OBJECTIVE To study the influence of glycosilated haemoglobin (HbA1c) on the therapies chosen by primary care doctors when treating type-2 diabetes (DM2) patients. DESIGN Retrospective observational study.Setting. Urban health district. SUBJECTS Randomised, representative sample of 152 DM2 patients, with a total of 265 documented clinical episodes which involved request and assessment of HbA1c result in 1999-2000. MEASUREMENTS AND INTERVENTIONS For each clinical episode we assessed the HbA1c and basal glycaemia figures and the therapeutic repercussion of the result on patient treatment: no change (NC), strengthening of hygienic-dietary measures (HDM), increase of dose (ID) or change of therapy (CT). RESULTS We found a mean overall figure of 0.87+/-0.03 requests for HbA1c per patient per year. Of the 265 HbA1c studied, 54.7+/-5.9% can be considered acceptable, 18.4+/-4.6% deficient, and 26.9+/-5.3% unacceptable. However, the overall result of HbA1c was NC in 40.3+/-5.8%, HDM in 18.7%, ID in 25.9+/-5.2%, and CT in 15.1+/-4.2% of cases. Therapy changes (CT or ID) occurred in 26.8% of cases when was acceptable, 60.4% of cases when the glycosilated HbA1c was deficient, and 60% when the glycosilated HbA1c was unacceptable (P<.001). CONCLUSIONS It should be emphasised that, even when the HbA1c was very deficient, the dose was only increased or the treatment changed in under two-thirds of the population. It is also paradoxical that, when HbA1c with acceptable results was found, therapy was changed in around 25% of patients.