- To describe the characteristics of patients with and without positive surgical margins (PSMs) and to analyse the impact of PSMs on secondary cancer treatment after radical prostatectomy (RP), with short-term follow-up.
It is projected that in 2020 > 25 000 new cases of prostate cancer will be diagnosed in Australia, which is more than the projections for any other cancer . Prostate cancer is estimated to account for 15% of the cancer burden and 3% of the overall health burden in Australian men, second only to lung cancer . Prostate cancer is principally a disease of older men, with the mean age of diagnosis being 68.4 years .
Over the past 15 years there has been a shift towards earlier diagnosis of prostate cancer and increasing use of surgery to treat the disease . Radical prostatectomy (RP) provides 15-year prostate cancer-specific survival of up to 93% . While there are a number of surgical approaches which can be used, including open, laparoscopic and robot-assisted techniques, all have the same goal: to resect the local tumour and cure the disease whilst minimizing side effects to the patient.
One measure of the success of RP is whether the cancer is contained within the resected prostate or has extended beyond the surgical margin. A review of the literature has identified the overall prevalence of positive surgical margins (PSMs) to be 24% for open retropubic RP, 20% for laparoscopic RP and 16% for robot-assisted LRP . The most well-known risk factor associated with PSM is a high tumour category . Other risk factors identified in the literature to varying extents include the skill of, and technique used by, the pathologist examining the specimen [7, 8], preoperative PSA level , surgical technique , surgical volume and surgeon experience . Surgical margin status in organ-confined disease is widely regarded as a measure of surgical quality.
There is evidence that PSMs after RP are independently associated with a greater risk of biochemical progression, even after accounting for various pathological features, including disease stage [6, 11]. The most striking effect is found for patients with intermediate- and high-risk disease . There is less clear evidence of a link between PSMs and metastases or death .
A Cochrane systematic review  and evidence-based guidelines from Canada , the USA , Europe  and Australia  suggest consideration of radiation treatment for men after RP, if found to have postoperative risk factors such as a PSM, pT3 disease or a detectable PSA level. A secondary analysis of one of the three randomized comparative trials suggesting the benefit of postoperative treatment with radiation for men with high-risk disease noted the particular risk posed by a PSM . A study of factors associated with the use of adjuvant therapies reported that patients with PSMs were three times more likely than those without PSMs to have additional treatment .
Little is known about PSM prevalence in Australia and whether PSMs influence the decision to perform additional treatment. No multicentre studies have evaluated the variation in PSM status and its association with subsequent care. The purpose of the present analysis was to describe the characteristics of patients with and without PSMs and to analyse the association of PSMs with secondary cancer treatment after RP.
Data on RP cases were obtained from the Prostate Cancer Registry, established in 2009 initially in four hospitals (accounting for 25% of all notifications to the Victorian Cancer Registry) to assess patterns of presentation, care and outcomes. An additional three hospitals joined the registry in late 2009, nine hospitals in 2010 and 22 hospitals in 2011, resulting in the registry accruing ∼72% of prostate cancer incidence cases in Victoria by February 2012. Registry recruitment is linked with mandatory notification of cancer status to the population-based Victorian Cancer Registry. Details of the recruitment strategy have been described previously .
Patients from recruiting hospitals with a diagnosis of prostate cancer confirmed by histopathology report at time of biopsy or RP were eligible for inclusion in the study. Evidence from data collected by the Victorian Cancer Registry indicates that, in 1997, ∼94% of prostate cancer diagnoses were histologically verified . All RPs with unequivocal determination of surgical margin status performed to 29 February 2012 were included in the analysis of risk factors for PSMs. Patients who underwent RP and received additional immediate or delayed treatment in the initial 12-month period after diagnosis were included in the analysis of factors associated with additional treatment. Patients were ineligible if their diagnosing or treating doctor informed the registry that they were not capable of providing consent.
