The aim of this study was to determine whether amniotic fluid insulin concentration (AFI) is a better parameter than mean maternal blood glucose values (MBG) for deciding about insulin therapy in patients with gestational diabetes. MBG's were calculated on the base of 9 blood glucose levels during a 24 hour period after one week of diet therapy. In a prospective trial between 1987 and 1989 in Karlsburg, 123 gestational diabetic patients were randomized into two groups. Treatment was either based on the concentration of AFI or MBG levels. In a second series in Berlin, 103 patients were offered amniocentesis. 81 patients agreed and 22 refused. Treatment was then analogous to that in Karlsburg. In both groups of the randomized population, strict metabolic control was achieved. There was no difference regarding pregnancy complications. Earlier labor induction and higher cesarean section rates were seen in the non-invasive group (p < 0.05). The incidence of diabetic fetopathy and neonatal hypoglycemia was significantly lower in the invasive group (p < 0.01), even though the metabolic control parameters did not differ between the two groups. The results in Berlin correspond to these findings. In conclusion, AFI enables the recognition of any hyperinsulinism reaction to the maternal metabolic situation. We recommend the additional measurement of the AFI concentration between 28 and 36 weeks as the direct fetal parameter for deciding about insulin treatment.