Dynamic prediction models improved the risk classification of type 2 diabetes compared with classical static models.

Area of Science:

• Epidemiology
• Biostatistics
• Medical Informatics

Background:

• Type 2 diabetes mellitus (T2DM) is a growing global health concern.
• Accurate prediction of T2DM incidence is crucial for timely intervention.
• Traditional static models often fail to utilize longitudinal data effectively.

Purpose of the Study:

• To develop and validate a dynamic prediction model for incident T2DM.
• To assess the performance of dynamic models using repeated measures of predictors.
• To compare dynamic prediction with static Cox regression models.

Main Methods:

• Utilized data from the Tehran Lipid and Glucose Study (n=8438).
• Employed joint modeling (JM) of longitudinal data and time-to-event analysis.
• Incorporated repeated measurements of fasting plasma glucose (FPG) and waist circumference (WC).

Main Results:

• Dynamic JM models demonstrated comparable discrimination to static Cox models.
• JM models showed superior calibration and higher clinical utility.
• Significant risk reclassification improvements were observed: 33% for FPG and 24% for WC models.

Conclusions:

• Dynamic prediction models offer substantial improvements over static models for T2DM risk.
• Repeated measures of FPG and WC enhance predictive accuracy.
• Decision support systems can help manage the complexity of dynamic models.