After obtaining approval from each hospital's human research ethics committee and the diagnosing or treating clinician (private patients) or Heads of Department (public patients), all patients diagnosed after the authorization date were progressively accrued. Hospitals provide details of each case to the registry to enable an explanatory statement to be sent ∼9 months after diagnosis, inviting participation and providing details of how patients can opt out of the registry if they choose not to participate. The statements are available in 12 common languages.
Histopathological data are captured through hospital information systems and pathology reports. Histopathological data are collected at the time of diagnosis and surgery, but not on biopsy specimens taken after diagnosis but before surgery (for active surveillance patients). Trained data collectors capture clinical information from medical records including treatment and PSA levels at diagnosis, before treatment and immediately preceding the 12-month follow-up date. In some instances a diagnosis is made in a notifying hospital and the patient subsequently undergoes RP at another hospital. In this situation, the surgery data are requested by the notifying hospital. Patients are telephoned 12 months after diagnosis to confirm the accuracy of treatment data and most recent PSA level and to collect quality-of-life information. Periodically, hospitals are asked to validate biopsies and RPs against International Classification of Disease-10 codes to confirm complete capture of cases.
The National Comprehensive Cancer Network (NCCN) risk criteria for disease progression were used to classify patients into low-, intermediate- and high-risk disease (Table 1) . Where the clinical T category was not recorded, if the Gleason score was ≤6 and the PSA concentration was <10 ng/mL, the patient was deemed to be at low risk for disease progression. Frequencies were used to describe the predictors of PSMs with a two-sided level of significance determined using a chi-squared test. PSMs were assessed by patient age at surgery using the Kruskal–Wallis equality-of-populations rank test. Significant predictors of PSMs at univariate level were investigated using unadjusted and adjusted Poisson regression, clustering by surgeon. The adjusted model was tested for significant interaction effects between model covariates.
|Clinically localized disease: very low risk for disease progression||T1a and Gleason ≤6 and PSA <10 ng/mL and <3 biopsy positive cores, ≤50% cancer in each core and PSA density <0.15 ng/mL/g|
|Clinically localized disease: low risk for disease progression||cT1–cT2a and Gleason score 2–6 and PSA <10 ng/mL|
|Clinically localized disease: intermediate risk for disease progression||cT2b–T2c or Gleason score 7 or PSA 10–20 ng/mL|
|Clinically localized disease: high risk for disease progression||cT3a or Gleason score 8–10 or PSA >20 ng/mL|
|Locally advanced disease: very high risk for disease progression||cT3b–T4, N0, M0|
|Locally advanced disease: metastatic disease||any cT, N1, M1|
Time from diagnosis to RP was compared according to hospital type (private/public) and location (metropolitan/regional hospitals) using the Kruskal–Wallis equality-of-populations rank test. To determine the predictors of further treatment, data were dichotomized into surgery only and surgery with additional treatment. A chi-squared test was used to determine significance at a univariate level, and logistic regression with clustering by surgeon was used to determine significance at a multivariate level. Radiotherapy was deemed to be immediate if it commenced <6 months after surgery and delayed if it commenced >6 months after surgery. Stata/ic 12.0 (StataCorp, College Station, TX, USA) was used for all analyses and a two-sided P value <0.05 was considered to indicate statistical significance.
Ethical approval was gained from participating hospitals, Monash University and the Cancer Council, Victoria.
Between 1 August 2008 and 29 February 2012, 6514 pathologically confirmed cases of prostate cancer were notified to the registry from recruiting hospitals. Of the 6514 cases notified from recruiting hospitals, 1164 were ineligible: in 619 (9.5%) cases the patient was diagnosed outside of the study time frame; in 243 (3.7%) cases the patient was both diagnosed and had treatment at a non-recruiting hospital; in 136 (2.1%) cases the patient had died; 110 (1.7%) cases were diagnosed before consent had been obtained from the patient's diagnosing doctor; in 31(0.5%) cases the patient was not recruited on the recommendation of the patient's doctor, in 22 (0.3%) cases the patient was too unwell or unable to consent because of cognitive deficit, and in three (0.05%) cases the patient could not be contacted. A further 105(1.6%) patients opted out because they were not interested in contributing to the registry. Of the 5245 patients (80.5% of notified prostate cancer cases) who were eligible and consented to contribute to the Prostate Cancer Registry, 2385 (45.5%) underwent surgery and were the focus of the present research. Treatment was confirmed by telephone interview with 2006/2385 (84.1%) patients. The reasons for loss to follow-up are shown in Fig. 1.
