Objectives There have so far been few papers studying the metabolic syndrome (MetS) prevalence rate in Japanese patients with schizophrenia. We studied the MetS prevalence rate in Japanese controls and inpatients with schizophrenia and compared the prediction factors for the occurrence of MetS. Methods The subjects were 319 inpatients with schizophrenia and 154 controls. The predictive utilities of body mass index (BMI) and the individual components of MetS for MetS diagnosis were evaluated. Results The prevalence of MetS did not differ between schizophrenia and control subjects. Subjects with schizophrenia showed higher prevalences of the MetS criteria for high-density lipoprotein cholesterol (HDL) (p<0.001) and waist circumference (WC) (p<0.001). In subjects with schizophrenia, the predictive power was found to be highest for HDL, followed by WC, BMI, triglyceride, diastolic blood pressure (BP), systolic BP and fasting plasma glucose. However, in control subjects, the predictive power was found to be highest for triglyceride, followed by WC, systolic BP, BMI, HDL, diastolic BP and fasting plasma glucose. HDL was the component most predictive of MetS in subjects with schizophrenia treated with antipsychotics. Conclusion Because, in normal clinical practice, it is difficult to obtain temporal measurements for all of the MetS criteria, measurement of HDL may be useful for predicting the MetS. Copyright (c) 2013 John Wiley & Sons, Ltd.