The aim of this study was to characterise longitudinal profiles of BMI from childhood and to examine the impact of level-independent childhood BMI trajectories on adult type 2 diabetes.

The longitudinal cohort consisted of 2449 adults (1613 white and 836 black) who had their BMI measured between four and 15 times from childhood (4–19 years) to adulthood (20–51 years) and fasting glucose measured in adulthood. Model-estimated levels and linear slopes of BMI at childhood age points were calculated in 1-year intervals using growth-curve parameters and their first derivatives, respectively.

BMI from childhood to adulthood fit cubic growth curves; linear and non-linear curve parameters differed significantly between race–sex groups. BMI showed race and sex differences from 15 years onwards. Individuals with hyperglycaemia had higher long-term BMI levels than those who were normoglycaemic in race–sex groups. Linear and non-linear slope parameters of BMI differed consistently and significantly between adult hyperglycaemia groups. The OR of childhood BMI levels for ages 4–19 years was 1.45–1.83 (p < 0.001 for all) for adult hyperglycaemia after adjustment for confounders. Level-adjusted linear slopes of BMI at ages 10–19 years showed significantly positive associations with adult hyperglycaemia (OR 1.17–1.50, p < 0.01 for all). The associations of childhood BMI linear slopes with adult hyperglycaemia were not significant during the age period 5–9 years. The trends in these associations were consistent across race–sex groups.

These observations indicate that childhood BMI trajectories have a significant impact on adult diabetes, independent of BMI levels. The adolescence age period is a crucial window for the development of diabetes in later life, which has implications for early-life prevention.

All data and materials are publicly available at the National Heart, Lung, and Blood Institute (NHLBI) Biologic Specimen and Data Repository and can be accessed at https://biolincc.nhlbi.nih.gov/studies/bhs .