The RPs were performed at 23 hospitals in Victoria. Preoperative patient and tumour characteristics are shown in Table 2. In all, 113 patients were classified as low-risk disease on the basis of having a PSA <10 ng/mL and a Gleason score of ≤6, but without a clinical T stage recorded. Based on the NCCN risk stratification criteria, 608/2385 patients (25.5%) were at low risk for disease recurrence, 1300/2385 patients (54.5%) were at intermediate risk, 423/2385 patients (17.7%) were at high risk, 32/2385 patients (1.4%) had locally advanced or metastatic disease, and for 22 patients (0.9%) a NCCN risk category could not be assigned. The median PSA level at diagnosis was 5.9 ng/L, the mean (sd; range) PSA level was 8.3 (31.6;1.2–1456) ng/L and the median Gleason score was 7.
|Age at RP (N = 2385)|
|Mean (sd) age, years||62.2 (6.7)|
|Median (range) age, years||62.9 (45.8–83.9)|
|First diagnosing hospital (N = 2257), n (%)|
|Notifying hospital (N = 2385), n (%)|
|Time from diagnosis to RP (N = 2256)|
|Mean (sd) days from diagnosis to RP||86 (100)|
|Median (IQR) days from diagnosis to RP||57 (39–92)|
|Surgical hospital location (N = 2357), n (%)|
|Surgical hospital type (N = 2359), n (%)|
|Hospital surgical volume (N = 2272), n (%)|
|≤10 cases/year||61 (2.7)|
|>10 cases per year||2211 (97.3)|
|Clinical T category at diagnosis (N = 1961), n (%)|
|≤ T1c||1052 (53.6)|
|N1, M1||21 (1.1)|
|Not assessed||6 (0.3)|
|Gleason score at biopsy (N = 2355), n (%)|
|PSA concentration at diagnosis (N = 2345), n (%)|
|0–10 ng/mL||1987 (84.7)|
|10.01–20 ng/mL||266 (11.3)|
|>20.0 ng/mL||92 (3.90)|
|NCCN risk of disease progression at diagnosis (N = 2363), n (%)|
|Low risk||608 (25.7)|
|Intermediate risk||1300 (55.0)|
|High risk||423 (17.9)|
|Locally advanced disease: very high risk||11 (0.5)|
|Locally advanced: metastatic disease||21 (0.9)|
|Surgical (RP) pathological T category (N = 2083), n (%)|
|pT2 disease||1384 (66.4)|
|pT3 disease||699 (33.6)|
Data collectors were instructed to record the surgical margin status as positive if it was explicitly stated on the pathologist report. Surgical margin status was reviewed in 2219/2385 patients (93.0%). Margin status was unequivocal for 2175/2219 patients (98.0%) but for 44 patients (2.0%) a definitive status was unable to be determined because of insufficient sample (n = 7), unclear documentation (n = 6) or margin status not being stated on the pathology report (n = 31). These cases were not included in further analyses.
