The aim of this study was to compare the allergy-preventive effect of a partially hydrolyzed formula with two extensively hydrolyzed formulas, in infants with a high risk for development of allergic disease. High-risk infants from four Danish centres were included in the period from June 1994 to July 1995. Five-hundred and ninety-five high-risk infants were identified. High-risk infants were defined as having bi-parental atopy, or a single atopic first-degree relative combined with cord blood immunoglobulin E (IgE) ≥ 0.3 kU/l. At birth all infants were randomized to one of three different blinded formulas. All mothers had unrestricted diets during pregnancy and lactation and were encouraged to breast-feed exclusively. If breast-feeding was insufficient, one of the three formulas, according to randomization, was given during the first 4 months. It was recommended not to introduce cow's milk, cow's milk products, and solid foods until the age of 4 months. After the age of 4 months a normal unrestricted diet and conventional cow's milk-based formula were given when needed. All infants were followed-up prospectively with interview and physical examination at the age of 6, 12, and 18 months, and if any possible atopic symptoms were reported. If food allergy was suspected, controlled elimination/challenge procedures were performed in a hospital setting. Of 550 infants included in the study, 514 were seen at all visits and 36 were excluded owing to non-compliance. Of 478 infants who completed the study, 232 were exclusively breast-fed, 79 received an extensively hydrolyzed casein formula (Nutramigen), 82 an extensively hydrolyzed whey formula (Profylac), and 85 a partially hydrolyzed whey formula (Nan HA), during the first 4 months of life. These four groups were identical in regard to atopic predisposition, cord blood IgE, birthplace, and gender. Exclusively breast-fed children were exposed less to tobacco smoke and pets at home and belonged to higher social classes, whereas the three formula groups were identical concerning environmental factors. The frequency of breast-feeding was high; only eight (2%) children were not breast-fed at all. The three formula groups were identical in regard to duration of breast-feeding and age at introduction of formula and solid foods. No significant differences were found in the three groups of infants receiving formula milk regarding the cumulative incidence of atopic dermatitis or respiratory symptoms. The cumulative incidence of parental-reported cow's milk allergy was significantly higher in children fed partially hydrolyzed formula (Nan HA) compared with extensively hydrolyzed formula (Nutramigen or Profylac) at 12 and 18 months (NanHA, 7.1%; Nutramigen, 2.5%; Profylac, 0%; p = 0.033). The cumulative incidence of confirmed cow's milk allergy was 1.3% (three of 232) in exclusively breast-fed infants, 0.6% (one of 161) in infants fed extensively hydrolyzed formula (Nutramigen or Profylac), and 4.7% (four of 85) in infants fed partially hydrolyzed formula (Nan HA). Partially hydrolyzed formula was found to be less effective than extensively hydrolyzed formula in preventing cow's milk allergy, 0.6% vs. 4.7% (p = 0.05), but because of the small number of cases the results should be interpreted with caution. Compared with other similar studies the frequency of atopic symptoms was low, even though the dietetic intervention did not include either maternal diet during lactation or dietary restrictions to the children after the age of 4 months.