Positive Surgical Margins in Radical Prostatectomy Patients Do Not Predict Long-term Oncological Outcomes: Results from SEARCH

Abstract

Purpose

To assess the impact of positive surgical margins (PSMs) on long-term outcomes after radical prostatectomy (RP), including metastasis, castrate-resistant prostate cancer (CRPC), and prostate cancer-specific mortality (PCSM).

Materials and Methods

Retrospective study of 4,051 men in SEARCH treated by RP from 1988–2013. Proportional hazard models were used to estimate hazard ratios of PSMs in predicting BCR, CRPC, metastases, and PCSM. To determine if PSMs were more predictive in certain patients, analyses were stratified by pathological Gleason score, stage, and pre-operative PSA.

Results

Median follow-up was 6.6 years (IQR 3.2–10.6) and 1,127 patients had over 10 years of follow-up. During this time, 302 (32%) men experienced BCR, 112 (3%) developed CRPC, 144 (4%) developed metastases, and 83 (2%) died of PC. There were 1600 (40%) men with PSMs. In unadjusted models, PSMs were significantly associated with all adverse outcomes: BCR, CRPC, metastases and PCSM (all p ≤ 0.001). After adjusting for demographic and pathological characteristics, margins were associated with increased risk of only BCR (HR 1.98, p<0.001), and not CRPC, metastases, or PCSM (HR ≤ 1.20, p>0.18). Similar results were seen when stratified by pathological Gleason score, stage, or PSA, and when patients who underwent adjuvant therapy were excluded.

Conclusions

PSMs after RP are not an independent risk factor for CRPC, metastasis, or PCSM overall or within any subset. In the absence of other high-risk features, PSMs alone may not be an indication for adjuvant treatment.

Introduction

While positive surgical margins after radical prostatectomy (RP) are consistently and independently associated with higher risk of biochemical recurrence (BCR), [1–3] the impact of positive surgical margins on long-term outcomes including metastasis, castrate-resistant prostate cancer (CRPC), and prostate cancer-specific mortality (PCSM) remains less clear with mixed results in prior studies. [4–7] Despite this, a positive surgical margin is often used as an indication for adjuvant radiotherapy (ART).[8–10] Furthermore, studies examining the impact of ART on mortality also have shown conflicting results. [11, 12]

In order to help clarify the prognostic value of positive surgical margins with regard to harder clinical endpoints and thereby shed light on their use as an indication for ART, we examined the long-term outcomes of men treated with RP within the Shared Equal Access Regional Cancer Hospital (SEARCH) cohort. Furthermore, we stratified by pathological grade, pathological stage, and pre-operative PSA to identify if there were subsets of patients for whom positive surgical margins were particularly strong predictors of systemic progression and PCSM.

Methods and Materials

After obtaining Institutional Review Board approval at each institution, data from patients who underwent RP between 1988 and 2013 at the Veterans Affairs Medical Centers in West Los Angeles, Palo Alto, and San Diego, California; Augusta, Georgia; and Durham and Asheville, North Carolina were collected and combined into the SEARCH database. [13] Patients treated with preoperative androgen deprivation or radiation therapy were excluded. Of the 5,073 patients in the SEARCH database, 4,937 (97%) patients had information available on surgical margins from the analysis of the RP specimen. The 122 patients who had positive lymph nodes were excluded from analysis because these patients are at high risk for poor disease outcomes, regardless of surgical margin status. Finally, we excluded patients with missing data on extracapsular extension or seminal vesicle invasion (n=169 patients), pathological Gleason score (n=251), pathological stage (n=149), PSA (n=107), time from surgery to BCR (n=79), race (n=4), and CRPC or recurrence status (n=5), resulting in a study population of 4,051 patients.

Patients were followed to determine clinical endpoints after surgery. BCR after RP was defined as a single PSA greater than 0.2 ng/ml, 2 values of 0.2 ng/ml, or secondary treatment for an elevated postoperative PSA. CRPC was defined as having a PSA increase of 2 ng/mL and 25% greater than the nadir after hormone treatment despite continuous therapy with an LHRH agonist, antagonist, or after orchiectomy. [14] Development of metastases was determined by bone scans or other imaging. PCSM was defined as having metastatic progressive CRPC at time of death with no obvious indication of another cause of death.

