Abstract
AIMS: We investigated the optimal cut-off points of nonfasting and fasting triglycerides in Japanese individuals with lower average triglyceride levels than westerners. METHODS: Residents aged 40-69 years without a history of ischemic heart disease or stroke were enrolled between 1980 and 1994 and followed. Serum triglyceride concentrations were measured from 10851 nonfasting (＜8 h after meal) and 4057 fasting (≥ 8 h) samples. As a prerequisite, we confirmed the shape of a receiver operating characteristic (ROC) curves, the area under ROC curves (AUC), and the integrated time-dependent AUC. We identified optimal cut-off points for incident ischemic heart disease based on C-statistic, Youden index, and Harrell's concordance statistic. We used dichotomized concentrations of triglycerides via the univariate logistic regression and Cox proportional hazards regression models. We also calculated multivariable hazard ratios and population attributable fractions to evaluate the optimal cut-off points. RESULTS: Nonfasting and fasting optimal cut-off points were 145 mg/dL and 110 mg/dL, with C-statistic of 0.594 and 0.626, Youden index of 0.187 and 0.252, and Harrell's concordance statistic of 0.590 and 0.630, respectively. The corresponding multivariable hazard ratios of ischemic heart disease were 1.43 (95%CI 1.09-1.88) and 1.69 (1.03-2.77), and the corresponding population attributable fractions were 16.1% (95%CI 3.3-27.2%) and 24.6 (-0.3-43.3). CONCLUSION: The optimal cut-off points of nonfasting and fasting triglycerides in the Japanese general population were 145 mg/dL and 110 mg/dL, respectively, lower than the current cut-off points recommended in the US and Europe.