Masanori Kaneko, Hiromi Suzuki, Hiroshi Watanabe, Eiji Oda and Yoshifusa Aizawa
Aims:To evaluated the usefulness of metabolic syndrome (MetS) as a predictor of incident diabetes comparing with other predictors of diabetes including hemoglobin A1c (HbA1c).
Materials and Methods: It is a retrospective five-year follow-up study from annual health examination data in a community including 1,997 Japanese subjects without diabetes at baseline. Logistic regressions using incident diabetes as a dependent variable were calculated for each predictor adjusted for age, sex and drinking status. Area under receiver operating characteristic curves (AUCs) and population attributable risk fractions (PAFs) of incident diabetes were calculated for each predictor.
Results:The odds ratios (95% confidence interval) of incident diabetes for MetS, fasting plasma glucose (FPG) ≥ 5.6 mmol/L and HbA1c ≥ 6.0% were 5.39 (2.72-10.7), 9.52 (5.08-17.9), and 33.5 (13.0-86.4), respectively. The AUCs (95% confidence interval) of diagnosing incident diabetes for FPG, HbA1c, and MetS were 0.82 (0.76-0.88), 0.89 (0.82-0.95) and 0.63 (0.53-0.72) respectively. The optimal cutoff points of FPG, HbA1c, and body mass index were 5.3 mmol/L, 6.0% and 23.5 kg/m2 respectively. The sensitivity and specificity of FPG, HbA1c and MetS for predicting diabetes were 0.67 and 0.80, 0.88 and 0.83, and 0.33 and 0.93, respectively. The PAFs of FPG ≥ 5.3 mmol/L, HbA1c ≥ 6.0%, and MetS were 59%, 86%, and 27%, respectively.
Conclusions: MetS was a poor predictor of diabetes compared with FPG or HbA1c and A1C was the best predictor of diabetes in a general Japanese population.