Trials with Irregular and Informative Assessment Times: A Sensitivity Analysis Approach

Many trials are designed to collect outcomes at pre-specified times after randomization. However, in practice, there is often substantial variability in the times at which participants are actually assessed, which poses a challenge to learning about the treatment effects at the targeted assessment times. Any method for analyzing a trial with such irregular assessment times relies on untestable assumptions. Therefore, conducting a sensitivity analysis is an important step that can strengthen conclusions from such trials. However, no sensitivity analysis methodology has been developed. We develop a methodology that accounts for possibly informative assessment times, where assessment at time t may be related to the outcome at that time, even after accounting for observed past history. We implement our methodology using a new augmented inverse-intensity-weighted estimator, and we apply it to a trial of low-income participants with uncontrolled asthma. We also evaluate the performance of our estimation procedure in a realistic simulation study.