Equity and representation
Why representation is a scientific question too
If an effect varies with a factor that is distributed differently in the study and the target population, the estimate does not transport. That makes representation an issue of validity for a defined target, not only of fairness, and it connects directly to the generalizability question rather than sitting beside it.
Where exclusion comes from
- Eligibility criteria that exclude comorbidity, age, pregnancy, renal impairment or non-fluent language speakers, often with no specific safety rationale.
- Recruitment concentrated at academic centres, which serve a particular population.
- Participation costs borne by the participant: travel, time off work, childcare.
- Trust, shaped by documented historical harms, which is addressed by conduct over time rather than by a leaflet.
- Digital-only recruitment or follow-up, which selects on access and literacy.
What to do at design time
State the target population and justify every exclusion against it. Design recruitment to reach the people the treatment is for, including community sites. Remove participation costs. Provide interpretation and translated materials. Involve the community in the design rather than at the recruitment stage. Where a subgroup matters scientifically, consider powering for it rather than reporting an underpowered subgroup analysis.
Reporting
Report enrolment against the eligible population, not against the general population, so the comparison means something. Report outcomes disaggregated where the numbers support it, and say plainly when they do not. And return results to the participating communities, which is part of respect for participants rather than an optional extra.
What to watch out for
Two opposite errors sit under this heading. Exclusion produces evidence that does not cover the patients who will receive the treatment, and it is usually the product of eligibility criteria and recruitment convenience rather than any decision to exclude. Extraction is the other: recruiting from a community that then gets no access to what the research produces. Note also that reporting a diverse enrolment is not the same as reporting whether the effect differed, and a trial is rarely powered for the second.
Clinical example
A trial excludes patients with an eGFR below 45 for no stated safety reason, and recruits at three tertiary centres. The drug will be prescribed largely in primary care to older patients with reduced renal function, which is the group the trial says nothing about.
- Strategies to improve diversity, equity, and inclusion in clinical trials (opens in a new tab) · Cancer · doi 10.1002/cncr.33905
- What makes clinical research ethical? (opens in a new tab) · JAMA · doi 10.1001/jama.283.20.2701