Tuesday, 19 August 2014: 11:15 AM
Kahtnu 2 (Dena'ina Center)
Mark S Gilthorpe, PhD , University of Leeds, Leeds, United Kingdom
Tao Jiang, MS , University of Leeds, Leeds, United Kingdom
Kate Tilling, PhD , University of Bristol, Bristol, United Kingdom
George T Ellison, PhD , University of Leeds, Leeds, United Kingdom
Paul D Baxter, PhD , University of Leeds, Leeds, United Kingdom
INTRODUCTION: Each year numerous studies evaluate longitudinal data within a lifecourse context with later-life health status (e.g. blood pressure) analysed with respect to repeated measures of early-life experiences (e.g. body mass index, BMI) using standard multiple linear regression.  Although more sophisticated methods are available, some have been shown to be problematic, hence there remains confusion around which is the most appropriate analytical strategy.  Standard multiple regression can suffer text-book errors in this lifecourse context that are sadly perpetuated.  We revisit these problems to provide insight and give guidance.

METHODS: We conducted a series of basic regression analyses, akin to those frequently seen in lifecourse research, on both the Avon Longitudinal Study of Parents and Children (ALSPAC) data and simulated data. The simulated data were designed to emulate the ALSPAC dataset to allow flexibility in sensitivity analyses.  Additionally, we extended our simulation to a much older cohort.

RESULTS: For our analyses, we defined “differential” to mean “difference between the value obtained and the value for the correct model that answers the research question”.  Analyses with birth weight as the exposure suffered differential when mediators were included, and this was persistent irrespective of which mediators were chosen, with mid to later mediators causing the larger differential overall.  Analyses with adult BMI as the exposure saw differentials vary according to which prior body mass measures were included as confounders: early-life confounders yielded large differentials whilst later-life confounders yielded the smaller differential. 

CONCLUSIONS: The common practice to analyse a multitude of longitudinal measures within a lifecourse context in a multiple regression model can lead to biased estimates of the main exposure.  More careful consideration of the use of multiple regression is required, with the distinction between genuine confounders and mediators becoming more widely understood.