Clinical outcomes for men with positive surgical margins after radical prostatectomy—results from the South Australian Prostate Cancer Clinical Outcomes Collaborative community-based registry

Abstract

Objective

Positive surgical margins (PSMs) after radical prostatectomy (RP) indicate failure of surgery to completely clear cancer. PSMs confer an increased risk of biochemical recurrence (BCR), but how more robust outcomes are affected is unclear. This study investigated factors associated with PSMs following RP and determined their impact on clinical outcomes (BCR, second treatment [radiotherapy and/or androgen deprivation therapy], and prostate cancer-specific mortality [PCSM]).

Methods

The study cohort included men diagnosed with prostate cancer (pT2-3b/N0/M0) between January 1998 and June 2016 who underwent RP from the South Australian Prostate Cancer Clinical Outcomes Collaborative database. Factors associated with risk of PSMs were identified using Poisson regression. The impact of PSMs on clinical outcomes (BCR, second treatment, and PCSM) was assessed using competing risk regression.

Results

Of the 2827 eligible participants, 28% had PSMs—10% apical, 6% bladder neck, 17% posterolateral, and 5% at multiple locations. Median follow-up was 9.6 years with 81 deaths from prostate cancer recorded. Likelihood of PSM increased with higher pathological grade and pathological tumor stage, and greater tumour volume, but decreased with increasing surgeon volume (odds ratio [OR]: 0.93; 95% confidence interval [CI]: 0.88–0.98, per 100 previous prostatectomies). PSMs were associated with increased risk of BCR (adjusted sub-distribution hazard ratio [sHR] 2.5; 95% CI 2.1–3.1) and second treatment (sHR 2.9; 95% CI 2.4–3.5). Risk of BCR was increased similarly for each PSM location, but was higher for multiple margin sites. We found no association between PSMs and PCSM.

Conclusion

Our findings support previous research suggesting that PSMs are not independently associated with PCSM despite strong association with BCR. Reducing PSM rates remains an important objective, given the higher likelihood of secondary treatment with associated comorbidities.

1. Introduction

Positive surgical margins (PSMs) are common after radical prostatectomy (RP), with rates reported in the range of 13%–35% [[1], [2], [3], [4], [5], [6], [7], [8]]. There is a general agreement that PSMs are pathologically defined as tumour cells abutting the inked surgical margin of the specimen [9]. PSMs can occur at single or multiple sites, in different locations across the prostate, and can be intra- or extra-prostatic.

Clinical and pathological variables are strongly associated with disease aggressiveness, but not solely associated with PSMs [[9], [10], [11], [12], [13], [14]]. Other factors impacting on likelihood of PSMs include surgeon experience [11,[15], [16], [17], [18]], surgical approach [10,15], hospital setting, and associated hospital volume [10,19]. Pre-planning technologies and intraoperative approaches may also impact on PSM rates. For example, increased use of multi-parametric magnetic resonance imaging fusion biopsy methods, which are becoming the standard for diagnostic and pre-treatment workup, provides better staging information to inform surgical planning [20]. Newer methods such as three-dimensional printing and intraoperative frozen section analysis also have potential to reduce PSM rates [21,22]. Definitive evidence from randomised trials that these technologies reduce the risk of PSM is still required.

Importantly, a PSM can be an independent trigger for radiotherapy as an adjunct treatment, with added treatment morbidity for the patient [23]. Studies have reported two- to five-fold increased risk of biochemical recurrence (BCR) in men with positive margins compared with those with negative margins [3,6,[12], [13], [14]]. How the more robust clinical endpoints are affected by a PSM is more controversial. While a few studies have reported that PSMs were associated with an increased risk of development of metastases [24] and prostate cancer-specific mortality (PCSM) [25,26], many have not found an effect on these hard clinical endpoints [11].

The objective of this study was to examine the cancer-specific factors associated with PSMs, and to compare outcomes according to sites of PSM in a population-based community cohort of men from South Australia who had undergone RP.

2. Methods

2.1. Data collection

The study cohort included all men in the South Australian Prostate Cancer Clinical Outcomes Collaborative (SA-PCCOC) database diagnosed with clinically localised or locally advanced prostate cancer (PCa), from January 1998 to June 2016, who were resident in South Australia and underwent RP as their primary treatment. SA-PCCOC prospectively enrolls patients diagnosed with PCa within the state of South Australia (population 1.7 million). The database includes patients from the three public tertiary hospitals in South Australia diagnosed from January 1998, with enrolment from private practice from January 2008, and currently recruits approximately 90% of newly diagnosed PCa cases in South Australia. Men who underwent salvage RP following radiotherapy, or who were clinically assessed or pathologically confirmed with T4, node-positive, or metastatic disease at diagnosis were ineligible.

