Introduction

Prostate cancer, the most common newly diagnosed cancer in men, is the second leading cause of cancer-related deaths (1). Biochemical recurrence (BCR), the most common pattern of disease relapse, is seen in nearly 30% of patients who have undergone radical prostatectomy (RP) (2). Numerous risk factors, including preoperative total prostate-specific antigen (PSA) level, pathological stage, Gleason grade (GG), perineural invasion (PNI), lymphovascular invasion (LVI) and positive surgical margin (PSM) have been identified in predicting BCR. PSM is seen in 10%-48% of patients after RP (3). Moreover, half of all patients with PSM develop BCR (4). The absence of BCR in a significant proportion of patients with PSM necessitates the examination of surgical margin features, including number, linear length, location and GG at the site of positive resection margin. This study investigates the prognostic role of surgical margin features in addition to well-defined risk factors in predicting BCR.

Materials and Methods

Study design

This study used the prostate cancer database from the Urooncology Association in Turkey, to which participating institutions submit data online. Data obtained from 984 patients who underwent RP with localised and locally advanced prostate cancer and met eligibility criteria for this study (Figure 1). Since our study was a retrospective study using a database, informed consent and ethics committee approval was not obtained.

Data Collection and Definitions

Clinical (age, preoperative total PSA), surgical (type of operation, lymphadenectomy status), pathological [International Society of Urological Pathology (ISUP) grade, extraprostatic extension (EPE), LVI, PNI, seminal vesicle invasion (SVI), surgical margin status, lymph node involvement (LNI)], PSM (number, linear length, location and GG) and follow-up (PSA outcome, BCR status) data were requested and recorded from the prostate cancer database. PSM features including linear length (<1 mm and ≥1 mm), number (single and multiple), location (apex, anterior, posterolateral, bladder neck and seminal vesicle) and GG (1-5) were recorded as submitted in the prostate cancer database. The ISUP grading system as identified in the 2014 ISUP consensus conference was used (5). BCR is defined as PSA >0.2 ng/mL after the RP (6). BCR-free survival (BCR-FS) is defined as the time from the date of RP to the date of BCR. In patients without BCR, BCR-FS is defined as the time from the date of RP to the date of the last follow-up visit.

Statistical Analysis

The normal distribution of continuous variables was evaluated through visual (histogram and probability plots) and analytical (Kolmogorov-Smirnov and Shapiro-Wilk tests) methods. For continuous variables, the statistical difference among the groups was determined using the Mann-Whitney U test. For categorical variables, statistically significant differences among groups were determined using chi-square tests. To assess differences in BCR-FS, Kaplan-Meier curves were generated and compared using the log-rank test. Univariate and multivariate Cox regression analyses were performed to identify the prognostic factors affecting BCR-FS. Statistical significance was accepted as p<0.05. IBM SPSS Statistics 15.0 was used for the statistical analysis.

Results

Clinicopathological Characteristics

Patients’ clinical, surgical and pathological data are summarised in Table 1. BCR was detected in 178 (18.1%) patients. The median follow-up time for all patients was 29 (6-210) months, and it was significantly higher in the BCR group (36 months vs 28 months). The total PSA was significantly higher in the BCR group (8.3 ng/mL vs, 6.7 ng/mL; p<0.001). Data analysis showed that a higher ISUP grade had a significant relationship with BCR (p<0.001). Moreover, EPE, SVI, LVI and LNI were all significantly higher in the BCR cohort (p<0.001).

Surgical Margin Features

The PSM was significantly higher in the BCR group (53.4% vs 29.5%; p<0.001). Surgical margin status was evaluated in terms of surgical margin features, including number, location, linear length and GG at PSM. There was no statistically significant difference between the BCR and disease-free group in terms of number, location and linear length of PSM (p>0.05). Patients were separated into two groups according to positive surgical margin Gleason grade (PSMGG) as PSMGG ≤3 and PSMGG ≥4. We found a significant relationship between higher PSMGG and BCR (p=0.043). We also analysed the PSMGG status in three groups-downgrade, same and upgrade-according to the difference from the index tumour GG. We found a significant relationship between the downgrade of PSMGG and disease-free status (Table 2).

Relationship Between BCR-FS and Clinicopathological Features

The median follow-up period was 28 (6-210) months in the disease-free group. According to Kaplan-Meier survival analysis, BCR-FS was found to be significantly lower in the group with the higher total PSA level, higher ISUP grade, EPE, SVI, LVI and LNI (log-rank p<0.001). Kaplan-Meier survival curves, including 2-year, 5-year and 10-year BCR-FSs, are shown in Figure 2.

