Bayesian sparse regression for exposure–response analyses of efficacy and safety endpoints to justify the clinical dose of valemetostat for adult T‐cell leukemia/lymphoma

Abstract

AbstractValemetostat is an oral inhibitor of enhancer of zeste homolog (EZH) 2 and EZH1 approved in Japan for the treatment of adult T‐cell leukemia/lymphoma (ATLL). To support the approved daily dose of 200 mg and inform dose adjustments in patients with ATLL, Bayesian exposure–response analyses were conducted using data from two clinical trials. The analyses included two efficacy endpoints, overall response by central and investigator assessments in patients with ATLL (n = 38, 150–200 mg), and six safety endpoints in patients with non‐Hodgkin lymphoma (n = 102, 150–300 mg), which included grade ≥3 laboratory values for anemia, absolute neutrophil count decreased, and platelet count decreased; any grade ≥3 treatment‐emergent adverse event (TEAE); and dose reductions and dose interruptions due to TEAEs. A slightly positive relationship was observed between unbound exposure and efficacy endpoints. A steeper relationship was observed in safety endpoints, compared with efficacy. Candidate covariate effects, except intercepts of the baseline laboratory values, were regularized via spike and slab priors in a Bayesian framework; only the laboratory values for corresponding hematologic TEAEs were shown to be of substantial impact. The target exposure range was established by defining a modified region of practical equivalence (184–887 ng·h/mL), which was expected to provide satisfactory efficacy and acceptable safety within the range of available exposure data. The simulated exposure range considering inter‐individual variability showed that 200 mg could reach target exposure in the overall population and across subpopulations of interest, supporting the use of valemetostat 200 mg in patients with ATLL.