Success rates with salvage radiotherapy (SRT) in men who have a postprostatectomy biochemical relapse are suboptimal. One treatment-intensification strategy includes elective irradiation of the pelvic lymph nodes with whole pelvis radiotherapy (WPRT).
An inter-institutional retrospective cohort study compared outcomes for patients who received SRT at 2 separate academic institutions with disparate treatment paradigms: almost exclusively favoring WPRT (n = 112) versus limiting treatment to the prostate bed (PBRT) (n = 135). Patients were excluded if they had lymph node involvement or if they received androgen-deprivation therapy. The Cox proportional hazards model was used to adjust for potential confounders.
In total, 247 patients were analyzed with a median follow-up of 4 years. The pre-SRT prostate-specific antigen (PSA) level (adjusted hazard ratio [HR], 1.58; P < .0001) and a Gleason score of 8 to 10 (adjusted HR, 3.21; P < .0001) were identified as independent predictors of increased risk of biochemical PSA progression after SRT. However, WPRT was not independently associated with biochemical progression-free survival in the multivariate model (adjusted HR, 0.79; P = .20). Neither low-risk patients nor high-risk patients (defined a priori by a preoperative PSA level ≥20 ng/mL, a pathologic Gleason score between 8 and 10, or pathologic T3 tumor classification) benefited from WPRT. Overall survival was similar between treatment groups. When restricting the analysis to patients with pre-SRT PSA levels ≥0.4 ng/mL (n = 139), WPRT was independently associated with a 53% reduction in the risk of biochemical progression (adjusted HR, 0.47; P = .031).
In the United States, approximately 20,000 patients are diagnosed each year with biochemical relapse after prostatectomy, and an estimated 41% to 99% of these patients may die from their disease without further treatment.1 Because the predominant risk of failure is local,2 salvage radiotherapy (SRT) is a rational treatment strategy that has been associated with improved overall survival (OS) rates compared with observation or salvage hormone therapies.3, 4 Although local salvage therapy may extend survival in some patients, outcomes remain suboptimal, and up to 14% of patients still die from their disease at 10 years despite SRT.3
SRT traditionally has been limited to targeting the prostate bed (prostate bed radiotherapy [PBRT]). However, up to 75% of radiation oncologists may consider intensifying treatment by electively irradiating the pelvic lymph nodes with whole pelvis radiotherapy (WPRT).5 The premise for this may be based on pathologic lymph node sampling series that demonstrate high rates of occult pelvic lymph node involvement at the time of prostatectomy or on more recent studies with lymphotropic nanoparticle-enhanced magnetic resonance imaging that demonstrates a >20% risk of occult lymph node involvement at the time of SRT in patients who have detectable levels of prostate-specific antigen (PSA).6, 7 Yet, retrospective data supporting a clinical benefit for irradiating the pelvic lymph nodes are scarce,8, 9 and the strength of evidence in currently available guidelines to address this issue is limited to expert opinion.10 While we await completion of the Radiation Therapy Oncology Group (RTOG) 0534 phase 3 trial, which is prospectively evaluating the role of WPRT during SRT, we report a retrospective comparison of outcomes from 2 institutions that favor different treatment approaches, offering either PBRT or WPRT.
METHODS AND MATERIALS
Patients undergoing SRT in the setting of a detectable PSA level after prostatectomy have long been treated at Virginia Commonwealth University (VCU) and Duke University according to institutional preferences for either WPRT or PBRT in 97% and 98% of all cases, respectively. Patients treated at the Hunter Holmes McGuire Veterans Affairs Medical Center (VAMC) in Richmond, Virginia are managed by VCU radiation oncologists who share a similar preference for WPRT and are thus included in the VCU patient data set. Using a retrospective cohort study design, 2 institutional review board-approved databases from VCU (1998-2008), Hunter Holmes McGuire VAMC (1998-2008), and Duke University (1988-2005) were combined to evaluate potential differences in outcomes with WPRT versus PBRT.
There were no defined policies at either institution regarding postprostatectomy PSA levels that would trigger a referral for SRT. Before SRT, all patients routinely underwent a history and physical examination, and patients who had Gleason scores ≥8, pathologic T3 (pT3) tumor classification, or positive surgical margins generally underwent a staging bone scan and/or pelvic computed tomography scan, at the discretion of the urologist and/or radiation oncologist. A restaging magnetic resonance imaging study with endorectal coil, magnetic resonance spectroscopy, or 111In-capromab pendetide scans (Prostascint; Cytogen Corporation, Princeton, NJ) rarely was performed at either institution during the study period. For purposes of this study, patients were excluded if they had either clinical or pathologic lymph node involvement, if they received androgen-deprivation therapy (ADT) at any time before post-SRT biochemical progression or if they received either WPRT (2%) or PBRT (3%) discordant with the institution's preference.