Positive surgical margins were identified for 592/2175 patients (27.2%) and were significantly less frequent for patients with pT2 disease than for those with pT3 disease (16.1 vs 50.7%, P < 0.001). Of the 542 RPs performed at public hospitals, margin status was known for 487 patients (90.0%) and of the 1817 RPs performed at private hospitals, margin status was known for 1684 patients (92.7%). There were 27 cases where the surgical hospital was not recorded. The total number of PSMs reported from public hospitals was 170 (overall PSM rate = 34.9%), of which 60 were for pT2 disease (pT2 PSM rate = 21.8%); 100 were for pT3 disease (pT3 PSM rate = 54.6%); and 10 (5.9%) were not classified. The total number of PSMs reported from private hospitals was 421 (overall PSM rate = 25.0%), of which 158 were for pT2 disease (pT2 PSM rate = 14.2%); 244 were for pT3 disease (pT3 PSM rate = 47.4%); and 19 (4.5%) were not classified. Overall, there were significantly more PSMs for cases from public hospitals compared with private hospitals (34.9 vs 25.0%, P < 0.001), and also for pT2 and pT3 cases (P = 0.001). Patients whose margin status was unknown were more likely to have surgery at a regional hospital than those whose status was known (21.0 vs 6.8%, P < 0.001), were more likely to be in the low (11.9%) or metastatic disease (19.1%) risk group than in the intermediate- (5.5%) and high- (8.3%) risk group, were more likely to have a low Gleason score on biopsy (11.3% missing if Gleason ≤ 6, 5.3% missing if Gleason sum = 7 and 8.7% missing if Gleason sum 8–10) and were more likely to be in a higher PSA category at diagnosis (7.2% missing in <10 ng/mL group, 9.0% in 10–20 ng/mL and 16.3% in >20 ng/mL group). No differences between known and unknown margin status were identified in relation to the patient's age, Gleason score at surgery, pathological T category or whether they underwent RP at a private or public hospital.
The date of RP was recorded for 2256/2385 patients (94.6%). Where the date of RP was known, the vast majority of patients had their RP within 12 months of diagnosis (n = 2182/2256 or 96.7%) and there was no difference in PSMs according to whether or not RP was performed within 12 months of diagnosis (PSM 27.1 vs 31.4%, P = 0.423). Table 3 shows the univariate factors associated with PSMs and Table 4 shows results of the multivariate analysis. Patients with PSMs were significantly more likely to be in the intermediate-, high- or very high-risk groups or in the metastatic disease group than in the low-risk group (incidence rate ratios [IRRs] 1.34, 95% CI 1.09–1.65, P = 0.007; IRR 1.96, 95% CI 1.57–2.45, P < 0.001; IRR 3.81, 95% CI 2.60–5.60, P < 0.001; and IRR 2.50, 95% CI 1.23–5.11. P = 0.012, respectively). PSMs were more prevalent in cases from hospitals performing <10 RPs per year compared with those from hospitals with higher surgical volumes (IRR 0.76, 95% CI 0.63–0.93, P = 0.006). Patients undergoing RP at private hospitals were less likely to have PSMs than those undergoing RP in public hospitals (IRR = 0.76, 95% CI 0.63–0.93, P = 0.006). Robot-assisted RP was associated with a significantly lower likelihood of PSMs than was open RP (IRR = 0.69, 95% CI 0.55–0.87, P = 0.002; Table 4).
|Variable||Surgical margin status||P|
|N = 592 (27.2%)||N = 1583 (72.8%)||N = 2175 (100%)|
|Age at surgery||0.368*|
|No. of patients||586||1568||2154|
|Mean (sd) age, years||62.5 (6.8)||62.1 (6.7)|
|Surgical hospital location, n (%)||0.862|
|No. of patients||591||1580||2168|
|Metropolitan||550 (27.3)||1467 (72.7)||2017|
|Regional||41 (26.6)||113 (73.4)||154|
