The predictor factors for positive surgical margins in patients of prostate cancer after radical prostatectomy: a systematic review and meta-analysis

Background and objectives The previous studies had demonstrate that positive surgical margins (PSMs) was independent predictive factor for biochemical and oncologic outcome in patients with prostate cancer (PCa). This study aimed to conduct a meta-analysis to identify predictive factors for PSMs after radical prostatectomy (RP). Methods We selected eligible studies via electronic database of PubMed, Web of Science and EMBASE from inception to February 2019. The risk factors for PSMs following RP were identicated. The pooled estimates of standardized mean differences (SMDs)/ odds ratios (ORs) and 95% condence intervals (CIs) were calculated. A xed-effect or random-effect was used to pool the estimates. Subgroup analyses were performed to explore the reasons for heterogeneity. Results Twenty two studies including 44,144 patients with PCa were eligible for further analysis. The results showed that PSMs were signicantly associated with preoperative PSA(pooled SMD=0.44; 95% CI:0.35–0.54; P<0.001), biopsy Gleason Score (< 6/ ≥ 7) (pooled OR=1.51; 95% CI:1.26–1.81; P<0.001), pathological Gleason Score (< 6/ ≥ 7) (pooled OR= 2.34; 95% CI:2.02–2.71; P<0.001), pathological stage (


Background
Prostate cancer (PCa) is the most common type of newly diagnosed malignancy and a leading cause of cancer-related death in males worldwide [1]. With the widely used of the prostate-speci c antigen (PSA) screening test, the majority of PCa patients are diagnosed in the early stages [2]. As a result, radical prostatectomy (RP) with bilateral pelvic lymph node dissection has been the gold standard for the treatment of patients with localized PCa [3]. The goal of RP for PCa is complete prostate extirpation, despite the favorable cancer control associated with RP, approximately 25% of all patients experience biochemical recurrence (BCR) [4]. A number of factors have been reported to be associated with BCR after RP, and one adverse risk factor is the presence of positive surgical margins (PSMs).
PSMs is de ned as an extension of the cancer cells to the inked cut surface of the RP specimen [5]. Our previous ndings have indicate that PSMs are signi cantly associated with BCR and poor survival outcome after RC [6,7]. However, no systematic research studies have shown which factors that may affecting the margin of PCa after RC. Conventional parameters for risk estimation of PSMs are mainly based on the following factors, including preoperative PSA (p-PSA), pathological T stage, pathological Gleason Score (GS) and multiple positive biopsy cores [8][9][10][11]. However, the prognostic value of these predictive factors is limited. Besides, PSMs may affected by remnant normal tissue and inadequate surgical skill [12]. Therefore, no consensus is reported regarding the above results. For these considerations, a comprehensive meta-analysis and systematic review were necessary to evaluate the predictive factors for PSMs in PCa patients following RP.

Literature and search strategy
We carried out this meta-analysis in accordance with the guidelines of the Preferred Reporting Items for Systematic Review and Meta-Analyses statement (PRISMA) [13]. A comprehensive literature search was conducted of the PubMed, Web of Science, and EMBASE databases. The search strategies were based on the combination of Medical Subject Headings (MeSH) and keywords as follows: 'prsotate cancer', 'radical prostatectomy', 'positive surgical margin', 'clinicopathological' and 'risk factors'. The last search was on February 2019. Meanwhile, to identify other eligible publications, reference lists were also manually screened. The language was restricted to English. Because we did not make clinical research in this study, no ethical approval needed and all analyses were based on previous published literatures.

Selection criteria and data extraction
Papers were included in this meta-analysis if they met the criteria as follows: (1) all patients diagnosis of PCa and PSMs were histopathologically con rmed; (2) treatment was limited to RP; (3) clinicopathological features were analyzed according to surgical margins status, and all studies were comparable study design; (4) standardized mean differences (SMDs)/ odds ratios (ORs) and 95% con dence intervals (CIs) were reported in paper or could be computed from given data; (5) if more than one articles from same cohort were identi ed, the most comprehensive and largest datasetwasadopted. Accordingly, studies with the following criteria were excluded:(1) case reports, review articles, editorials and non-original articles;(2) papers published not in English; (3) studies that did not analyze the PSMs and the clinical features;(4) lacking su cient data to acquire SMDs/ ORs and 95% CIs. The literature search was performed by two investigators independently. Disagreement was resolved by discussion.
Data extraction and quality assessment Two researchers (ZLZ and WQ) assessed the titles and abstracts of the searched studies, respectively. Any disagreements were reconciled by a third researcher (HZ). The following information was extracted from the included studies: publication information ( rst author's last name, publication year, country of origin and study design), patients' characteristics (mean age, p-PSA, followup time) and PCa outcomes (tumor stage, GS, oncologic outcomes). According to the Newcastle-Ottawa quality assessment scale (NOS) [14], two researchers independently assessed the quality of each study. According to its criteria, the NOS estimates studies based on 3 parts: selection, comparability and outcome assessment. For quality assessment, scores ranged from 0-9, and studies with scores of 6 or more were rated as being of high quality.