Relationship Between BCR-FS and Surgical Margin Features

In 333 (33.8%) patients, PSMs were observed. Since PSMGG was not routinely reported in every patient with a PSM, 120 patients were included in the analysis involving PSMGG. According to Kaplan-Meier survival analysis, BCR-FS was found to be significantly higher in the group with a negative surgical margin (NSM) (log-rank p<0.001) (Figure 3a). The NSM group was compared separately from the PSMGG ≤3 and PSMGG ≥4 groups in terms of BCR-FS. It was shown that patients with NSM had a significantly better BCR-FS than patients with PSMGG ≥4 (log-rank p=0.005). However, there was no significant difference between patients with NSM and PSMGG ≤3 in terms of BCR-FS (log-rank p=0.662). On account of this finding, we combined NSM and PSMGG ≤3 groups into a single cohort, and then compared with PSMGG ≥4 group. According to Kaplan-Meier survival analysis, BCR-FS was still significantly lower in the PSMGG ≥4 group (Figure 3b). Kaplan-Meier survival curves, including 2-year and 5-year BCR-FSs, are shown in Figure 3.

Prognostic Factors in Predicting BCR-FS

Univariate Cox regression analysis indicated that total PSA, ISUP grade, EPE, SVI, LVI, LNI, PSM and PSMGG are significantly associated with BCR-FS (p<0.001, p<0.001, p<0.001, p<0.001, p<0.001, p<0.001, p<0.001 and p=0.007, respectively). Multivariate Cox regression analysis was performed to identify the independent prognostic factors for predicting BCR. For this purpose, we created two different predictive models. We included total PSA, ISUP grade, EPE, SVI LVI and LNI in both models. In addition to these predictive factors, Model 1 included surgical margin status, and Model 2 included PSMGG. When evaluating PSMGG in multivariate Cox regression analysis, we combined patients with NSM and PSMGG ≤3 into the same group, and then compared them with PSMGG ≥4. In multivariate Cox regression analysis of Model 1, ISUP grade, EPE and PSM were found to be independent prognostic factors in predicting BCR-FS [Hazard ratio (HR): 1.89, p=0.035 and HR: 4.65, p<0.001, HR: 1.82, p=0.030, HR: 1.77, p=0.042, respectively]. However, in Model 2, only ISUP grade was found to be an independent prognostic factor in predicting BCR-FS (HR: 10.04, p<0.001, Table 3).

Discussion

Patients with BCR after RP have a worse prognosis. In these patients, the risk of metastatic disease and cancer-related death increases significantly (2). Cancer-specific survival declines significantly in patients with a shorter PSA doubling time and shorter interval to BCR after RP as well as in patients with higher Gleason scores (GSs). A PSM and higher pathologic tumour stage were also found to increase the risk of metastatic disease (7). Prediction of BCR-FS plays an important role in determining which patients would benefit from adjuvant treatment, especially radiation therapy (ART). In a meta-analysis including three randomised-controlled trials (EORTC22911, SWOG8794, ARO96-02/AUO-AP09/95) in which ART was compared with a wait-and-see strategy, metastasis-FS was found to be significantly higher in the ART group (Odds ratio=0.77, p=0.02). Clinical progression-FS was found to be significantly lower in patients with SVI or PSM (HR=0.73, p=0.0003) (8). In another multi-institutional study, no significant difference was found in terms of metastasis-FS and overall survival when early salvage radiotherapy and ART were compared (92% vs 91%, p=0.9, and 89% vs 92%, p=0.9, respectively) (9).

Preoperative PSA, pathological GS, EPE, SVI, PSM and LNI are included in nomograms as independent predictive factors for BCR (10). Kattan et al. (11) created the first nomogram in 1999, including these predictors, and external validations of this nomogram have been performed subsequently (12,13). In our study, we found a statistically significant relationship between high preoperative PSA, high pathological ISUP grade, EPE, SVI, PSM, LVI and LNI and a shorter BCR-FS in univariate analysis. In a meta-analysis evaluating the prognostic significance of six clinicopathological features including PSM, EPE, SVI, LVI, PNI and LNI, all these factors were found to be statistically significant for BCR-FS (HR: 1.79, p<0.001; HR: 2.03, p<0.001; HR: 1.97; p<0.001; HR: 1.85, p<0.001; HR: 1.59, p<0.001; HR: 1.88, p<0.001, respectively) (14). LVI and PNI are not involved in current prostate cancer nomograms. The prognostic role of PNI in predicting BCR is controversial. In a large multicentre study, Kraus et al. (15) found that PNI was not an independent predictor for BCR. However, it can be an indicator of unfavourable histology, such as high G. In our study, we found no statistically significant relationship between PNI and BCR. However, LVI was found to have a significant association with a higher BCR risk in a meta-analysis evaluating its prognostic value (16). We found a statistically significant relationship between high total PSA, high ISUP grade, PSM, EPE, SVI, LVI and LVI and BCR. Kaplan-Meier analyses also showed shorter BCR-FS in these groups. However, PNI was not statistically associated with BCR.

