A new 'mechanistic-practical" framework for designing and interpreting randomized trials.

Methodologists have traditionally categorized randomized controlled trials (RCTs) as explanatory (representing the ideal setting) and pragmatic (representing the real-world setting). Although this framework has greatly advanced the design and interpretation of RCTs, current interpretations of the explanatory-pragmatic framework suffer from two major limitations. First, they confound purpose with structure. Second, they ignore the varying perspective of those using RCT results to make clinical and policy decisions in the real world. The purpose of a trial should determine researchers' choices regarding the trial's structure and the structure of a trial determines the extent to which a decision maker will find the results useful. In this article, we introduce two terms that refer explicitly to the purpose of a trial: A trial is mechanistic to the extent that it addresses a biological relationship. In contrast, a trial is practical to the extent that it provides comprehensive information that bears directly on specific health care decisions. This revised framework facilitates investigators' choice of optimal trial design, and clinicians' optimal interpretation of RCT results. If our goal is clinical trials most relevant to individual patient decision making, we will eschew the use of trials that enroll patients unlikely to benefit (e.g., those with uncertain diagnosis); those likely to be noncompliant; treated by practitioners whose differing expertise is likely to result in differing outcomes; and permitting cointerventions that are likely to influence treatment effectiveness-i.e., the conventional pragmatic trial. Instead we will design, implement, and apply the results of practical trials to our patients.