We examined the determinants of glycaemic control in a consecutive cohort of 562 newly-referred Chinese type 2 diabetic patients (57% women) during a 12-month period. All patients underwent a structured assessment with documentation of clinical and biochemical characteristics. Pancreatic beta-cell function was assessed by fasting plasma C-peptide concentration. Insulin deficiency was defined as fasting plasma C-peptide <0.2 pmol/ml. Insulin resistance (IR) was calculated using the homeostasis model assessment (HOMA) based on a product of fasting plasma glucose and insulin concentrations. Treatment was considered appropriate when insulin-deficient patients were treated with insulin and non-insulin-deficient patients were treated with oral agents or diet. Mean (+/-SD) age was 54.3+/-13.8 years (range 17-87 years) and disease duration was 5.0+/-5.9 years. At the time of referral, 70.5% (n=396) were on drug therapy (9% on insulin and 62.8% on oral agents), 20.6% (n=116) were on diet and 9% (n=50) had not received any form of treatment. The mean HbA(lc) was 8.4+/-2.3%. The geometric mean (x// antilog SD) of IR was 4.62x//2.51 (range 0. 63-162.7) and correlated only with waist : hip ratio (WHR, p=0.008). The geometric mean of plasma C peptide was 0.47x//2.89 nmol/l and correlated with BMI (p<0.001). Glycated haemoglobin was correlated positively with age (p=0.013), disease duration (p<0.001), IR (p<0. 001) and negatively with BMI (p<0.001). Glycated haemoglobin was lower in patients who had seen a dietitian (7.9% vs. 8.7%, p<0.001) or diabetes nurse (7.8% vs. 8.7%, p<0.001) or who performed self blood glucose monitoring (7.9% vs. 8.6%, p=0.001) and higher among smokers (8.9% vs. 8.2%, p=0.003). Compared to insulin-deficient patients (n=118), non-insulin-deficient patients (n=413) had features resembling that of the Metabolic Syndrome with increased WHR (p=0.005), blood pressure (p<0.001), BMI (p=0.001) and were older (p=0.04). Amongst the insulin-deficient patients, 27% were treated with oral agents or diet. Patients receiving appropriate therapy (n=362) had a lower HbA(lc) than those treated inappropriately (n=173) (8.2% vs. 8.7%, p=0.02). On multivariate analysis, short disease duration (p<0.001), low IR (p<0.001), high BMI (p=0.001), diabetes education (p<0.001), lack of smoking (p=0. 014) and choice of appropriate treatment (p=0.009) were the independent determinants of good glycaemic control.