Poor dose optimization in cancer trials leads to poor patient outcomes in tumor suppression and drug tolerance as well as failures in the drug development process. Most phase I clinical cancer trials still use traditional dose-finding methods, which are inadequate for evaluating novel cancer therapies, such as molecularly targeted agents and immunotherapies. Traditional approaches include using rule-based designs instead of model-based designs, assuming one dose should be recommended to all patients, and assuming the higher the dose, the better. This dissertation aims to address each of the inefficiencies that exist in phase I trials to optimize patient and trial outcomes in oncology, specifically in settings where the patient population is heterogeneous, i.e., settings where eligibility criteria have been expanded or settings evaluating a therapy that targets multiple tumor types and mutations.
In the first part of this work, we address the inefficiencies of rule-based designs and the barrier to implementation of model-based designs. We use published phase I trials that used the most common rule-based method, the 3+3 design, and compare the trial outcomes to those obtained with novel model-based designs. In the second part of the work, we propose a broadened eligibility dose-finding design to address the situation of unknown patient heterogeneity in phase I cancer trials where eligibility is expanded, and multiple eligibility criteria could potentially lead to different optimal doses for patient subgroups. Lastly, we address patient heterogeneity in efficacy by developing a dose-optimization design that accounts for patient-specific characteristics, toxicity, pharmacokinetic data, and efficacy to identify the target population and inform the optimal dose for each subpopulation.
The findings in each work highlight the advantages of model-based designs, particularly when tailored for the therapy and patient population in question. Using published dose-finding trials, we show that novel designs would recommend different doses about 40% of the time and confirm the advantages of these designs compared with the 3 + 3 design, as suggested previously by simulation studies. When accounting for heterogeneity in toxicity, the broadened eligibility design identified when the expanded subpopulation should be recommended a lower dose due to their tolerance and identified the criteria affecting toxicity at least 60% of the time in simulation studies. The dose-optimization design, focusing on heterogeneity in efficacy, demonstrated that a model-based approach to identifying the target population can be effective. Further, in the presence of heterogeneity, patient characteristics relating to molecular tumor characteristics were identified correctly, and a different optimal dose was recommended for each identified target subpopulation. The simulation studies of all proposed designs show that accounting for heterogeneity, even when the source of heterogeneity is unknown, is beneficial. In addition, the simulation studies highlight the poor performance of a naive method that recommends one dose for all.
Our findings in this dissertation reveal the large proportion of the patient population that will be incorrectly dosed if inappropriate dose-finding designs are used. While we cannot directly understand the effect of dose selection on cancer trial outcomes, it is likely that not handling characteristics of novel cancer therapies early on contributes to the high attrition rates of cancer trials and the toxicity burden encountered in later trials and post-approval studies.
| Identifer | oai:union.ndltd.org:columbia.edu/oai:academiccommons.columbia.edu:10.7916/45rg-4p32 |
| Date | January 2023 |
| Creators | Silva, Rebecca Bryn |
| Source Sets | Columbia University |
| Language | English |
| Detected Language | English |
| Type | Theses |
Page generated in 0.002 seconds