Extension and application of the partial continual reassessment method to a Phase-IIa dose-finding Tuberculosis trial with a continuous endpoint

Novel regimens for treating Tuberculosis (TB) are currently under clinical evaluation. Often, these consist of regimens of multiple novel agents that are compared to a standard regimen. Selecting a minimum efficacious dose of a regimen can be challenging, as dosing of two agents is investigated and the combination-response relationship is unknown. Generally, in Phase IIa TB trials, early efficacy of novel drugs is investigated by examining a continuous measure of a marker over time.In this work, which is motivated by the desire to identify the best regimen based on data from a planned trial in conjunction with a recently completed study, we propose an extension of the partial continual reassessment method (POCRM) in order to estimate the response at each regimen when the outcome is a continuous endpoint. Using simulations, we will compare the proposed method with two other approaches, namely one that considers different regression models and another that models the interactions between doses of two agents, in the setting of this trial. We will show that the proposed method can lead, on average, to a lower estimation bias of the efficacy response and a higher proportion of correct selections of the appropriate dose compared to competing approaches.