We report 18 consecutive phenformin-treated diabetic patients admitted to this Medical Service acutely ill with metabolic acidosis. Lactic acidosis was anticipated, and documented, in all. Also, however, though most of the patients had only weakly positive, or even negative, serum reactions with the nitroprusside reagent, all were found to have coexisting ketoacidosis, plasma 3-hydroxybutyrate averaging 7.1 mmol/L. +/- 3.9 (S.D.). This finding suggest that treatment of these patients should include insulin, and often also glucose, because most do not have marked hyperglycemia and some have hypoglycemia. The lactic acidosis in the nine patients who survivied was, on average, less severe than in the nine who died, but the difference was not statistically significant. Surivival correlated closely with the absence of shock on arrival. Only eight patients had a identifiable acute illness other than the metabolic acidosis. The other 10 patients had no discernible cause for the acute illness apart from their treatment with phenoformin. This finding raises serious doubts about whether phenformin should be used to treat patients with diabetes.