It is common knowledge that a diet rich in ultra-processed foods is one of many factors that can increase the risk of developing chronic non-communicable diseases. However, that hasn’t stopped these products from gaining more space in our homes and, consequently, becoming essential parts of our daily lives. A recent study published in The Journal of Nutrition found that more than half of the calories consumed at home by adults in the US come from ultra-processed foods.
Do ultra-processed foods kill?
As provocative as this section title may sound, it accurately reflects the focus of a recent meta-analysis published in the American Journal of Preventive Medicine, which estimated the dose-response relationship between ultra-processed food (UPF) consumption and all-cause mortality. It also sought to calculate population-attributable fractions [1], and the total number of premature deaths (ages 30-69) from all causes attributable to UPF consumption in eight countries.
These countries were grouped by level of UPF intake:
● Low (Colombia and Brazil)
● Intermediate (Chile and Mexico)
● High (Australia, Canada, the United Kingdom, and the United States).
Data on UPF consumption were obtained from the most recent national food surveys, such as the 2017–2018 National Health and Nutrition Examination Survey in the United States. The proportion of UPFs in the diet was measured based on their percentage of total calories in the average diet. Data on premature deaths were sourced from the Global Burden of Disease study. This large-scale epidemiological project estimates the impacts of diseases, injuries, and risk factors across 204 countries and territories between 1990 and 2021.
The population attributable fractions (PAFs) and the number of attributable deaths are calculated by reducing UPF consumption to the theoretical minimum risk level, assumed to be 0% of total energy intake.
Among the 239,982 participants included in the analysis, 14,779 deaths were recorded. The average contribution of UPFs to total energy intake varied substantially across the countries analyzed. From Colombia (15%) and Brazil (17.4%) to significantly higher intakes in Australia (37.5%) and the United States (54.5%).
The estimated dose-response to the consumption of UPFs showed an increased all-cause mortality of 2.7% for each 10% increase in consumption. Unsurprisingly, the PAFs reflected these consumption differences, with UPFs contributing to 3.9% of deaths in Colombia, as high as 13.7% and 13.8% in the US and UK, respectively. The absolute number of premature deaths [2] ranged from 1,874 in Chile to 124,107 in the United States.
Based on these findings, the authors conclude:
“These findings highlight that adherence to ultra-processed dietary patterns represents a relevant public health concern in middle- and high-income countries. The findings support the need for reducing the consumption of UPF through creation of healthy environments using regulatory and fiscal policies.”
Limitations and Perhaps an Unwarranted Assumption?
Among the limitations acknowledged by the authors are the relatively small number of cohort studies evaluated, and despite adjustments for potential confounding factors, unmeasured variables may still have influenced the results. Although this is a meta-analysis, the included studies are observational, only identifying associations and not establishing causality. The estimates of ultra-processed food (UPF) consumption relied on tools such as food frequency questionnaires and 24-hour recalls, prone to recall bias and the Hawthorne effect. Finally, self-reported dietary patterns likely change over time, potentially compromising the accuracy of exposure estimates and the attributed risks.
Perhaps the most significant limitation is the authors’ assumption that a safe level of UPF intake is 0% - any caloric contribution from UPFs would correspond to an increased risk of all-cause mortality. While this approach aligns with the principle of primary prevention, which assumes no exposure level is entirely safe, it carries important methodological drawbacks. For instance, it does not account for ultra-processed foods with very low or zero calorie content, such as diet gelatin or diet soft drinks.
Furthermore, by relying solely on energy contribution, their model groups foods with vastly different nutritional profiles but identical calorie counts.
For example, YoPRO, a popular protein drink in Brazil, contains approximately 157.5 kcal, which is equivalent to roughly one ounce of a Kinder Joy chocolate egg. Although the nutritional inferiority of Kinder Joy, a candy with greater sugar, saturated fats, and caloric density than a protein shake, both would account for 7.88% of the daily caloric intake in a 2,000 kcal diet. In other words, whether someone consumes a ready-to-drink protein shake or a Kinder Joy alongside whole-wheat bread for breakfast, their theoretical risk of mortality would increase by 2.7%.
Although this example simplifies complex methodology — since the study doesn’t assess the impact of individual foods and relies on population-level averages — it effectively illustrates how a methodology based solely on caloric contribution can yield statistically significant yet nutritionally misleading results.
Thus, the findings remain within a theoretical framework that, while plausible given current trends in UPF consumption, becomes entangled with confounding variables such as sedentary lifestyles, elevated body mass index (BMI), smoking, and stress.
In writing a rebuttal to this study, Dr. Nerys Astbury, Associate Professor at the Nuffield Department of Primary Health Care Sciences at the University of Oxford, notes:
“In my view, the NOVA system, which defines foods according to different levels of food processing, has many limitations, including arbitrary definitions and overly broad food categories, the over-emphasis of food ingredients as opposed to the processing per se, and the difficult practical application of the system in accurately classifying foods. This is especially notable when attempting to classify foods from dietary data collected in large cohort studies, as in this study. More research is needed to ascertain a causal link between UPF and disease and to establish the mechanisms involved.”
At the end of the day, moderation remains key, even when it comes to the “horrific” ultra-processed foods.
[1] PAF is an epidemiological measure indicating the proportion of all cases of a disease or adverse outcome in a population that can be attributed to a specific cause.
[2] A value based on an equation combining PAFs, total premature deaths, and population size.
Sources: Trends in Adults’ Intake of Unprocessed/Minimally Processed, and Ultra-processed foods at Home and Away from Home in the United States from 2003-2018. The Journal of Nutrition. DOI: 10.1016/j.tjnut.2024.10.048
Premature Mortality Attributable to Ultraprocessed Food Consumption in 8 Countries. American Journal of Preventive Medicine. DOI: 10.1016/j.amepre.2025.02.018