The routine delivery of WPRT during the study period for patients who received SRT traditionally consisted of a 4-field technique extending to the upper half of the sacroiliac joint for inclusion of lymph nodes to the bifurcation of the common iliac vessels (Fig. 1). The standard prescription delivered 50.4 Gray (Gy) in 1.8-Gy daily fractions to the pelvis before using a shrinking field technique to boost the prostate bed and seminal vesicle remnants to a final dose between 61.2 Gy and 72 Gy. For patients who received PBRT, results from that institution have been previously reported, and the treatment technique traditionally consisted of a 4-field box technique limited to the prostate bed and periprostatic tissue with field reductions after 46 Gy, if deemed necessary, for a final prostate bed boost dose ranging between 59.4 Gy and 74 Gy.11-15 At both institutions, 3-dimesional conformal planning was used for the final boot dose, and no patients in this study received intensity-modulated radiation therapy (IMRT).
Definitions of Biochemical and Overall Survival
Biochemical progression was scored at the first documented serum PSA rise of ≥0.2 ng/mL above the postradiotherapy nadir and was confirmed by a subsequent rise. Patients without biochemical stabilization or declines in PSA levels after SRT were coded for biochemical progression on the first day of SRT. For all patients, the Social Security Death Index was accessed in December 2009 to assess the date of death. All patients who were alive at that time were assumed to have been alive as of December 31, 2008, allowing a 1-year lag for entry into the national database.
Baseline characteristics, including age, level of PSA, radiation dose, and time intervals, were evaluated as continuous variables; whereas all the other characteristics were categorized. For comparison between the 2 radiotherapy treatment groups, the 2-sample t test and the Fisher exact test were used for continuous and categorical variables, respectively. The assumption of a normal distribution was checked for each continuous variable, and data transformations were performed for symmetry and normality when highly skewed, including both preoperative and pre-SRT PSA levels, which were log transformed. In the absence of adequate data transformation, the Wilcoxon rank-sum test was used.
Dates of biochemical progression and death were used to calculate biochemical progression-free survival (bPFS) and OS intervals, respectively, starting from the beginning of SRT. The Kaplan-Meier method was used to generate survival curves, and log-rank testing was used to compare the 2 radiotherapy groups. The Cox proportional hazards model was used to adjust for potential confounding covariates, and the multivariate Cox model included both unbalanced characteristics and prognostic covariates that had a significance level of P < 0.1 in a univariate model. A “final” multivariate Cox model excluded covariates that were insignificant in the multivariate model (P ≥ .1).
To account for significant imbalances that were identified in patient characteristics between the 2 groups, patients were matched 1-to-1 using a propensity score nearest-neighbor matching method.16 Propensity scores were calculated with a logistic regression model using well accepted prognostic variables of extracapsular extension, seminal vesicle invasion, and margin status, as well as all covariates that were unbalanced between the 2 SRT groups, including a persistently positive postoperative PSA level, natural logarithm of pre-SRT PSA(ln), Gleason score (8-10 vs ≤7), radiation dose, and time interval from surgery to SRT.
There were 2 subgroup analyses. The first subgroup analysis was planned before study initiation and stratified patients a priori by risk groups reported in a similar study,8 defining “high risk” according to the presence of a preoperative PSA level ≥20 ng/mL, a Gleason score between 8 and 10, or pT3 disease, which has been reported to confer a >20% risk of pelvic lymph node involvement.6 The second subgroup was explored after the initial analysis was completed to exclude patients who may have never progressed and, thus, included only patients who had pre-SRT PSA levels ≥0.4 ng/mL.17 All computations were performed using SAS 9.2 software (SAS Institute, Inc., Cary, NC). The lead author was solely responsible for the integrity of the combined database.
There were 247 patients available for comparison—112 who received WPRT, and 135 who received PBRT—with a median follow-up of 50.4 months and 46.5 months, respectively (P = .13). There were significant imbalances in patient characteristics between patients who received treatment at the separate institutions, as indicated in Table 1. Whereas patients who were referred for SRT at the institution offering WPRT were more likely to have a Gleason score between 8 and 10, patients who were referred at the institution offering PBRT were more likely to have had pre-prostatectomy PSA levels ≥20 ng/mL, persistently positive post-prostatectomy PSA levels, positive surgical margins, a shorter interval between prostatectomy and SRT, higher pre-prostatectomy and pre-SRT PSA levels, and received lower cumulative doses of radiotherapy.
