A priority of nutrition science is to identify dietary determinants of health and disease to inform effective public health policies, guidelines, and clinical interventions. Yet, conflicting findings in synthesizing evidence from randomized trials and observational studies have contributed to confusion and uncertainty. Often, heterogeneity can be explained by the fact that seemingly similar bodies of evidence are asking very different questions. Improving the alignment within and between research domains begins with investigators clearly defining their diet and disease questions; however, nutritional exposures are complex and often require a greater degree of specificity. First, dietary data are compositional, meaning a change in a food may imply a compensatory change of other foods. Second, dietary data are multidimensional; that is, the primary components (ie, foods) comprise subcomponents (eg, nutrients), and subcomponents can be present in multiple primary components. Third, because diet is a lifelong exposure, the composition of a study population's background diet has implications for the interpretation of the exposure and the transportability of effect estimates. Collectively clarifying these key aspects of inherently complex dietary exposures when conducting research will facilitate appropriate evidence synthesis, improve certainty of evidence, and improve the ability of these efforts to inform policy and decision-making.
Keywords: causal inference; nutrition epidemiology; nutrition science.
© The Author(s) 2024. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.