The long-term complications of diabetes have major consequences for individual subjects and growing healthcare delivery and cost implications for society. Evidence for the benefits of good glycaemic control, as monitored by glycated haemoglobin measurements, has been developed in the 25 years since they were introduced to the point where HbA(1c) assays play central roles in patient management, clinical guidance and audit, and clinical trial design. In this review this evidence is examined and three classes of uncertainty identified that diminish confidence in the effectiveness of these roles for HbA(1c). 1 Analytical variability between different methods for HbA(1c) has restricted the application of clinical targets and this problem has recently been addressed by reference method standardization. There are two approaches to this which result in different HbA(1c) values and this discrepancy needs to be resolved. 2 Biological variability in HbA(1c) values between individuals also restricts its predictive role when applied to populations. The correlations between HbA(1c) measurements and various components of glycaemia (overall, fasting, postprandial) are still uncertain and differences in protein glycation and de-glycation are greater between subjects than often thought. The influence of variability in erythrocyte life span is an area where research is needed, especially in diabetic subjects. 3 Clinical variability is the most important and complex area of uncertainty. A predictive link between HbA(1c) and clinical outcomes is not as clear-cut as often stated. The correlation with the development of microvascular disease is well established in Type 1 diabetes, but in Type 2 subjects (90% of those with diabetes) the evidence that HbA(1c) monitoring is of value in predicting or preventing macrovascular disease is not strong, although it is the major cause of morbidity and early death in this group. It is recommended that, as a matter of urgency, these issues be examined, particularly within the context of self-care in diabetes.