Data on margin status and location were collected through ongoing review of pathology reports. Margin status was grouped as negative or positive, subdivided into single or multiple sites, and categorised by location—apex, bladder neck, or posterolateral (the latter category includes margins at all other sites other than apex or bladder neck). Since Gleason score at the margin is not recorded and length of positive margin only recently recorded in the SA-PCCOC database, these margin characteristics could not be examined.

Outcomes assessed in this study were BCR, subsequent treatment (any second treatment and salvage therapy), and PCSM. BCR after surgery was defined as a detectable prostate-specific antigen (PSA) level of ≥0.2 ng/mL with a second confirmatory level of ≥0.2 ng/mL [27]. Date and cause of death were determined via linkage with state and national death registers. “Any secondary treatment” was defined as receipt of radiotherapy or androgen deprivation therapy commencing at any time after the date of surgery.

Data were also extracted for the following potential confounders: age at diagnosis, National Comprehensive Cancer Network diagnostic risk category, pre-treatment PSA, number of positive cores at diagnosis, final Gleason score of prostatectomy specimen, pathological stage, tumour volume at pathological assessment (assessed via Chen et al.’s method [28]), and whether nerve-sparing surgery was undertaken (any vs. none). An indicator of surgeon volume was provided by SA-PCCOC, based on the cumulative number of prostatectomies performed by the relevant urologist prior to the date of an individual's surgery. Other surgical and hospital measures were not available due to potential attribute disclosure, given the small number of urologists and institutes undertaking prostatectomies within the South Australia.

2.2. Statistical analysis

Factors associated with margin status, location, and multiplicity were identified using multivariable logistic regression, with mutual adjustment for other clinical and demographic characteristics. The impact of PSMs on clinical outcomes (BCR, subsequent treatment, and PCSM) was examined using multivariable competing risk regression analyses (Fine and Gray [29]), with death from other causes as the competing risk. Survival time for BCR, any second treatment, and PCSM were calculated from the date of surgery until the date of the event of interest, date of death, or censoring date (December 31, 2019), which ever occurred first. Multivariable models were adjusted for age (<60 years old, 60–69 years old, and 70–79 years old), period of surgery (<2007, 2008–2010, 2011–2013, and 2014–2016), pre-treatment PSA (<5 ng/mL, ≤5–10 ng/mL, ≤10–20 ng/mL, and >20 ng/mL), pathological tumor stage (pT2, pT3a, and pT3b), and pathological grade (grouped as International Society of Urological Pathology grades 1–5). Separate models were run for surgical margin status categorised as: i) positive or negative, ii) negative, single or multiple sites, and iii) each positive margin location (apex, bladder neck, and posterolateral), mutually adjusted for any other positive sites. To explore whether margin status conferred different levels of risk according to extent of disease, we also undertook subgroup analysis for BCR within pathological stage groups, with similar adjustment for covariates.

Statistical analyses used Stata v.14.1 (StataCorp, College Station, TX, USA). Ninety-five percent confidence intervals (CIs) are reported throughout.

2.3. Ethical review

The SA-PCCOC database has been reviewed and approved by the Southern Adelaide Clinical Human Research Ethics Committee (protocol 307.14). The approval includes permission to use an opt-out consent process, consistent with the National Statement on Ethical Conduct in Human Research. Ethics approval for this project was also provided by South Australia Human Research Ethics Committees (protocol 3746).

3. Results

Of the 9481 men recruited to SA-PCCOC during the study period, 3091 underwent RP as primary treatment. Of the men, 251 were ineligible and 13 were missing data on margin status or date of surgery, leaving a total of 2827 for analysis. Demographic and clinical characteristic of the study participants are presented in Table 1, for the whole cohort and by surgical margin status. PSMs were reported in 28% of men in the cohort. Posterolateral site was the most common (17% of men in the total cohort), followed by apical margins (10%), and then bladder neck margins (6%). The pattern of location of PSMs was consistent across all stages. The exception was for pT3b disease, where bladder neck and apical margins were similar in percentage terms. Margins were at a single site in 23% of men undergoing RP, and at multiple sites in 5%. Single site PSM rates increased with pT, occurring in 16% of pT2, 29% of pT3a, and 37% of pT3b cases. Similarly, PSM rates at multiple sites increased with higher pT, occurring in 2% of pT2 cases, 6% of pT3a cases, and 16% of pT3b cases.

