Abstract
PURPOSE: This study compared two measurement error-correction approaches-linear mixed-effects approach to measurement error-correction (MEM) and simulation-extrapolation (SIMEX)-for assessing the association between choline intake and coronary heart disease (CHD) prevalence among United States (US) community-dwelling adults. METHODS: Simulations were conducted to evaluate the performances of five estimation approaches: benchmark analysis, 1-day method, average method (AveMethod), MEM, and SIMEX. Data from the National Health and Nutrition Examination Survey (NHANES) were analyzed to determine the relationship between choline intake and CHD prevalence using these methods. RESULTS: Both MEM and SIMEX effectively corrected for measurement error-induced biases; MEM generally outperformed SIMEX except when the standard deviation of true exposure (σ(X)) exceeded the standard deviation of random measurement error (σ(U)). Analysis of NHANES data revealed that choline intake was significantly and inversely associated with CHD prevalence using the 1-day method (β= -0.39; 95 % confidence interval: -0.72, -0.05; odds ratio: 0.68). Other approaches did not reveal statistically significant associations. CONCLUSIONS: MEM and SIMEX mitigated most measurement error-related biases in the simulations, although MEM demonstrated better overall performance. After correction for measurement error, choline intake was not significantly associated with CHD prevalence.