Outcomes with Whole Pelvis Versus Prostate Bed Radiotherapy
Nontreatment clinicopathologic factors that were associated with a higher risk of biochemical progression after SRT on univariate analysis included a higher pre-SRT PSA level, a Gleason score between 8 and 10, and the presence of extracapsular extension (see Table 2). After multivariate testing, only the pre-SRT PSA level (adjusted hazard ratio [HR], 1.58; 95% confidence interval [CI], 1.30-1.92; P < .0001) and a Gleason score between 8 and 10 (adjusted HR, 3.21; 95% CI, 1.94-5.31; P < .0001) retained a significant association with an increased risk of biochemical PSA progression.
Table 2. Biochemical Progression-Free Survival for All 247 Patients
In the evaluation of treatment techniques, both the log-rank test and the univariate Cox proportional hazards model identified a crude association between WPRT and improved bPFS (P = .04 with both tests), with an estimated 19% relative improvement in bPFS at 4 years (82% vs 69%, respectively) (see Fig. 2). However, this benefit was not statistically significant after controlling for potentially confounding factors in the multivariate model (adjusted HR, 0.79; 95% CI, 0.431-1.198; P = .20) (see Table 2). The receipt of WPRT was also not associated with improved OS in either univariate or multivariate analyses.
Subgroup Analysis of Patients Stratified by A Priori Risk Groups
When patients were stratified by risk groups (low risk, n = 99; high risk, n = 148), there appeared to be an association between WPRT and improved bPFS among high-risk patients (univariate Cox proportional hazards model: HR, 0.59; P = .058), as demonstrated in Figure 3. However, the association of WPRT and bPFS in high-risk patients remained insignificant after multivariate testing (data not shown).
Subgroup Analysis of Patients With Presalvage Radiotherapy Prostate-Specific Antigen Levels ≥0.4 ng/mL
When the analysis was restricted only to those who had pre-SRT PSA levels ≥0.4 ng/mL (n = 139), an independent association was observed between WPRT and improved bPFS (see Fig. 4). The benefit of WPRT in this subgroup remained significant in the final multivariate model after excluding variables that were insignificant in univariate testing (adjusted HR, 0.47; 95% CI, 0.24-0.94; P = .0313) as demonstrated in Table 3. However, an OS benefit still was not observed with WPRT in this subgroup either (adjusted HR, 1.46; 95% CI, 0.50-4.24; P = .48).
Table 3. Biochemical Progression-Free Survival for 139 Patients With Presalvage Radiotherapy Prostate-Specific Antigen Levels ≥0.4 ng/mL
Abbreviations: CI, confidence interval; ECE, extracapsular extension; Gy, grays; HR, hazard ratio; PBRT, prostate bed radiotherapy; PSA, prostate-specific antigen; PSA(ln), natural logarithm prostate specific antigen; SRT, salvage radiotherapy; SVI, seminal vesicle invasion; WBRT, whole pelvis radiotherapy. Although only variables with a p < .1 are included in the multivariate model, preoperative PSA(ln) and Persistently elevated PSA are included here as they are significantly unbalanced (see Table 1).
To our knowledge, this analysis represents the largest series to date evaluating the role of elective pelvic lymph node irradiation during postprostatectomy SRT. This report updates a previous presentation of these data18 with notable differences in methodology, including longer PSA follow-up, exclusion of patients who received ADT, and selection of a biochemical endpoint focusing on PSA progression; whereas, previously, we used an absolute post-SRT PSA threshold of 0.1 ng/mL to define a biochemical failure event.
The study design sought a balanced group of patients who received treatment according to institutional preferences, as opposed to physician preference, to avoid selection bias. We excluded patients with clinical or pathologic lymph node involvement to assess the role of “elective” lymph node irradiation. In addition, to minimize the challenges of scoring the time to biochemical progression after SRT, often confounded by variable testosterone-recovery rates, we excluded all patients who received ADT.
