Rethinking about the dose limiting toxicities (DLTs): They can be equivocal!

Abstract

3064 Background: The primary goal of oncology phase I trials is to determine the maximum tolerated dose (MTD) in a small number of patients. Within the standard design methodology currently employed there remains many challenges and one pertains to the variability around defining dose limiting toxicities (DLTs). The pre-specified DLT criteria may not well reflect the clinician’s practical experience with some adverse events, while others may actually be related to the disease status more than the investigational drug and could be misclassified. DLT misclassification seen in testing a small sample size from each dose group could generate a large bias on dose finding and the MTD estimation. Methods: To mitigate the risk of dichotomizing and misclassifying DLTs, we proposed a strategy that introduced the new concept of “equivocal” DLT or AE. A novel dose escalation approach is applied to increase the variability associated with less interpretable AEs so that the model recommendations are more weighted towards the unequivocal AEs/DLTs. To evaluate this novel approach, we established a framework incorporating two types of systematic measurement errors on DLT misclassification, one for the misclassified DLT that is not related to the drug treatment while the other for the non-DLT AE that should be considered severe and relevant to dose finding. In our simulation studies, the Bayesian logistic regression model (BLRM) was used to guide dose escalation in simulated trials to compare the novel weighting approach with the traditional approach. A few numerical examples were also included for method illustration. Results: For different types of measurement errors, simulation studies showed that the weighting approach could successfully improve the trial performance, with higher chance of finding the correct MTD and treating more patients at the MTD level. Conclusions: The DLT weighting strategy provides a flexible but powerful tool that may incorporate the clinician’s valuable experience on some specific DLTs/AEs and improve MTD estimation in oncology phase I dose-escalation trials.