To examine differences in the performance of HbA1c for diagnosing diabetes in Arabs compared with Europeans.Methods
The PubMed, Embase and Cochrane library databases were searched for records published between 1998 and 2015. Estimates of sensitivity, specificity and log diagnostic odds ratios for an HbA1c cut-point of 48 mmol/mol (6.5%) were compared between Arabs and Europeans, using a bivariate linear mixed-model approach. For studies reporting multiple cut-points, population-specific summary receiver operating characteristic (SROC) curves were constructed. In addition, sensitivity, specificity and Youden Index were estimated for strata defined by HbA1c cut-point and population type. Database searches yielded 1912 unique records; 618 full-text articles were reviewed. Fourteen studies met the inclusion criteria; hand-searching yielded three additional eligible studies. Three Arab (N = 2880) and 16 European populations (N = 49 127) were included in the analysis.Results
Summary sensitivity and specificity for a HbA1c cut-point of 48 mmol/mol (6.5%) in both populations were 42% (33–51%), and 97% (95–98%). There was no difference in area under SROC curves between Arab and European populations (0.844 vs. 0.847; P = 0.867), suggesting no difference in HbA1c diagnostic accuracy between populations. Multiple cut-point summary estimates stratified by population suggest that Arabs have lower sensitivity and higher specificity at a HbA1c cut-point of 44 mmol/mol (6.2%) compared with European populations. Estimates also suggest similar test performance at cut-points of 44 mmol/mol (6.2%) and 48 mmol/mol (6.5%) for Arabs.Conclusions
Given the low sensitivity of HbA1c in the high-risk Arab American population, we recommend a combination of glucose-based and HbA1c testing to ensure an accurate and timely diagnosis of diabetes.