference in outcome between groups (all P N .2). Therefore, due to relative similarities between groups, patients with all 11 variables available were used for the remainder of analysis and make up the cohort for this study.

Treatment

Methods and materials Patient selection and nomogram variables An institutional review board–approved retrospective analysis identified 575 patients who received RP for PCa and then later received salvage external beam radiation therapy (EBRT) with or without ADT from 1986 to 2010. Salvage patients were defined as any patient treated with EBRT for the following: (1) a persistently elevated serum PSA, defined as a PSA of 0.1 ng/mL or greater on 2 subsequent PSA readings; or (2) biochemical failure (BF) following RP, defined as an initial serum PSA of ≥ 0.2 ng/ mL with a second confirmatory level of N 0.2 ng/mL. 3,4 Nomogram variables were defined as indicated in the Stephenson nomogram 4 and primary external validation set 6 and included prostatectomy PSA ng/mL (n = 394),

Salvage EBRT was administered with either 3dimensional conformal therapy (3DCRT) or intensity modulated RT. The prescribed radiation dose ranged from 63.0 to 73.8 Gy in daily fractions of 1.8-2.0 Gy. Treatment targeted the prostate fossa with 0.5-2.0 cm planning target volume margins. The pelvic lymphatics were also treated in 8% of patients while 18% of patients received ADT during RT for a median 6.0 months (interquartile range [IQR], 4.2-13.3).

Follow-up and endpoints Following treatment, patients were typically seen 3 to 4 times per year during the first 2 years, biannually for the next 3 years, and then annually thereafter. Freedom from biochemical failure (FFBF) following salvage radiation was defined as no serum PSA value that exceeded N 0.2 ng/

Practical Radiation Oncology: Month 2014

mL than the postradiation nadir, followed by a second higher serum PSA value, 4 or any PSA value of at least 0.5 ng/mL greater than the post-SRT nadir. 4,13 FFM was defined as freedom from clinical, radiologic, or pathologic evidence of metastasis. Prostate cancer–specific survival was defined as freedom from death attributed to PCa or death in any patient with either hormone-refractory PCa or evidence of metastatic disease before death.

Statistical analysis The total number of Stephenson nomogram points (range, 85-274) was categorized into risk tertiles (low [b 157], intermediate [157-187], and high [N 187]). Oneway analysis of variance was used for comparison of continuous variables and the χ 2 test was used to compare categoric variables. Actuarial 6-year FFBF, FFM, and PCSS by risk group were calculated using univariate survival analyses that were performed using the log-rank test and Kaplan-Meier methods. Multivariate analysis was performed using Cox proportional hazards regression using all variables in the nomogram and radiation treatment time (years) to account for potential timedependent confounders (changes in treatment, staging, grading, etc). Variables in the model were used as categoric or continuous variables as demonstrated in the Stephenson nomogram 4,6; prostatectomy PSA, pre-RT PSA, PSADT, and RT dose were analyzed as continuous variables and Gleason score (2-6, 7, 8-10), persistently elevated PSA post-RP (yes vs no), neoadjuvant ADT (yes vs no), and presence or absence of SVI, ECE, SM, and LNI, were analyzed as categoric variables. Additionally, to account for multiple testing corrections, a second series of analyses were completed using a stepwise linear regression model. Only P values b .01 were chosen to remain in the model to decrease the chance of falsely rejecting the null hypothesis. To quantify the discriminatory ability of the models, c-index values were calculated as the proportion of usable pairs which were concordant. Pairs which were tied in the predictor were included as 0.5. All statistical analysis was performed using MedCalc, version 11.5.1 (MedCalc Software, Mariakerke, Belgium), SAS, version 9.2 (SAS Institute, Cary, NC), or STATA, version 11.1 (StataCorp, College Station, TX).

