Keywords: Bayesian logistic regression, single ascending dose, multinomial logistic regression
Bayesian logistic regression model (BLRM) with escalation with overdose control (EWOC) has been widely used in Phase I oncology dose-finding trials. We adopt this model-based approach to a non-oncology first in human (FIH) single ascending dose (SAD) trial in healthy volunteers. We model the probability of dose limiting event (DLE) to guide dose escalation decision. Clinical trial simulation is used to compare BLRM with traditional SAD design, which having a fixed cohort size of 8 (6 on active drug and 2 on placebo). We show that BLRM is more efficient with higher accuracy to identify target dose. We also extend the BLRM to a Bayesian multinomial regression model, which estimate probability of severe, moderate, and mild adverse event simultaneously. Through simulation, we show that this model performs well under different scenarios.