Quantifying Heterogeneity in Phase III Oncology Clinical Trials

Basket trials have large potentials to save costs during the drug development process, through testing  novel drugs on cancers located in different parts of the body at once. Despite having similar biomarkers and underlying biological mechanisms, grouping treatments not based on the locational differences of the cancer is considered novel.  The heterogeneity between indications in a basket trial may be a concern when pooling results in a confirmatory set. As a result, the subgroup analysis conducted in this study aimed to outline that heterogeneity has always been present in traditional and well reputed phase III trials. If the level present can be quantified, it could influence the acceptable level of heterogeneity in confirmatory basket trials potentially tolerable by the health authorities. Through quantifying the level of heterogeneity present in traditional trials, future inquiries comparing the levels of heterogeneity to basket trials could be possible.The study focused on 30 randomized phase III oncology clinical trials that were approved recently (2020-2023), located through looking through recent FDA approvals. The studies had to meet a filter criteria for subgroups and time to event endpoints before a  meta analysis was conducted. Heterogeneity was defined as how much of the overall benefit was retained (retained delta) for each subgroup in the forest plot relative to the delta of the overall hazard ratio. Benefit of the treatment (delta) was defined as 1 - hazard ratio. This was also applied to upper and lower bound 95% confidence intervals of the hazard ratio. The number of events and patients for each subgroup was also recorded. The studies were also categorized based on the type of intervention being tested: Chemotherapy, Immunotherapy,  Targeted therapy, or a combination of the three Additionally, subgroups that are N/4 or larger will be investigated to determine if heterogeneity is still present in large amounts or not. The results and analysis will be presented at the conference.