An intent-to-treat method for enhancing analysis of clinical trials with rescue medication: a mixed-model approach

An intent-to-treat (IT) analysis includes all observed data, even those obtained after the primary treatment assignment is changed or rescue treatment is substituted. The so-called "pragmatic" approach accounts for this type of treatment change or rescue treatment. The advantage of the pragmatic approach to IT analysis is that the timing and duration of the rescue treatment can be included in the analysis with mixed-effect modeling, which estimates the difference between the randomized treatment arms while adjusting for the use (timing and duration) of the subsequent treatment. We illustrate the inclusion of information on rescue treatment timing and duration in the analysis of a two-arm randomized clinical trial comparing antidepressant medications in an elderly depressed population. In this study, the randomized treatments are found to be similar. This enhanced analysis, where information about rescue treatment is implicitly added in the model, corroborates our previous two-piece spline analysis, where information about rescue treatment was not included. In addition, a significant drug by rescue interaction indicates a differential effect of the timing of rescue treatment between the two treatments. This analytic approach illustrates the importance of collecting and incorporating data after rescue, including the timing of rescue.