Model-Based Designs for Safety

Traditional designs often utilize dichotomous toxicity outcomes for dose estimation, i.e., adverse events (AEs) are classified as either dose-limiting toxicities (DLTs) or non-DLTs based on prespecified criteria. While this is the most common choice for categorizing toxi-cities, an obvious flaw of dichotomizing outcomes is the loss of valuable information. Thus, ordinal grading and continuous toxicity scores have been proposed and developed for early-phase dose estimation algorithms.