Results Patient cohort Of the 179 evaluable patients, median age was 63.5 years (IQR, 57.7-68.8) and median follow-up was 85.0 months (IQR, 52.8-113.2). Eighty-seven percent of patients had Gleason score of 7 or higher. Other adverse risk features existed in 52% for ECE, 12% for SVI, and 45% for positive

Stephenson nomogram for salvage RT

3

Table 1 Pretreatment- and treatment-related characteristics for analysis cohort Variables Total points, median (IQR) Age, median (IQR) Follow-up, median (IQR) mo Prostatectomy PSA, median (IQR) ng/mL Gleason score 2-6 7 8-10 SVI positive ECE positive SM positive LNI positive Persistently elevated post-RP PSA Pre-RT PSA median (IQR) ng/mL PSA doubling time, median (IQR) mo Neoadjuvant ADT Radiation dose, median (IQR) Gy

Total Stephenson nomogram (n = 179) 174 (148-197) 63.5 (57.7-68.8) 85.0 (52.8-113.2) 8.0 (5.7-13.7)

13% (23) 65% (117) 22% (39) 12% (22) 52% (93) 45% (80) 0.6% (1) 14% (24) 0.6 (0.4-1.3) 7.6 (4.3-15.1) 18% (33) 68.4 (64.8-68.4)

ADT, androgen deprivation therapy; ECE, extracapsular extension; IQR, interquartile range; LNI, lymph node invasion; PSA, prostatespecific antigen; RP, radical prostatectomy; RT, radiation therapy; SM, surgical margins; SVI, seminal vesicle invasion.

SM. Additionally, patients had a median PSA prior to prostatectomy of 8.0 ng/mL, persistently elevated post-RP PSA in 14% of the cohort, median pre-RT PSA 0.6 ng/ mL, and median PSADT of 7.6 months (Table 1).

Stephenson nomogram and clinical endpoints The median total points from the Stephenson nomogram was 174. For the entire cohort 6-year FFBF was 38% (95% confidence interval [CI], 30-46), FFM 79% (95% CI, 73-85), and PCSS 96% (95% CI, 92-100). The median nomogram score for those with at least 2 years of followup post RT and no BF was 156 (IQR, 133-177) and the median duration of follow-up was 62 months (IQR, 3198). The median nomogram score for those with BF but without metastasis was 175 (IQR, 151-198), where the median time to BF was 32 months (IQR, 16-49) and the median follow-up post RT was 95 months (IQR, 62-134). Similarly, in those with BF and metastasis the median nomogram score was 195 (IQR, 175-210), where BF occurred on average 13 months (IQR, 7-22) post RT and metastasis occurred a median of 49 months (IQR, 25-78) post RT. There was not a significant difference in mean nomogram score between those who had metastasis (192;

4

S. Johnson et al Table 2

Practical Radiation Oncology: Month 2014

Pretreatment- and treatment-related characteristics by nomogram tertile

Variables

Low risk (n = 60)

Total points, median (IQR) a Age, median (IQR) a Follow-up, median (IQR) mo a Prostatectomy PSA, median (IQR) ng/mL a Gleason score b 2-6 7 8-10 SVI positive b ECE positive b SM positive b LNI positive b Persistently elevated post-RP PSA b PreRT PSA median (IQR) ng/mL a PSA doubling time, median (IQR) mo a Neoadjuvant ADT b Radiation dose, median (IQR) Gy a

139 (130-148) 61.6 (56.1-68.5) 80.3 (51.0-107.8) 8.6 (5.4-14.0)

Intermediate risk (n = 59)

High risk (n = 60)

P

174 (163-181) 65.0 (58.4-69.5) 94.3 (56.3-118.3) 8.1 (6.2-12.5)

205 (197-222) 63.7 (58.2-69.3) 86.5 (51.3-113.0) 7.8 (5.7-14.2)

b .001 .5 .9 .6

7% (4) 78% (46) 15% (9) 10% (6) 56% (33) 44% (26) 0% (0) 10% (6) 0.6 (0.4-1.0) 7.9 (5.6-13.5) 10% (6) 68.4 (64.8-68.4)

2% (1) 57% (34) 41% (25) 20% (12) 62% (37) 30% (18) 0% (0) 17% (10) 1.1 (0.6-1.8) 4.1 (2.3-6.0) 15% (9) 68.4 (64.8-68.4)

b .0001 30% (18) 62% (37) 8% (5) 7% (4) 38% (23) 60% (36) 1.6% (1) 13% (8) 0.4 (0.3-0.7) 15.2 (9.3-18.0) 30% (18) 66.3 (64.8-68.4)

.03 .01 .001 .2 .6 .002 b .001 .03 .008

ADT, androgen deprivation therapy; ECE, extracapsular extension; IQR, interquartile range; LNI, lymph node invasion; PSA, prostate-specific antigen; RP, radical prostatectomy; RT, radiation therapy; SM, surgical margins; SVI, seminal vesicle invasion. a Analysis of variance (ANOVA) test. b χ2 test.