Table 1.

Cohort profile by margin status^a.

Characteristic	Margin status		Margin locationb			Multiplicity
	Negative	Positive	Apical	Bladder neck	Posterolateral	Single site	Multi-sites
Total, n (%)	2030 (72)	797 (28)	289 (10)	172 (6)	470 (17)	652 (23)	145 (5)
NCCN risk category at Dx, n (%)c
Low	568 (78)	159 (22)	57 (8)	30 (4)	85 (8)	142 (18)	17 (2)
Intermediate	1232 (71)	493 (29)	189 (12)	87 (5)	314 (19)	391 (24)	102 (6)
High	230 (61)	145 (39)	43 (11)	55 (15)	71 (19)	119 (32)	26 (7)
Gleason score at RP, n (%)c
≤6	502 (80)	126 (20)	52 (8)	19 (3)	60 (10)	115 (18)	11 (2)
3+4	932 (72)	356 (28)	133 (10)	58 (5)	215 (17)	296 (23)	60 (5)
4+3	416 (68)	199 (32)	61 (10)	52 (8)	127 (21)	157 (26)	42 (7)
8	103 (67)	51 (33)	19 (12)	18 (12)	31 (20)	35 (23)	16 (10)
9–10	70 (52)	64 (48)	24 (18)	24 (18)	37 (28)	48 (36)	16 (12)
Missing	7	1	0	0	0	1	0
Pathological stage, n (%)c
pT2	1163 (82)	257 (18)	105 (7)	38 (3)	143 (10)	223 (16)	34 (2)
pT3a	743 (65)	396 (35)	130 (11)	84 (7)	241 (21)	329 (29)	67 (6)
pT3b	124 (46)	144 (54)	54 (20)	50 (19)	86 (32)	100 (37)	44 (16)
Age at Dx, mean±SD, year	62 ± 7	63 ± 7	63 ± 7	63 ± 6	62 ± 7	63 ± 7	63 ± 6
Pre-Rx PSA, median (IQR), ng/mL	7 (5–9)	8 (6–11)	8 (6–12)	9 (6–13)	8 (6–11)	8 (6–11)	9 (6–13)
Number of positive margin status core, median (IQR)	3 (2–6)	4 (2–7)	5 (3–7)	5 (3–7)	4 (2–7)	4 (2–6)	5 (3–8)
Tumour volume, median (IQR), cm3	2.3 (1.2–4.2)	4.3 (2.3–8.3)	4.6 (2.8–8.5)	8.0 (3.3–12.7)	4.0 (2.0–7.9)	4.0 (2.2–7.7)	7.6 (3.4–12.5)
Nerve-sparing surgery, n (%)c	1086 (53)	420 (53)	150 (10)	71 (5)	256 (17)	354 (24)	66 (4)

NCCN, National Comprehensive Cancer Network; Dx, diagnosis; IQR, interquartile range; Pre-Rx PSA, pre-treatment prostate-specific antigen levels; RP, radical prostatectomy; SD, standard deviation; pT, pathological tumor stage; pN, pathological lymph node stage.

^aRP cases among men from South Australia with non-metastatic prostate cancer between January 1998 and December 2016, excluding N1, M1, pT4, and pN1, n=2827.

^bLocation includes cases with multifocal positive margins.

^cThe percentages in columns 4–8 are only for percentages with positive margins.

3.1. Factors associated with margin status and site

Table 2 shows results of multivariable analyses of factors associated with margin status, for any PSM, specific PSM sites, and multiple PSM sites. Men with a greater tumour volume had higher likelihood of any PSM (odds ratio [OR] 2.9, 95% CI 2.0–4.0), as did those with higher pT (pT3a: OR 2.3, 95% CI 1.9–2.8; pT3b OR 3.4, 95% CI 2.5–4.7; compared to pT2) and higher RP Gleason score 9–10 (OR 1.7, 95% CI 1.0–2.7; compared to Gleason score of ≤6). Higher surgeon volume was protective (OR 0.93, 95% CI 0.88–0.98 per 100 previous prostatectomies) and risk of PSMs was lower for surgeries conducted during 2011–2013 compared with before 2008 (OR 0.6, 95% CI 0.5–0.8). No difference was found in relation to nerve-sparing surgery.

Table 2.

Factors associated with any PSM, specific PSM sites, and multiple PSM sites derived from multivariable logistic regression models (1998–2016).

