To assess the effect of phosphate replacement therapy on the course of diabetic coma, 24 patients with severe diabetic ketoacidosis and 16 patients with nonketotic hyperosmolar coma were randomly assigned either to standardized conventional treatment alone or combined with phosphate infusions. Insulin and fluid therapy produced a rapid fall of plasma phosphorus; almost all patients not receiving phosphate infusions developed marked hypophosphatemia within 12 h. Hypophosphatemia was prevented by administration of 62 ± 5 mmol (range 35-140) sodium phosphate. Initial red blood cell 2,3-diphosphoglycerate (2,3-DPG) concentrations were markedly decreased in ketoacidotic patients. The recovery of 2,3-DPG upon institution of therapy was accelerated when phosphate replacement infusions were given. The increase in 2,3-DPG during the first 48 h was 56% greater (P < 0.02) when phosphate was administered, but later the difference between the two treatment groups disappeared. Nonketotic hyperosmolar coma patients revealed normal 2,3-DPG concentrations on admission, and a similar decline of plasma phosphorus occurred, as in ketoacidosis, during treatment. 2,3-DPG levels remained unaffected by phosphate therapy. While plasma calcium levels declined during the initial 48 h in both groups, transient postinfusion hyperphosphatemia was noted in 7 of 17 patients. A favorable effect of phosphate therapy on the clinical course of diabetic ketoacidosis or hyperosmolar coma could not be demonstrated. DIABETES 29: 87-95, February 1980.T he occurrence of excessive urinary phosphorus excretion in patients with uncontrolled diabetes mellitus has been recognized for more than 100 yr. 1 Severe phosphorus depletion ensued when insulin was withdrawn from diabetic subjects, 2 and marked hypo-