Surgical margin status and its impact on prostate cancer prognosis after radical prostatectomy: a meta-analysis

Background and purpose Positive surgical margins (PSMs) correlate with adverse outcomes in numerous solid tumours. However, the prognostic value of PSMs in prostate cancer (PCa) patients who underwent radical prostatectomy remains unclear. Herein, we performed a meta-analysis to evaluate the association between PSMs and the prognostic value for biochemical recurrence-free survival (BRFS), cancer-specific survival (CSS), overall survival (OS), cancer-specific mortality (CSM) and overall mortality (OM) in PCa patients. Materials and methods According to the PRISMA statement, online databases PubMed, EMBASE and Web of Science were searched to identify relevant studies published prior to February 2018. The hazard ratios (HRs) and 95% confidence intervals (95% CIs) were calculated to evaluate the relationship between PSMs and PCa. Results Ultimately, 32 cohort studies that met the eligibility criteria and involved 141,222 patients (51–65,633 per study) were included in this meta-analysis. The results showed that PSMs were significantly predictive of poorer BRFS (HR = 1.35, 95% CI 1.28–1.48, p < 0.001), CSS (HR = 1.49, 95% CI 1.16–1.90, p = 0.001) and OS (HR = 1.11, 95% CI 1.02–1.20, p = 0.014). In addition, PSMs were significantly associated with higher risk of CSM (HR = 1.23, 95% CI 1.16–1.30, p < 0.001) and OM (HR = 1.09, 95% CI 1.02–1.16, p = 0.009) in patients with PCa. Conclusions Our study suggests that the presence of a histopathologic PSM is associated with the clinical outcomes BRFS, CSS, OS, CSM and OM in patients with PCa, and PSMs could serve as a poor prognostic factor for patients with PCa. Electronic supplementary material The online version of this article (10.1007/s00345-018-2333-4) contains supplementary material, which is available to authorized users.


Introduction
In 2016, prostate cancer (PCa) was the most common newly diagnosed cancer in males, with 1.6 million new cases per year, and 26,730 men died from PCa, which was the third leading cause of cancer death in males [1]. With the wide use of prostate-specific antigen (PSA) screening and increased public awareness of PCa, 90% of patients are being diagnosed with localised PCa [2]. Despite effective treatments with curative intent such as radical prostatectomy (RP), up to 30% of patients will experience biochemical recurrence (BCR), of which 20-30% will progress to clinical metastasis or death [3]. To date, there have been a number of studies performed to identify histological parameters associated with prognostic outcomes after RP, which might lead to more informative prognostic information in patient monitoring.
A positive surgical margin (PSM) is determined by the stained areas of soft tissue on the RP specimen. The incidence of PSMs is influenced by the presence of extra-prostatic extension, with a rate that ranges from 10 to 48% [4]. Despite improvements in surgical techniques and standardisation of the RP procedure, PSMs remain an active area of investigation regarding the variability among surgeons and institutions. Several studies have shown that PSMs can predict metastatic progression [5] and/or local recurrence and distant metastasis [6,7], whereas other studies have shown no such relationship [8,9]. Therefore, to further clarify the prognostic value of PSMs in PCa, we performed this meta-analysis based on all published epidemiological studies to evaluate whether the presence of a PSM has a prognostic impact on biochemical recurrence-free survival (BRFS), cancer-specific survival (CSS), overall survival (OS), cancer-specific mortality (CSM) and overall mortality (OM) in patients with PCa.

Literature search
According to the guidelines of the preferred reporting items for systematic reviews and meta-analyses (PRISMA) [10], we searched PubMed, EMBASE and Web of Science from their inception to February 2018. Because the studies included in this meta-analysis have been published, no ethical approval was required. MeSH terms and free words searched for were as follows: 'prostate cancer OR prostate neoplasm', 'radical prostatectomy', 'positive surgical margin', 'survival outcome', 'prognosis' and their combinations.  The reference lists of previous relevant reviews were also manually checked to identify all available studies. The language of the publications was restricted to English.

