Abstract P044: Repeated Measurement of Insulin Sensitivity is Increasing the Precision of Incident Dyslipidemia Prediction Compared to Single Baseline Measurement: The Ansan-Ansung Community Cohort Study

Abstract

Introduction: Medical data are commonly measured repeatedly. However, there is a lack of studies of the usability of trajectory as a repeated measurement on the prediction model compare to the single measurement. Objectives: This study aimed to evaluate whether repeated measurement of insulin sensitivity can improve the prediction of incident dyslipidemia beyond the single measurement. Methods: Participants (n=3899; aged 40-69 years, during visit 1 to 7, mean follow-up 12 years) were recruited from the Ansan-Ansung cohort study, a subset of Korean Genome Epidemiology Study. Oral glucose tolerance test (OGTT) were measured at every visit with a 2 year interval. Matsuda index of each visit was calculated by the formula 10 4 /((GLU0xINS0xGLU60xINS60) 0.5 ). We used total seven times of OGTT (28 OGTT index) for trajectories. Latent mixture modeling was used to identify trajectories in Matsuda index for 12 years. We used tertiles of baseline Matsuda index for single measurement. Age, body mass index, alcohol drinking, cigarette smoking, c reactive protein, education level, and income were covariates. We compared two model’s discrminatory power using area under the curve of the receriver operating characteristics curve (AUROC). Results: In the trajectory model (repeated measurement), three distinct insulin sensitivity trajectories were identified: low-stable (Group 1; 88.8%), steady decreasing (Group 2; 9.7%), and fast decreasing (Group 3; 1.5%). In the baseline single measurement model, tertile groups were devided: low tertile(<6.25, N=1306, 33.50%), tertile2 (6.25-10.89, N=1277, 32.75%), and tertile 3(≥10.89, N=1316, 33.75%). Trajectory model showed better discrimination power compared to single measurement model in the total population (AUROC 0.6303 vs. 0.6234) and men (AUROC 0.6671 vs. 0.6470) but not in women (AUROC 0.6171 vs. 0.6176). Conclusions: Repeated measurement of insulin sensitivity can improve the prediction of incident dyslipidemia in the general population.