‘Racial’ differences in genetic effects for complex diseases John P A Ioannidis 1–3 , Evangelia E Ntzani 1 & Thomas A Trikalinos 1,3 ‘Racial’ differences are frequently debated in clinical, epidemiological and molecular research and beyond 1,2 . In particular, there is considerable controversy regarding the existence and importance of ‘racial’ differences in genetic effects for complex diseases 3–6 influenced by a large number of genes 7 . An important question is whether ancestry influences the impact of each gene variant on the disease risk. Here, we addressed this question by examining the genetic effects for 43 validated gene-disease associations across 697 study populations of various descents. The frequencies of the genetic marker of interest in the control populations often (58%) showed large heterogeneity (statistical variability) between ‘races’. Conversely, we saw large heterogeneity in the genetic effects (odds ratios) between ‘races’ in only 14% of cases. Genetic markers for proposed gene-disease associations vary in frequency across populations, but their biological impact on the risk for common diseases may usually be consistent across traditional ‘racial’ boundaries. ‘Race’ is difficult to define 8–11 and inconsistently reported in the literature 12 . Thus we focused on major, distinct ‘racial’ groups that are unlikely to be influenced by reporting and in which most gene-disease association research has been done so far. We compared frequencies of genetic markers of interest and their genetic effects across European, East Asian, African and other populations. We screened 134 meta- analyses of genetic association studies for various diseases (Supple- mentary Note and Supplementary Table 1 online) compiled by updating a previous database 13,14 . Of those, 12 were covered by another more comprehensive meta-analysis, 36 had no sufficient data from at least two ‘racial’ groups, and in 3, o80% of the available studies could be classified by ‘race’. Of the remaining 83 studies, 40 had no statistically significant results either overall or for any of the included ‘racial’ groups. We analyzed the remaining 43 gene-disease associations for which either the overall meta-analysis showed statis- tically significant results (P o 0.05; n ¼ 32) or there were statistically significant results for at least one ‘racial’ group without the overall meta-analysis reaching formal significance (n ¼ 11). We included the latter group, even though their validity is somewhat less certain, to try to avoid biasing the results towards our null hypothesis of no ‘racial’ differences in genetic effects. If anything, these gene-disease associa- tions may show larger differences in genetic effects between ‘racial’ groups than truly exist. The 43 eligible meta-analyses (Table 1) included ‘race’-specific data on 697 gene-disease association studies (European, n ¼ 479; East Asian, n ¼ 139; African, n ¼ 47; other, n ¼ 32) with total sample size of 297,411. All meta-analyses included data on individuals of European descent, and 34, 24 and 18 also had data on subjects of East Asian, African and other descent (Jewish, n ¼ 12; Turkish, n ¼ 7; Arabic, n ¼ 3; Hindu, n ¼ 8; Native Indian, n ¼ 1; Hawaiian, n ¼ 1), respectively. Heterogeneity between studies was almost ubiquitous for the frequencies of the genetic markers of interest in the control popula- tions and relatively common for the magnitude of the genetic effects (odds ratios) when all studies were considered (Table 2). Thirty-six (84%; 95% binomial confidence interval (c.i.) ¼ 69–93%) meta- analyses had statistically significant heterogeneity (based on w 2 or Fisher’s exact test, as appropriate) in the frequencies of the genetic markers of interest in the control populations across all studies. Seven (16%) meta-analyses had statistically significant between-study het- erogeneity across all studies, but not within any of the descent groups. Significant between-study heterogeneity in one descent group, but not across all studies, was never seen (Table 2). Twenty-two (51%; 95% c.i. ¼ 35–67%) meta-analyses had statistically significant hetero- geneity in the odds ratios across all studies based on the w 2 -based Q statistic. Only two meta-analyses had statistically significant between-study heterogeneity across all studies, but not within any descent groups; another two meta-analyses had statistically signifi- cant between-study heterogeneity in some descent groups but not across all studies (Table 2). ‘Racial’ group–specific frequencies of the genetic markers of interest in control populations are shown in Figure 1 (ref. 15). The between-group variance was larger than the within-group variance in 24 meta-analyses, and the opposite was true in 19 cases (Supplementary Table 2 online). In pairwise comparisons (Supple- mentary Table 3 online), we observed statistically significant differ- ences in 20 of 34 (59%) comparisons of European versus East Asian descent, 12 of 19 (63%) comparisons of European versus African descent and 10 of 21 (48%) comparisons of East Asian versus African descent. Heterogeneity may be influenced by the number of available studies in each assessment. Therefore, we also evaluated the I 2 statistic, Published online 14 November 2004; doi:10.1038/ng1474 1 Clinical and Molecular Epidemiology Unit, Department of Hygiene and Epidemiology, University of Ioannina School of Medicine, Ioannina 45110, Greece. 2 Biomedical Research Institute, Foundation for Research and Technology-Hellas, Ioannina, 45110, Greece. 3 Institute for Clinical Research and Health Policy Studies, Tufts-New England Medical Center, Tufts University School of Medicine, Boston, Massachusetts 02111, USA. Correspondence should be addressed to J.P.A.I. (jioannid@cc.uoi.gr). 1312 VOLUME 36 [ NUMBER 12 [ DECEMBER 2004 NATURE GENETICS LETTERS © 2004 Nature Publishing Group http://www.nature.com/naturegenetics