Evaluating the associations between the baseline and other exposure variables with the longitudinal trajectory when responses are measured with error

2013-11-26Evaluating the associations between the baseline value and other exposure variables with the longitudinal trajectory constantly occurs in the medical research. Traditional analyses using the observed measurements can produce biased estimates of the true relationships when responses are measured with error. Bias-correction formulas are developed using the linear regression for studies with a baseline and one follow-up visit. When the data has multiple follow-up visits, random intercept and random slope mixed effects models are always used. However, current bias-correction methods developed from aforementioned mixed effects model have their limitations. Motivated by a real data example, we developed a new method to correct the bias in association estimates using the mixed effects model. The validity of formulas was evaluated by the simulation study. Two methods were proposed to test the bias-corrected estimates. Finally, two real study examples were provided to demonstrate the application of our methods