What's known on the subject? and What does the study add?
It has been suggested that a very short positive margin does not confer additional risk of BCR after radical prostatectomy.
This study shows that even very short PSM is associated with increased risk of BCR.
To re-evaluate, in a larger cohort with longer follow-up, our previously reported finding that a positive surgical margin (PSM) ≤1 mm may not confer an additional risk for biochemical recurrence (BCR) compared with a negative surgical margin (NSM).
Patients and Methods
Margin status and length were evaluated in 2866 men treated with radical prostatectomy (RP) for clinically localized prostate cancer at our institution from 1994 to 2009.
We compared the BCR-free survival probability of men with NSMs, a PSM ≤ 1 mm, and a PSM < 1 mm using the Kaplan–Meier method and a Cox regression model adjusted for preoperative prostate-specific antigen (PSA) level, age, pathological stage and pathological Gleason score (GS).
Compared with a NSM, a PSM ≤ 1 mm was associated with 17% lower 3-year BCR-free survival for men with pT3 and GS ≥ 7 tumours and a 6% lower 3-year BCR-free survival for men with pT2 and GS ≤ 6 tumours (log-rank P < 0.001 for all).
In the multivariate model, a PSM ≤ 1 mm was associated with a probability of BCR twice as high as that for a NSM (hazard ratio [HR] 2.2), as were a higher PSA level (HR 1.04), higher pathological stage (HR 2.7) and higher pathological GS (HR 3.7 [all P < 0.001]).
In men with non-organ-confined or high grade prostate cancer, a PSM ≤ 1 mm has a significant adverse impact on BCR rates.
Recent studies have established an association between the length of positive surgical margin (PSM) and risk of biochemical recurrence (BCR) after radical prostatectomy (RP) [1-5]. These findings have resulted in a recommendation by the International Society of Urologic Pathology to use margin length, rather than the qualitative terms ‘focal’ or ‘extensive’, as the preferred metric of margin extent . The attention to margin length stems from the need for more accurate risk stratification of patients with a PSM after RP. A common clinical dilemma in this setting is whether to treat patients with adjuvant radiation or whether to observe until the time of BCR. Indeed, adjuvant radiation lowers rates of BCR in patients with a PSM ; however, as many as one third of such men may be over-treated as they will never experience a recurrence . This number may be even higher in men with a short PSM, given their relatively low risk of recurrence compared with men with a longer PSM. It remains unclear, however, if there is a safe threshold for margin length that is not associated with a significant additional risk of recurrence.
Previously, we reported that a relatively short (≤1 mm) PSM may confer a risk of BCR similar to a negative surgical margin (NSM) ; however, this finding may have been confounded by our relatively small cohort size and limited follow-up. In the present study, we have evaluated a significantly larger cohort of men after RP with longer follow-up, in order to address this question more definitively.
Materials and Methods
Data were collected for 2866 men who underwent RP for clinically localized prostate cancer between 1994 and 2009 at our institution. Of these, 979 (34%) underwent open RP from 1994 to 2003 and 1887 (64%) underwent robot-assisted RP from 2003 to 2009. The data elements included age, preoperative PSA level, biopsy Gleason score (GS), clinical stage, pathological GS and stage, margin status (positive or negative), location of PSM, length of PSM in mm and number of PSMs. A PSM was defined as glands present at the inked resection surface. If multiple glands were present at the margin, with uninvolved surface in between them, only the length of glands at the margin was measured and then added up. For patients with multiple sections with PSMs, margin length was determined by the total length of all the PSMs added together. BCR was defined as a serum PSA level >0.10 ng/mL on two consecutive measurements. TNM 6 was used for staging classification. Two hundred patients were excluded from the analysis because of neoadjuvant hormonal therapy (n = 34), salvage RP (n = 6), aborted surgery (n = 29) or lack of follow-up (n = 131). Patients treated with adjuvant radiation, hormonal, or both therapies (n = 61) were censored from further analysis at the time of initiation of adjuvant therapy.