Inclusion and exclusion criteria
The eligible studies were included only if they met the following criteria: (1) clinical trials that reported patients with PCa; (2) PSM status that was assessed by pathologists; (3) survival outcomes (BRFS, CSS, OS, CSM and OM) of patients with PSMs that were compared with those of patients with negative surgical margins; (4) results that were reported as risk estimate hazard ratios (HRs) with corresponding 95% confidence intervals (95% CIs), or sufficient data that was provided to estimate these measures; and (5) the adoption of only the more well-designed, recent and informative publication in this meta-analysis when more than one study analysed the same patient cohort. Accordingly, studies with the following criteria were excluded: (1) reviews, meeting abstracts, letters, case reports, author replies and articles not on humans; (2) studies not related to PCa; (3) studies that did not analyse the presence of a PSM and the clinical features and survival outcomes; and (4) studies lacking sufficient data to acquire HRs and 95% CIs.

Data extraction and quality assessment
The following data of the eligible studies were extracted independently by two reviewers (ZLZ and HZ): first author, publication year, country, sample size, recruitment period, age of patients, preoperative PSA, histopathological subtype, follow-up time, and survival end point. All discrepancies in data extraction were resolved by discussion between the two reviewers or consultation with a third reviewer (BW). The quality of the included studies was assessed using the Newcastle-Ottawa scale (NOS) [11] for nonrandomized studies. Each study was assessed by eight methodological items with a score ranging from 0 to 9. Studies with scores of six or higher were graded as high quality. Only high-quality studies were included for further analysis to assure the quality of this meta-analysis.

Statistical analysis
Pooled HRs with 95% CIs were used to evaluate the association of a PSM with PCa prognosis and clinicopathological characteristics. An observed HR > 1 indicated a poor prognosis for patients with PSMs. Heterogeneity between studies was assessed using the Q and I 2 statistics. p < 0.10 or I 2 > 50% were used to indicate heterogeneity. A random-effect (RE) model was used when heterogeneity was observed (p < 0.1); otherwise, a fixed-effect (FE) model was used. To obtain a more precise evaluation of heterogeneity, subgroup analysis was performed for BRFS, CSS, OM and OS based on geographical region, date of publication, mean age, sample size, mean preoperative PSA (p-PSA) concentration, median follow-up and adjuvant radiotherapy (aRT). Sensitivity analysis was performed to test the reliability of the total pooled results by sequential omission of individual studies. In addition, publication bias was assessed using funnel plots and Egger's test. All statistical tests in this meta-analysis were undertaken using Stata 14.0 software (Stata Corporation, College Station, TX). All statistical tests were two-tailed, and p < 0.05 was considered statistically significant.

Study characteristics and quality assessments
The detailed characteristics of the studies are listed in Table 1. All studies were published between 2010 and 2017, with the mean duration of follow-up varying from 18.1 to 174 months. A total of 141,222 patients (ranging from 51 to 65,633) underwent RP for PCa management, of which 31,421 patients were reported to have PSMs. Nine studies [8,17,19,20,30,32,34,36,37,42] reported the use of radiotherapy as an adjuvant therapy after RP, and the proportion of patients who received aRT was 0.2-69%. Of the 32 studies, 11 were conducted in North America, 10 in Asia, 8 in Europe and 3 at international multi-centres. All articles included were published in English. The NOS was applied to assess the quality of the included studies, and the results showed all the studies were of high quality, with an NOS score ≥ 7 (Supplementary Table S1).