|Hospital type, n (%)||<0.001|
|No. of patients||591||1580||2171|
|Public||170 (34.9)||317 (65.1)||487|
|Private||421 (25.0)||1263 (75.0)||1684|
|Hospital surgical volume†, n (%)|
|≤10 cases/year||19 (47.5)||21 (52.5)||40|
|>10 cases/year||555 (26.4)||1544 (73.6)||2099||0.003|
|Surgical approach, n (%)||<0.001|
|No. of patients||591||1582||2173|
|Open RP||376 (32.6)||776 (67.4)||1152|
|Robot-assisted RP||172 (20.4)||672 (79.6)||844|
|Laparoscopic RP||43 (24.7)||131 (75.3)||174|
|Clinical T category, n (%)||<0.001|
|No. of patients||455||1311||1757|
|≤ T1c||190 (20.2)||749 (79.8)||939|
|T2a||123 (28.7)||306 (71.3)||429|
|T2b||60 (30.0)||140 (70.0)||200|
|T2c||39 (33.6)||77 (66.4)||116|
|T3a||34 (48.6)||36 (51.4)||70|
|T3b||9 (75.0)||3 (25.0)||12|
|PSA at diagnosis, n (%)||<0.001|
|No. of patients||585||1565||2150|
|0–10.0 ng/mL||453 (24.7)||1378 (75.3)||1831|
|10.1–20.0||89 (26.8)||153 (63.2)||242|
|>20||43 (55.8)||34 (44.2)||77|
|Biopsy Gleason score||<0.001|
|No. of patients||590||1578||2155|
|≤6||131 (19.8)||530 (80.2)||661|
|7||342 (28.6)||852 (71.4)||1194|
|8–10||117 (37.4)||196 (62.6)||313|
|NCCN risk rating, n (%)||<0.001|
|No. of patients||590||1577||2168|
|Low||102 (19.1)||432 (80.9)||534|
|Intermediate||317 (26.0)||901 (74.0)||1218|
|High||153 (39.5)||234 (60.5)||387|
|Locally advanced-very high risk||9 (81.8)||2 (18.2)||11|
|Metastatic disease||9 (52.9)||8 (47.1)||17|
|Hospital surgical volume|
|Open RP (reference)||1.00|
|NCCN risk of disease progression|
|Very high risk||3.81||2.60–5.60||<0.001|
Patients treated at private hospitals were more likely than those treated at public hospitals to receive earlier RP: median (interquartile range [IQR]) time from diagnosis to RP 50 (35–77) vs 90 (64–133) days (P < 0.001). Patients at metropolitan hospitals were more likely than those at regional hospitals to receive earlier RP: median (IQR) time from diagnosis to RP 56 (39–91) vs 66 (45–103) days (P < 0.001). A total of 1987 patients (91.1%) received RP as a monotherapy in the initial 12-month period after diagnosis, 123 patients (5.6%) received RP and radiotherapy; 47 patients (2.1%) received RP and androgen deprivation therapy (ADT), 23 patients (1.1%) received RP, ADT and radiotherapy, one patient (0.1%) received RP, ADT and chemotherapy and one patient (0.1%) received RP, radiotherapy, ADT and chemotherapy. Of the 147 patients who received radiotherapy, with or without other treatment, in the initial 12 months after RP, 32 (21.7%) received it within 6 months of surgery (‘immediate’). The median (IQR) PSA concentration before immediate radiotherapy (n = 23) was 0.23 (0.05– 0.70) ng/mL, before delayed radiotherapy (n = 111) it was 0.14 (0.09–0.3) ng/mL, before ADT (n = 72) it was 1.06 (0.17–10.6) ng/mL and before chemotherapy (n = 2) it was 11 (8.3–13.7) ng/mL.
Independent factors associated with additional treatment after RP are shown in Table 5. Table 6 shows the multivariate analysis of factors associated with additional treatment. Significant predictors included a PSM compared with a negative surgical margin (odds ratio [OR] 5.61, 95% CI 3.82–8.22, P < 0.001), a pathological category of T3 compared with T2 (OR 4.72, 95% CI 2.69–8.23, P < 0.001) and high-risk or very-high-risk disease compared with low-risk disease (OR 4.36, 95% CI 2.24–8.50, P < 0.001 and OR 4.50, 95% CI 1.34–15.17, P = 0.015, respectively). Patients receiving robot-assisted RP were significantly less likely to receive additional treatment in the initial 12 months after diagnosis compared with those receiving open RP (OR 0.59, 95% CI 0.39–0.88, P = 0.010).
