Ability of patients with type 1 diabetes mellitus to predict changes in HbA1c levels

Forty patients with Type 1 diabetes mellitus were examined on five occasions, three months apart. Their mean (+/- SD) age was 38.9 (13.2) years. At each visit they participated in a teaching program and also filled out a questionnaire on average morning and daytime blood glucose levels, weekly number of blood glucose measurements and hypoglycaemias. They were then told their last HbA1c value and asked whether they believed the present one would be unchanged (+/- 0.3%), slightly higher/lower (0.4-1.0%), or much higher/lower (more than 1.0%). They did infrequent home blood glucose monitoring [mean weekly number of measurements 6.9 (6.7)] and did not do systematic self-adjustment of insulin doses. Their HbA1c at inclusion was 9.0 (1.6)% and 8.4 (1.3)% at the end of the study (p < 0.05). Generally, the changes in HbA1c were few, and only more than 1.0% in 22.5% of tests. The kappa values for agreement between predicted and actual HbA1c values at the five visits were between 0.27 and 0.12. The ability to predict HbA1c changes did not improve during the study. A correct prediction (direction of change or no change) was seen in only 45.0% at the start and 42.5% at the end of the study. Those predicting correctly did no more home blood glucose measurements than the others. In conclusion, the patients' own perception of glycaemic control is a poor indicator of HbA1c changes over time.