Reporting and representation of race/ethnicity in published randomized trials

Abstract

Background: Although adequate representation of specific subgroups (eg, women and the elderly) in randomized controlled trials (RCTs) has been under intense scrutiny, there are few data on representation by race.

Methods: Using all RCTs cited by the 2007 American Heart Association guidelines for cardiovascular disease prevention in women (although trials were included whether or not there were female participants), we explored the extent to which information on race was reported in the baseline characteristics. Race/ethnicity categories were whites, blacks, Asians, Hispanics, and "others."

Results: Overall, 156 trials were analyzed. Demographic data on race/ethnicity were reported in 55 (35%) trials and increased significantly over time (1970s, 12.5%; 1980s, 25%; 1990s, 30.5%; 2000s, 46.2%; P for trend = .011). Among the 55 trials reporting any race/ethnicity information, trial inclusion of whites, blacks, Asians, Hispanics, and "others" was reported in 27%, 13%, 14%, 5%, and 10% of trials, respectively, and increased over time (P for trend < .05 for all). Trials enrolling subjects only in the United States or globally, including the US, were more likely to report race composition than trials that included no US sites (US only 64% vs global 62% vs non-US 21%, P < .01). Industry- and federal/foundation-funded RCTs reported race with similar frequency (industry 36% vs federal 34% vs both 24%, P = .44). When we isolated our analyses to trials that were funded by the National Institutes of Health, 12 (67%) of 18 RCTs reported race/ethnicity as a baseline characteristic.

Conclusion: Although reporting the race/ethnic composition of study populations is increasing over time, two thirds of all RCTs supporting a recent American Heart Association () guideline failed to publish any information on race. A necessary first requirement in translating RCT evidence to patients of all races is information regarding racial demographics. Such information should be strongly encouraged in future publications.