DISCUSSION
Post-RP BCR is a condition that can be encountered in clinical practice due to heterogeneity in prostate cancer. Although RP is the gold standard treatment option for localized PCa, radiotherapy and local ablative treatments are also available, and BCR prediction is important for optimal treatment in patients diagnosed with localized PCa. Occurrence of BCR depends on clinical stage, pre-operative total PSA value and pre-operative prostate biopsy GS 13. Predictive models for BCR after RP have been developed to date. D’Amico risk classification and CAPRA scoring are frequently used models3,4. To best of our knowledge, there is no study in the literature that compares GP scoring and PRIX scoring which are relatively newly developed and are not frequently used in clinical practice with D’Amico risk classification and CAPRA scoring. In our study, we evaluated these four predictive models in terms of their predictive value of BCR for the first time in Turkish patients.
According to our results Harrell’s concordance index results of all predictive models were range from 0,693 to 0,831. It means that all predictive models performed well both in short-term and long-term prediction of BCR. In the study conducted by Lughezzani et al. the Harrell’s concordance index for 3-yr BCR predictions was 70.4%, 74.3%, and 75.2% for the D’Amico, CAPRA, and Stephenson models, respectively. Similarly, 5 yr after RP, the Harrell’s concordance index of the three BCR predictive models was 67.4%, 72.9%, and 73.5%, respectively8. In an another study conducted by Tamblyn et al. the Harrell’s concordance index was 0.791 and 0.787 for the Stephenson and CAPRA models respectively 14. Also, in a study conducted by Yoshida et al indicated that the concordance index of the PRIX score and the D’Amico classification to predict BCR was 0.719 and 0.730, respectively 6. To best of our knowledge there is no study evaluating concordance index for GP score in literature. In our study the Harrell’s concordance index for 3 year BCR predictions was 79,8%, 79,1%, 72,3% and 71,4% for the GP score, PRIX, CAPRA and D’Amico respectively. In addition, the Harrell’s concordance index for 5 year BCR predictions was 77,8%, 77,1%, 70,2% and 69,3% for the GP score, PRIX, CAPRA and D’Amico respectively. Our results for D’Amico, CAPRA and PRIX was similar with literature in respect of the Harrell’s concordance index for 3-yr and 5-yr BCR predictions. This can indicate us that the Harrell’s concordance index for 3-yr and 5-yr BCR predictions of GP score was reliable. In addition, Harrell’s concordance index for 3-yr and 5-yr of GP score was slightly higher than D’Amico and CAPRA.
Pre-operative total PSA and GS score which are parameters of GP score are important predictive factors for BCR. In a study conducted by Mithal et al. shown that the greatest improvement in accuracy over the BCR was GS. Specifically, for the outcomes of BCR, c-indices for Gleason score were 0.66 15. In addition, Acimovic et al stated that increasing level of preoperative Gleason score, higher level of preoperative PSA and higher percent of positive biopsies were independently associated with occurrence of BCR but clinical stage of disease, number of biopsies and Free/Total PSA ratio did not affect the occurrence of BCR 16. Literature indicates that for BCR pre-operative total PSA level and GS score are more important factors than others like clinical stage. In our study all predictive models other than GP score use clinical stage in scoring or classification of patient. Although multiparametric MRI (mpMRI) has been increasingly used in clinical practice, digital rectal examination (DRE) is still mostly used tool for clinical staging. In our study we also used DRE in clinical staging of patients. We think that more accurate clinical staging can be performed with mpMRI but its contribution to prediction of BCR is not certain. In a study conducted by Capogrosso et al indicated that the accuracy of the Kattan nomogram (c-index, 0.724) and the D’Amico risk classification (c-index, 0.651) was not significantly improved by adding the mpMRI score (Model 1: c-index, 0.725; Model 2: c-index, 0.674) 17. On the other hand, Manceau et al defined mpMRI Imaging-Based Risk Classification for recurrence. They concluded that this classification was significantly correlated with the risk of BCR (p < 0.001) and the area under curve (AUC) for predicting BCR was 0.714 for the imaging-based classification compared with 0.710 for the D’Amico classification18. It is still controversial that routinely available mpMRI information is a potential marker to add to preoperative prediction models to stratify patients’risk and inform treatment planning.
Nonetheless, there are some limitations to this study. Firstly, our study population was not large. Secondly, we couldn’t perform decision-curve analyses. Thirdly, our study has short median follow-up. Certainly, longer median follow-up could affect our results. All prostate biopsy pathology reports were not evaluated by same one pathologist. This could result in differences in GS assignment and thought affect score or classifications of predictive models. Lastly, BCR prediction was performed according to pre-operative models. Surely after RP, more precise BCR predictions can be acquired using models that depend on pathologic RP 19.