The surgical techniques of both open and robot-assisted RP have been well described . Pathological specimens were processed according to the Johns Hopkins protocol . Prostate specimens were coated in ink and fixed in formalin. The prostatic apex was transversely shaved and then longitudinally sectioned in 1-mm increments. The remainder of the specimens were sectioned transversely in 3-mm blocks and divided into quarters. Slides from each quarter were prepared, stained, and microscopically examined. A PSM was defined as tumour cells present at the inked margin. Among cases with a PSM (n = 402), in 80% (n = 323) the slides were available for secondary review by two dedicated uropathologists (H.A., T.A.) for margin length measurement. The non-available cases did not differ from the rest of the PSM group in terms of PSA level, pathological stage, GS or age (all P > 0.2; data not shown). On exploratory multivariate model analyses, PSM location, surgical approach and year of surgery were not independently associated with the outcome, and therefore they were omitted from the final model.
The impact of margin status and length (≤1 mm and >1 mm vs NSM as a reference) on BCR was modelled using Cox regression univariate and multivariate analyses, adjusting for preoperative PSA level, pathological stage (pT3 vs pT2 as a reference) and pathological GS (≥7 vs ≤6 as a reference). The decision to choose these stage and grade strata was made after exploratory analyses showing no significant impact of more detailed stratification on model output. The analyses were performed with Stata 11.0 (College Station, TX, USA). A P value of 0.05 was considered to indicate statistical significance.
The median patient age was 60 years and the median PSA level was 5.4 ng/mL (Table 1). Seventy-five percent had pathologically organ-confined disease and 47% had GS ≤ 6 on final pathology. Fourteen percent (n = 402) had a PSM and 10% (n = 277) experienced BCR. The median (interquartile range [IQR]) follow-up was 20 (8–37) months.
Table 1. Characteristics of the study cohort (N = 2866)
Approach, n (%)
Median (IQR) age, years
Median (IQR) PSA, ng/mL
Clinical stage, n (%)
Biopsy GS, n (%)
PSM, n (%)
Pathological GS, n (%)
Pathological stage, n (%)
Lymph nodes, n (%)
Median (IQR) follow-up, months
The median (IQR) PSM length was 0.8 (0.3–2.1) mm for the 323 PSM cases available for review (Table 2). Of those, 61% (n = 169) were ≤1 mm. In patients with a NSM, a PSM ≤ 1 mm and a PSM > 1 mm, 3-year BCR-free survival was 92%, 78% and 60% (log-rank P < 0.001), respectively (Fig. 1A). In patients with pathological stage pT2 and GS ≤6 disease, PSM ≤ 1 mm conferred a 6% lower 3-year BCR-free survival compared with a NSM (log-rank P < 0.001 [Fig. 1B]). By contrast, for patients with pT3 and GS ≥7 disease, the 3-year BCR-free survival was 17% lower for a PSM ≤ 1 mm compared with a NSM (all log-rank P < 0.001 [Fig. 1B and C]).
Table 2. Characteristics of men with a PSM (n = 402)
Location, n (%)
Stage, n/N (%)
Median (IQR) length: continuous, mm
Length: categorical, n/N (%)
In the multivariate model, a higher PSA level (hazard ratio [HR] 1.04), higher pathological stage (HR 2.7) and higher pathological GS (HR 3.7) were associated with a higher probability of BCR (all P < 0.001 [Table 3]). In the same model, a PSM ≤ 1 mm conferred a BCR probability twice as high as that for a NSM (HR 2.2, P < 0.001) and was almost fourfold higher in patients with a PSM > 1 mm (HR 3.7, P < 0.001).
Table 3. Cox regression analysis
GS ≥ 7
>1 mm negative ref.
The presents results show that a PSM ≤ 1 mm more adversely affects BCR-free survival in men with extraprostatic, high grade disease compared with men with organ-confined, low grade tumours. Furthermore, on multivariate analysis, a PSM ≤ 1 mm is independently associated with a probability of BCR twice as high. While the follow-up of the present cohort is not mature enough to determine whether this translates into higher rates of metastases, a higher BCR rate will lead to higher rates of salvage radiation therapy.
