Background: Methods for the assessment of changes in dietary intake across the life course are underdeveloped.
Methods: We demonstrate the use of linear-spline multilevel models to summarize energy-intake trajectories through childhood and adolescence and their application as exposures, outcomes, or mediators. The Avon Longitudinal Study of Parents and Children assessed children’s dietary intake several times between ages 3 and 13 years, using both food frequency questionnaires (FFQs) and 3-day food diaries. We estimated energy-intake trajectories for 12,032 children using linear-spline multilevel models. We then assessed the associations of these trajectories with maternal body mass index (BMI), and later offspring BMI, and also their role in mediating the relation between maternal and offspring BMIs.
Results: Models estimated average and individual energy intake at 3 years, and linear changes in energy intake from age 3 to 7 years and from age 7 to 13 years. By including the exposure (in this example, maternal BMI) in the multilevel model, we were able to estimate the average energy-intake trajectories across levels of the exposure. When energy-intake trajectories are the exposure for a later outcome (in this case offspring BMI) or a mediator (between maternal and offspring BMI), results were similar, whether using a two-step process (exporting individual-level intercepts and slopes from multilevel models and using these in linear regression/path analysis), or a single-step process (multivariate multilevel models). Trajectories were similar when FFQs and food diaries were assessed either separately, or when combined into one model.
Conclusions: Linear-spline multilevel models provide useful summaries of trajectories of dietary intake that can be used as an exposure, outcome, or mediator.
From the aMRC Centre for Causal Analyses in Translational Epidemiology, University of Bristol, Bristol, United Kingdom; and bSchool of Social and Community Medicine, University of Bristol, Bristol, United Kingdom.
E.L.A. is funded by a UK Medical Research Council (MRC) studentship (Grant ref: G0600705). A.F., C.M.-W., and L.D.H. are funded by UK MRC fellowships (Grant refs: 0701594, MR/J011932/1, and G1002375, respectively). The research leading to these results has received funding from the European Union’s Seventh Framework Program (FP7/2007–2013), under grant agreement n° HEALTH-F2-2009–241762 for the project FLIP, and the UK MRC (Grant ref: G1000726). The UK MRC (Grant ref: G074882), the Wellcome Trust (Grant ref: WT076467), and the University of Bristol provide core funding support for ALSPAC. The UK MRC (Grant ref: G0600705) and the University of Bristol provide core funding for the MRC Centre of Causal Analyses in Translational Epidemiology. K.N. and P.E. have received research funding from the Arthritic Association, United Kingdom, Danone Baby Nutrition (Nutricia Ltd), and Pfizer Nutrition, United Kingdom, and P.E. has received consultancy funding from Plum Baby United Kingdom. This work has been carried out independently.
P.E. and K.N. have received occasional support from commercial infant food manufacturers and have spoken at invited lectures. None of the remaining authors had a conflict of interest to declare.
Supplemental digital content is available through direct URL citations in the HTML and PDF versions of this article (www.epidem.com). This content is not peer-reviewed or copy-edited; it is the sole responsibility of the author.
Correspondence: Emma L. Anderson, MRC Centre for Causal Analyses in Translational Epidemiology School of Social and Community Medicine Oakfield House, Oakfield Grove Bristol BS8 2BN, United Kingdom. E-mail: firstname.lastname@example.org.
Received July 26, 2012
Accepted February 19, 2013