Factor	Any PSM			Apex			Bladder neck			Posterolateral			Multiple PSM sites
	OR	95% CI	p-Value	OR	95% CI	p-Value	OR	95% CI	p-Value	OR	95% CI	p-Value	OR	95% CI	p-Value
Age at diagnosis, year
<60	1.0	Reference	0.76	1.0	Reference	0.61	1.0	Reference	0.94	1.00	Reference	0.09	1.0	Reference	0.41
60–69	0.9	0.8–1.1	0.46	1.1	0.8–1.4	0.59	0.9	0.6–1.4	0.72	0.8	0.6–1.0	0.08	0.9	0.6–1.4	0.63
70–79	1.0	0.7–1.3	0.75	1.1	0.7–1.7	0.66	1.0	0.6–1.7	0.93	0.7	0.5–1.1	0.10	0.7	0.4–1.4	0.34
Treatment period
≤2007	1.0	Reference	0.17	1.0	Reference	0.22	1.0	Reference	0.002	1.0	Reference	0.020	1.0	Reference	0.97
2008–2010	1.0	0.8–1.2	0.79	0.7	0.5–1.0	0.06	2.0	1.1–3.5	0.020	1.0	0.8–1.4	0.810	1.0	0.6–1.7	0.98
2011–2013	0.6	0.5–0.8	0.01	0.7	0.4–1.1	0.10	2.6	1.4–4.8	0.003	0.5	0.3–0.7	<0.001	0.6	0.3–1.1	0.09
2014–2016	0.8	0.6–1.2	0.31	0.7	0.4–1.1	0.15	3.1	1.7–5.9	0.001	0.7	0.5–1.1	0.140	1.2	0.6–2.1	0.63
Pre-treatment PSA, ng/mL
<5	1.0	Reference	0.39	1.0	Reference	0.20	1.0	Reference	0.89	1.0	Reference	0.78	1.0	Reference	0.73
≥5–<10	0.9	0.7–1.2	0.63	1.1	0.7–1.5	0.79	0.6	0.4–0.9	0.02	1.0	0.7–1.3	0.86	0.7	0.4–1.1	0.14
≥10–20	1.3	0.9–1.7	0.08	1.7	1.1–2.6	0.02	0.9	0.5–1.5	0.73	1.2	0.9–1.7	0.25	1.4	0.8–2.5	0.22
≥20	1.3	0.8–2.3	0.31	1.4	0.8–2.3	0.19	1.2	0.5–2.6	0.68	0.9	0.5–1.7	0.67	1.2	0.5–2.9	0.74
RP Gleason score
≤6	1.0	Reference	0.08	1.0	Reference	0.35	1.0	Reference	0.03	1.0	Reference	0.010	1.0	Reference	0.010
3＋4	1.3	1.0–1.7	0.05	1.3	0.9–1.8	0.23	0.8	0.4–1.3	0.33	1.8	1.3–2.6	<0.001	2.8	1.4–5.7	0.010
4＋3	1.4	1.0–1.9	0.05	1.0	0.7–1.6	0.83	1.2	0.6–2.2	0.63	2.1	1.4–3.1	<0.001	3.6	1.7–7.6	0.001
8	1.5	0.9–2.3	0.10	1.3	0.7–2.5	0.36	1.7	0.8–3.7	0.17	2.1	1.2–3.5	0.006	4.9	2.0–11.4	<0.001
9–10	1.7	1.0–2.7	0.02	1.6	0.8–2.9	0.16	1.4	0.7–3.1	0.35	2.2	1.3–3.9	0.003	3.4	1.4–8.5	0.010
pT
2	1.0	Reference	<0.001	1.0	Reference	0.01	1.0	Reference	<0.001	1.0	Reference	<0.001	1.0	Reference	<0.001
3a	2.3	1.9–2.8	<0.001	1.4	1.0–1.9	0.03	1.9	1.2–3.0	0.004	2.5	2.0–3.2	<0.001	2.1	1.3–3.4	0.001
3b	3.4	2.5–4.7	<0.001	1.9	1.2–2.8	0.003	3.4	2.0–5.8	<0.001	3.4	2.4–4.9	<0.001	4.0	2.3–7.0	<0.001
Tumour volume, per 10 cm3
Q1 (≤1.4)	1.0	Reference	<0.001	1.0	Reference	<0.001	1.0	Reference	<0.001	1.0	Reference	<0.001	1.0	Reference	<0.001
Q2 (>1.4–2.7)	1.3	0.9–1.9	0.10	2.3	1.2–4.5	0.01	2.1	0.8–5.6	0.14	1.2	0.8–1.7	0.47	2.3	0.7–7.2	0.170
Q3 (>2.7–5.4)	1.7	1.2–2.5	<0.001	3.7	2.2–7.5	<0.001	3.0	1.2–7.6	0.02	0.9	0.6–1.4	0.62	1.7	0.5–5.5	0.380
Q4 (>5.4)	2.9	2.0–4.0	<0.001	5.3	2.8–9.9	<0.001	8.9	3.7–20.4	<0.001	1.5	1.0–2.3	0.05	7.8	2.5–20.6	<0.001
Surgeon volumea	0.93	0.88–0.98	0.004	0.96	0.89–1.03	0.26	0.90	0.82–0.97	0.010	0.91	0.87–0.97	0.003	0.85	0.76–0.95	0.003
Nerve-sparing surgery	0.9	0.7–1.2	0.45	1.1	0.7–1.5	0.78	0.7	0.4–1.0	0.070	0.9	0.6–1.1	0.32	0.7	0.4–1.1	0.110