Statistical analysis
For this meta-analysis, pooled SMDs/ ORs with 95%CIs were used to describe the relationship between risk factors and PSMs.
An OR >1 or SMD > 0 suggested a close relationship for PSMs in patients with PCa. Heterogeneity among studies was evaluated by using Cochran's Q test and Higgins I-squared statistic. If the I 2 value is > 50% or the P heterogeneity is < 0.1, which suggest a statistically signi cant heterogeneity in the included studies, a random-effects (RE) model was adopted; otherwise, xed-effects (FE) model was used. To consider potential reason for heterogeneity, subgroup analysis was conducted. To test the stability of the result, we performed the sensitivity analysis by excluding one study in turn. Visual inspection of asymmetry in funnel plots was carried out to assess the potential publication bias. Furthermore, we performed Begg's tests to provide quantitative evidence of publication bias. Those statistical analyses or data syntheses were calculated using STATA version 12.0 (Stata Corporation, College Station, TX, USA). All statistical tests were two sided, and P < 0.05 was considered to be statistically signi cant.

Literature search
A flowchart of the literature selection process is shown in Figure 1. The initial search of electronic databases identified 1,284 records according to the searching criteria; after duplicates were removed, 611 papers remained. Four hundred and sixteen papers were then excluded by screening titles and abstracts. Ten 195 full-text articles were further examined and 173 articles were excluded because 18 same cohort of patients and 155 lacking enough data for further research. At last, 22 articles [8,  published between 2009 and 2018 were included in this meta-analysis.

Features of included studies
Summary of major characteristics of these studies are shown in Table 1 and Table 2. All the studies were of retrospective study design. The sample size ranged from 144 to 12,515, and a total of 44,144 patients were included. A total of 10.457 PCa patients with PSMs were included in our study, which accounts for 23.7% of all patients. Geographically, 8 studies were conducted in North America, 6 in Asian, 5 in Europe, 2 in Australia and 1 in Multi-center. All patients had received RP as primary treatment for PCa.
According to NOS quality assessment, all studies in this study were categorized as of high quality. (Supplementary Table S1)

Subgroup analysis
Considering that no signi cant heterogeneity in p-PSA and BMI, besides, the number of studies that evaluated EPE, SVI and prostate volume was relatively small, we only conducted subgroup analysis for biopsy GS, pathological GS, pathological stage, PLN, age and nerve sparing ( Table 3) Meanwhile, no correlation was founded for p-PSA, biopsy GS, PNI and PLN. The inconsistent results of the above studies may due to small sample size, single-center design and inhomogeneous population. To the best of our knowledge, no studies have systematic addressed the preoperative predictive factors for PSMs after RP. In the present study, we identi ed 22 studies involving 44,144 patients, and the PSMs rate was 23.7%, which is comparable to previous reports. The meta-analysis showed that p-PSA, biopsy GS (< 6/ ≥7), pathological GS (< 6/ ≥7), pathological stage (<T2/ ≥T2), PLN, EPE and SVI have statistically signi cant association with PSMs. Moreover, the pooled OR/SMD of the results suggested that age, BMI, prostate volume and nerve sparing were not independent prognostic factors for PSMs in patients after RP. Subgroup analyses revealed similar result despite different geographical region, publication year, sample sizes and median follow-up. Further sensitivity analysis and publication bias test were also taken, and the overall results showed that our data were stable and reliable. This is the first comprehensive study to investigate the pathological features of PSMs and predictive factors for PSMs in patients treated with RP, and the results of this analysis are meaningful. Two strengths of this study are as follows: First, large sample size of PCa patients from different geographic areas were include, and the findings in our study may more robust than those of individual study. Second, a summary OR/SMD were conducted to compare the difference between PSMs and NSMs in PCa patients categorized by several confounders. Therefore, our findings could provide solid evidence for prognostic factors in PCa patients with PSMs.Nevertheless, the present study has some limitations that should be acknowledged. First, all the studies were retrospectively performed. Second, a substantial heterogeneity was detected, while sensitivity analysis and subgroup analysis failed to identify the potential heterogeneity. Third, the number of inclued studies were limited in publication bias, subgroup and sensitivity analyses, which could lead to unpersuasive conclusions. Finally, all articles are in English, which could cause publication bias, although no bias was detected in the present study.

Conclusions
The meta-analysis demonstrates that p-PSA, biopsy GS, pathological GS, pathological stage, PLN, EPE and SVI were independent factors for predicting PSMs after RP, and a combination of these factors might be useful for predicting PSMs in patients with PCa   Supplemental Figure Legends Supplementary Figure S1. Sensitivity analysis (pooled ORs) of the association between the predictive factors and PSMs risk.
Supplementary Figure S2. Sensitivity analysis (pooled SMDs) of the association between the predictive factors and PSMs risk.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.