Understanding ethnic diversity in open dementia neuroimaging datasets

Abstract Ethnic differences in dementia are increasingly recognised in epidemiological measures and diagnostic biomarkers. Nonetheless, ethnic diversity remains limited in many study populations. Here we provide insights into ethnic diversity in open access neuroimaging dementia datasets. Datasets comprising dementia populations with available data on ethnicity were included. Statistical analyses of sample and effect sizes were based on the Cochrane Handbook. 19 databases were included, with 17 studies of healthy groups or combination of diagnostic groups if breakdown was unavailable, and 12 of MCI and dementia groups. Combining all studies on dementia patients, the largest ethnic group was Caucasian (20,547 participants) with the next most common being Afro-Caribbean (1958), followed by Asian (1211). The smallest effect size detectable within the Caucasian group was 0.03, compared to Afro-Caribbean (0.1) and Asian (0.13). Our findings quantify the lack of ethnic diversity in openly available dementia datasets. More representative data would facilitate the development and validation of biomarkers relevant across ethnicities.