95% CI, 184.2-199.8) and those who died of prostate cancer (198; 95% CI,186.2-209.8) (P N .1). These results suggested differences in nomogram score in those with or without the clinical failure. This was confirmed on univariate analysis using Cox proportional hazards models, which demonstrated that the Stephenson nomogram, as a continuous variable, was prognostic of FFBF (P b .0001), FFM (P b .0001), and PCSS (P = .0001). Clinical, pretreatment, and treatment characteristics are shown by low (n = 60, median points, 139 [IQR, 130-148]), intermediate (n = 59, median points, 174 [IQR, 163-181]), and high (n = 60, median points, 205 [IQR, 197-222]) risk groups (Table 2). Not surprisingly, there were statistical differences among groups with a larger proportion of the high-risk features presenting in the intermediate and high-risk groups when compared with the low-risk group. The actuarial 6-year FFBF was 59% (95% CI, 4573), 38% (95% CI, 24-52), and 16% (95% CI, 6.2-26) Table 3

for the low-, intermediate-, and high-risk groups, respectively. This was similar to the estimated 6-year progression-free probability by the Stephenson nomogram for these risk groups, which were N 57%, 35%57%, and b 35% for the low-, intermediate-, and highrisk groups (supplementary Table 2; available online only at www.practicalradonc.org). Six-year results for FFBF, FFM, and PCSS are demonstrated for the entire cohort and by risk groups (Table 3) while Fig 1 (A-C) shows Kaplan-Meier survival curves by risk group for each endpoint (all, P ≤ .0005).

Independent predictors of outcome Given the close association of variables and to further establish independent predictors of outcome following SRT, multivariate Cox proportional hazards models were completed for all 3 endpoints using the variables incorporated into the Stephenson nomogram and year of

Freedom from clinical progression as a function of Stephenson nomogram and risk group

Outcome at 6-y

Entire cohort

Low risk

Intermediate risk

High risk

P

Freedom from biochemical failure

38% (95% CI, 30-46) 79% (95% CI, 73-85) 96% (95% CI, 92-100)

59% (95% CI,45-73) 95% (95% CI,89-100) 100%

38% (95% CI, 24-52) 81% (95% CI, 70-93) 96% (95% CI, 90-100)

16% (95% CI, 6.2-26) 62% (95% CI, 48-76) 90% (95% CI, 82-98)

b .0001

Freedom from metastasis Prostate cancer–specific survival CI, confidence interval.

b .0001 .0005

Practical Radiation Oncology: Month 2014

Stephenson nomogram for salvage RT

5

FFBF. Positive SM (P = .02; HR, 0.4; 95% CI, 0.2-0.8) and pre-RT PSA ng/mL (P = .0001; HR, 1.6; 95% CI, 1.3-2.0) were prognostic for FFM, while the only prognostic variable for PCSS was pre-RT PSA ng/mL (P = .03; HR, 1.2; 95% CI, 1.0-1.4). When correcting for multiple testing corrections by using stepwise linear regression modeling, positive SM (P b .0001; HR, 0.4; 95% CI, 0.2-0.6), persistently elevated post-RP PSA ng/ mL (P = .01; HR, 2.2; 95% CI, 1.3-4.0) and pre-RT PSA ng/mL (P = .0001; HR, 1.2; 95% CI, 1.1-1.3) were prognostic for FFBF. The only variable that remained in the model which was prognostic for FFM was pre-RT PSA ng/mL (P b .0001; HR, 1.5; 95% CI, 1.3-1.8), while the only 2 variables which were prognostic for PCSS were prostatectomy PSA ng/mL (P = .004; HR, 1.0; 95% CI, 1.0-1.1) and pre-RT PSA ng/mL (P = .0006; HR, 1.2; 95% CI, 1.1-1.3). Additionally, for these endpoints overall prognostic value as evaluated by the c-index indicated adequate prognostic significance for clinical endpoints BF (0.71), DM (0.75), and PCSM (0.75). These values were superior when compared with each of the individual variables in the nomogram (Table 5).

