SSM Research Summary Competition 2023 Pre-doctoral winner

The Society for Medicine and Public Health ran this year a writing competition for the best research summary based on work presented at SSM in the past two years.

Congratulations to Georgia Tomova for winning one of the two places offered for the SSM Research Summary Competition 2023! Georgia Tomova is a PhD student of the Alan Turing Institute, based at the Leeds Institute for Data Analytics, University of Leeds and her Twitter name is @GeorgiaTomova

Read her winning entry here:

Research Summary Title: Challenges of Estimating Causal Effects in Nutritional Epidemiology: A Summary of Talks Presented at SSM Annual Scientific Meeting 2021

Estimating the causal effect of a dietary exposure is challenging. The best evidence is
thought to come from randomised controlled trials, but dietary experiments are difficult to
perform and do not represent the conditions that people normally consume food.
Consequently, there is considerable interest in estimating causal effects from observational
data. Alas, this too is extremely difficult, partly due to two major challenges: 1) how to
control for the rest of the diet, and the many unmeasurable determinants of diet, and 2)
how to estimate the effect of substituting one food with another.
We examined these two challenges using novel causal diagrams and data simulations in two
papers presented at the SSM Annual Meeting in 2021.
The first study showed that the best way to isolate the causal effect of individual dietary
exposure and minimise confounding from common dietary determinants is to separately
adjust for all sources of energy intake. Unfortunately, the use of this ‘all-components’ approach
is extremely rare in most applied nutritional research. Instead, it is most common for
researchers to control for ‘total energy intake’ as an overall measure of energy intake
alongside a selection of other components deemed important to the outcome under study.
In the second study, we examined the implications of this common approach. Since ‘total
energy’ is a collider for the exposure and all other sources of energy intake, the routine
practice of adjusting for energy intake leads to inadvertent, and potentially radical, changes
in the research question. At best, if the exposure and all other components are measured in
calories, then a study that adjusts for total energy will end up inadvertently estimating a
substitution effect, e.g. of refined carbohydrates instead of whole grains. At worst, if all
foods are measured in grams, then adjustment for total energy intake will introduce an
obscure substitution with limited real-world meaning.
These results have some important practical implications. In original research, when seeking
to estimate the causal effect of a dietary exposure, it is extremely important to establish a
clear research question and use an analytical method that is appropriate for that aim, e.g.
the all-components approach. In evidence synthesis, when seeking to pool the estimates
from multiple studies, it is important to identify what effect was estimated by each study
and only combines estimates for the same effects. Some of the heterogeneity in existing
meta-analyses may stem from pooling estimates of radically different effects.

Oral Presentations linked to the research summary:
1. Adjustment for energy intake in nutritional research: a causal inference perspective
2. Performance of food substitution models

Publications linked to the research summary:
1. Tomova GD, Arnold KF, Gilthorpe MS, Tennant PWG. Adjustment for energy intake in
nutritional research: a causal inference perspective. 2022. Am J Clin Nutr 115(1):189-
98. DOI: 10.1093/ajcn/nqab266
2. Tomova GD, Gilthorpe MS, Tennant PWG. Theory and performance of substitution
models for estimating relative causal effects in nutritional epidemiology. 2022. Am J
Clin Nutr 116(5):1379-88. DOI: 10.1093/ajcn/nqac188* * Editor’s Choice article