American Statistical Association
Recent oncology clinical trials are saturated with the development of targeted therapies. This has led to a growing interest in a class of designs called "basket trials", whereby treatment allocation is biomarker-driven rather than disease-driven. In these studies, investigators are essentially screening for specific populations, i.e. baskets, that respond to an experimental drug or combination of drugs. Previously, we developed a model-free, aggregation approach for such a trial setting, whereby baskets are either treated as independent or aggregated after an interim assessment of heterogeneity of the current response rates across all baskets (Statistics in Medicine, 2017). In this work, we evaluated a popular modeling approach using Bayesian hierarchical modeling, as presented by Berry et al. (Clinical Trials, 2013). During our efforts to calibrate and compare the operating characteristics of the two designs, we found some interesting artifacts of the proposed prior. Most notably, our results suggest the proposed inverse gamma prior is strongly favorable toward declaring homogeneity of the response rates across all baskets; thus, producing a very high overall false positive rate when the drug works in some or even most baskets, but not in all or none of the baskets. In this talk, I will present the results from this investigation and discuss the advantages and disadvantages of model-free versus model-based designs.
|Date:||Wednesday, February 15, 2017|
|Time:||4:00 - 5:00 P.M.|
Memorial Sloan Kettering Cancer Center
Department of Epidemiology and Biostatistics
485 Lexington Avenue
(Between 46th & 47th Streets)
2nd Floor, Conference Room B
New York, New York
**Outside visitors please email firstname.lastname@example.org for building access.
You must be on the security list to enter the floor.