Discussion
Despite diverse multimodality treatment options and extensive studies, PCa remains a major health burden in men, and its diverse clinical outcomes regarding progression is a challenge to be addressed. As a result, various factors, including pathologic features and novel molecular biomarkers, are currently regarded as being useful for predicting  the prognostic outcomes of RP. Nevertheless, PCa has been shown to be characterised by unique biological features and heterogeneous genetic backgrounds, indicating the limitations for predicting postoperative prognostic outcomes in patients with localised PCa [43]. RP with pelvic lymph node dissection is the standard of care for localised PCa, with the goals of providing good oncologic and functional outcomes, especially in patients with good life expectancy. However, a proportion of patients inevitably demonstrate adverse pathologic features such as PSMs, seminal vesicle invasion [44], lymph node metastasis [45] and perineural invasion [46]. The reported incidence of PSMs, notwithstanding its significant decrease with RP because of the advances in surgical techniques, signifies locally adverse pathology, and PSMs remain an ominous prognostic factor [47,48]; moreover, the management of patients with PSMs remains challenging. Furthermore, the impact of PSMs on control of PCa has been controversial. For example, in an analysis of the pathological reports of 65,633 specimens from RPs, Wright et al. demonstrated the independent role of a PSM in PCa [26]. Subsequently, Alkhateeb et al. [49] reported that a PSM was an independent predictor of BRFS in patients with intermediate-and high-risk PCa. However, Mithal et al. [8] reported that a PSM was significantly associated with all adverse outcomes in unadjusted models, although PSMs were only associated with increased risk of BCR (HR = 1.98, p < 0.001) and not with castration-resistant disease, metastases, or CSM (HR ≤ 1.29, p > 0.18) after adjusting for demographic and pathological characteristics.
Patients with BCR following RP have been shown to be at increased risk for subsequent metastases and death. However, BCR represents an early event in the natural history of PCa with heterogeneous outcomes, and BCR does not systematically translate into clinical progression [42]. Although previous studies have found that PSMs are associated with an increased risk of BCR, their association with more clinically robust endpoints is still controversial [50]. The prognostic heterogeneity may often have been incompletely characterised due to limitations in sample size, and only a large study with enough events can evaluate whether a PSM is an independent predictor of clinical outcome. In this meta-analysis, we synthesised 32 studies with a large sample of 75,589 patients, including 31,421 PSM patients (22.2%), to explore the relationship between PSMs and oncologic outcomes in localised PCa.
To the best of our knowledge, the present study was the first to systematically evaluate the prognostic value of PSM in patients with PCa, and the data showed that a PSM was a predictor for BRFS (HR = 1.35, p < 0.001), CSS (HR = 1.49, p = 0.001), OS (HR = 1.11, p = 0.014), CSM (HR = 1.23, p < 0.001) and OM (HR = 1.09, p = 0.009). The findings were consistently independent of geographical region, publication year, age, sample size, p-PSA, follow-up duration and aRT (yes/no). Sensitivity analyses indicated that the findings were reliable and robust. In addition, there was no evidence of significant publication bias in these analyses according to Begg's tests. Although there was no evidence of heterogeneity in terms of CSM, significant heterogeneity was detected in the analysis of the BRFS, CSS, OS and OM models. To further explore the source of heterogeneity, subgroup analyses were conducted. Our data showed that the significant variations were reduced within some items.
Although we used a systematic method to perform the present study, the following limitations also should be taken into account. First, the applied methods for detecting PSMs in the pathologic specimen were varied in the included studies, which may cause heterogeneity among the studies. Second, substantial heterogeneity was observed in the meta-analysis; although we chose the RE model according to heterogeneity, it still existed in our studies. The heterogeneity was probably caused by differences in factors such as the patients' characteristics and different durations of follow-up. Third, we only included published studies written in English, and grey literature was not included, which may cause selection bias. Fourth, all the included studies were retrospective cohort studies, and data extracted from those studies may have led to inherent potential bias.
Nevertheless, the present study has several key strengths. First, the meta-analysis included 32 studies with a large sample size to detect more stable associations between PSMs and clinical outcomes of PCa patients. Second, with the strict inclusion and exclusion criteria, we extracted available data from relevant studies. Furthermore, the results were found to be reliable and robust through subgroup and sensitivity analyses. Therefore, PSM determination, with excellent accessibility and low costs, warrants wider application in patients with PCa for risk stratification and decisionmaking of individualised treatment.
In conclusion, the results of this meta-analysis demonstrated that the finding of PSMs by histopathology is closely associated with poor survival in patients with PCa. Due to limitations in this study, large-scale, multicentre prospective studies with standardised methods and long-term follow-up are needed to verify our results.

Availability of materials and data
All data generated or analysed during this study are included in this published article (and its supplementary information files).