Discussion

Figure 1 Freedom from (A) biochemical failure, (B) metastasis, and prostate cancer–specific survival (C) as a function of the Stephenson risk group.

radiation treatment (Table 4). Gleason score 7 (P = .05; HR, 2.4; 95% CI, 1.0-5.6) and 8-10 (P = .02; HR, 3.5; 95% CI, 1.3-9.5), positive SM (P b .0001; HR, 0.3; 95% CI, 0.2-0.6), persistently elevated post-RP PSA ng/mL (P = .05; HR, 1.8; 95% CI, 1.0-3.4), and pre-RT PSA ng/mL (P = .006; HR, 1.2; 95% CI, 1.0-1.3) were prognostic for

The Stephenson nomogram was developed and validated to predict BF following SRT. However, this is the first analysis of clinical failure following SRT evaluated using this commonly utilized nomogram. In particular, for 179 patients treated with salvage EBRT following BF after radical prostatectomy the 6-year FFBF was 38%. Although 62% of men in this group had BF after SRT only half of these (32% overall) went on to develop metastasis and a smaller number (7% overall) went on to die of prostate cancer. Despite these less frequent events nomogram score was associated with both FFM (P b .0001) and PCSS (P = .0001) with c-indices of 0.75 and 0.75, respectively, which were similar in prognostic significance as for BF in this cohort where the c-index was 0.71. These findings support the use of the Stephenson nomogram as a prognostic tool to help counsel patients, not only about their chances of remaining free from PSA progression following treatment with SRT but their chances of remaining free of metastasis or PCa death; however, these result will need to be externally validated before they can be applied to clinical practice. Following its development the Stephenson nomogram was initially externally validated by Moreira et al. 6 They identified 102 patients from the Shared Equal Access Regional Cancer Hospital (SEARCH) database and demonstrated a 6-year FFBF of 57% (42%-69%). This rate is considerably higher than what was seen in the

6

S. Johnson et al

Practical Radiation Oncology: Month 2014

Table 4 Cox proportional hazards model for biochemical failure, metastasis, prostate cancer–specific mortality, and overall mortality Variables

Prostatectomy PSA Gleason score 2-6 7 8-10 SVI positive ECE positive SM positive LNI positive Persistently elevated post-RP PSA Pre-RT PSA PSA doubling time Neoadjuvant ADT Radiation dose Radiation treatment time, y

FFBF

FFM

PCSS

HR (95% CI)

P

HR (95% CI)

P

HR (95% CI)

P

1.0 (0.9-1.0)

.8

1.0 (1.0-1.0)

.4

1.0 (1.0-1.1)

.10

Reference 2.4 (1.0-5.6) 3.5 (1.3-9.5) 1.5 (0.8-2.9) 1.1 (0.7-1.8) 0.3 (0.2-0.6) 1.1 (0.1-9.7) 1.8 (1.0-3.4) 1.2 (1.0-1.3) 1.0 (0.9-1.0) 0.4 (0.2-0.7) 1.2 (1.0-1.4) 0.8 (0.6-1.1)

— .05

Reference 5.0 (0.7-37.9) 4.7 (0.5-40.5) 1.7 (0.7-4.0) 1.8 (0.9-3.6) 0.4 (0.2-0.8) 0.0 (0.0-42.5E+174) 2.2 (0.9-5.4) 1.6 (1.3-2.0) 1.0 (0.9-1.0) 0.5 (0.2-1.2) 1.2 (0.9-1.4) 0.9 (0.5-1.5)

— .12

Reference 1.9E+5 (0.0-2.1E+201) 8.4E+4 (0.0-937E+198) 2.3 (0.6-9.4) 1.0 (0.3-3.3) 0.5 (0.1-1.9) 0.0 (0.0-10.1E+303) 2.1 (0.6-6.8) 1.2 (1.0-1.4) 0.9 (0.8-1.0) 1.5 (0.4-6.4) 1.0 (0.7-1.5) 0.4 (0.1-1.4)

