The results of three controlled trials performed on children with insulin-dependent diabetes mellitus were examined for evidence of seasonal variation in concentrations of glycosylated haemoglobin (HbA1). All three studies showed lower levels during the summer months. Multiple regression analysis showed that the month of sampling accounted for a significant proportion of the total variance in HbA1 levels (P less than 0.001 in all three studies). We suggest that exercise, dietary changes and the frequency of minor illnesses may all contribute to this fluctuation which has important implications for the design of clinical trials in childhood diabetes.