Processed meats, sugar-sweetened beverages, and trans-fatty acids all have been associated with chronic diseases, frequently type 2 diabetes, ischemic heart disease, and colorectal cancer. A new study in Nature Medicine, widely reported in the mainstream and social media, examines the associated risks. It is another meta-analysis, but with a difference. It employs a different analytic methodology, aiming to enhance confidence in its conclusions by reducing the inevitable variability in aggregating multiple study designs and findings – the acknowledged Achilles' Heel of meta-analysis.
Old Wine in A New Container
In summarizing the findings of meta-analysis and systematic review, it is challenging to combine the wide variety of study types, treatments, and clinical outcomes into a coherent, quantifiable conclusion. Heterogeneity refers to the extent to which the results or findings from different studies investigating the same risk-outcome relationship differ from one another. The presence and quantification of heterogeneity profoundly impact the certainty of a risk-outcome relationship – the uncertainty of the conclusions rises with the heterogeneity of the data.
Heterogeneity across a "very mixed evidence landscape" is a significant reason for many scientific controversies. Burden of Proof (BoP) studies attempt to quantify and integrate the complexity of heterogeneity into the assessment of evidence; greater heterogeneity reduces the certainty that an actual effect exists. Furthermore, BoP studies mathematically estimate the level of risk or benefit closest to the null hypothesis — that there is no risk or benefit — consistent with the available data, providing a more conservative and arguably more realistic measure of certainty for risk-outcome relationships.
- Burden of Proof, unlike other expressions of aggregate data, does not assume that risk and health outcomes increase in a straight line, at a constant proportional rate. It is the assumption of linearity that allows researchers to conclude and the media to report that eating one hot dog will decrease your life expectancy by 30 minutes. BoP recognizes that biological systems and clinical outcomes may take many forms, from the S-curve observed with growth to the J-shaped curves associated with the harms of excessive and insufficient salt intake. BoP allows the data to reveal its true shape, fitting the curve to the data rather than forcing the data to fit a predetermined straight line.
- The Burden of Proof removes 10% of the outliers, the “least coherent observations,” thereby reducing heterogeneity and stabilizing risk estimates, which generates smooth, plausible data curves.
Using this approach, researchers believe that Burden of Proof studies more accurately capture the genuine dose-response relationship, leading to more reliable and robust estimates of risk.
The findings of a BoP study are communicated through a 1- to 5-star rating, intended to inform individual choice, health policy, and practice. One Star indicates no association; individuals need not concern themselves with these risks or benefits. Five stars, a “very strong evidence of association,” where average exposure increases excess risk by 85% or more, e.g., smoking and lung cancer or high blood pressure and ischemic heart disease.
For our purposes, it is essential to consider two stars, where there is a weak or minimal association, on the verge of being considered without association. Here, the average exposure increases excess risk by 0-15%. Two-star ratings indicate a potential, weakly supported risk that warrants further investigation. In the original study demonstrating the methodology and utility of BoP, the relationship between the consumption of vegetables or unprocessed red meat and ischemic heart disease, where the risk was lowered by 12% and 1%, respectively, was rated two stars. It was the “very high heterogeneity between studies” that prompted that rating.
The current study
All the results demonstrated nonlinear, monotonic increases in risk with greater consumption, with the steepest increases in risk at low levels of intake, roughly one serving/day or less. Most studies were adjusted for age, sex, BMI, energy intake, and physical activity. The ratings were, with one exception, low, with only two stars, primarily due to the significant heterogeneity of the data, which included administrative records, registries, death certificates, self-reported incidences, biomarkers, and physician diagnoses. This weak rating reflects high heterogeneity, likely due to small effects, inconsistent findings, and unaccounted differences such as genetics or confounding factors.
Processed meat consumption and type 2 diabetes – The analysis involved 16 studies, with over 1 million participants and 64,000 type 2 diabetes outcomes. For those exposed, i.e., consuming between the 15th and 85th percentile of consumption, which we consider average, the risk of type 2 diabetes was 11%, compared to a theoretical scenario of no consumption at all. Trimming outliers significantly affects the risk–outcome score, suggesting the association is relatively sensitive to outliers- Processed meat consumption and ischemic heart disease – One Star. The analysis of 11 studies, with over 1 million participants, and 31,000 IHD outcomes. For the average exposure, the risk of IHD was 0.1%, with no change in the result after removing outliers.
