Site of positive surgical margins influences biochemical recurrence after radical prostatectomy


Herbert Lepor, Department of Urology, New York University School of Medicine, 150 East 32nd Street, 2nd Floor, New York, NY 10016, USA.



To determine whether the number and location of positive surgical margins (PSMs) in radical prostatectomy (RP) surgical specimens affect biochemical recurrence (BCR) rates.


The locations of PSMs were recorded for 1308 consecutive men who underwent RP between October 2000 and December 2006. BCR was defined as three consecutive prostate-specific antigen (PSA) level rises with the peak level ≥0.15 ng/mL. Multivariate regression analyses were used to identify preoperative predictors of PSMs and BCR. The estimated 5-year risk of BCR was calculated using the Kaplan–Meier method.


In all, 128 (9.8%) men had one or more PSMs. The mean body mass index, mean preoperative serum PSA level, the distributions of clinical stage and biopsy Gleason scores, and the presence or absence of biopsy perineural invasion were significantly different between men with or with no PSMs. In multivariate analysis, baseline serum PSA level, Gleason score and perineural invasion were independent preoperative predictors of PSMs. The 5-year actuarial BCR rates were dependent on the site of the PSM (P = 0.035) and not the number of PSMs (P = 0.18). The rank order of estimated 5-year BCR rates according to the site of PSMs were base > anterior > posterolateral > apex ≈ posterior.


About half of the men with PSMs in the RP surgical specimen in our prospective series did not develop BCR. The risk of BCR was dependent on the site and not the number of PSMs. Adjuvant therapy should be considered in cases with anterior and basilar PSMs due to the very high risk of BCR.


positive surgical margin


radical prostatectomy


biochemical recurrence


body mass index.


The finding of a positive surgical margin (PSM) after radical prostatectomy (RP) implies that the cancer was not completely resected. It is therefore not surprising that numerous reports have consistently reported that a PSM represents an independent predictor of biochemical recurrence (BCR) after RP [1,2]. These studies have also shown that most men with PSMs do not develop BCR. Because of the unpredictable clinical course of cases with PSMs, there is no consensus of how to manage these individuals. The management options for cases with PSMs include surveillance, adjuvant radiotherapy or androgen-deprivation therapy.

The present study was designed to determine if the number or the site of PSMs is predictive of BCR in a large cohort of men prospectively followed for BCR.


Between October 2000 and December 2006, 1308 men underwent open retropubic RP by a single surgeon using a previously described technique. All men provided informed consent granting permission to take and enter related data into an Institutional Review Board-approved database.

All TRUS-guided prostate biopsies were performed at outside institutions and were internally reviewed using a standardized protocol and reporting methodology. The surgical specimens were also examined using a standardized protocol. The pathological stage (pT2 vs pT3), Gleason score, estimated percentage of tumour volume in the surgical specimen, and the number and site (anterior, posterior, posterolateral, apex, base) of PSMs were prospectively recorded. Surgical margins were considered positive when tumour was seen at the inked surface of the surgical specimen [1]. A unifocal PSM was defined as a PSM seen in only one of the five anatomical locations of the surgical specimen and a multifocal PSM was defined when two or more of the locations had PSMs. The serum PSA level was measured at 3, 6, 12, 18, 24, 30 and 36 months and annually thereafter after RP. In men with pT3 disease or a PSM, additional PSA testing was done every 6 months between 3 and 6 years after RP. BCR was defined as three consecutive rising PSA levels with a peak level of >0.15 ng/mL.

Continuous and categorical data were compared using Student’s t-test and Pearson’s Chi-square test, respectively. Logistic regression models were used to determine, in univariate and multivariate analyses, whether preoperative factors such as age, body mass index (BMI), serum PSA level, clinical stage, biopsy Gleason score and the presence of perineural invasion on biopsy were predictors of PSMs. Multivariate analyses based on Cox’s proportional hazards models were used to ascertain pathological variables that were independent predictors of BCR and also to verify the relationship between the location and extent of PSMs with BCR. The estimated 5-year risk of BCR for unifocal vs multifocal PSMs, and the individual sites of unifocal PSMs (apex, anterior, base, posterolateral, and posterior) were determined using the Kaplan–Meier method and compared by log-rank tests. All statistical tests were performed as two-sided with P < 0.05 considered to indicate statistical significance.


In all, the mean (sd) follow-up was 33.5 (20.1) months. Follow-up of >5 years was available in 246 men.

The baseline characteristics are summarized in Table 1. These characteristics are compared for men with and with no PSMs. The subgroup with PSMs had significantly higher Gleason score distribution, incidence of perineural invasion in the biopsy, proportion of clinical T2 disease and mean serum PSA levels.

