Intention to treat analysis, compliance, drop-outs and how to deal with missing data in clinical research: a review

Abstract

Background: Pragmatically, intention to treat (ITT) analysis has become the 'gold standard' for analysing the results of clinical trials. Despite its popularity and wide use, ITT is not without its critics, controversies and misunderstandings. To perform an ideal ITT requires a full set of data, where all patients providing data are followed, independent of any protocol deviation. However, most of the time, clinicians and researchers are faced with non-compliance and drop-outs. Thus, researchers should be familiar with the concepts associated with ITT and strategies to deal with missing data. Objectives: The objective of this review is to clarify and summarise the important aspects of ITT limitations, and contributions to clinical research. In addition, the concepts of effectiveness and efficacy will be discussed in the context of randomised controlled trial (RCT) analysis. This will help clinicians and researchers to have a greater understanding of ITT and apply this knowledge when designing, evaluating, and reporting RCTs. Conclusions: Depending on the objective of the trial, different approaches to data analysis could be used. If the trial's objective is purely explanatory, an 'as treated' or 'per protocol' analysis could be a reasonable option. Nevertheless, it is advised that these approaches are not reported on their own because of their susceptibility to bias when estimating treatment effects. If the objective of the trial is pragmatic, that is, it addresses the effectiveness of a specific treatment in real life (clinical setting), an ITT should be the primary analysis.