- Processed meat consumption and colorectal cancer – The analysis of 18 studies involving over 2.6 million participants and 30,000 colorectal cancer outcomes. For the average consumption, the risk of colorectal cancer was 7% compared to no consumption. While the absolute score was sensitive to outliers, with a significantly decreased risk, it did not change to star scoring.
- Sugar-sweetened beverage (SSB) consumption and type 2 diabetes – The analysis involved 19 studies, encompassing 500,000 participants and 39,000 type 2 diabetes events. For the average consumption, the risk of type 2 diabetes was 8% compared to no consumption. Trimming the data had little impact.
- Sugar-sweetened beverage (SSB) consumption and ischemic heart disease – The analysis included eight studies, involving 900,000 participants and 24,000 IHD events. As with type 2 diabetes, the steepest risk was seen at very low consumption of an SSB, roughly 9 ounces of soda; that increased risk compared to no consumption was 2%. Trimming data had no impact.
- Trans Fatty Acid Consumption and Ischemic Heart Disease – The analysis involved six studies, representing 220,000 individuals and 12,000 IHD events. Average consumption was associated with a 3% elevated risk of IHD compared to no consumption. The trimming of data significantly reduced the risk to -0.12% and reduced the star rating to one, no association.
Drawing Conclusions
“The monotonic increases in health risk with increased consumption of processed meat suggest that there is not a ‘safe’ amount of processed meat consumption with respect to diabetes or colorectal cancer risk.”
The authors have provided the data dots along with their interpretation. For all the objectivity and quantifiability of the Burden of Proof methodology, this is the moment when bias re-enters either silently or aloud. While not explicitly cited, the recommendations reflect the spirit and underlying rationale of the precautionary principle, prioritizing the avoidance of potential harm over waiting for definitive proof of risk.
The precautionary principle is explicitly found in the paper presenting the BoP methodology.
“The precautionary principle implies that public policy should pay attention to all potential risks.”
That, of course, includes one and two-star ratings. Although they suggest that these ratings “should be investigated further through more robust, well-powered research, especially for those risks where exposure and outcome are common,” they call for study. The current researchers call for policy implementation.
“This study found evidence—under a conservative interpretation of the available data—to justify robust efforts and policies to promote the reduced consumption of processed meat, SSBs, and TFAs, particularly industrially produced TFAs, to reduce the risk of chronic diseases. Our finding supports the recent initiative of the WHO to ban industrially produced trans fats and their call to tax SSBs to reduce diet-related noncommunicable diseases. Our observation that the greatest increases in disease risk occurred at low intake levels suggests that even lower levels of habitual consumption of these dietary risk factors are not safe.”
A proactive stance to prevent harm, even in the presence of a weak association and the conclusion that “even lower levels of habitual consumption of these dietary risk factors are not safe,” clearly embodies the spirit of taking preventive measures in the face of scientific uncertainty about the precise magnitude or consistency of harm.
The Burden of Proof methodology introduces a welcome layer of restraint in risk assessment, tempering sweeping claims with mathematical humility. While researchers’ “conclusions” and media headlines may trumpet the risk of a single soda or hot dog, the real take-home message from this study and methodology reminds us that weak associations, especially in the presence of high heterogeneity, demand caution, not crusades. A one- or two-star rating should prompt curiosity, not panic. Before MAHA overhauls diets or public policy, let’s be sure we’re acting on solid ground, not statistical shadows.
Sources: Health Effects Associated With Consumption Of Processed Meat, Sugar-Sweetened Beverages And Trans Fatty Acids: A Burden Of Proof Study Nature Medicine DOI: 10.1038/s41591-025-03775-8 While we can unfortunately only provide a link to the abstract, a copy of the study was provided to us by the authors to base our reporting.
The Burden of Proof studies: assessing the evidence of risk Nature Medicine DOI: 10.1038/s41591-022-01973-2
Chuck Dinerstein, MD, MBA
Director of Medicine
Dr. Charles Dinerstein, M.D., MBA, FACS is Director of Medicine at the American Council on Science and Health. He has over 25 years of experience as a vascular surgeon.