|Surgery only||Surgery + other treatment||Total no.|
|1987 (91.1%)||195(8.9%)||2182 (100%)|
|Age at RP||1973||195||2168||0.532KW|
|No. of patients|
|Mean (sd) age, years||62.1 (6.7)||62.5 (6.5)|
|Median (IQR) age, years||62.8 (57.8–67.1)||63.2 (57.5–67.6)|
|<65 years, n (%)||1260 (91.2)||121 (8.8)||1381||0.616|
|≥65 years, n (%)||713 (90.6)||74 (9.4)||787|
|Surgical hospital location, n (%)||0.769|
|No. of patients||1982||195||2177|
|Metropolitan||1831 (91.1)||179 (8.9)||2010|
|Regional||151 (90.4)||16 (9.6)||167|
|Hospital type, n (%)||1982||195||2177||<0.001|
|Public||425 (86.7)||65 (13.3)||490|
|Private||1557 (92.3)||130 (7.7)||1687|
|Hospital surgical volume, n (%)||0.587|
|No. of patients||1948||187||2177|
|≤10 cases/year||43(93.5)||3 (6.5)||46|
|>10 cases/year||1905 (91.2)||186 (8.8)||2089|
|Surgical approach, n (%)||<0.001|
|No. of patients||1980||195||2175|
|Open radical RP||1026 (88.5)||134 (11.5)||1160|
|Robot-assisted RP||783 (94.1)||49 (5.9)||832|
|Laparoscopic RP||167 (93.8)||11 (6.2)||178|
|Other||4 (80.0)||1 (20.0)||5|
|Clinical T category, n (%)||<0.001|
|No. of patients||1606||151||1757|
|≤T1c||884 (95.4)||43 (4.6)||927|
|T2a||396 (91.0)||39 (9.0)||435|
|T2b||176 (86.7)||27 (13.3)||203|
|T2c||102 (86.4)||16 (13.6)||118|
|T3a||51 (69.9)||22 (30.1)||73|
|T3b||6 (54.5)||5 (45.5)||11|
|PSA at diagnosis, n (%)||1964||194||2158||<0.001|
|0–10.0 ng/mL||1705 (92.8)||133 (7.2)||1838|
|10.1–20.0 ng/mL||209 (86.4)||33 (13.6)||242|
|>20 ng/mL||50 (64.1)||28 (35.9)||78|
|RP Gleason score, n (%)||<0.001|
|No. of patients||1961||194||2155|
|≤6||304 (97.4)||8 (2.6)||312|
|7||1512 (93.0)||114 (7.0)||1626|
|8–10||145 (66.8)||72 (33.2)||217|
|NCCN risk rating, n (%)||<0.001|
|No. of patients||1977||195||2172|
|Low||500 (97.1)||15 (2.9)||515|
|Intermediate||1160 (94.1)||73 (5.9)||1233|
|High||299 (75.3)||98 (24.7)||397|
|Very high||5 (50.0)||5 (50.0)||10|
|Metastatic disease||13 (76.5)||4 (23.5)||17|
|Pathological T category, n (%)||<0.001|
|No. of patients||1826||183||2009|
|pT2 disease||1298 (97.4)||35 (2.6)||1333|
|pT3 disease||528 (78.1)||148 (21.9)||676|
|PSM, n (%)||<0.001|
|No. of patients||1916||184||2100|
|No PSM||1484 (96.9)||47 (3.1)||1531|
|PSM||432 (75.9)||137 (24.1)||569|
|NCCN risk category|
The purpose of the present analysis was to describe the prevalence of PSMs and their influence on subsequent additional treatment in the 12-month period after diagnosis, using data from a contemporary cohort of patients enrolled in a prostate cancer registry. We found that PSMs were present in the resections in 27% of patients treated by RP and that patients treated at private hospitals, and with a robot-assisted technique were less likely to have PSM compared with those treated at public hospitals using an open technique. There was a volume–quality relationship, with those hospitals performing <10 RPs per year having a higher rate of PSMs compared with those performing >10 RPs per year. The three most significant factors associated with patients receiving additional treatment in the initial 12-month period after diagnosis were pathological T category, denoting invasion of the seminal vesicle or beyond (nearly fivefold increased risk), having high- or very-high-risk disease (a fourfold increased risk) and having PSMs (a fivefold increased risk).
Notwithstanding the observed association between postoperative risk factors and subsequent treatment, only a subset of men with adverse pathology after RP received additional treatment. This suggests that other factors besides pathology influence the use of additional treatments. North American population registry-based studies in the period 2000–2007 have also shown that small proportions of patients (11.5%) with adverse pathology go on to undergo postoperative radiation treatment .
The prevalence of PSMs after RP in Victoria may have declined over recent years; in a review of all RPs in Victoria in the period 1995–2000, the overall PSM prevalence was 31% . The overall PSM prevalence identified in the present study, 27%, is less than the 34% identified in a study of 1383 patients contributing to the CaPSURE database , yet is significantly higher for the subgroup of patients with a diagnosis PSA of 0–4 ng/mL (63/344 or 18.3% vs 34/450 or 7.6%, Fisher's exact test: P < 0.001). The PSM prevalence rates of 25% for laparoscopic and 20% for robot-assisted RP in the present study are high compared with those from a single-site study of 200 laparoscopic and 200 robot-assisted RPs, which reported prevalence rates of 12 and 13.5%, respectively . The PSM prevalence rate of 33% in the present study for open RP was marginally less than the 35% reported by Smith et al. , yet the robot-assisted RP PSM prevalence of 15% reported by the same study was lower than ours (20%).
Of interest was the finding that robot-assisted RP was associated with a 30% lower risk of PSMs compared with an open approach, even after taking into account factors such as the stage of disease at diagnosis (including in the model PSA level, clinical T stage and Gleason score). There have been contradictory results from comparisons of open and robot-assisted techniques [9, 26]. Smith et al.  examined PSMs using a single-institution study of 200 robot-assisted operations with 200 open radical retropubic operations between 2002 and 2006 and reported that the robot-assisted approach was associated with fewer PSMs (9.4 vs 24.1% for pT2 disease, P < 0.001, and 50 vs 60% for pT3 disease). This finding was contrary to that of another single-institution study conducted between 2003 and 2008, which showed that when propensity matching of cases based on age, race, preoperative PSA level, biopsy Gleason score and clinical stage was used, both open and laparoscopic approaches yielded a lower PSM prevalence compared with the robot-assisted technique . This difference may be attributable to the fact that the study by Smith et al. reviewed the cases of only four surgeons who were proficient at robot-assisted surgery, while the study by Magheli et al.  included only surgeons who had completed their urological training, but the majority had performed <150 robot-assisted RPs in total. Magheli et al. stated that the open technique provided tactile information which assisted the surgeon in reducing PSMs. It may be that the present study found a lower PSM prevalence for cases treated by the robot-assisted technique because of a learning curve phenomenon. Urologists performing robot-assisted surgery were experienced in the technique. In contrast, many open RPs at public hospitals are performed by training surgeons. Had the registry commenced data collection at a time when the urologists were learning the robotic technique, then the PSM prevalence rate may have been higher.
Similarly, we found that patients treated at private hospitals were less likely to have PSMs than those treated at public hospitals. Reasons for this are unknown. It may reflect surgeon experience, as operations performed at private hospitals are usually performed by a consultant surgeon while many undertaken at public hospitals are performed by trainees under supervision. Surgeon experience has been shown to affect the PSM rate when laparoscopic , open , and robot-assisted  approaches are used. It may also indicate a difference in patient profile and specimen review processes between public and private hospitals. Notably, there was no association between PSM prevalence and whether patients were treated in regional or metropolitan hospitals.
Our finding that patients having surgery performed at a low-volume hospital (defined as <10 cases/year) were more likely to have PSMs compared with those treated at higher-volume hospitals, even after taking into account surgical approach and disease stage, is at odds with the study by Lawrentschuk et al. , which was confined to cases with pathological stage T2 disease. Chun et al.  similarly found that the PSM rate was not affected by surgical volume for men with low- and intermediate-risk disease, but that surgeons who operated on higher volumes of patients with high-risk disease could expect to have a significantly lower PSM prevalence rate compared with that of surgeons who operated on lower volume. A limitation of both the present study and that by Lawrenschuk et al. is that neither could analyse PSMs based on surgical volume without knowledge of individual surgeon caseload outside the reporting hospitals. Sammon et al.  did not examine PSM prevalence but found that high-volume centres had more favourable outcomes for both open and robot-assisted RP with regard to most intraoperative and postoperative complications compared with low-volume hospitals.
Given that PSMs are a known independent predictor of biochemical recurrence and cancer-specific mortality, a common goal of surgeons should be to avoid PSMs where possible. In the present paper we provide some indication of the factors associated with high PSMs in an Australian prostatectomy sample; however, there are a number of limitations to this study which may affect its interpretation. Importantly, we did not determine how specimens were sectioned, nor did we assess the experience of the pathologist undertaking the review, both of which have been associated with PSM status [7, 33]. A study by van der Kwast  found that review of specimens by an expert urological pathologist after an initial review by a general pathologist resulted in only 69.4% agreement (κ = 0. 45) with regard to PSM status. Ekici et al.  found that partial sectioning of specimens resulted in good agreement between pathologists regardless of their level of experience, but when an expert pathologist employed a complete sectioning technique there was a significant difference in PSM status. Evans et al.  found that when 12 expert urological pathologists compared their results, there was good sensitivity and specificity in determining PSM status (83.3 and 97.5%, respectively). These studies suggest that both the technique and the experience of pathologist will affect interpretation of PSM status. Having a centralized pathology review system would provide greater understanding of whether the PSM differences identified in the present study are the result of variation in the quality of the surgery or pathology review across sites.
The Prostate Cancer Registry does not collect the site and number of PSMs so we were unable to report on the location of the PSM. There is evidence that the prostate apex is the most common site of PSMs and that postolateral margins place the patient most at risk of disease recurrence ; however, Grossfeld et al.  found that the number of PSMs and their location had no significant impact on PSA recurrence nor on non-adjuvant secondary treatment. Stephenson et al.  found that neither the number of PSMs nor the location had enhanced capacity to predict biochemical recurrence over a simple model containing only whether the patient had positive or negative margins. In addition, the PCR does not collect details of the patient's body mass index, which has also been shown to be associated with PSMs .
As discussed earlier, another known limitation of the present study is that, while the registry collects details of surgeon volume in contributing sites, it does not assess surgeon volume at non-participating sites, so the impact of the individual surgeon volume–quality relationship on PSMs and subsequent treatment could not be assessed.
Finally, as this was an observational study and not a randomized controlled clinical trial, we cannot exclude other patient variables, which were not collected but might account for differences in PSM prevalence and subsequent treatments.
In conclusion, patients who have undergone RP and who have a PSM are significantly more likely to receive additional treatment than those with clear surgical margins. Being treated at a private hospital using a robot-assisted technique appears to be associated with a lower risk of PSMs, even after taking into account the patient's age and disease stage. This needs further investigation as it may simply reflect surgeon experience and case selection. Any effort to reduce the frequency of PSMs will undoubtedly also result in better outcomes, both in terms of patient psychological wellbeing and financial burden to the health system.
This project was funded by Cancer Australia (Project ID1010384). The authors would like to acknowledge the support provided to the Prostate Cancer Registry by all contributing urologists, radiation oncologists and medical oncologists in contributing hospitals.
S.E. and J.M. have received grants from Cancer Australia, Prostate Cancer Foundation of Australia, Australian Department of Health and Ageing, Victorian Department of Health, the Victorian Cancer Agency and the National Health and Medical Research Council. J.D.'s position is funded by a grant from Cancer Australia.
positive surgical margin
incidence rate ratio
androgen deprivation therapy
National Comprehensive Cancer Network