Prediction of type 2 diabetes (T2D) in children and adolescents with obesity is complex. We have previously shown that impaired fasting glycemia (IFG) in this group has a limited predictive value for T2D in young adults. Hence, better models for estimating probability of T2D is of great interest in order to tailor treatment and allocate recourses to high-risk individuals.


We have prospectively followed children and adolescents aged 5-18 years with obesity and data of fasting glucose and HbA1c by using the Swedish Childhood Obesity Register (BORIS). In adulthood we identified T2D in national registers. Cox-Proportional Hazard models were used to evaluate combinations of prediabetic levels of fasting glucose (100-110 [i-IFG ADA] and 110-125 mg/dl [IFG WHO]) and HbA1c (≥39mmol/mol).


In total 3 243 individuals (45% female) were followed for a mean ± SD of 9.3±4.1 years. At baseline the age and BMI Z-score were 13.3 ± 2.7 years and 2.91 ± 0.42 Z-score units respectively. Twenty percent had fasting glucose levels above 100 mg/dl including 5.4% who had fasting glucose over 110 mg/dl, and 12.3% hade elevated HbA1c (≥39mmol/mol). In total did 6.0% develop T2D in young adulthood. A fairly good prediction model included sex, age, BMI Z-score, fasting glucose and HbA1c in childhood as well as time to follow-up, AUC=0.76. In Cox regression models adjusting for sex, age and BMI Z-score, i-IFG ADA in combination with HbA1c < 39 was not associated with future T2D, HR=0.96 [0.57-1.62], whereas i-IFG ADA in combination with prediabetic HbA1c was, HR=4.63 [2.66-8.06]. The highest risk was observed among subjects with IFG WHO and HbA1C levels ≥39, HR=10.90 [6.58-18.05].


The combination fasting glucose and HbA1c in childhood is superior in predicting adult T2D compared with individually assessed biomarkers.