Assessing the Causal Effect of Treatment in the Presence of Self-Selection of Dosage
*Xin Gao, FDA
Keywords: adverse event, causal modeling, dose tolerance, principal stratification, potential outcome.
To make drug therapy as effective as possible, patients in clinical trials are often put on an escalating dosing schedule. But patients may choose to take a lower dose because of side effects. Therefore, even if the dose schedule is randomized, the dose level received is a post-randomization variable, and comparison between the treatment arm and the control arm for a specific dose received may no longer have a causal interpretation. We use the potential outcomes framework to define pre-randomization “principal strata" from the distribution of dose tolerance under treatment arm, with the goal of estimating the causal effect of treatment within the subgroups of the population who tolerated a given level of treatment dose. Adverse events are included in the model to help identify subjects' principal strata membership. Inference is obtained by treating the outcomes, doses, and adverse events under the unobserved randomization arm as missing data and using multiple imputation in a Bayesian framework. Results from simulation studies imply that the proposed causal model provided valid inferences of the causal effect of treatment within each principal stratum under certain reasonable assumptions. We apply the proposed model to a randomized clinical trial with escalating dosing schedule for the treatment of interstitial cystitis.
Important Dates & Deadlines
- October 9 - 11, 2013