A treatment shown to work is shown to work in the population studied. Where enrolment is uneven, the evidence supporting everyday prescribing is thinner for some groups than others.
Trials describe their participants
A trial estimates an effect within the group recruited. Extending that estimate to different populations is an inference, not a direct finding.
The inference is usually reasonable, since biological mechanisms often generalise. It is weaker where differences in metabolism, comorbidity or concurrent medication are plausible.
Regulators and guideline bodies note these limits, but the caveat rarely travels as far as the headline result, so it seldom reaches the point of prescribing.
Eligibility criteria narrow the pool
Protocols exclude participants with additional conditions, those taking particular medicines, and often those outside a defined age band, to reduce variability in the result.
Those exclusions remove exactly the people who most commonly receive the treatment afterwards, since real patients frequently have several conditions at once.
Pragmatic trial designs relax criteria to recruit a population resembling ordinary practice, at the cost of a noisier estimate requiring larger numbers.
Practical barriers filter who can take part
Participation typically requires repeated visits during working hours, travel to a research site and time away from work or caring responsibilities.
Sites are concentrated in large academic centres, so proximity itself shapes who is invited. People living further away are less likely to be approached at all.
Reimbursing costs, offering evening appointments and running trials through community clinics all widen recruitment, though each adds expense to the study budget.
Trust is shaped by history
Some communities have documented experience of research conducted without consent or without benefit returning to them, and that history informs current willingness to participate.
Recruitment approaches that begin with community organisations and explain the study before enrolment opens tend to work better than late outreach.
Returning results to participating communities also matters, since a study that takes data and reports nothing back reinforces the original reluctance.
Reporting requirements are tightening
Regulators and journals increasingly expect enrolment to be reported by demographic group and expect sponsors to explain recruitment plans in advance.
Reporting alone does not change composition, but it makes gaps visible and gives guideline committees a basis for stating where evidence is limited.
Requirements differ between regulatory systems and are being revised, so what a sponsor must collect and publish depends on where approval is sought.