— .9

.02 .2 .7 b .0001 .9 .05 .006 .3 .004 .01 .1

.16 .2 .18 .02 .9 .08 .0001 .2 .13 .12 .7

.9 .3 .9 .3 .9 .2 .03 .08 .6 .8 .2

CI, confidence interval; ADT, androgen deprivation therapy; ECE, extracapsular extension; FFBF, freedom from biochemical failure; FFM, freedom from metastasis; HR, hazard ratio; LNI, lymph node invasion; PCSS, prostate cancer–specific survival; PSA, prostate-specific antigen; RP, radical prostatectomy; RT, radiation therapy; SM, surgical margins; SVI, seminal vesicle invasion.

original Stephenson nomogram and in the current cohort, which were 32% and 38%, respectively. However, the cohort in Moriera et al included a lower risk population than that which was demonstrated in the current analysis. For example, the median PSADT in the study of Moriera et al was 10.3 months versus 7.6 months in this dataset. Also, 30% of patients had a Gleason score of b 6 compared with 13% herein. These variances in population not only account for differences in FFBF but likely account for the differences seen in concordance index. We demonstrated a slightly higher c-index for FFBF than the original Stephenson nomogram (0.71 vs 0.69); however, this value was more similar than the external validation set of Moriera et al (0.65). Additionally, independent prognostic variables of FFBF demonstrated by Moriera et al 6 were slightly different than those seen in the much larger dataset of Stephenson et al. 4 Of the multiple variables which were demonstrated by Stephenson et al to be independently

prognostic of FFBF following SRT, Moriera et al validated only pre-RT PSA level and surgical margins. However, after completion of the model accounting for other variables they were able to show trends for Gleason score 7 (P = .069), Gleason score 8-10 (P = .065), and persistently detectable PSA following RP (P = .065). A second external validation study by Hugen et al 5 evaluated 87 patients treated with SRT following RP and demonstrated that pre-RT PSA and Gleason score were the only independent prognostic variables associated with FFBF. Similar to Stephenson et al, we demonstrated that prognosticators of FFBF were Gleason score, positive SM, persistently elevated post-RP PSA, pre-RT PSA, and neoadjuvant ADT. Unlike that which was shown by Stephenson et al, 4 we demonstrated that higher doses of radiation were directly associated with higher (not lower) risks of biochemical failure; however, this finding is likely indicative of practice patterns, wherein higher risk individuals received

Practical Radiation Oncology: Month 2014

Stephenson nomogram for salvage RT

Table 5 Concordance indices for Stephenson nomogram and individual variables for biochemical failure (BF), DM, and prostate cancer–specific mortality (PCSM) c-index

Variables

Prostatectomy PSA Gleason risk group SVI ECE SM LNI Persistently elevated post-RP PSA Pre-RT PSA PSA doubling time Neoadjuvant ADT Radiation dose Nomogram

BF

DM

PCSM

0.506 0.577 0.526 0.541 0.596 0.496 0.538 0.654 0.609 0.529 0.588 0.713

0.532 0.624 0.551 0.560 0.531 0.495 0.569 0.654 0.666 0.507 0.621 0.747

0.506 0.611 0.572 0.556 0.547 0.495 0.474 0.714 0.712 0.452 0.585 0.753

ADT, androgen deprivation therapy; DM, distant metastasis; ECE, extracapsular extension; LNI, lymph node invasion; PSA, prostatespecific antigen; RP, radical prostatectomy; RT, radiation therapy; SVI, seminal vesicle invasion.

higher doses of radiation. RT dose was typically prescribed based on a sliding scale between 64.8 and 70.2 Gy as a function of PSA with lower PSA (b 0.4 ng/ mL) given lower doses and higher PSA (N 1.0 ng/mL) given higher doses (see Table 2). As such, in this cohort it is not unexpected that higher dose RT correlated with worse BF. Of note, pre-RT PSA was shown to be an independent predictor of FFBF in the current analysis as well as the studies of Moriera et al, 6 Hugen et al, 5 and Stephenson et al. 4 In fact, pre-RT PSA was also the only variable which showed significant prognostic capability for all 3 clinical endpoints in the current analysis (see Table 4). Moriera et al demonstrated that persistently elevated PSA after surgery may be a good predictor of outcome; however, this variable was not significant in the original Stephenson nomogram 4 but was included based upon a previous study from the same group. 2 We support the findings of Moriera et al and others that have demonstrated that persistently elevated PSA is strongly prognostic of FFBF. 2,10 In regard to selection of treatment, on multivariate analysis the use of neoadjuvant ADT was associated with a 60% improvement in FFBF (HR, 0.4; 95% CI, 0.2-0.8) and appeared to also impact FFM (although not significantly so [P = .13]). Past retrospective studies also identified improved FFBF when evaluating ADT use with SRT following RP. 14-19 In addition, preliminary results from Radiation Therapy Oncology Group (RTOG) 9601 demonstrate a reduction in PSA failure and metastasis following treatment with ADT (2 years) and RT for patients post RP, where BF within 7 years was 60% in those treated with RT alone and 43% in those treated with RT and ADT. 20 Interestingly, the median duration of