PSM, positive surgical margin; OR, odds ratio; RP, radical prostatectomy; Q, quartile; CI, confidence interval; PSA, prostate-specific antigen; pT, pathological tumor stage.

Note: the “Reference” indicates the reference group for each variable considered.

^aPer 100 previous prostatectomies.

For specific PSM sites, men with pT3b compared with pT2 disease had increased PSM risk at all locations, with the highest risk for PSM at the bladder neck (OR 3.4, 95% CI 2.0–5.8) and posterolateral sites (OR 3.4, 95% CI 2.4–4.9). Those with higher Gleason score (≥4+3) had increased risk of PSMs at posterolateral sites, but not at the apex and bladder neck. Tumour volume was associated with increased risk of PSMs at the apex (OR 5.3, 95% CI 2.8–9.9) and bladder neck (OR 8.9, 95% CI 3.7–20.4). Higher surgeon volume was associated with decreased risk of bladder neck (OR 0.90, 95% CI 0.82–0.97 per 100 previous prostatectomies) and posterolateral (OR 0.91, 95% CI 0.87–0.97 per 100 previous prostatectomies) margins. Likelihood of PSMs at the bladder neck had increased during the most recent period, while it decreased at posterolateral sites. No differences were seen in relation to nerve-sparing approach.

Clinical factors found to be associated with multiple PSMs included pathological grade (RP Gleason score 4+3: OR 3.6, 95% CI 1.7–7.6; RP Gleason score 8: OR 4.9, 95% CI 2.0–11.4; compared with RP Gleason score 6), pT (pT3a: OR 2.1, 95% CI 1.3–3.4; pT3b: OR 4.0, 95% CI 2.3–7.0; compared with pT2), and higher tumour volume (OR 7.8, 95% CI 2.5–20.6). Surgeon volume was protective for multiple PSM sites (OR 0.85, 95% CI 0.76–0.95). Nerve-sparing surgery were not associated with multiple PSM sites.

3.2. Clinical outcomes associated with margin status

Median follow-up time was 9.6 (interquartile range 6.4–12.1) years. Among the 2640 men with follow-up PSA measures, crude five-year cumulative incidence of BCR was 21%, and 40% for positive compared with 13% for negative margins. The proportion receiving any secondary treatment was 17%, 31% and 10% for positive and negative margins, respectively. Over the total period of follow-up, 81 (2.87%) men had died from PCa and 196 (6.93%) had died from other causes. Five-year cumulative incidence of PCa death was 0.5% overall, 1% in men with PSMs and 0.3% in those with negative margins. Cumulative incidence plots for BCR, second treatment, and PCSM by surgical margin status (any positive vs. negative) are shown in Fig. 1.

Figure 1.

Cumulative incidence plots with 95% confidence intervals by surgical margin status (any positive vs. negative). (A) Cumulative incidence of biochemical recurrence; (B) Cumulative incidence of second treatment; (C) Cumulative incidence of prostate cancer death.

Table 3 presents results of multivariable competing risk regression analyses for BCR, second treatment, and PCa mortality. Risk of BCR was higher for positive margins compared with negative margins (adjusted sub-distribution hazard ratio [sHR] 2.5, 95% CI 2.1–3.1). A single PSM site presented less of a risk for BCR (sHR 2.3, 95% CI 1.9–2.8) than multiple PSM sites (sHR 3.6, 95% CI 2.6–4.8). Risk of BCR was increased for PSMs at all locations, including PSMs at the bladder neck (sHR 1.9, 95% CI 1.4–2.5), apex (sHR 1.7, 95% CI 1.3–2.2), and posterolateral sites (sHR 2.1, 95% CI 1.7–2.5), compared to negative margins.

