OBJECTIVE Although metabolic syndrome (MS) is considered to be a predictor of diabetes and cardiovascular disease, its value for predicting diabetes beyond the levels of fasting glucose (FG) and hemoglobin A1c (HbA1c) remains questionable. In this study, we evaluated the usefulness of MS and Japanese MS (JMS) as predictors of diabetes. METHODS This study is a longitudinal study using data for 2,034 subjects who visited our Medical Check-up Center in both 2008 and 2011 and were free from diabetes at baseline. The odds ratios (ORs) of developing diabetes in 2011 for MS and JMS were calculated after being adjusted for FG, age, sex and antihypertensive and antihyperlipidemic medication use, then further adjusted for HbA1c. The optimal cutoff points for FG, HbA1c and body mass index (BMI) to discriminate the development of diabetes and their sensitivities/specificities were obtained from receiver operating characteristic curves. These sensitivities/specificities were compared with the sensitivities/specificities of MS and JMS. RESULTS The ORs (95% confidence intervals (CIs)) of developing diabetes for MS and JMS were 1.36 (0.57-3.23) (p=0.484) and 3.23 (1.25-8.36) (p=0.016), respectively, after being adjusted for FG and 1.06 (0.39-2.86) (p=0.915) and 2.34 (0.79-6.95) (p=0.126), respectively, after being further adjusted for HbA1c. The sensitivities/specificities of FG, HbA1c, BMI, MS and JMS were 0.625/0.897, 0.828/0/919, 0.625/0.734, 0.375/0.911, and 0.375/0.941, respectively. CONCLUSION MS and JMS were found to be poor predictors of diabetes in a Japanese health screening population.