To compare abdominal visceral fat with subcutaneous fat in relation to their association with type 2 diabetes. Abdominal fat distribution was measured using Computed Tomography in 60 subjects (30 diabetics and 30 non-diabetics). Computed tomography images obtained at two intervertebral locations L2-L3 and L4-L5 were used to measure areas of total fat, visceral fat and subcutaneous fat using slice thickness of 5mm and attenuation range of -190 to -30 Hounsfield units. Data were analyzed using logistic regression. At L2-L3 level, taking visceral fat and subcutaneous fat as predictor variables, diabetes was correctly classified at 78.0% and 66.10% respectively. At L4-L5 level, taking visceral fat and subcutaneous fat as predictor variables, diabetes was correctly classified at 72.88% and 67.80% respectively. Regardless of the measurement site, visceral fat has significantly stronger association with diabetes, compared to subcutaneous fat. Visceral fat at L2-L3 level alone may be a better predictor of diabetes. Abdominal fat distribution, visceral fat, subcutaneous fat, type 2 diabetes.