SummaryFrom June 1960 to June 1961, one hundred and seven diabetic adults were submitted to a treatment with a French Riguanide (dimethylbiguanide, DMBG).These outpatients were carefully selected in order to fulfil the following conditions : no obesity, diabetes not controlled by diet alone, patients co-operative enough to follow a steady diet during the whole trial, to perform daily their own urine analysis and write down the results in a note book.They were submitted to an average of six periods of alternate treatment either with diet alone, diet + DMBG or diet + placebos (similar in all respects to the DMBG tablets). Most patients followed also one or more periods of sulfonamide treatment and some cases were treated with insulin. Transfer from insulin to DMBG was done progressively.Control of the diabetes was assessed by body weight, daily urine analysis and frequent titrations of glucose in the blood and urine at various moments of the day. Results of treatment with the drug were considered good when effective control of the diabetes was obtained without insulin. Thirty-five cases (>3o %) were discarded because they showed no clcarcut response to the drug or because their diabetes happened to be unexpectedly mild and had been brought under control by diet alone. Of the remaining 72 cases, 49 responded well to DMBG treatment as subsequent trials with placebo tablets conclusively proved.In many cases, Biguanide appeared to be as effective as the sulfonamides and in some cases as effective as insulin. Response to DMBG seemed to be correlated (P<0,001) with the mildness of the diabetes and with a successful response to the sufonamides (r = o,83, P<0,001) but not with either age, duration of the diabetes or a small excess of weight.No toxic or allergic effect has been observed. Gastro-intestinal discomfort occurred in one third of the cases, serious disturbances in only four cases.The effective dose extends from 2x500 mg to 3x1000 mg daily. In some cases, even with severe diabetes, DMBG induced a remission and could be dispensed with entirely. This circumstance, and the fact that even a careful selection was unable to exclude a few cases of unsuspectedly mild diabetes, emphasizes the necessity of using placebos in therapeutic trials of this type.