Trabecular bone score (TBS), a noninvasive textural analysis of the lumbar spine dual-energy X-ray absorptiometry (DXA) image, has been shown to predict fractures in Caucasian (CA) populations but has not been thoroughly studied in African-American (AA) populations. The aim of this study was to compare the TBS in AAs and CAs and to assess whether TBS can be used to refine fracture risk stratification in AA patients. Eight hundred twenty-five women (390 AAs, 435 CAs) referred for bone mineral density (BMD) as part of their clinical care had measurements of the TBS, the BMD of the lumbar spine, total hip, and femoral neck, and vertebral fracture assessment for detection of vertebral fractures. Unadjusted TBS was higher in CA than AA (1.275 vs 1.238, p < 0.001), but this was no longer true after adjusting for age and tissue thickness. Interestingly, differences in TBS were still highly significant in those under 60 yr of age even with adjustment for tissue thickness, but not in older subjects. There were 74 CAs and 56 AAs with vertebral fractures on vertebral fracture assessment (17% vs 14%, p = 0.30). In CA, the odds ratio (OR) for prevalent vertebral fracture per SD decrease in TBS was 2.33 (p < 0.001), whereas in AA, the OR was 1.43 (p = 0.02). In a multivariate logistic regression model that also included age, BMD T-score, and glucocorticoid use, the association between TBS and prevalent vertebral fractures was still highly significant in CAs (OR 1.54, p = 0.008) but not in AAs (OR 1.23, p = 0.21). Our results suggest that TBS may be less discriminatory in regard to fracture risk in AAs than in CAs and that TBS may need to be used differently in these 2 ethnic groups.