Accounting for Carryover Toxicity in Phase I Clinical Trials with Intra-Patient Dose Escalation

Abstract

Intra-patient dose escalation (IPDE) provides a strategy for more efficient Phase I clinical trials. However, IPDE poses additional challenges due to the need to account for carryover toxicity from previous dosings a patient has received. To that end, we propose two CRM-based approaches to IPDE that incorporate potential carryover toxicity. We compare these methods to the CRM without IPDE and the AIDE-CRM, an existing Bayesian adaptive approach to IPDE. In simulations across a range of scenarios, we show that our approaches have similar operating characteristics to the CRM without IPDE, but with a 20% reduction in time and participants needed to complete the trial; these results hold even in the presence of strong carryover toxicity that hinders the performance of the AIDE-CRM.

Keywords: Bayesian adaptive design; Phase I; dose finding; intra‐patient dose escalation.

FIGURE 1

Effective dose after an initial dose of 10 mg and a second dose of 15 mg is 20 mg when $\alpha =0.5$.

FIGURE 2

Example trial following escalation procedure.

FIGURE 3

True dose–response curves of simulation scenarios.

FIGURE 4

Endpoints in each scenario across levels of $\alpha$. Correct selection percent is the percent of trials that selected the true MTD in Scenarios 1–5 and the percent of trials that stopped early for safety in Scenario 0.