PSM is a well-known predictor for BCR (17,18,19). However, there is limited evidence about PSM subgroups, including number, location, linear length and GG/GS. Apical and posterolateral prostate were the most common locations in terms of PSM (20). Although the prognostic effect of location and number of PSM on BCR is controversial (20,21,22,23), the relationship between both extended positive surgical margin linear length (PSMLL) and higher PSMGG/SMGS with BCR has been previously reported. Mainly PSMLL ≥3 mm and PSMGG ≥4 have been shown to be the most prominent factors in predicting BCR (24,25,26,27,28,29). Unlike most studies, we grouped PSMLL as <1 mm and ≥1 mm since the database was designed in this way. We found no significant relationship between number, linear length and location of PSM and BCR. However, we found a statistically significant relationship between PSMGG and BCR. PSMGG ≥4 was found to be associated with a significantly shorter BCR-FS. Iremashvili et al. (27) designed a study in which they provided different “PSM GS” definitions (GS at the margin, high-grade Gleason pattern present at the margin, predominantly high-grade GS at the margin, GS at the margin higher than overall GS and GS at the margin lower than overall GS) and were used to evaluate BCR-FS. Their results showed that all the definitions of GS at PSM were independently associated with the risk of BCR. It was stated that a “high-grade Gleason pattern at a PSM” could be used as the most useful definition since it provided at least as much prognostic benefit as the others. We also found a statistically significant relationship between a lower PSMGG than the overall GG and disease-free status.

We developed two predictive models by using total PSA, ISUP grade, EPE, SVI, LVI, LNI, surgical margin status and PSMGG. In Model 1, which included surgical margin status instead of PSMGG, ISUP grade, EPE and surgical margin status were found to be independent prognostic factors in predicting BCR-FS. However, in Model 2, which included PSMGG instead of surgical margin status, only ISUP grade was found to be an independent prognostic factor in predicting BCR-FS. In Model 1, the expected results were determined in accordance with the current literature except for the SVI (10,12,14). Since there were only 120 patients with PSMGG data, we had to evaluate patients with NSM and PSMGG ≤3, which have similar BCR-FS, in the same group against patients with PSMGG ≥4. A possible explanation for the unexpected result in Model 2 may have been due to the smaller number and the shorter follow-up time of the patients who have the PSMGG data in contrast to previous studies which identified the independent prognostic role of PSMGG in predicting BCR-FS (24,25,26,27,30).

Study Limitations

The study is retrospective. There were limited data, including LND and PSM features. Since the type of LND was not always recorded, they were all evaluated in the same group. Not performing LND in almost half of the patients led to these patients not being able to be included in the multivariate analyses. Although the study included 333 patients with PSM, PSMGG data were only available for 120 patients. Since the PSMGG data had been reported relatively recently, these patients had a shorter follow-up. We thought that the lack of PSM data caused less accurate results in multivariate analysis, which included PSMGG. Moreover, there might have been a difference in pathologic interpretation, especially when evaluating PSM features, due to the lack of central pathologic review. We also used a cut-off value of 1 mm for PSM linear length, since it was categorised in the database as such. Continuous linear length data could produce more accurate results.

Conclusion

The present study shows that ISUP grade, EPE and PSM as independent prognostic factors in predicting BCR-FS. Although BCR-FS was significantly shorter in the patients with PSMGG ≥4, in multivariate analysis, PSMGG was not found as an independent prognostic factor in predicting BCR-FS. There is a need for large, randomised prospective studies to identify the role of PSMGG to be used in nomograms as an independent predictor in order to determine patients who would benefit from ART.

Acknowledgements

Publication: The results of the study were not published in full or in part in form of abstracts.

Contribution: There is an other contributor who may not be listed as authors: Saadettin Eskiçorapçı, Acıbadem University Faculty of Medicine, Department of Urology, İstanbul, Turkey, Çağ Çal, Ege University Faculty of Medicine, Department of Urology, İzmir, Turkey, Volkan İzol, Çukurova University Faculty of Medicine, Department of Urology, Adana, Turkey.

Conflict of Interest: No conflict of interest was declared by the authors.

Financial Disclosure: The authors declared that this study received no financial support.

Ethics

Ethics Committee Approval: Since our study was a retrospective study using a database, informed consent and ethics committee approval was not obtained.

Informed Consent: Retrospective study.

Peer-review: Externally and internally peer-reviewed.

Authorship Contributions

Concept: T.S.S., L.T., Design: T.S.S., L.T., Data Collection or Processing: G.A., S.B., E.S., T.M., B.A., Analysis or Interpretation: M.Y.K., Literature Search: M.Y.K., Writing: M.Y.K., T.S.S.