Table 3.

Risk of BCR, second treatment, and PCa mortality, associated with any PSM, PSM site, and multiplicity of sites, derived from competing risk regression models.

Margin statusa	BCRb			Any secondary treatmentc			PCa deathd
	sHR	95% CI	p-Value	sHR	95% CI	p-Value	sHR	95% CI	p-Value
Any PSM vs. negative margin	2.5	2.1–3.1	<0.001	2.9	2.4–3.5	<0.001	1.3	0.8–2.0	0.18
PSM at apex vs. all other casese	1.7	1.3–2.2	<0.001	2.0	1.6–2.5	<0.001	1.3	0.7–2.4	0.39
PSM at bladder neck vs. all other casese	1.9	1.4–2.5	<0.001	1.8	1.3–2.5	<0.001	1.7	0.9–3.5	0.11
PSM at posterolateral vs. all other casese	2.1	1.7–2.5	<0.001	2.3	1.9–2.8	<0.001	0.9	0.6–1.6	0.83
Single PSM site vs. negative	2.3	1.9–2.8	<0.001	2.6	2.1–3.1	<0.001	1.4	0.9–2.2	0.25
Multiple PSM sites vs. negative	3.6	2.6–4.8	<0.001	4.7	3.5–6.3	<0.001	1.2	0.5–2.9	0.66
Intra-prostatic PSM vs. negativef	4.5	3.2–6.3	<0.001	4.8	3.3–7.0	<0.001	0.8	0.3–2.2	0.59

CI, confidence interval; BCR, biochemical recurrence; PCa, prostate cancer; PSM, positive surgical margin; pT, pathological tumor stage.

Note: sHR is the adjusted sub-distribution hazard ratio from competing risk regression models adjusted for age, treatment year, pre-treatment prostate-specific antigen levels, radical prostatectomy Gleason score, and pT.

^aNumber of events per total cases in model (number with competing risk).

^bBCR assessed among men with prostate-specific antigen follow-up recorded, 615 BCR/2640 cases (172 other deaths).

^c551 treated/2818 cases (190 deaths).

^d81 PCa deaths/2818 cases (196 other deaths).

^eMutual adjustment for other margin locations.

^fIntra-prostatic PSMs defined as any PSM in men with pT2 disease, models restricted to pT2 cases.

Risk of any second treatment was significantly higher for positive compared with negative margins (sHR 2.9, 95% CI 2.4–3.5). Multiple PSM sites conferred a greater risk (sHR 4.7, 95% CI 3.5–6.3) of a second treatment than single PSM site (sHR 2.6, 95% CI 2.1–3.1) compared with negative margins; and risk of second treatment was equally increased across each of the specific margin sites. In contrast, no association was found between margin status, location or multiplicity, and PCa death in multivariable analysis.

The presence of intra-prostatic positive margins (PSM among men with pT2 disease) was associated with increased risk of BCR (sHR 4.5, 95% CI 3.2–6.3) and secondary treatment (sHR 4.8, 95% CI 3.3–7.0) compared to negative margins with pT2 disease (Table 3). The effect of PSMs on risk of BCR decreased with increasing stage, as indicated in Table 4. No association was observed in relation to intra-prostatic margins and risk of PCa death.

Table 4.

Risk of biochemical recurrence by margin status and pT derived from competing risk regression models.

Margin status and pT	Relative to pT2 clear margin			PSM vs. clear margina
	sHR	95% CI	p-Value	sHR	95% CI	p-Value
pT2
Clear margins	1.0	Reference	–	1.0	Reference	–
Positive margin	4.3	3.2–5.9	<0.001	4.6	3.2–6.2	<0.001
pT3a
Clear margin	2.2	1.7–2.9	<0.001	1.0	Reference	–
Positive margin	5.5	4.2–7.2	<0.001	2.6	2.0–3.3	<0.001
pT3b
Clear margin	4.9	3.4–7.0	<0.001	1.0	Reference	–
Positive margin	7.1	5.0–10.0	<0.001	1.4	1.0–2.0	0.06

CI, confidence interval; sHR, sub-hazard ratio; pT, pathological tumor stage; PSM, positive surgical margin; –, not available.

Note: the “Reference” indicates that this category is the reference category.

^aStratified by stage category.