The present findings are consistent with previous studies showing that BCR rates after RP are less affected by capsular incision in organ-confined tumours compared with PSMs associated with extraprostatic extension [12, 13]. Notably, in the subgroup of low grade, low stage patients, the difference in 3-year BCR-free survival between a NSM and a PSM ≤ 1 mm was only 6%.
In a similar study, Cao et al.  found a stronger association between PSM length and BCR rates in men with pT2 prostate cancer than in men with pT3 disease; however, these authors did not group patients with both pT2 and GS≤ 6 tumours together in a combined low-risk category and they also stratified patients differently, using a threshold of 3 mm for PSM length. Differences in measurement technique may also affect the results and a direct comparison between the present study and theirs is therefore difficult. Indeed, most studies have found that men with pT2 disease and PSMs resulting from capsular incision have an intermediate BCR-free rate between that of men with NSMs and men with pT3 disease.
Accurate BCR risk stratification in men undergoing RP with PSMs is necessary to identify appropriate candidates for adjuvant radiation therapy and thereby to avoid unnecessary treatment of men who may never experience a BCR. Previous studies have shown that PSM length is an accurate metric since it is easily measurable and reproducible. Indeed, in a recent multi-institutional study, Stephenson et al.  showed that the extent of margin involvement was a significant predictor of BCR, although it did not contribute additional accuracy to the nomograms derived from their analysis. Disagreement with our results may reflect our linear measurement of PSM compared with their subjective stratification into either focal or extensive categories that may be confounded by inter-observer variability. Non-apical margin location has also been linked to a higher probability of BCR, but we did not incorporate PSM location into the present analysis .
The present results suggest men with both pT2 and GS ≤ 6 disease with PSMs resulting from capsular incision have a low risk of 3-year BCR-free survival and these men should therefore be followed expectantly. By contrast, men with either pT3 or GS ≥ 7 disease with a PSM of ≤1 mm or shorter should be monitored closely for BCR and should strongly consider adjuvant external beam radiation therapy or salvage radiation at the earliest signs of BCR. Ongoing trials are currently accruing to address this clinical dilemma .
The present study has some limitations that should be considered when interpreting our findings. Firstly, follow-up was limited and it is conceivable that more recurrence events would occur in low risk patients with a PSM ≤ 1 mm over time, but this would be true also for the high grade tumours. Obviously, despite this limitation we had enough events in our cohort to show that a PSM < 1 mm is associated with an increased risk of BCR, and the risk ratio is both clinically and statistically significant. Secondly, during our study period both stage and grade migration occurred, which could affect our results, although we believe the impact would be relatively minor. Thirdly, there could be inter-observer variability for the margin evaluation . PSM length may be uncertain because adjacent consecutive areas of cancerous glands intercalated between normal glands may be interpreted either as several short PSMs or as a single longer PSM. In our experience, these occasions are rare. Finally, other than BCR, there are more clinically important outcome points, e.g. measurable local recurrence, distant metastases and cancer-specific survival. To achieve these endpoints, much larger series and follow-up of a decade or longer would be needed, a limitation of most single-institution series.
In summary, the present report shows that even a short PSM (<1 mm) negatively impacts BCR. In addition, it shows the importance of a critical appraisal of preliminary results and the need for re-assessment with an adequate sample size and length of follow-up. After RP, a PSM ≤ 1 mm confers an independent risk of BCR which is twice as high as that for a NSM. Furthermore, men with a PSM ≤ 1 mm and GS ≥ 7 or pT3 disease have a 20% higher risk of BCR at 3 years than those with GS ≤ 6 and organ-confined tumours. Therefore, these men, despite having a short PSM, should be monitored closely and should strongly consider adjuvant external beam radiation therapy or salvage radiation at the first evidence of BCR.