Internal vs External Validity

Internal validity asks whether the result is true within the studied sample — free of bias, confounding and chance, so the measured effect really is caused by the intervention. External validity (generalisability) asks whether that true result transfers to other patients, settings and time. They are sequential: a study must first be internally valid before external validity is even worth asking — but high internal validity never guarantees it.

A two-gate pipeline. Gate one (internal validity) checks the result is real inside the trial. Gate two (external validity) checks it travels to the patient in front of you. A trial can sail through the first gate — flawless randomisation, blinding, follow-up — and still fail the second because its participants look nothing like your ED population.

Read it top-down and never skip a gate. If internal validity fails, external validity is irrelevant — a biased result generalises nothing. If internal validity holds, then scrutinise the inclusion/exclusion criteria, setting and case-mix: were the elderly, the comorbid, the “real” ED patients excluded? The narrower the entry door, the more cautiously you extrapolate.

It is the bridge from a “positive trial” to a bedside decision. Tight explanatory RCTs maximise internal validity but, by enrolling a narrow, ideal population, can sacrifice external validity — efficacy under perfect conditions need not equal effectiveness in your department. Forgetting the second question is how evidence is misapplied to patients it was never tested on.

• Internal = true HERE (no bias / confounding / chance)
• External = applies THERE (other patients & settings)
• Order: internal first — then ask external
• Explanatory RCT: high internal, often lower external validity

Explanatory vs pragmatic trials — an explanatory RCT (does this drug work under ideal conditions?) uses strict inclusion criteria, expert centres and rigid protocols. That buys strong internal validity but a narrow, unrepresentative sample — modest external validity. A pragmatic trial deliberately enrols a broad, real-world population across ordinary departments, trading a little internal control for results that generalise to everyday practice. A landmark efficacy trial in a hand-picked population can be entirely internally valid yet still not answer “will this help the patient in resus tonight?”

• The internal–external trade-off — tightening entry criteria raises internal validity but narrows generalisability; the two pull against each other.
• Over-generalising efficacy trials — treating “works in the trial” as “works in my unselected patients”.
• Narrow selection — heavy exclusions (elderly, comorbid, pregnant) limit who the result can apply to, however clean the methods.