4. Discussion

Our study reported clinically important outcomes including BCR, additional treatment, and mortality according to surgical margin status, using data from a prospective population-based cohort of men undergoing RP with long-term follow-up, providing an evidence base for discussion by community urologists. Our findings showed a strong positive association between PSMs and risk of BCR following RP, and similarly between PSMs and time to second treatment, but we found no association between PSMs and risk of PCa death. While the latter may be due to the low number of PCa deaths among men selected for RP, strong associations were seen for other tumour characteristic (PSA, pathological grade, and stage). This suggested that margin status is not independently associated with risk of PCa death, despite its strong independent association with BCR.

In our cohort, posterolateral PSMs were the most common margin site and bladder neck PSMs the least common margin site. These findings are not entirely consistent with previous literature [1,5]. For example, Smith et al. [5] reported that the apex was the most common location of PSMs in both robotic and open retro-pubic RP, and Patel et al. [1] found that apex was the most common PSM site followed by posterolateral in their robotic RP series.

The association between PSMs and other adverse pathology is consistent with previous reports [6,7,11]. PSMs are more frequently reported in association with higher clinical and higher pathological stage, higher Gleason score [6,7,11], and increased number of positive cores [14]. Likewise, our finding of decreasing risk of PSMs with increasing surgical volume is consistent with previous reports indicating an inverse association for open surgery [21], robotic surgery [15,18,30], different pTs [15], single [15,31] and multiple surgeon series [14], and from large databases [8]. The fact that surgeon volume was frequently found to be inversely associated with PSM rates indicated that factors other than disease severity can influence PMS rates [6].

While there was a trend toward decreased apical and posterolateral PSMs over the study period, we noted an increase in the likelihood of PSM at the bladder neck. More widespread use of bladder neck preserving surgery to improve continence outcomes may account for the upward trend in PSMs at the bladder neck [32]. On the other hand, increasing use of multi-parametric magnetic resonance imaging in diagnostic and pre-surgery work was likely to have led to a reduction in PSM rates in more recent periods, through more accurate assessment of extra-capsular disease to inform surgical planning [20]. Other techniques such as three-dimensional printed models for intraoperative guidance [21,33] and intraoperative frozen section analysis allowing real-time histological assessment during robotic RP [22,34] offer further potential to reduce PSM rates. However, neither three-dimensional visualisation nor intraoperative frozen section analysis method has been used in the South Australian region. Use of multi-parametric magnetic resonance imaging fusion biopsy only becomes part of routine practice in the latter period of this study (2014 onwards).

While any PSM was associated with increased risk of BCR, we found little difference according to margin site. Evidence for the relative strength of effects for different margin locations is inconsistent. Several studies reported that apical margins were more important predictors of BCR than posterolateral margins [35,36]. However, O'Neil et al. [37] reported that PSMs at the bladder neck were more likely to be associated with BCR than other locations, while Roder et al. [38] reported that non-apical margins were a stronger predictor of BCR in pT2 patients. Others have reported no independent effect of margin location [7]. The reason for these differences is unclear. Overall, it may be that the sub-classifications of margin location are of limited utility compared to the presence or absence of any PSM.

Our findings of substantially increased risk of BCR among those with multiple PSMs were consistent with reports indicating that PSMs at multiple sites were a more powerful predictor of recurrence than unifocal margins [35,36,39]. Likewise, having more extensive (≥3 mm) margins has also been reported to increase risk of BCR compared to negative surgical margin [2,34,[40], [41], [42]], even in pT2 disease [43]. In contrast, Huang et al. [43] found that no specific margin derived variable (focality, linear length, diathermy artefact, or plane of tumour) consistently predicted risk of BCR, even though overall margin status (positive or negative) was a predictor. The stronger effects for PSM and BCR in men with pT2 diseases than among men with pT3a or pT3b disease suggested that PSMs may have greater relative impact on oncological outcomes in the absence of other major clinical risk factors. Other studies have also reported a more substantial impact of PSMs in patients with less adverse pathology [7,44].

Our findings also showed a positive association between PSMs and risk of secondary treatment, as reported by others [45]. While some men who had PSMs may have received adjuvant radiotherapy for other indications (e.g., seminal vesical invasion or extracapsular extension), our data indicated an independent effect of margin status on receipt of second treatment, with similar findings according to location and multiple sites.

