Personalized cancer treatment has long been constrained by a fundamental problem: clinicians cannot reliably predict which patients will benefit from a given strategy before it is too late to course-correct. For adaptive therapy in prostate cancer — a paradigm that deliberately cycles treatment on and off to prevent resistant cell populations from dominating — this gap has been especially consequential, given that responders and non-responders often look identical at baseline.
Published in JAMA Oncology, this retrospective modeling study applied a two-population differential equation framework to PSA dynamics collected during the very first treatment cycle, then used the derived parameters as predictive biomarkers for time to progression, mean daily drug dose, and overall survival. Two independent cohorts were analyzed: 40 patients with castrate-sensitive prostate cancer receiving intermittent androgen deprivation therapy, drawn from a trial spanning 1996–2006, and 13 patients with metastatic castrate-resistant disease treated with adaptive abiraterone acetate between 2015 and 2022. The model separately characterized the growth and competition kinetics of drug-sensitive versus drug-resistant cell populations, translating early PSA trajectories into mechanistic parameters that outperformed simple empirical PSA metrics as prognostic signals.
This work sits at a productive intersection of mathematical oncology and precision medicine, building on prior evolutionary game-theory-based adaptive therapy trials — most notably the Moffitt Cancer Center's landmark prostate studies. The mechanistic framing is theoretically compelling: by estimating the ratio of sensitive to resistant cells from first-cycle dynamics, the model attempts to quantify the very substrate that adaptive therapy is designed to exploit. That said, the cohorts are small — 40 and 13 patients respectively — and the retrospective, nonrandomized design limits causal inference. The mCRPC cohort in particular is underpowered. Prospective validation in larger, preferably randomized trials is essential before these biomarkers could guide clinical scheduling. Nevertheless, if externally validated, this approach could shift adaptive therapy from a population-level strategy to a genuinely individualized one — a meaningful step forward.