People often wonder why one nutritional study tells them that eating too many eggs, for example, will lead to heart disease, while another tells them the opposite. The answer to this and other controversial food studies may lie in using statistics, according to a report published today in American Journal of Clinical Nutrition.
Research led by scientists from the University of Leeds and the Alan Turing Institute – National Institute data science and artificial intelligence—shows that the standard and most common statistical approach to studying the relationship between food and health may produce misleading and meaningless results.
Lead author Georgia Tomova, Ph.D. The researcher, from the Institute for Data Analytics at the University of Leeds and the Alan Turing Institute, said: “These findings have implications for everything we think we know about the effects of food on health.
“It is common knowledge that different nutritional studies tend to produce different results. One week the food seems bad, and the next it seems good.”
The researchers found that the widespread practice of statistical control or accounting for total energy intake can lead to dramatic changes in the interpretation of results.
Controlling for other foods eaten can further bias the results, making unhealthy foods appear healthy or vice versa.
Ms Tomova added: ‘Because of the wide variation between individual studies, we tend to rely on review articles to provide an average estimate of whether and to what extent a particular food causes a particular health condition.
“Unfortunately, because most studies have different approaches to controlling for the rest of the diet, it’s likely that each study estimates a completely different amount, making the ‘average’ meaningless.”
The research identified the problem using the new techniques of “causal inference” popularized by Judea Pearl, author of The Book of Why.
Senior author Dr Peter Tennant, associate professor in the Department of Health Data at Leeds Medical School, explained: “Unless you can do an experiment, it’s very difficult to tell whether something causes something else and to what extent.
“That’s why people say, ‘correlation does not equal causation.’ These new methods of ‘causal inference’ promise to help us discover the causal effects of correlation, but they also highlight quite a few areas that we do not fully understand.”
The authors hope that this new study will help nutrition scientists better understand the problems with inadequate control of total energy intake and overall diet and gain a clearer understanding of the effects of diet on health.
Dr. Tennant added: “Different studies can produce different estimates for a number of reasons, but we believe that this one statistical problem may explain much of the disagreement. Fortunately, this can be easily avoided in the future.”
Georgia D. Tomova et al. Theory and effectiveness of substitution models for estimating relative causal effects in nutritional epidemiology, American Journal of Clinical Nutrition (2022). DOI: 10.1093/ajcn/nqac188
Provided
University of Leeds
Citation: Statistical Oversight May Explain Inconsistencies in Nutritional Research (2022, October 13) Retrieved October 13, 2022, from https://medicalxpress.com/news/2022-10-statistical-oversight-inconsistencies-nutritional.html
This document is subject to copyright. Except in good faith for the purpose of private study or research, no part may be reproduced without written permission. The content is provided for informational purposes only.