While there is consistent evidence of increase risk of BCR with PSMs and secondary treatment, there is considerable uncertainty in relation to whether PSMs are associated with clinical endpoints such as development of metastatic or castration resistant disease, PCa death, or overall survival [46]. Our findings did not suggest any strong association between PSMs and the most definitive clinical endpoint—PCa death. However, we observed differences in PCSM according to pathological grade and stage (Supplementary Table 1) indicating sufficient power regarding our sample size to detect differences. Part of the uncertainty in the literature over the clinical implications may relate to variations in definitions and reporting of PSMs, pathologist's experience and technique, patient risk factors, length of follow-up, and presence of other pathological risk modifiers. For example, some reported PSMs may arise from inexperience handling of the specimen rather than being true PSMs [[47], [48], [49]]. However, these issues are unlikely to completely explain the lack of association with PCSM. As Yossepowitch et al. [11] concluded in their review, it was likely that the “long-term impact of PSM on survival is largely influenced by other risk modifiers”. Despite the lack of association with PCSM, we observed a higher rate of initiation of secondary treatment in patients with PSMs. Therefore, minimising the rate of PSMs to avoid the comorbidity associated with additional treatments for PCa remains an important goal of surgery [50].

Questions have been raised about the utility of PSMs in measuring surgical competence or quality compared to clinical outcome measures such as BCR rates, since not all patients with a PSM developed BCR [51]. However, intra-prostatic margins (pT2 with PSM) have been considered a measure of surgical competence [8,52]. Furthermore, the stronger relative effect of PSMs on risk of BCR in men with pT2 disease compared with pT3 disease and the increased risk of secondary treatments reinforce the case for including intra-prostatic margins as a quality indicator.

4.1. Limitations and strengths

Data on additional margin characteristics (Gleason grade at the margin and length of margin) were unavailable, which is an important limitation given these characteristics are likely to have a major impact on clinical outcomes. In addition, we were unable to differentiate multiple margins if they occurred at the same site, e.g., multifocal margins at the base would be classified as a single site. In addition, the database does not record metastatic disease as an endpoint; therefore, we used time to second treatment as a surrogate endpoint. While this measure does not indicate development of metastatic disease, initiation of a second treatment any time after RP is likely to have major consequences on functional outcomes and subsequent comorbidities. Therefore, subsequent treatment represents a meaningful outcome measure from a patient perspective. Despite having median follow-up of nearly 10 years, assessing differences in risk of PCSM may require an even longer period of follow-up.

Another limitation is that several changes in grading and staging classification for PCa have occurred during the study period. For example, in the 7th edition of the American Joint Committee on Cancer TNM staging schema (2010), microscopic bladder invasion was reclassified as T3a, rather than T4. Likewise, major changes in grading PCa have occurred with revisions in 2005 and again in 2014 to International Society of Urological Pathology grade groups. Consequently, categories for pathological stage and grade groups were not consistent across the entire study period. While we have attempted to minimise the impact of these changes by excluding all men with pT4 disease and by including treatment year in all our models, these measures may not have adequately accounted for grade or stage shifts throughout the study period. In addition, there was no central review of pathology. Our study included pathological assessments from multiple pathologists, not all uro-pathology specialists, across several different pathology centres. Therefore, differences in technique and interpretation were likely to lead to variation in assessment of margin status. However, we believe that it is important to report on real-world practice given that this variability directly influences post-surgical monitoring and decisions about further treatment among the broader community of clinicians.

Regarding strengths of this study, ascertainment of date and cause of death are likely to be accurate given linkages to state and national death registrars, and verification through the central cancer register. Other strengths include the large sample size, and multi-institutional nature of SA-PCCOC, which offers a “real world” perspective.

5. Conclusion

This large multi-institutional, community-wide study shows that PSMs following RP are associated with both extent of disease and surgeon experience. The presence of any PSM was found to be an independent risk factor for BCR and secondary treatment. While risk of BCR is equivalent across different PSM sites, multiple PSM sites pose a greater risk. Despite the higher risk of BCR, PSMs do not appear to be associated with increased risk of PCSM. Given the higher likelihood of secondary treatment with its associated comorbidities, reducing PSM rates remains an important objective. Further development of pre-treatment and intraoperative technologies to improve surgical precision is likely to lead to lower PSM rates.

Author contributions

Study concept and design: Nicholas R. Brook, Kerri R. Beckmann, Michael E. O'Callaghan, Kim L. Moretti.

Data acquisition: Michael E. O'Callaghan.

Data analysis: Kerri R. Beckmann, Andrew D. Vincent.

Drafting of manuscript: Kerri R. Beckmann.

Critical revision of the manuscript: Michael E. O'Callaghan, Andrew D. Vincent, Kim L. Moretti, Nicholas R. Brook.

Conflicts of interest

The authors declare no conflict of interest.

Appendix A. Supplementary data

The following is the Supplementary data to this article: