Better Understanding of Heterogeneous Treatment Effects
A 2018 symposium panel with researchers from FDA, Johns Hopkins, and two private companies say the design and analysis of clinical studies play critical roles in evaluating and characterizing heterogeneous treatment effects. A special issue paper in the journal Pharmaceutical Statistics says researchers “can do more or better at understanding heterogeneous treatment effects and providing the best information on heterogeneous treatment effects.” The paper includes information from the several presentations made at the panel session, reactions to the panel presentations, audience questions and panelists’ responses, and conclusion by panel moderator Mark Rothmann from the CDER Office of Biostatistics.
“When we have statistical models to predict individual patient outcomes,” Rothmann said, “we use the data from all subjects. It would make sense that if we wanted to predict treatment effects for individuals or subgroups, we would use the data from all subjects…. More complicated and better models could also be used that include effect modification…. I think we must move away from this dichotomy that we have had of either using the overall population estimate and its corresponding 95% confidence interval or just use the raw subgroup estimate and its corresponding 95% confidence interval…. We should expect different effects across subgroups and individuals.”