Demographic and pathological information between patients with and without positive surgical margin were compared using t-tests and Wilcoxon rank sum tests for normally and non-normally distributed continuous variables, respectively, and chi-square tests for categorical variables. Cox proportional hazards models were used to estimate hazard ratios and 95% confidence intervals of positive surgical margins in predicting BCR, CRPC, metastases, and PCSM. All models were adjusted for age (continuous), race (white vs. black vs. other race), pre-operative PSA (continuous, log-transformed), pathological Gleason score (2–6 vs. 3+4 vs. 4+3 vs. 8–10), seminal vesicle invasion (yes vs. no), extracapsular extension (yes vs. no), year of surgery (continuous), surgical center, and receipt of adjuvant therapy (yes vs. no). We repeated the analysis of the adjusted models, stratifying by pathological Gleason score (2–6, 3+4, 4+3, 8–10), pathological stage (T2, T3/T4), and pre-operative PSA (<4, 4–9.9, 10–19.9, ≥20) to determine if positive surgical margins were more strongly tied to long-term outcomes in certain subsets of patients. Analyses were also repeated using competing risks regression, treating non-prostate cancer death as a competing risk for each of the endpoints (except overall survival). All analyses were performed using Stata 13.0 (Stata, Corp., College Station, TX). Statistical significance was 2-sided with P<0.05.

Results

The clinical and pathological characteristics of patients separated by surgical margin status are shown in Table 1. There were 1,600 (40%) patients with positive margins. Positive surgical margins were significantly associated with younger age, race, less recent year of surgery, higher pre-operative PSA, worse pathological stage, higher Gleason score, extracapsular extension, seminal vesicle invasion, higher likelihood of receiving adjuvant or salvage therapy, and shorter time until recurrence (all p<0.01).

Table 1.

Demographic, clinical, and pathological characteristics of patients by surgical margin status

	Positive Margins (N=1600; 40%)	Negative Margins (N=2451; 60%)	P-value
Age, mean ± SD	61.8 ± 6.4	62.3 ± 6.1	0.010
Race, n (%)			<0.001
White	862 (54)	1510 (62)
Black	645 (40)	769 (31)
Other	93 (6)	172 (7)
Year of Surgery, median (Q1–Q3)	2004 (1999, 2009)	2005 (2000, 2010)	0.002
PSA, median (Q1–Q3)	7.4 (5.1, 11.4)	6.0 (4.5, 8.7)	<0.001
Pathological Stage, n (%)			<0.001
T2	960 (40)	2109 (86)
T3/T4	640 (40)	341 (14)
Pathological Gleason score, n (%)			<0.001
2–6	381 (24)	1026 (42)
7 (3+4)	715 (45)	818 (33)
7 (4+3)	251 (15)	370 (15)
8–10	253 (16)	237 (10)
Extracapsular Extension, n (%)	516 (32)	263 (11)	<0.001
Seminal Vesicle Invasion, n (%)	246 (15)	135 (6)	<0.001
Adjuvant Therapy, n (%)	92 (6)	19 (1)	<0.001
Salvage Radiotherapy, n (%)	235 (15)	81 (3)	<0.001
Months from RP to BCR, median (Q1–Q3)*	11.4 (3.6, 34.3)	21.8 (6.6, 49.8)	<0.001

SD=standard deviation; Q1=25^th percentile; Q3=75^th percentile; RP=radical prostatectomy; BCR= biochemical recurrence

^*Among patients who experienced biochemical recurrence

The median follow-up time among all patients who did not develop a BCR was 6.6 years (IQR 3.2–10.6). There were 1,127 (28%) patients who had greater than 10 years of follow-up. During the follow-up period, 1,302 (32%) patients experienced a BCR, 112 (3%) developed CRPC, 144 (4%) developed metastases, and 83 (2%) died of PC. Altogether, 942 (23%) died during follow-up from all causes. In the unadjusted models, positive surgical margins were significantly associated with increased risk of all PC outcomes – BCR, CRPC, metastases and PCSM (all HR ≥ 2.0, p ≤ 0.001; Table 2), but not overall survival (p=0.37). After adjusting for demographic and pathological characteristics, margins remained significantly associated with increased risk of BCR (HR 1.98, 95% CI 1.75–2.23, p<0.001). However, they were not independently associated with CRPC (HR 1.20, 95% CI 0.96–1.83, p=0.41), metastases (HR 1.29, 95% CI 0.88–1.88, p=0.19), PCSM (HR 1.28, 95% CI 0.78–2.11, p=0.33), or overall survival (HR 0.96, 95% CI 0.83–1.11, p=0.588). These results were consistent even after adjusting for receipt of salvage radiotherapy as a time-dependent covariate (Table 2). After stratifying by pathological Gleason score, pathological stage, or PSA group, positive surgical margins remained associated with increased risk of BCR on multivariable analysis (Table 3). The lack of association between positive surgical margins and long-term outcomes held after stratification by pathological Gleason score, pathological stage, or PSA group (Table 3), All results held when patients receiving adjuvant therapy were excluded (Supplementary Tables 1 & 2). Results were also largely unchanged when the analysis was repeated with competing risks regression (Supplementary Table 3).

