Pancreatic beta cell function was evaluated in 30 insulin treated non-insulin dependent diabetic patients, by estimating serum C-peptide after meal stimulation. C-peptide response was low (less than 0.6 pmol/ml) in 15 patients and it was significant (greater than 0.6 pmol/ml) in the other 15 patients. All the patients were then started on a high carbohydrate, high fibre diet and a combination of glibenclamide and metformin. In 18 patients, optimal regulation of hyperglycemia was achieved in one week and the others required insulin treatment. Among the 15 with low C-peptide values, 8 patients responded to oral hypoglycaemic agents and their C-peptide responses improved (from 0.17 +/- 0.16 to 0.78 +/- 0.2 pmol/ml). Among the 15 with significant C-peptide values, 10 responded well to oral drugs and their C-peptide values improved further (from 0.79 +/- 0.21 to 1.17 +/- 0.44 pmol/ml); but the others required insulin despite the good beta cell reserve. This study shows that the beta cell response in insulin treated NIDDM varies widely as it is influenced by the exogenous insulin and hyperglycaemia a random estimation of C-peptide will be of limited value in predicting the response to therapy.