External validation of the NADIR model for early prostate‐specific antigen response prediction in enzalutamide‐treated metastatic hormone‐sensitive prostate cancer: An analysis of the ARCHES trial

Abstract Background Early favorable prostate‐specific antigen (PSA) response to ≤0.2 ng/mL by 6 months is associated with superior overall survival (OS) in metastatic hormone‐sensitive prostate cancer (mHSPC). The nadir androgen receptor pathway inhibitor (ARPI)‐derived integrative response (NADIR) model is a multivariable model developed to predict this early favorable PSA response in patients receiving ARPI. The authors report an independent external validation of the locked NADIR model in the ARCHES trial. Methods Of 574 enzalutamide‐treated patients, 558 with complete predictors were included. Discrimination, Brier score/index of prediction accuracy (IPA), and calibration were assessed. Predicted probability tertiles were used descriptively for PSA response. Six‐month landmark Cox models explored associations with OS. A sensitivity analysis restricted the cohort to <3 months of prior androgen deprivation therapy (ADT). Results Early favorable PSA response occurred in 359 patients (64%). The area under the receiver operating characteristic curve was 0.80 (95% confidence interval [CI], 0.76–0.83), the Brier score was 0.17, and the IPA was 25%. Recalibration intercept was 0.026 and slope was 0.97. Response rates across decreasing predicted probability tertiles were 92%, 64%, and 37%. In the landmark analysis ( n = 547; 177 deaths), each 10‐percentage‐point increase in predicted probability was associated with lower mortality (hazard ratio, 0.90; 95% CI, 0.85–0.96). In the <3‐month prior‐ADT subgroup ( n = 401), the area under the receiver operating characteristic curve and IPA were 0.78 and 23%, respectively. Conclusions The NADIR model demonstrates robust and reproducible discrimination, calibration, and prognostic stratification in a distinct ARPI‐treated mHSPC cohort, supporting its potential utility for early response risk stratification and trial enrichment pending prospective validation of model‐guided treatment adaptation.