Post-SABR early PSA density kinetics in androgen deprivation therapy-naïve patients: early landmark models for long-term biochemical recurrence-free survival.

Abstract

PurposeTo identify early post-treatment prostate-specific antigen density (PSA-D) models associated with long-term biochemical recurrence-free survival (bRFS) after ultrahypofractionated prostate radiotherapy without androgen deprivation therapy (ADT), and to compare early kinetic composite models with single post-treatment PSA-D classifiers.Methods and materialsProspectively collected patients receiving 45 Gy in 5 consecutive fractions without ADT were analyzed. This analysis used a prespecified, clinically driven landmark approach evaluating PSA-D change across early post-treatment intervals while prioritizing interpretability over model complexity. Candidate 2-factor models combined ISUP grade group with PSA-D kinetics in the baseline→3-month and 3 → 6-month windows and were screened by Kaplan-Meier separation, cohort size, and clinical usability, with fixed-time 7-year ROC analysis used descriptively. An exploratory 3-feature composite score was also assessed.ResultsThe final no-ADT cohort comprised 294 patients with 39 biochemical relapses. Seven-year bRFS was 100.0%, 96.2%, 83.4%, and 73.7% for low-, favorable intermediate-risk (FIR), unfavorable intermediate-risk (UIR), and high-risk disease, respectively. PSA and PSA-D decline rates from baseline to 12 months differed significantly across NCCN risk groups, with steeper declines in UIR and high-risk disease than in FIR disease (both P 0.138, separated 7-year bRFS at 96.7% versus 62.7% (P 0.033 separated 7-year bRFS at 93.5% versus 53.4% (P P = 0.016) and the adverse 3 → 6-month PSA-D term (HR, 3.41; 95% CI, 1.69-6.85; P ConclusionsEarly PSA-D kinetics may support early risk stratification for long-term bRFS after prostate SABR without ADT. Composite models combining baseline pathologic risk with early PSA-D kinetics provide a treatment-response-based layer of risk stratification that complements baseline NCCN classification. Single PSA-D cutoffs remain informative but are less satisfactory than kinetic models because they do not capture response velocity.