Table 2.

Hazard ratios for positive surgical margin outcomes following radical prostatectomy

	BCR			CRPC			Metastases			PC-specific mortality			Overall Survival
	HR	95% CI	P	HR	95% CI	P	HR	95% CI	P	HR	95% CI	P	HR	95% CI	P
	Margins
Crude	2.58	2.31–2.88	<0.001	2.01	1.38–2.93	<0.001	2.00	1.43–2.78	<0.001	2.21	1.43–3.44	<0.001	1.06	0.93–1.21	0.370
Adjusted*	1.98	1.75–2.23	<0.001	1.20	0.96–1.83	0.408	1.29	0.88–1.88	0.186	1.28	0.78–2.11	0.327	0.96	0.83–1.11	0.588
Adjusted**	-	-	-	1.19	0.78–1.82	0.425	1.29	0.89–1.89	0.184	1.28	0.78–2.11	0.328	0.97	0.84–1.13	0.700

HR are vs. negative surgical margins

BCR=biochemical recurrence; CRPC=castrate-resistant prostate cancer; CI=confidence interval

^*Adjusted for age, race, preoperative PSA, pathological Gleason score, seminal vesicle invasion, extracapsular extension, year of surgery, surgical center, and receipt of adjuvant therapy

^**Adjusted for age, race, preoperative PSA, pathological Gleason score, seminal vesicle invasion, extracapsular extension, year of surgery, surgical center, receipt of adjuvant therapy, and receipt of salvage radiation therapy (time-dependent)

Table 3.

Hazard ratios for positive surgical margins predicting prostate cancer outcomes, stratified by pathological Gleason score, pathological stage, and PSA groups

	BCR			CRPC			Metastases			PC-specific mortality
	HR	95% CI	P-value	HR	95% CI	P-value	HR	95% CI	P-value	HR	95% CI	P-value
	Pathological Gleason score*
2–6	2.14	1.69–2.71	<0.001	1.42	0.43–4.75	0.568	1.18	0.56–2.49	0.669	1.81	0.45–7.25	0.401
7 (3+4)	1.92	1.59–2.34	<0.001	0.82	0.36–1.89	0.647	1.12	0.50–2.54	0.780	1.58	0.51–4.84	0.427
7 (4+3)	1.92	1.42–2.58	<0.001	1.26	0.40–3.97	0.694	1.13	0.40–3.19	0.815	0.70	0.20–2.50	0.582
8–10	1.99	1.47–2.69	<0.001	1.22	0.67–2.85	0.579	1.30	0.67–2.51	0.431	1.24	0.57–2.69	0.588
Pathological Stage**
T2	2.43	2.08–2.84	<0.001	1.65	0.83–3.28	0.151	1.69	0.95–3.01	0.074	2.05	0.83–5.06	0.121
T3/T4	1.28	1.06–1.55	0.010	0.90	0.52–1.54	0.693	0.99	0.61–1.63	0.981	1.08	0.59–1.97	0.803
PSA***
<4	2.39	1.40–4.10	0.001	0.32	0.02–4.83	0.408	1.26	0.30–5.33	0.752	N/A	-	-
4–9.9	1.98	1.69–2.33	<0.001	1.28	0.67–2.46	0.454	1.37	0.80–2.34	0.254	1.31	0.60–2.89	0.496
10–19.9	2.04	1.62–2.58	<0.001	1.01	0.47–2.15	0.985	1.01	0.50–2.03	0.978	1.04	0.96–1.10	0.931
!	2.22	1.43–3.44	<0.001	1.46	0.46–4.61	0.521	1.55	0.44–5.47	0.496	1.05	0.23–4.84	0.950

HR are vs. negative surgical margins

^*Adjusted for age, race, preoperative PSA, seminal vesicle invasion, extracapsular extension, year of surgery, surgical center, and receipt adjuvant therapy

^**Adjusted for age, race, preoperative PSA, pathological Gleason score, year of surgery, surgical center, and receipt adjuvant therapy

^***Adjusted for age, race, pathological Gleason score, seminal vesicle invasion, extracapsular extension, year of surgery, surgical center, and receipt adjuvant therapy

N/A indicates that model did not converge because of low event counts

Discussion

We found that among over 4,000 men undergoing RP with a median follow-up of 6.6 years and over 1,100 having greater than 10 years of follow-up, positive surgical margins were predictive of BCR, but not the harder clinical end points of CRPC, metastasis, or PCSM. All patients with positive lymph node status were excluded from our analysis as these patients are already at very high risk of recurrence. These results suggest that while positive surgical margins predict BCR they are not independent predictors of long-term outcomes. If confirmed, these results question the value of using positive surgical margins alone to decide on the need for ART if the goal of ART is to reduce PCSM.

