Definitions of Validity Terms for Use in Discussions of Randomized Controlled Trials.

Abstract

We review existing definitions and usages of validity terms and propose a single definition for each term for use in communicating inferences from Randomized Controlled trials (RCTs).

Methods: Two trialists and a statistician reviewed definitions in various dictionaries and literature to identify confusions and propose unified definition for each term.

Results: We propose the following disambiguated and mutually coherent set of definitions for validity terms: TARGET POPULATION: A well-defined population, for whom inferences from an RCT are asserted as valid by the investigators or by other users.

Internal validity: An assertion that an inference from an RCT is at low risk of confounding (Distortion of the effects of the intervention by differences in prognostic factors between arms of the trial other than the difference in intervention exposure).

External validity: An umbrella term asserting that an internally valid inference from an RCT applies to a target population specified by the investigators or by other users, requiring any of three justifications: representativeness, applicability or extrapolatability.

Representativeness: An assertion of external validity of an inference based on the statistical representativeness of the participants, as a random sample of the source (and target) population.

Applicability: An assertion of the external validity of an inference based on subjective assessment of the similarity in context and in distribution of known predictors of outcome between the participant sample and the target population.

Extrapolatability: An assertion of the external validity of an inference from an RCT to a target population based on a common mechanism of action. BIAS: Bias is the opposite of validity and may be internal, due to confounding from systematic error in design, measurement, and analysis; or external, due to mismatch between the population represented by the RCT participants and their setting, and the target population and its context.

Conclusion: With wide uptake, this coherent set of definitions for key terms related to validity could improve understanding and design of RCTs.