This study evaluated the efficacy of rosiglitazone in non-obese and obese Korean type 2 diabetic patients of long duration. A total of 125 patients (M:F = 44:81, mean age: 58.4 +/- 9.1 years, BMI: 24.2 +/- 2.7kg/m 2 , duration of diabetes: 11.0 +/- 6.4 years) were randomly allocated to 12 weeks of rosiglitazone treatment (4mg per day) or a control group. Responders were defined as patients who experienced fasting plasma glucose (FPG) reduction of >20% or HbA 1 c reduction of >1 (%). Rosiglitazone significantly improved glycemic control by reducing FPG and HbA 1 c (-3.4mmol/l and -1.1%, P < 0.001, respectively). It also significantly increased HOMA β - c e l l f u n c t i o n (+9.7, P < 0.01) and QUICKI (+0.029, P < 0.001), and decreased HOMA I R (-1.73, P < 0.001). Females and those with higher waist-hip ratio made up a greater portion of rosiglitazone-responders. Responders (45 patients, 75%) also showed significantly higher FPG, HbA 1 c , systolic blood pressures, fasting insulin levels and HOMA I R , and lower QUICKI than nonresponders. Among these parameters of responders, waist-hip ratio of non-obese subgroup, initial glycemic control of obese subgroup, and systolic blood pressure of both subgroups lost their significance after subdivision analysis. However, the baseline HOMA I R and QUICKI were significantly correlated with the response rate to rosiglitazone. Moreover, in multiple logistic regression analysis, HOMA I R and QUICKI retained their significance as the independent predictors. Even in Korean type 2 diabetic patients of long duration but with relatively preserved β-cell function, rosiglitazone improved glycemic control, insulin sensitivity, and β-cell function. In this ethnic group, female gender, central obesity, and especially severe insulin resistance were identified as predictive clinical parameters of rosiglitazone-responders.