The Association Between Ultra-processed Food Consumption and the Risk of Mortality – What is the Real Deal?
Posted:
2019-06-05
Originally Posted February 15, 2019. Reposted June 5, 2019
Another observational nutrition study is making headlines around the world (1-3) but is also raising concerns about the failure of news stories to address the
significant limitations of the study. The observational prospective cohort study looked at the association between ultra-processed food consumption and the risk of mortality in middle-aged adults in France (4). The authors concluded that an association existed based on the data of more than 44,000 participants.
A PEN Evidence Analyst offered this analysis:
The results portray a very weak relationship between ultra-processed foods and mortality. The effect size, which was equal to a hazard ratio of 1.14, is so small that it is not likely of importance. GRADE guidelines encourage that for an observational study an effect size should be greater than 2, or better yet 5, before we should assume that there is an important effect (5). This hazard ratio of 1.14 was statistically significant, which is not surprising because there were over 44,000 participants in the study. When the sample size is very large, statistical significance is highly likely even when the effect is so small that it is not likely of any real importance.
Additional concerns are that the categorization of foods in the ultra-processed category included sugar-sweetened beverages and highly processed snack foods but also included ready-made meals and breads, which could be quite nutritious. Further, the researchers allowed the participants to select which 24-hour periods they reported their food intake, so we don’t know how well their selections reflect typical intakes.
The results observed could be due to the fact that those that ate the most of these ultra-processed foods were those most likely to have higher mortality rates for reasons beyond dietary intake. The individuals consuming the most of these ultra-processed foods were more likely to be of lower income, smoke and/or be single obese males with low levels of physical activity. Some additional variable or variables related to these variables that were controlled for in the analysis could be the actual causal factor for mortality. Therefore, residual confounding may explain this study’s results.
The researchers did mention that reverse causation was possible; that is those people who are at higher risk of death from a chronic disease may have been selecting more processed foods. This type of study design cannot rule out reverse causation.
The NHS and Science Media Centre reviews also outline the limitations and put the study and its findings into perspective.
References:
- Scutti S. Eating 'ultraprocessed' foods accelerates your risk of early death, study says. CNN. 2019 Feb 12. Available from: https://edition.cnn.com/2019/02/11/health/ultraprocessed-foods-early-death-study/index.html
- Eating ultra-processed foods can increase risk of early death: study. CTV. 2019 Feb 12. Available from: https://www.ctvnews.ca/health/eating-ultra-processed-foods-can-increase-risk-of-early-death-study-1.4293071
- Donnelly L. Modern diets could be killing us, suggests major study on ultra-processed foods. The Telegraph News. 2019 Feb 11. Available from: https://www.telegraph.co.uk/news/2019/02/11/modern-diets-could-killing-us-suggests-major-study-ultra-processed/
- Schnabel L, Kesse-Guyot E, Allès B, Touvier M, Srour B, Hercberg S, et al. Association between ultraprocessed food consumption and risk of mortality among middle-aged adults in France. JAMA Intern Med. 2019 Feb 11. doi:10.1001/jamainternmed.2018.7289. [Epub ahead of print]. Abstract available from: https://www.ncbi.nlm.nih.gov/pubmed/30742202
- Guyatt GH, Oxman AD, Sultan S, Glasziou P, Akl EA, Alonso-Coello P, et al. GRADE guidelines: 9. Rating up the quality of evidence. J Clin Epidemiol. 2011 Dec;64(12):1311-6. doi: 10.1016/j.jclinepi.2011.06.004. Epub 2011 Jul 30. Abstract available from: https://www.ncbi.nlm.nih.gov/pubmed/?term=21802902