Optimizing D’Amico risk groups in radical prostatectomy through the addition of magnetic resonance imaging data

Abstract Objectives To improve the predictive efficacy of the D’Amico risk classification system with magnetic resonance imaging (MRI) of the pelvis. Materials and methods We studied 729 patients from a series of 1310 radical prostatectomies for T1–T2 prostate cancer who underwent staging pelvic MRI. Each patient was classified with T2, T3a or T3b MRI, and N (+) patients were excluded. We identified the therapeutic factors that affected the biochemical progression-free survival (BPFS) time (prostate specific antigen [PSA] levels >0.4 ng/mL) using a univariate and multivariate study with Cox models. We attempted to improve the predictive power of the D’Amico model (low risk: T1; Gleason 2–6; PSA levels 20 ng/mL). Results In the univariate study, the clinical factors that influenced BPFS were the following: Gleason 7 (HR: 1.7); Gleason 8–10 (HR: 2.9); T2 (HR: 1.6); PSA levels 10–20 (HR: 2); PSA levels >20 (HR: 4.3); D’Amico intermediate (HR: 2.1) and high (HR: 4.8) risk; T3a MRI (HR: 2.3) and T3b MRI (HR: 4.5). In the multivariate study, the only variables that affected BPFS were the following: D’Amico intermediate risk (HR: 2; 95% CI 1.2–3.3); D’Amico high risk (HR: 4.1; 95% CI 2.4–6.8); T3a MRI (HR: 1.9; 95% CI 1.2–2.9) and T3b MRI (HR: 3.9; 95% CI 2.5–6.1). In predictive model using the multivariate Cox models, we assessed the weight of each variable. A value of 1 was given to D’Amico low risk and T2 MRI; a value of 2 was given to D’Amico intermediate risk and T3a MRI and a value 3 was given to D’Amico high risk and T3b MRI. Each patient had a marker that varied between 2 and 6. The best model included 3 groups, as follows: 494 (67.7%) patients in group 1, with a score of 2–3 points (HR, 1), a BPFS of 86% ± 2% and 79% ± 2% at 5 and 10 years, respectively; 179 (24.6%) patients in group 2, with a score of 4 points (HR, 3), a BPFS of 60% ± 4% and 54% ± 5% at 5 and 10 years, respectively; and 56 (7.7%) patients in group 3, with a score of 5–6 points (HR, 9.3), a BPFS of 29% ± 8% and 19% ± 7% at 5 and 10 years, respectively. The median BPFS time was 1.5 years. Conclusion MRI data significantly improve the predictive capacity of BPFS when using the D’Amico model data.