Our results are consistent with prior studies showing that positive surgical margins are predictive of BCR. [1, 2, 15] However, less than half of patients experiencing a BCR are likely to progress to systemic disease and rates of 10-year PCSM are consistently reported below 5% [16] calling into question the use of higher BCR risk alone as an indication for immediate treatment. Indeed, after a median follow-up of 6.6 years in the current study, the risk of PC death was only 2%. As such, it is noteworthy that few studies have examined the impact of positive surgical margins on the harder clinical endpoints of CRPC, metastasis and PCSM. Of the larger studies, Wright et al used the Surveillance Epidemiology and End Results (SEER) database to conduct a population based study of over 65,000 men undergoing RP. Positive surgical margins significantly predicted PCSM only in patients in the high grade and stage groups.[4] Limitations of this study were the short median follow-up time of 50 months as well as a lack of data on PSA. A recent single-institutional study of 1,712 patients by Mauermann et al from Laval University in Quebec, had a slightly longer median follow up time of 75 months and compared single and multiple positive surgical margins with negative surgical margins. [7] Though our study confirmed their finding that positive surgical margins predicted only BCR and no later endpoints, they were unable to look at subgroups due to low numbers of metastasis and PCSM. A study of comparable follow-up time to our own was done in 2010 by Boorjian et al, while Eggener et al conducted the largest study to date (>11, 500 patients) using longer follow up times. [5, 6] In both cases, results were similar to our own in that positive surgical margins predicted BCR but not PCSM. However, no stratification was done to assess for whether certain subgroups were at increased risk.

The importance of stratification becomes apparent when looking at the outcomes of high versus lower-risk groups. Both intermediate and lower risk groups have excellent survival profiles, regardless of treatment. [16] As Spahn et al speculated, one could argue that the lack of significance of positive surgical margins in the prior studies mentioned may in part be due to large proportions of low and intermediate risk patients. [17] Ours is the first sizable study with at least intermediate follow-up times, which stratifies across multiple risk groups. As such, it is an important observation that we did not find any association between positive surgical margins and CRPC, metastasis or PCSM in even the highest risk patients. This is unlikely to have been limited by the study size as we had over 100 patients undergoing systemic progression and 80 patients dying from PC.

Importantly, all our results held true even when all patients receiving ART were excluded, reinforcing the notion that a positive surgical margin alone is not an adequate reason to initiate ART as these men are not at increased risk of long-term poor outcomes. In the three major trials examining the impact of ART versus observation for men at high-risk of progression after RT, whereas a major reduction in BCR risk has consistently been shown, a reduction in PCSM risk has generally not been seen with the exception of one trial. [11, 12, 18–20] Thus, while achieving negative surgical margins remains a critical goal for the urologic surgeon in order to avoid the stress and morbidities of BCR and the further treatments that come along with it, the strong likelihood of overtreatment resulting from positive surgical margins in the absence of other adverse features predictive of PCSM must be considered. It is possible that most patients may be able to adequately delay disease progression with salvage radiotherapy (SRT); however, the results of ongoing trials comparing SRT and ART are yet to be known.

This study has several limitations. This study was conducted in a population of patients who underwent only RP within the VA health system. However, all VA hospitals in SEARCH are academically affiliated and surgeries are performed by attending academic surgeons reducing concerns about surgical technique. Nonetheless, the majority of surgeries in our study involved a resident participating in the surgery. The degree to which this influenced the rate of positive margins is unclear. Indeed our surgical margin rate was high, though this may also reflect the relatively high risk population being treated at the VA. Whether similar results would be seen in a population with a lower overall rate of positive surgical margins is unknown. As we only include men who had complete data, this may have introduced a bias. However, we had complete data on 80% of patients, this likely not a major concern. Furthermore, there was no central pathological review and thus the interpretation of a positive surgical margin likely varied across institutions. We also could not comment on the number, location and extent of surgical margins, nor the tumor volume, each of which would be useful in further qualifying our results. Lastly, given the long natural history of PC, longer follow-up times are needed to confirm our findings.

Conclusions

In our cohort, positive surgical margins after RP were not an independent risk factor for CRPC, metastasis or PCSM overall or within any subsets of RP patients, even when patients undergoing ART were excluded. These findings suggest that in the absence of other high-risk features, positive surgical margins may not worsen long-term outcomes and may not be an indication for adjuvant treatment.