The relationship between clinic-measured random blood glucose and glycated haemoglobin was investigated in 204 non-insulin-dependent diabetic patients to determine the value of random blood glucose on management decisions in these patients. Treatment was with diet alone in 104 patients (51%: Group 1), and diet and oral hypoglycaemic agents in 100 patients (49%: Group 2). Random blood glucose and glycated haemoglobin were lower in Group 1 than Group 2 (10.9 +/- 4.0 vs 13.0 +/- 4.0 mmol l-1, p < 0.001, and 8.5 (7.8-10.2) vs 9.6 (8.1-11.0)%, p < 0.01, respectively), and there was a positive correlation between random blood glucose and glycated haemoglobin in both groups (rGroup1 = 0.76, and rGroup2 = 0.54, both p < 0.001). In Group 1, 48 (46%) patients had a random blood glucose < 10 mmol l-1 and all but 2 of these had a glycated haemoglobin of < 10%. Thus, random blood glucose < 10 mmol l-1 was 96% sensitive for glycated haemoglobin < 10%. In Group 2 the same sensitivity was 92%. These data suggest that clinic-measured random blood glucose levels below 10.0 mmol l-1 predict acceptable overall glycaemic control in non-insulin-dependent diabetic patients, particularly in those on diet alone. However, a clinic-measured random blood glucose above 10 mmol l-1 was of limited value in predicting glycated haemoglobin values above 10% and a blood glucose cut-off of 14 mmol l-1 appeared more useful. Where resources are limited, clinic random blood glucose estimation may allow clinicians to use glycated haemoglobin measurements more discriminately.