We studied retrospectively, 219 episodes of diabetic ketoacidemia in 119 patients aged 13 months to 30 years, to determine the trend of the concentration of sodium in serum as glucose declined during treatment of uncomplicated episodes and of episodes with complications attributable to brain swelling. Of 20 complication, 13 were minor (headache only) and 7 major (death or near death). The concentration of sodium in serum failed to rise as that of glucose declined in 82 (54%) of 164 uncomplicated episodes and in 18 (95%) of 20 complicated episodes (p less than 0.01). Hence complications were more likely to occur among patients with a failure of the concentration of sodium to rise as glucose declined. Fifty-eight episodes of diabetic ketoacidemia in 40 patients aged 1 1/2 to 20 years were then studied prospectively on a 48-hour treatment plan to provide the volume of deficit evenly, with half the deficit of sodium in the first 42 hours. Sodium concentration in serum rose in 55 (95%) of 58 episodes as that of glucose declined. No patient had a major complication. We conclude that failure of the sodium concentration measured in serum to rise as glucose concentration declines is a marker for excessive administration of free water. An expanded repair period, with repair fluid containing an average of 125 mmol/L Na+ early in therapy, will usually protect against a downward trend in the concentration of sodium in serum and therefore against a rapid decline in effective serum osmolality. This regimen may be protective against near-death episodes and brain herniation during treatment.