Table 1.  Preoperative characteristics of the study population
Baseline characteristicTotalSurgical marginsP
  • *

    Statistically significant, two-sided.

Number of patients13081281180 
Mean (sem)    
 Age, years58.42 (0.19)58.99 (0.60)58.36 (0.20)0.321
 BMI, kg/m227.10 (0.11)27.80 (0.39)27.03 (0.11)0.040*
 PSA level, ng/mL6.37 (0.14)8.48 (0.71)6.14 (0.13)0.002*
% of patients    
 Gleason score   <0.001*
 Clinical stage   0.010*
 Perineural invasion   <0.001*

The univariate analysis showed that BMI, serum PSA level, clinical stage, Gleason score and perineural invasion were significantly associated with PSMs. In multivariate analysis, serum PSA level, Gleason score and the presence of perineural invasion were statistically significant independent predictors of PSMs (Table 2).

Table 2.  Preoperative factors predicting PSMs
Preoperative characteristicUnivariateMultivariate
OR (95% CI)POR (95% CI)P
  1. OR, odds ratio; *statistically significant, two-sided.

Age, years1.014 (0.987–1.041)0.3210.997 (0.967–1.029)0.861
BMI, kg/m21.055 (1.002–1.110)0.041*1.051 (0.995–1.109)0.074
PSA level, ng/mL1.057 (1.030–1.084)<0.001*1.053 (1.023–1.083)<0.001*
Clinical stage 0.014* 0.366
 ≤T2a1.000 (reference) 1.000 (reference) 
 T2b2.353 (1.297–4.270)0.0051.319 (0.641–2.715)0.452
 ≥T2c1.977 (0.428–9.134)0.3832.930 (0.535–16.032)0.215
Gleason score <0.001* <0.001*
 ≤61.000 (reference) 1.000 (reference) 
 >61.616 (1.344–1.943)<0.001*2.337 (1.506–3.627)<0.001*
Perineural invasion <0.001* 0.005*
 no1.000 (reference) 1.000 (reference) 
 yes2.790 (1.824–4.269)<0.001*2.027 (1.234–3.331)0.005*

Using a multivariate Cox’s proportional hazards regression analysis, Gleason score, pathological stage, presence of PSMs and estimated tumour volume were all significant independent pathological predictors of BCR (Table 3).

Table 3.  Multivariate analysis of pathological factors predicting BCR and BCR in cases in PSMs
CharacteristicMultivariate HR (95% CI)P
  1. HR, hazard ratio; *Statistically significant, two-sided; †Controlled for pathological Gleason score (>6 vs ≤ 6), pathological stage (pT3 vs pT2), and tumour volume at specimen (>5% vs ≤ 5%).

Pathological factors
Gleason score (>6 vs ≤6)5.397 (2.894–10.065)<0.001*
Pathological stage (pT3 vs pT2)2.167 (1.387–3.388)0.001*
PSMs (positive vs negative)3.401 (2.212–5.231)<0.001*
Tumour volume (>5% vs ≤5%)3.150 (1.640–6.051)0.001*
Cases with PSMs
Extent of PSM (multiple vs single)1.535 (0.707–3.333)0.278
Location of PSM (anterior/base vs apex/posterior/posterolateral)2.076 (1.041–4.140)0.038*

In all, 128 (9.8%) men had PSMs. Of these cases, 101 and 27 exhibited unifocal and multifocal PSMs, respectively (Table 4). The risks of BCR in the subsets of men with unifocal and multifocal PSM in the present series were 32.7% and 34.6%, respectively. The estimated 5-year risk of BCR for unifocal and multifocal PSMs were 48.2% and 53.8%, respectively. The Kaplan–Meier plots for BCR were not significantly different for unifocal and multifocal PSMs (P = 0.18).

Table 4.  Impact of the number of PSM sites and a unifocal PSM site on BCR
PSMN% patients with BCRLog-rank P*% 5-year risk of BCR*
  • *

    Derived from Kaplan–Meier plot.

Number of sites  0.18 
 Single (unifocal) 10132.7 48.2
 Multiple (multifocal) 2734.6 53.8
Unifocal  0.035 
 Apex 2832.1 41.5
 Anterior  580.0 80
 Base  1154.5 100
 Posterolateral 2818.5 49.7
 Posterior 2931.0 38.3

The anatomical location of the 101 cases with unifocal PSM sites and their associated incidences of BCRs are given in Table 4. The posterior, apex and the posterolateral margins of the prostate were the most common sites of unifocal PSM. The rank order (highest to lowest) for the risk of developing BCR was anterior > base > apex ≈ posterior > posterolateral.

