OBJECTIVE: To compare a simple measure - age of onset of obesity - to an obese-years construct (a product of duration and magnitude of obesity) as risk factors for type 2 diabetes. METHOD: Participants from the Framingham Heart Study who were not obese and did not have diabetes at baseline were included (n=4,320). The Akaike Information Criterion (AIC) was computed to compare four Cox proportional hazards models with incident diabetes as the outcome and: (i) obese-years; (ii) age of onset of obesity; (iii) body mass index (BMI); and (iv) age of onset of obesity plus magnitude of BMI combined, as exposures. RESULTS: AIC indicated that the model with obese-years provided a more effective explanation of incidence of type 2 diabetes compared to the remaining three models. Models including age of onset of obesity plus BMI were not appreciably different from the model with BMI alone, except in those aged ≥60. CONCLUSIONS: While obese-years was the optimal obesity construct to explain risk of type 2 diabetes, age of onset may be a useful, practical addition to current BMI in the elderly. IMPLICATIONS: Where computation of obese-years is not possible or impractical, age of onset of obesity combined with BMI may provide a useful alternative. © 2016 Public Health Association of Australia.