7

ADT in the current cohort was 6.0 months, and it is not clear what the optimal duration of ADT use. The current prospective trial being performed in North America (RTOG 0534 (SPPORT [short-term androgen deprivation with pelvic lymph node or prostate bed only radiation therapy]) randomizes patients to either no ADT or 6 months of ADT with and without pelvic lymph node RT while in Great Britain the RADICALS (Radiation therapy and Androgen Deprivation in Combination after Local Surgery) trial randomizes patients to no ADT or ADT of 6 versus 24 months. These prospective trials may provide additional information on both the benefit of ADT use and the appropriate duration of ADT use in post-RP RT patients. Interestingly, although we demonstrated that high Gleason score is independently associated with worse FFBF it was not significantly associated with other clinical endpoints (FFM and PCSS). One issue with the Stephenson nomogram and other risk classification systems (D’Amico, American Joint Committee on Cancer, and National Comprehensive Cancer Network) 21 is that patients with a Gleason score of 8 are assigned the same score or risk grouping as patients with a primary Gleason pattern of 5 (Gleason score 9 or 10), which has been found to be an independent prognostic factor in multiple surgical and radiation series including patients treated with SRT. 22 Breaking Gleason score 8-10 to separately evaluate those with and without Gleason pattern 5 might, therefore, improve prognostic capacity. These findings should be interpreted in light of specific limitations, which include the retrospective study design and the inability to obtain all variables included in the Stephenson nomogram for those who received SRT. However, these same issues were encountered in the first validation set, which resulted in analysis of a smaller subset (56%) than the original cohort. 6 We evaluated differences in our cohort by comparing treatment, pretreatment, and clinical factors between those with and without all variables and demonstrated that there was no difference in 6-year outcomes (all, P N .2) between the 2 groups, suggesting that these findings may be generalizable to our original cohort. In conclusion, these results suggest that the Stephenson nomogram is an appropriate prognostic tool for estimating the response to SRT following RP for relevant clinical endpoints such as metastasis and prostate cancer death. Although the Stephenson nomogram has modest prognostic ability for clinical endpoints in this cohort, it is similar and indeed slightly greater in prognostic significance for these endpoints as for FFBF. Additionally, the multivariate models and c-indices demonstrate that although there are a few variables that play the greatest prognostic value in the Stephenson nomogram (namely positive surgical margins and PSA prior to RT) the overall prognostic value of the nomogram is superior for all clinical endpoints when including all variables. We appreciate that it is possible

8

S. Johnson et al

that some of these variables add very little to the model, and indeed likely some of them may even decrease the prognostic accuracy of the model. Therefore, further validation of this instrument for these endpoints is warranted and there remains a need for improvement in the nomogram to identify those who will most likely achieve a durable response to SRT following biochemical recurrence after RP.

