External validation of the European association of urology biochemical recurrence risk groups to predict mortality after radical prostatectomy or radiation therapy in a North American cohort
Carlo Silvani, Alfonso Santangelo, Alex Stephens, Jack Considine, Shane Tinsley, Bassel Salka, Akshay Sood, Sebastiano Nazzani, Alberto Briganti, Andrea Salonia, Francesco Montorsi, Nicola Nicolai, Emanuele Montanari, Craig Rogers, Firas AbdollahAbstract
Objectives
This study aimed to externally validate the European Association of Urology (EAU) biochemical recurrence (BCR) risk stratification in a North American population after radical prostatectomy (RP) and radiation therapy (RT), where validation remains lacking despite prior European and Asian validation.
Materials and methods
We identified all patients with BCR after RP or RT between 1995 and 2023 from a North American institutional database and classified them by EAU criteria. Primary outcome was prostate cancer‐specific mortality (CSM). We calculated Harrell's concordance indices (C‐index) and used competing‐risk regression to assess associations between EAU risk groups and CSM, comparing performance to multivariable models including age, clinical stage, Gleason grade, PSA doubling time and time to BCR.
Results
Among the 940 patients (646 RP, 294 RT; 40.5% African American), 563 (59.9%) had low‐risk and 377 (40.1%) high‐risk BCR. The 10‐year cumulative incidence of CSM was 3.6% versus 12% for low‐risk versus high‐risk RP patients and 18.4% versus 49.5% for low‐risk versus high‐risk RT patients. EAU high‐risk BCR was associated with increased CSM in both groups (RP: HR 2.83, 95% CI 1.47–5.46; RT: HR 3.98, 95% CI 2.43–6.53). The EAU classification showed moderate discrimination (Harrell's C‐index 0.62 for RP, 0.69 for RT). Multivariable models including clinical variables demonstrated a Harrell's C‐index of 0.76 for both RP and RT.
Conclusions
This first North American validation confirms moderate EAU discriminative ability. For RP patients, low 10‐year CSM in low‐risk BCR (3.6%) supports surveillance. However, low‐risk RT BCR showed substantial CSM (18.4%), exceeding high‐risk RP (12%), suggesting current criteria inadequately stratify risk after RT.