LAUSR

Phase 1 oncology studies focus on safety of novel treatments and identifying a dose associated with acceptable toxicity level. Various model-based designs have been proposed for guiding dose escalation and estimating maximum tolerated dose in dose-finding studies. However, these methods are either excessively conservative or imprudent by allowing overly toxic doses. Transparent and easy to implement model-assisted designs have also received increasing attention but require pre-set rules including perceived dose levels. Therefore, we propose a hybrid model-based design that has a high probability to select MTD with a good balance of overdose control by disentangling in two separate models, which is flexible and easy to implement. Extensive simulations show the model to have real promise.