The Kaplan–Meier plots showing the site-specific risk of BCR are shown in Fig. 1. The 5-year BCR rates according to the site of the unifocal PSM, are shown in Table 4. PSMs located at the base and anterior aspects of the specimen were associated with a significantly higher 5-year risk of BCR than the other sites. Less than half of men with PSMs at the posterolateral, apex and posterior aspects of the prostate have developed a BCR at 5 years.

Figure 1.

Kaplan-Meier plots representing the PSM site-specific risk of BCR.

To further validate these findings, Cox’s proportional hazards models were used to analyse the relationship between the extent and site of PSMs vs BCR. This multivariate analysis was used to determine if the extent or site of PSMs influenced BCR independently of the pathological variables listed in Table 3. The extent of the PSM (multiple vs single site of PSM) was not a significant independent predictor of BCR (Table 3). The site of a PSM (anterior and base vs apex, posterior and posterolateral) was a significant independent predictor of BCR (Table 3).


The reported incidence of PSMs varies widely in large RP series, ranging from 8.8% to 37%[2–8]. The variation is most probably due to multiple factors including case selection, surgical technique, methodology of specimen processing and the year surgery was performed. The incidence of PSMs has been decreasing over time, presumably due to stage migration and development of surgical technique [8–10].

In the present study, the overall PSM rate was 9.8%. Our subgroup with PSMs had a significantly greater proportion of cancers with Gleason score ≥7, a higher proportion of clinical T2 disease and higher mean PSA levels at baseline. Using univariate analysis, the preoperative factors predictive of a PSM included BMI, serum PSA level, clinical stage, biopsy Gleason score and biopsy perineural invasion. Using multivariate analysis, serum PSA level, biopsy Gleason score and biopsy perineural invasion were the significant independent predictors of PSMs. All of these preoperative factors have previously been shown to be independent predictors of PSMs [4,5,8,11,12]. The BMI has been shown to be a significant independent predictor of PSM [11] presumably due to the technical challenges of performing RP in obese surgical candidates and an associated lower serum testosterone that probably lowers PSA levels and delays cancer detection. In the present study, BMI approached statistical significance as an independent predictor of PSM.

The present study also examined a PSM and other pathological findings as risk factors for BCR. One of the strengths of the present study was that PSA follow-up was prospective and robust as a full-time data manager was responsible for maximizing follow-up. In addition, it was our strong recommendation not to offer adjuvant therapy in all cases with PSMs until there was a demonstrable BCR.

In multivariate analysis, specimen Gleason score, pathological stage, percentage tumour volume in the surgical specimen and PSMs were all significant risk factors for BCR. All of these factors have been previously associated with BCR [13].

We had adequate follow-up information to present our 5-year BCR rates in men with PSMs. Overall, the present BCR rate at 5 years for men with a PSM was 49.4%. The reported 5-year BCR rates for PSMs in contemporary RP series ranges between 29.4% to 64%[5,14]. The present BCR rate is considerably higher than the 29.4% 5-year actuarial BCR rate reported by Orvieto et al. [5] and the 33% reported by Blute et al. [14]. The higher BCR rate in the present series suggests that our pathologists may be more accurately designating true PSMs.

Interestingly, in the present series there was not a higher BCR rate in men with multifocal vs unifocal PSMs. We do not think this failure to show statistical significance was due to the relatively few men with multifocal PSMs because the BCR rates between unifocal and multifocal PSMs were virtually identical. This finding is consistent with the report by Eastham et al. [4] showing no difference in BCR for one or two PSMs. The BCR rates were significantly higher in their subgroup with two or more PSMs; however, this group represented only 6% of all cases with PSMs and only 0.7% of their total series [4]. Grossfeld et al. [15] and Kausik et al. [16] also suggested that the simple presence of a PSM predicts a more ominous outcome, regardless of the extent of the PSM. However, other groups have reported that more extensive PSMs are associated with higher rates of BCR [3,7,17,18]. We did not prospectively record the extent of PSMs in the present database.

Several other investigators have reported on the location of PSMs. The range of reported percentages of PSMs at the apex, base, anterior, posterolateral and posterior surfaces of prostate RP specimens are 29–53%, 1–14%, 1–18%, 9–35% and 5–16%, respectively. While the rank order for sites of PSMs varies, the more common sites are generally the apex and posterolateral surfaces [3,4,6,7,15,17,18]. In the present series, there were PSMs at the apex, base, anterior, posterolateral and posterior in 27.7%, 10.9%, 5.0%, 27.7% and 28.7% of cases, respectively. The present site-specific incidence of PSMs is generally consistent with published data.

