Childhood obesity has increased significantly in recent decades. The objective was to examine the perinatal risk factors related to childhood obesity. In a prospective study, 89 women with normal glucose tolerance (NGT) or gestational diabetes mellitus (GDM) and their offspring were evaluated at birth and at 8.8 +/- 1.8 y. At birth, obstetrical data, parental anthropometric measures, and neonatal body composition were assessed; at follow-up, diet and activity were assessed and laboratory studies were conducted. Weight was classified by using weight for age and sex, and body composition was measured by using dual-energy X-ray absorptiometry. In childhood, data were analyzed as tertiles and prediction models were developed by using logistic and stepwise regression. No significant differences in Centers for Disease Control and Prevention weight percentiles, body composition, and most metabolic measures were observed between children of mothers with NGT and GDM at follow-up. Children in the upper tertile for weight had greater energy intake (P = 0.02), skinfold thickness (P = 0.0001), and leptin concentrations (P < 0.0001) than did those in tertiles 1 and 2. Children in the upper tertile for percentage body fat had greater waist circumference (P = 0.0001), insulin resistance (P = 0.002), and triglyceride (P = 0.009) and leptin (P = 0.0001) concentrations than did children in tertiles 1 and 2. The correlation between body fat at birth and follow-up was r = 0.29 (P = 0.02). The strongest perinatal predictor for a child in the upper tertile for weight was maternal pregravid body mass index (BMI; kg/m(2)) >30 (odds ratio: 3.75; 95% CI: 1.39, 10.10; P = 0.009) and for percentage body fat was maternal pregravid BMI >30 (odds ratio: 5.45; 95% CI: 1.62, 18.41; P = 0.006). Maternal pregravid BMI, independent of maternal glucose status or birth weight, was the strongest predictor of childhood obesity.