‘All of Us’ Research Program’s Diversity Is Off to a Good Start, Data Suggest

The federally funded All of Us research program—at around a third of the way toward its target enrollment of 1 million people—appears to be pretty successful thus far at recruiting participants from racial/ethnic groups that have, over the years, been vastly understudied.

The good news comes from an outside analysis, by investigators unaffiliated with All of Us, that was published online recently in JAMA Cardiology. The data suggest non-Hispanic Black or African American individuals are overrepresented compared with the broader US population. However, non-Hispanic Asian individuals continue to be underrepresented, as do Hispanic or Latino individuals, though to a lesser degree.

Senior author Pradeep Natarajan, MD (Broad Institute of MIT and Harvard, Cambridge, MA, and Massachusetts General Hospital, Boston), speaking with TCTMD, said the research program stands out both for seeking a cohort that better reflects the United States’ diversity and for its engagement with communities that historically have been excluded.

“There’s been a major issue in medical research, and particularly with genetics research, that the populations being studied are not diverse. The problem with that is that has been an important driver of disparities and also a barrier to implementation of findings from [this] research back into the clinic,” he noted.

Natarajan, as an example in the cardiology space, cited the story of PCSK9 inhibitors—drugs that were only developed thanks to the discovery of certain genetic mutations that are more common in Americans of African descent than those of European descent. Polygenic risk scores also are an area where diversity is integral, he added.

“Diversity among participants matters. It allows researchers to get a better, more-comprehensive view of the factors contributing to health and disease,” said Martin Mendoza, PhD, director of health equity for the All of Us research program.

“If we want to change the one-size-fits-all approach to healthcare and advance health equity, we need a more diverse representation in medical research,” he told TCTMD in an email. “All of Us aims to enroll at least 1 million people who reflect the diversity of the United States. This includes racial and ethnic populations, sexual and gender minorities, seniors, those from rural communities, people with disabilities, and people with lower education and economic status.”

The current results, Mendoza said, reflect the program’s enrollment goals.

Natarajan said that, with All of Us ongoing, the study authors hope their analysis can inspire even more “targeted recruitment with initiatives, resources, and support” that can boost enrollment of underrepresented groups, at a minimum matching US Census data but ideally exceeding those proportions. “It’s a tremendous program. . . . Now we have the opportunity to actually help inform [its] generalizability, and we should do the best that we can to improve representation,” he commented.

Some Over, Some Under

Led by Nina Kathiresan (Broad Institute of MIT and Harvard), the investigators analyzed representation of racial/ethnic groups in the All of Us program between May 2017 and June 2022, comparing it to the information collected by the US Census Bureau in 2020. Among the participants with data on sex and age, 61.6% were female (mean age 53.1 years), 38.3% were male (mean age 56.7 years), and 0.1% were nonbinary (mean age 55.4 years).

There were 358,705 adults participating in All of US during that time. Compared with the 2020 Census, the All of US cohort had a higher proportion of non-Hispanic Black or African American individuals but lower proportions of non-Hispanic white, non-Hispanic Asian, and uncategorized or multiracial individuals. The Hispanic or Latino population was just slightly less represented in All of US versus the 2020 Census.

Proportions of Various Racial/Ethnic Groups

There were some geographic variations. The underrepresentation of Hispanic or Latino individuals was most pronounced in Puerto Rico, for example, and while non-Hispanic Black or African American individuals were overrepresented nationwide, they were underrepresented in the District of Columbia.

Additionally, a quarter of the All of US participants had genetic information that could be used to estimate ancestry. When the researchers looked within each self-reported race/ethnicity category, they found that “genetically inferred background did not fully cluster into continentally defined genetic ancestry clusters,” with the greatest heterogeneity among people who reported being Hispanic or Latino.

If we want to change the one-size-fits-all approach to healthcare and advance health equity, we need a more diverse representation in medical research. Martin Mendoza

Mendoza said that overall, the data and trends seen here are similar to what can be accessed on the All of Us website. In terms of outreach, community partnerships at the national, regional, and local levels have been key. “Our partners emphasize medical research as a two-way process where participants not only contribute data but provide guidance and help advise program leaders. Participants also receive valuable information regarding their own genetic health history through their participation,” he explained.

All of Us, a long-term program, is aiming to meet its recruitment goal by the end of 2026. Already, said Mendoza, their data are being used in more than 6,500 projects, and they plan to start enrolling not just adults, but also children.

Natarajan and Mendoza agreed it’s important not to just make the cohort nationally representative, but to go above and beyond by overrepresenting historically understudied communities. With an adequate sample size of these populations, researchers can use the All of Us data to address a broad range of questions, said Mendoza.

For his part, Natarajan highlighted the OurHealth study, set to look at CVD risk in people of South Asian descent, as well as the PRIMED Consortium, a National Institutes of Health-funded program gathering diverse datasets in order to improve the use of polygenic risk scores.