Effect of preoperative risk group stratification on oncologic outcomes of patients with adverse pathologic findings at radical prostatectomy

Abstract

Purpose Current National Comprehensive Cancer Network (NCCN) guideline recommends initial therapy for non-metastatic prostate cancer (PC) according to the preoperative risk groups. But adjuvant therapy after radical prostatectomy (RP) is recommended only based on the adverse pathologic findings (APFs) irrespective of these risk groups. We assessed whether the model incorporates preoperative risk group and APFs can predict the long-term oncologic outcomes better than that only based on APFs.

Patients and Method We retrospectively reviewed the clinical data of 4,404 men who underwent RP at our institution between 1992 and 2014. After exclusion of patients who received neoadjuvant therapy or those with incomplete pathological and follow-up data, 3,092 men were included in the final analysis. All patients were stratified into low-, intermediate-, and high-risk groups according to the NCCN guideline and APFs were defined as extraprostatic extension (EPE), seminal vesicle invasion (SVI), or positive surgical margin (PSM). Baseline characteristics of men and pathologic outcomes were compared using χ2-tests for categorical data, and Student’s t-test or analysis of variance (ANOVA) test for continuous data. The adequacy of model fit to the data was compared between the models of APFs with and without risk groups using the likelihood-ratio test and model discrimination was compared with the concordance index (c-index) for predicting biochemical recurrence (BCR) and PC-specific mortality (PCSM). Kaplan-Meier estimates of BCR-free survival (BCRFS) and cumulative incidence estimates of PCSM were compared between the models using log-rank test for BCRFS and Gray's modified log-rank test for PCSM. A multivariate Cox regression analysis was used to identify factors predictive of BCR, whereas a multivariate competing risk regression analysis was performed for PCSM with death from other causes as the competing event.

Results There were significant differences in age, preoperative prostate-specific antigen (PSA) level, biopsy Gleason score (GS), clinical stage, and APFs across the risk groups (p <0.001 for all). Of 3,092 patients, 899 men experienced BCR and 85 men died due to PC at a median follow-up of 66 months (interquartile ranges [IQR] 65-96). Adding risk groups to the model only with APFs significantly improved the fit to the data (likelihood ratio test, p <0.001) and the c-index increased from 0.693 to 0.732 for BCR and from 0.707 to 0.747 for PCSM. The BCRFS rate for men with APFs was worse than those without APFs in not only overall patients but also each risk groups (overall: p <0.001, low: p = 0.027, intermediate: p <0.001, and high: p <0.001), but there was no difference in the cumulative incidence estimates of PCSM between men with and without APFs in low and intermediate risk groups (p = 0.903 and p = 0.253, respectively). Although RP GS ≥8 and PSM were independently associated with BCR in not only overall patients but also each risk groups (overall: GS ≥8 [HR 4.66, p <0.001], PSM [HR 1.93, p <0.001], low: GS ≥8 [HR 2.94, p = 0.007], PSM [HR 1.87, p = 0.010], intermediate: GS ≥8 [HR 1.85, p = 0.022], PSM [HR 2.42, p <0.001], and high: GS ≥8 [HR 4.63, p <0.001], PSM [HR 1.71, p <0.001]), only RP GS ≥8 and SVI were associated with PCSM in overall patients (GS ≥8: HR 5.39, p <0.001, SVI: HR 3.36, p <0.001) and high risk group (GS ≥8: HR 6.31, p = 0.010, SVI: HR 4.05, p = 0.001). The major limitation was that we did not perform the competing risk analysis for PCSM in low and intermediate risk groups because of small number of events.

Conclusion Our results show that the postoperative estimation of oncologic outcomes in men with APFs at RP is improved by considering preoperative risk group stratification. Although PSM was independent predictor for BCR, only RP GS ≥8 and SVI were associated with PCSM in overall patients and high risk group.