Validating the Framingham Offspring Study equations for predicting incident diabetes mellitus

Area of Science:

• Epidemiology
• Biostatistics
• Preventive Medicine

Background:

• Investigators proposed a simple point score from the Framingham Offspring Study (FOS) to estimate 8-year diabetes mellitus (DM) risk.
• Validation of prediction models and comparison of estimated DM risk with observed incidence were needed.

Purpose of the Study:

• To validate DM risk prediction models.
• To compare DM risk estimated by the FOS point score with observed DM incidence in a new cohort.

Main Methods:

• A longitudinal observational cohort study was conducted using 20,644 Kaiser Permanente Northwest (KPNW) members without prior DM.
• Logistic regression models were re-estimated, and the FOS point score was compared against 8-year observed DM incidence in the KPNW sample.

Main Results:

• Prediction models showed similar performance in FOS and KPNW samples, with comparable odds ratios and areas under the receiver operating characteristic curves.
• The FOS point score substantially underestimated DM incidence in the general KPNW sample but was highly accurate in a subsample with incidence matching the FOS.
• Observed DM incidence was 16.5% in the total KPNW sample and 5.1% in the FOS-matched subsample.

Conclusions:

• The accuracy of the FOS point score is dependent on the similarity of the underlying DM incidence in the population to which it is applied.
• Accurate adaptation of the point score for different populations requires recalculating DM incidence at each point level.