Bayesian re-analysis of ERA223 and PEACE-3 temporal dynamics and bone protection in radium-223-treated metastatic castration-resistant prostate cancer.
Chen W, Yoshida S, Yajima S, Fujiwara M, Sato H, Hirakawa A (+4 more)
Abstract
ObjectiveTo define the time course of fracture risk with radium-223 in combination with androgen receptor pathway inhibitor (ARPi) therapy and assess the clinical impact of concomitant bone-protective agent (BPA) use on mitigating this risk.MethodsAn exploratory person-time approximation approach was used to perform a Bayesian re-analysis of the landmark ERA223 and PEACE-3 clinical trials. In ERA223, we estimated interval-specific fracture incidence rate ratios (IRRs) during follow-up to characterize temporal trajectories. Interval-specific analyses were based on an exploratory person-time approximation approach. In PEACE-3, we used published hazard ratios to re-derive posterior distributions of risk ratios, risk differences, and posterior distributions comparing pre- vs post-mandatory BPA fracture incidences. Posterior distributions were modeled with beta-binomial and gamma-Poisson conjugate priors.ResultsIn the ERA223 population, the fracture excess risk peaked during the first year of treatment, with elevated interval-specific IRRs in the 0-6 month and 6-12 month intervals. In PEACE-3, the treatment-induced fracture burden for the radium-223 combination treatments was lower following requirement of BPA, with the risk ratio (RR) falling from 2.54 (95% credible interval [CrI] = 1.53-4.60) prior to the mandate to 1.46 (95% CrI 0.88-2.49) post-implementation. The addition of BPA prophylaxis demonstrated a synergistic protective effect, yielding a more pronounced reduction in fracture risk within the radium-223 combination arm relative to the ARPi-monotherapy group.ConclusionsFracture risk was confined to early periods of ERA223 treatment and diminished after the compulsory addition of BPAs in PEACE-3. Due to the limitations of using estimated person-time data and observational comparisons, the findings are best interpreted as descriptive and hypothesis-generating.