External Validity (Generalisability)
External validity is the extent to which a study's result transfers to patients, settings and comparators outside the one studied, a separate question from whether the result is internally correct.
Internal validity asks whether the estimate is right for the sample that produced it, free of bias from confounding, measurement, attrition and analytic choices. External validity asks whether that estimate is a useful guide anywhere else. The two trade against each other: tightening eligibility, standardising adherence and using a placebo comparator all protect internal validity while shrinking the population the answer applies to.
The incretin programmes show the mechanics. STEP 1 studied adults with obesity and excluded type 2 diabetes; that population was tested separately in STEP 2, where the weight loss achieved was smaller. SELECT enrolled only people with established cardiovascular disease and a body mass index of at least 27, so its result speaks to that group. Trials also supply titration support and monitoring that ordinary care does not, and real-world persistence is markedly lower than in-trial retention.
Relative effects transport better than absolute ones, so applying a trial's relative risk reduction to a new population's own baseline risk beats importing the trial's absolute reduction wholesale. That breaks down where the mechanism interacts with a feature of the new population, such as renal function, age, or background therapy already occupying the same pathway.
Two symmetrical errors follow. The first is assuming that a result from rodents, or from a twelve-week split-face study in twenty volunteers, describes what a different population will experience; this is routine in peptide marketing, where a mouse tendon model becomes a human repair claim without comment. The second is rejecting a relevant trial because the participants were not exactly like this patient, when the question is whether there is a specific reason to expect the effect to differ.