We examined the subset of cases with unifocal PSM to gain insights into the clinical relevance of site-specific PSMs. This is a highly relevant subgroup, as 79.5% of the present PSMs and 56–87% of PSMs in other series were unifocal [3,4,6,7,15,18]. The rank order (highest to lowest incidence) of sites with unifocal PSM in the present series was: posterior, apex, posterolateral, base and anterior aspect of the prostate. The log-rank test analysis suggested that the overall risk of BCR was site specific (P = 0.035). The rank order of sites with the highest to lowest risk of developing a BCR was: anterior, base, apex, posterior and posterolateral. As the posterolateral was a common single site of PSM, we examined the present data to determine if injudicious nerve sparing was contributing to PSMs. Overall, 2376 neurovascular bundles were preserved in the present 1308 men undergoing RP. Only 56 of the 2376 (2.36%) nerve-sparing procedures were associated with a PSM. Of these 56 cases, the posterolateral region was the single site of a PSM in only 28 cases (1.18%), which provides reassurance that our selection of candidates and technique of nerve-sparing surgery is appropriate. The present findings are consistent with other reports suggesting that nerve sparing in appropriately selected cases does not compromise margin status [7,19].

The 5-year actuarial BCR risk for all cases with a unifocal PSM was 48.2%. The risk of BCR was site-dependent. The probability of BCR at 5-years was 100%, 80%, 49.7%, 41.5% and 38.3%, for solitary PSMs at the base, anterior, posterolateral, apex and posterior, respectively. While the sample sizes for anterior and base sites were small, it does appear that a PSM at these sites is an ominous indicator of residual disease. When we analysed only those cases with PSMs, and controlled for the other significant pathological variables, the presence of PSMs at the base and/or anterior aspects of the prostate was associated with a significantly higher likelihood of BCR, even in cases with multiple margins at different sites of the prostate. Obek et al. [17] reporting on a cohort of 546 cases also reported that PSMs identified at the base carried the highest risk of BCR. Blute et al. [3] examined the impact of site-specific PSMs on recurrence-free survival in 2712 men undergoing RP. They found that a PSM at the base of the prostate was associated with the lowest rate of recurrence-free survival. As prostate cancer spreads through the perineural sheath, Villers [20] postulated that the base of the prostate would offer a preferred pathway for cancer dissemination because of the large communications between intraprostatic and periprostatic neural tissue close to the superior neurovascular bundle. Unlike other sites, the ability to achieve a negative margin at the base is limited by the proximity of the trigone. Due to the high rate of BCR it is reasonable to consider adjuvant intervention in these cases. The anterior aspect of the prostate is usually involved by direct extension from cancers originating in other regions of the gland. Therefore, anterior PSMs are often associated with larger and more aggressive tumours, which probably explain the higher risk of BCR.

The sites with the lowest 5-year risk of BCR were the apex and posterior. The apex is the most common site for PSMs and one of the most frequent sites in the present series [3,4,7,18]. However, both the paucity of connective tissue around the prostate in this area, and the surgical technique of apical dissection favour the reporting of false-PSMs. Moreover, the reduced vascularity and the lack of tissue support for residual apical tumour cells in comparison with the bladder neck is another possible explanation for the lower risk of BCR associated with PSMs at the apex. We feel the relatively low risk of BCR associated with apical PSMs justifies an observation strategy.

One limitation of the present study was the relatively short mean follow-up of only 33.5 months. However, most BCRs are reported to occur within the first 2–3 years after RP [4]. In all, 70% of the present series was followed for >2 years and 246 cases were evaluable at 5 years.

The experience of the surgeon has been reported to predict the risk of PSM [10]. Individual surgeons should examine their PSM rate independently for pathologically organ confined disease (pT2) and those cancers associated with extracapsular extension (pT3). If individual PSM rates exceed those published, efforts should be directed towards improving surgical technique.

In conclusion, the present study confirms previous reports that PSM is an independent predictor of BCR after RP performed for the cure of clinically localized prostate cancers. The BCR rates were similar for cases with unifocal and multifocal PSMs. The risk of BCR was highly dependent on the site of the PSM. Men with PSMs at the base and anterior margins of the prostate have a worse prognosis and should be considered candidates for adjuvant treatment due to their very high likelihood of BCR. Although more common, PSMs at the apex, posterior and posterolateral sites, are associated with lower BCR rates. In these cases, it is reasonable to follow PSA levels and offer salvage treatment at the time of BCR. In the present series, PSMs rarely occurred posterolaterally, which suggests that we are appropriately selecting cases and performing the nerve-sparing technique without compromising local disease control.


Guilherme Godoy and Basir U.Tareen received support from the Bruce and Cynthia Sherman Fellowship in Urologic Oncology.


None declared.