Although crude analysis suggested improved bPFS with WPRT, this finding was not statistically significant after controlling for potential confounding variables in multivariate testing. In a planned subset analysis, we identified a trend toward improved outcomes with WPRT in high-risk patients, as reported in a similar study; however, this trend was not statistically significant in either the “low-risk” group or the “high-risk” group, as defined a priori in accordance with that earlier publication.8
Once it was determined that WPRT was not independently associated with improved bPFS, a second, but unplanned, subset analysis was explored. We hypothesized that analyses of the entire group, if diluted with patients who may have never developed biochemical progression, may obfuscate a potential therapeutic benefit from WPRT in others. On the basis of previous reports demonstrating that up to 31% of patients with a single pre-SRT PSA value <0.4 ng/mL may not develop progression without SRT even 10 years later,17 our analysis was restricted to patients who had pre-SRT PSA levels ≥0.4 ng/mL. This subset (n = 139) excluded 60% of patients who received WPRT and 30% of patients who received PBRT patients from the initial data set. The multivariate model included statistical strategies of both including and excluding variables that are insignificant as univariate factors (both findings are reported in Table 3). Because factors other than pre-SRT PSA level and/or Gleason score have rarely been associated significantly with bPFS and, furthermore, were insignificant in the current analysis, we highlight findings from the final model in Table 3, which demonstrates that patients who received WPRT had a 53% reduction in their risk of bPSA progression (adjusted HR, 0.47; P = 0.031). When we explored a potentially lower pre-SRT threshold, including all 195 patients who had pre-SRT PSA levels ≥0.2 ng/mL, the benefit of WPRT was no longer significant (adjusted HR, 0.65; 95% CI, 0.33-1.30; P = .22).
We acknowledge the well known limitations associated with any retrospective study, specifically because imbalances in treatment groups may reflect disparate referral patterns, which may introduce unintended selection bias. Thus, a matched-pair analysis was sought to better control for these imbalances16; however, only 42 patients could be matched, precluding a more meaningful analysis. In addition, the potential confounding effect of pre-SRT PSA velocity was not included in the multivariate analyses, because PSA data points between surgery and SRT were limited for this analysis.
We are aware of only 2 other publications that have explored the potential benefit of WPRT during SRT. A smaller series from the University of California, Los Angeles (n = 46), which also did not include any patients who received ADT, also failed to demonstrate any differences in outcomes with WPRT versus PBRT.9 Meanwhile, a larger study from Stanford University (n = 160) reported an independent association between WPRT and improved bPFS.8 It is important to note that patients in the latter study who were treated with either WPRT or PBRT also received ADT (68% and 33%, respectively) and had known lymph node involvement (17% and 5%, respectively). Collectively, these findings suggest a potential interaction between WPRT and ADT, which is also being explored in the ongoing prospective phase 3 RTOG 0534 trial.
The authors urge caution in uniformly applying the finding of improved biochemical control associated with WPRT in patients with pre-SRT PSA levels ≥0.4 ng/mL, because an increased risk of bladder and bowel toxicity is well known whenever larger pelvic fields are used. Unfortunately, the disparate methodologies used for data collection introduced significant observation bias in our collection of toxicity data, and thus preclude a meaningful analysis. More specifically, treatment-related side effects were coded prospectively at the institution offering PBRT by a single investigator (M.S.A.) but were coded retrospectively by multiple investigators where WPRT was offered.
If considering elective lymph node irradiation, 1 strategy to potentially reduce the increased toxicities associated with WPRT is consideration of IMRT. In a recent international survey of nearly 1000 radiation oncologists, over 78% reported the use of IMRT during postprostatectomy SRT.5 Interested readers are encouraged to reference a recent report from the University of Pennsylvania by Deville et al, who compared WPRT with PBRT using IMRT during postprostatectomy SRT.19 In that study, the use of larger fields with IMRT was not associated with any significant increase in grade 3 gastrointestinal or genitourinary toxicity, although grade 2 acute gastrointestinal toxicities were more frequent.
In conclusion, this report fails to demonstrate a clear benefit from WPRT in all men who received postprostatectomy SRT for biochemical failure. However, a benefit for WPRT may exist for patients who have pre-SRT PSA levels ≥0.4 ng/mL. The independent association of improved biochemical control with WPRT in patients with pre-SRT PSA levels ≥0.4 ng/mL suggests the biologic plausibility that certain patients with postprostatectomy biochemical-only relapse may harbor occult disease limited to the pelvic lymph nodes and may benefit from WPRT by delaying the time to biochemical PSA progression and initiation of ADT. However, we did not demonstrate an OS benefit, and a risk of increased toxicity with larger treatment fields remains a concern. The results of the ongoing phase 3 RTOG 0534 trial are awaited to provide more meaningful information for clinical decision making.