Practical Considerations in Correspondence to Emerging Regulatory Feedback on Dose Escalation in Early Hematology Clinical Development

• Published in: Blood 2024-11, Vol.144 (Supplement 1), p.7522-7522
• Main Authors: Huang, Baisong, Lu, Cindy, Slade, Daniel, Kurz, Hazel, Jahn, Thomas, Vicente, Sergio, Medeiros, Bruno, Marshall, Jayne
• Format: Article
• Language: English
• ISSN: 0006-4971; 1528-0020
• DOI: 10.1182/blood-2024-201703

Introduction: Bayesian model assisted designs such as Modified Toxicity Probability Interval (mTPI-2) have recently gained popularity in early phase hematology dose escalation (DE) clinical trials with the objective of reliably estimating maximum tolerated dose (MTD) efficiently. Key to the mTPI-2 design is the target dose limiting toxicity (DLT) rate and the equivalence interval (EI), defining a toxicity rate that is considered acceptable and a range in which toxicity rates are considered clinically indifferent, respectively. A target DLT rate of 30% and an EI of (25%, 35%) under mTPI-2 design has generated a prespecified dose escalation/de-escalation framework that has been deemed acceptable from a patient, clinical and regulatory perspectives across oncology indications. Recent regulatory feedback, particularly in studies related to hematologic malignancies have proposed design parameters that resulted in more stringent dose escalation criteria. This includes a target DLT rate of 25% or lower, along with an accordingly adjusted EI and these changes posed a practical consideration on whether the conventional cohort size of three DLT-evaluable participants remains optimal for these alternate design parameters. Methods: To evaluate the effect of cohort size, we conducted simulations of thousands of dose escalation trials employing the mTPI-2 design with a common configuration of 5 dose levels (DLs), a maximum total N = 36 or 48 participants, and a maximum cohort size of n=12. The target DLT rate was set at 25% to align with recent regulatory feedback, and the EI was defined as (20%, 33%). These simulations encompassed various scenarios with monotonically increasing DL vs. toxicity relationship. We compared the performance of the mTPI-2 design under cohort sizes of n=3, 4, or 5, focusing on the following key performance metrics: 1. Reliability: probability of identifying the true MTD 2. DE Efficiency: reliability divided by sample size used for each simulated trial 3. Safety: proportion of patients experiencing DLT or receiving a dose higher than the MTD Results:Our simulation results indicate that, under common scenarios where MTD exists among the defined candidate dose levels (DLs), implementing a cohort size of n=4 resulted in a noticeable improvement in reliability of 5% - 9% in absolute value compared to a cohort size of n=3 (cross-scenario reliability range: 44% - 61%; cross-scenario average sample size range:14.1 - 28.9 for cohort size of n=3), with