The intraindividual variability of fasting triglyceride--a challenge for further standardization

The extraordinarily high intraindividual variability of fasting triglyceride levels often leads to substantial bias in epidemiologic studies and hinders the use of the former as potential markers of coronary heart disease risk in individual patients. The magnitude of the latter problem is illustrated applying the concept of the intraclass correlation coefficient to repeated fasting triglyceride measurements in 2227 white men aged 20 to 69 in the Lipid Research Clinics Prevalence Study. A one-way random effects analysis of variance model with the natural logarithm of triglycerides as the outcome variable is used to calculate 95% confidence intervals for a person's 'true' or 'steady-state' triglyceride level, depending on the number and results of measurements taken. For example, a 95% confidence interval corresponding to a single triglyceride measurement of 200 mg dl-1 (a commonly used threshold value for hypertriglyceridaemia) would be as wide as 124 to 323 mg dl-1. Although there is a substantial gain in precision if a second triglyceride measurement is available, the confidence intervals remain very large. The gain in precision from each additional measurement becomes progressively smaller and has to be weighed against considerations of costs and inconvenience.