*Geek Box: Pragmatic RCTs

*Geek Box: Pragmatic RCTs

When we talk about “randomised controlled trials”, it is often as if this term refers to a single type of design. And if you were to ask anyone who reads research what a “good” RCT looks like, they would likely list off certain methodological qualities, e.g., double-blinding and placebo-control.

However, RCTs exist on a spectrum. The design characteristics we typically associate with RCTs – randomisation, double-blinding, placebo-control – and the related assumptions [e.g., independence of effects] are what are known as “explanatory RCTs”. These RCT designs are aimed at high levels of internal validity, which are the design characteristics considered important to demonstrate efficacy, i.e., that the treatment A produces outcome B under ideal circumstances. In short, explanatory trials that test efficacy are asking, “does this work?”

On the other end of the spectrum are what are known as “pragmatic RCTs”. These trials are aimed at demonstrating external validity, which are the characteristics required to demonstrate effectiveness, i.e., how effective the treatment is in the real-world setting in which it will be applied. In short, pragmatic trials that test effectiveness are asking, “will this work to the same extent in real life?”

Trials are not simply dichotomised along explanatory/efficacy – pragmatic/effectiveness lines, and so it is important to note that the word ‘spectrum’ here is not semantic. For example, a tightly controlled metabolic ward study may demonstrate that a certain diet, when specifically prepared and catered to participants, results in a particular outcome. Would that effect be observed, or observed to the same extent, if participants had to prepare and adhere to the diet in their day-to-day life? Such a study would not answer that question and would therefore be more to the purely explanatory end of the spectrum.

However, a study in which the intervention was delivered through primary care where the participants knew they were receiving the treatment would be more to the pragmatic end of the spectrum, as it would be the exact “real world” scenario in which the treatment would be applied.

Where a trial is designed to be a pragmatic RCT, factors like double-blinding are less of a concern to the methodological quality of the trial; single-blind or blinding of assessors may suffice. However, the quality of a pragmatic RCT will always be bolstered by taking extra steps to minimise bias. The Lyon-Diet Heart Study is a great example of a pragmatic RCT in nutrition; participants knew they were being asked to follow the intervention diet, but were unaware they were in a comparative trial. The researchers also kept the participant’s doctors blinded to the fact that their patient was taking part in the trial, to avoid any undue changes to medications or routine care. Both explanatory and pragmatic RCTs have value for nutrition research, but with current knowledge the field would benefit from an emphasis on more pragmatic designs.