References 1. 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;299:2760-2769. 2. Stephenson AJ, Shariat SF, Zelefsky MJ, et al. Salvage radiotherapy for recurrent prostate cancer after radical prostatectomy. JAMA. 2004;291:1325-1332. 3. Cooperberg MR, Hilton JF, Carroll PR. The CAPRA-S score: A straightforward tool for improved prediction of outcomes after radical prostatectomy. Cancer. 2011;117:5039-5046. 4. 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;25:2035-2041. 5. Hugen CM, Polcari AJ, Quek ML, Garza RP, Fitzgerald MP, Flanigan RC. Long-term outcomes of salvage radiotherapy for PSArecurrent prostate cancer: Validation of the Stephenson nomogram. World J Urol. 2010;28:741-744. 6. Moreira DM, Jayachandran J, Presti Jr JC, et al. Validation of a nomogram to predict disease progression following salvage radiotherapy after radical prostatectomy: Results from the SEARCH database. BJU Int. 2009;104:1452-1456. 7. Boorjian SA, Thompson RH, Tollefson MK, et al. Long-term risk of clinical progression after biochemical recurrence following radical prostatectomy: The impact of time from surgery to recurrence. Eur Urol. 2011;59:893-899. 8. Tosoian JJ, Trock BJ, Landis P, et al. Active surveillance program for prostate cancer: An update of the Johns Hopkins experience. J Clin Oncol. 2011;29:2185-2190. 9. Denham JW, Steigler A, Wilcox C, et al. Why are pretreatment prostate-specific antigen levels and biochemical recurrence poor predictors of prostate cancer survival? Cancer. 2009;115:4477-4487. 10. Song DY, Thompson TL, Ramakrishnan V, et al. Salvage radiotherapy for rising or persistent PSA after radical prostatectomy. 2002;60:281-287.

Practical Radiation Oncology: Month 2014 11. Freedland SJ, Humphreys EB, Mangold LA, et al. Risk of prostate cancer-specific mortality following biochemical recurrence after radical prostatectomy. JAMA. 2005;294:433-439. 12. Pound CR, Partin AW, Eisenberger MA, Chan DW, Pearson JD, Walsh PC. Natural history of progression after PSA elevation following radical prostatectomy. JAMA. 1999;281:1591-1597. 13. Stephenson AJ, Kattan MW, Eastham JA, et al. Defining biochemical recurrence of prostate cancer after radical prostatectomy: A proposal for a standardized definition. J Clin Oncol. 2006;24: 3973-3978. 14. Cheung R, Kamat AM, de Crevoisier R, et al. Outcome of salvage radiotherapy for biochemical failure after radical prostatectomy with or without hormonal therapy. Int J Radiat Oncol Biol Phys. 2005;63: 134-140. 15. Eulau SM, Tate DJ, Stamey TA, Bagshaw MA, Hancock SL. Effect of combined transient androgen deprivation and irradiation following radical prostatectomy for prostatic cancer. Int J Radiat Oncol Biol Phys. 1998;41:735-740. 16. Jang JW, Hwang WT, Guzzo TJ, et al. Upfront androgen deprivation therapy with salvage radiation may improve biochemical outcomes in prostate cancer patients with post-prostatectomy rising PSA. Int J Radiat Oncol Biol Phys. 2012;83:1493-1499. 17. Katz MS, Zelefsky MJ, Venkatraman ES, et al. Predictors of biochemical outcome with salvage conformal radiotherapy after radical prostatectomy for prostate cancer. J Clin Oncol. 2003;21: 483-489. 18. King CR, Presti Jr JC, Gill H, et al. Radiotherapy after radical prostatectomy: Does transient androgen suppression improve outcomes? Int J Radiat Oncol Biol Phys. 2004;59:341-347. 19. Soto DE, Passarelli MN, Daignault S, et al. Concurrent androgen deprivation therapy during salvage prostate radiotherapy improves treatment outcomes in high-risk patients. Int J Radiat Oncol Biol Phys. 2012;82:1227-1232. 20. Shipley WU, Hunt D, Lukka HR, Major P, Heney NM, Grignon D, et al. Initial report of RTOG 9601, a phase III trial in prostate cancer: Effect of anti-androgen therapy (AAT) with bicalutamide during and after radiation therapy (RT) on freedom from progression and incidence of metastatic disease in patients following radical prostatectomy (RP) with pT2-3,N0 disease and elevated PSA levels. Orlando, Florida: American Society of Clinical Oncology Genitourinary Cancers Symposium. 2011. 21. D'Amico AV, Whittington R, Malkowicz SB, et al. Biochemical outcome after radical prostatectomy, external beam radiation therapy, or interstitial radiation therapy for clinically localized prostate cancer. JAMA. 1998;280:969-974. 22. Jackson W, Hamstra DA, Johnson S, et al. Gleason pattern 5 is the strongest pathologic predictor of recurrence, metastasis, and prostate cancer-specific death in patients receiving salvage radiation therapy following radical prostatectomy. Cancer. 2013;119:3287-3294.