A number of 13,700 participants (2962 men and 10,738 women) aged 36 to 55 years were followed for up to 5 years using data from annual health checkups. Incident cases of diabetes were identified from self-reports or single fasting plasma glucose measurements (鈮?.0 mmol/l). Weight/BMI change was calculated from participants鈥?weight/BMI values at age 20 years and weight/BMI values at a given point during follow-up and used as a time-dependent variable in age-stratified multivariate Cox proportional hazards models.
During the 5 year follow-up, 408 participants (137 men and 271 women) developed diabetes. Even after adjusting for BMI during follow-up and other possible confounders, weight/BMI gain since age 20 years was significantly associated with an increased risk of developing diabetes. The hazard ratios were: 2.30 (95% confidence interval (CI): 1.31-4.04) for those who gained 6.0 to <10.0 kg and 3.09 (95% CI: 1.79-5.34) for those who gained 鈮?0.0 kg [reference: <2.0 kg change]; and 2.61 (95% CI: 1.58-4.31) for those who gained 3.0 to <5.0 kg/m2 and 3.70 (95% CI: 2.22-6.16) for those who gained 鈮?.0 kg/m2 [reference: <1.0 kg/m2 change].
The results indicate that long-term weight/BMI gain in adulthood is a significant predictor for the development of diabetes, independent of attained weight status. Because weight gain within the normal weight range could increase the risk of diabetes, non-obese people should also be warned against possible weight gain.