Bartlett · Health technology assessment (Winchester, England) 2005 · Systematic review, meta-analysis, and record-linkage cohort study · n=52 RCTs and 134,598 cohort participants

The causes and effects of socio-demographic exclusions from clinical trials.

Cited 252 times in the scientific literature.

Level 1 - systematic review of randomized trials

Systematic review and meta-analysis of randomized controlled trials combined with cohort analyses

PubMed 16181564 · doi:10.3310/hta9380 · record verified 2026-08-29

What was done

The authors examined exclusions of women, older people, and minority ethnic groups in healthcare research using two exemplars: statins for secondary prevention of coronary heart disease (CHD) and non-steroidal anti-inflammatory drugs (NSAIDs) for osteoarthritis (OA). They reviewed literature, held three stakeholder workshops, and analyzed 27 randomized controlled trials (RCTs) of statins and 25 RCTs of NSAIDs. Trial demographics were compared against epidemiological need and usage estimated from English routine data, the Somerset and Avon Survey of Health (1996-97), and a Scottish record-linkage cohort (3,188 CHD patients and 131,410 individuals dispensed NSAIDs). Evidence synthesis and risk modelling evaluated the impact of trial exclusions on relative and absolute outcomes.

What was found

In statin trials, mean participant age was 58.5 years and 16.3% were women, whereas in England women represented 45% of the ~537,000 annual incident CVD cases needing statins, and two-thirds were aged 65 or older. Statins reduced CVD incidence by about 25% in both men and women, with benefits extending to age 75. In NSAID trials, mean age was 61.9 years and 68.5% were women; gastrointestinal (GI) side effects were commonly reported, but renal adverse effects were not. In the Scottish cohort, high NSAID exposure increased GI side-effect risk by ~50% and renal impairment risk by ~140%; side-effect risk increased with age and was lower in women. Ethnicity was poorly reported in both trial sets. US trials were more inclusive than UK/European trials.

Why it matters

While demographic under-representation in trials did not alter the relative risk reduction of statins, baseline event rates differ substantially by age, sex, and ethnicity, meaning that excluding these groups directly biases absolute effectiveness, harm, and cost-effectiveness estimates.

Limits

The analysis was limited to two specific clinical indications (statins for secondary CHD prevention and NSAIDs for OA pain). Inadequate trial reporting of race and ethnicity precluded quantitative synthesis of ethnic exclusions, and adverse event patterns for socio-demographic subgroups were detectable in observational data but omitted or underreported in RCTs.

Cited by