Can "realist" randomised controlled trials be genuinely realist?

Abstract

In this paper, we respond to a paper by Jamal and colleagues published in Trials in October 2015 and take an opportunity to continue the much-needed debate about what applied scientific realism is. The paper by Jamal et al. is useful because it exposes the challenges of combining a realist evaluation approach (as developed by Pawson and Tilley) with the randomised controlled trial (RCT) design. We identified three fundamental differences that are related to paradigmatic differences in the treatment of causation between post-positivist and realist logic: (1) the construct of mechanism, (2) the relation between mediators and moderators on one hand and mechanisms and contexts on the other hand, and (3) the variable-oriented approach to analysis of causation versus the configurational approach. We show how Jamal et al. consider mechanisms as observable, external treatments and how their approach reduces complex causal processes to variables. We argue that their proposed RCT design cannot provide a truly realist understanding. Not only does the proposed realist RCT design not deal with the RCT's inherent inability to "unpack" complex interventions, it also does not enable the identification of the dynamic interplay among the intervention, actors, context, mechanisms and outcomes, which is at the core of realist research. As a result, the proposed realist RCT design is not, as we understand it, genuinely realist in nature.