Treatment of type I diabetes mellitus is hindered by the often large fluctuations in blood glucose concentration experienced by affected individuals. To determine to what extent day-to-day variation in blood glucose levels can be reduced if insulin is injected in the same anatomic region rather than in different regions using a rotational scheme, as is commonly recommended, 12 type I diabetic subjects were studied. Insulin injections were given in the abdomen for 3 days and rotated among arms, abdomen, and thighs for 3 days using a crossover design with random assignment of treatment order. Blood samples for measurement of plasma glucose levels were obtained at nine scheduled times on each day. Insulin dose, diet, and physical activity were held constant for each subject. During the abdominal injection period, the mean SD of plasma glucose levels and the mean variance of plasma glucose levels were both less at all nine time points than during the rotating injection period. Overall values for the SD of plasma glucose levels were 2.7 ± 0.2 mmol/L for the abdominal injection period and 3.7 ±0.3 mmol/L for the rotating injection period. Overall values for the variance of plasma glucose levels were 9.2 ± 1.4 mmol2/L2 for the abdominal injection period and 17.4 ±2.2 mmol2/L2 for the rotating injection period. We conclude that the common clinical practice of rotating the anatomic regions used for insulin injections increases day-to-day variation in blood glucose concentration. Use of a single anatomic region, eg, the abdomen, for all insulin injections may reduce this variation and allow greater precision in the adjustment of insulin doses.