BACKGROUND: In epidemiological studies, within-person variability in measured values of a risk factor may underestimate the association between prolonged or 'usual' levels of a risk factor with risk of disease - 'regression dilution'. The importance of regression dilution for high-density lipoprotein (HDL)-cholesterol and the extent to which this may differ from that for total cholesterol is not known. The aim of this study was to assess the magnitude of regression dilution bias for HDL-cholesterol, total cholesterol and blood pressure after varying intervals of follow-up in two prospective cohort studies. METHODS: Regression dilution ratios were estimated for each risk factor using the correlations between baseline and re-survey values in the Glostrup Population Studies and the NHLBI Framingham Heart Study after various time intervals. The regression dilution ratios in each cohort after a fixed interval between measurements were compared. RESULTS: The regression dilution ratios after 10 years were 0.51 and 0.56 for systolic blood pressure in Glostrup and Framingham, respectively; 0.52 and 0.54 for diastolic blood pressure; and 0.68 and 0.63 for total cholesterol. In both studies, the regression dilution ratios for these risk factors became more extreme with increasing intervals between measurements. The regression dilution ratio for HDL-cholesterol after 10 years in Glostrup was 0.72, which suggests that the importance of regression dilution for HDL-cholesterol was similar to that for total cholesterol. CONCLUSION: Failure to correct for increasing regression dilution with longer follow-up may account for some of the discrepant results obtained for the importance of these risk factors in epidemiological studies at varying intervals of follow-up.

Original publication




Journal article


J Cardiovasc Risk

Publication Date





143 - 148


Adult, Aged, Bias (Epidemiology), Blood Pressure, Cholesterol, Cholesterol, HDL, Cohort Studies, Female, Health Surveys, Humans, Male, Middle Aged, Prospective Studies, Regression Analysis, Risk Assessment, Risk Factors, Sensitivity and Specificity, Time Factors