Why are observational studies on animal-sourced foods and chronic disease inconclusive?

Observational evidence has linked 'plant-based' eating to a lower risk of an impressive series of health issues, including obesity, cardiovascular disease, diabetes, diverticular disease, eye cataract, cancer, and vulnerability to coronavirus infection. However, such data are largely insufficient from a methodological perspective to support dietary recommendations away from animal source foods. This applies to the case against red and/or processed meats as well, despite reported associations between their intake and various illnesses, including intestinal disease, asthma, pneumonia, kidney disease, type-2 diabetes, cardiometabolic disease, and cancer. These associations do not necessarily imply that animal source foods are intrinsically unhealthy, and more rigorous research is needed to draw definitive conclusions.

Observational evidence linking animal-source foods, particularly red and processed meats, to negative health outcomes is generally weak. Systematic evaluations, including those using the GRADE protocol, classify this evidence as having ‘low’ to ‘very low’ certainty. An umbrella review characterizes the connection between red and processed meats and colorectal cancer as merely ‘suggestive’, while the ‘burden of proof’ methodology found only weak evidence linking unprocessed red meat to certain diseases and no association with stroke. The Bradford-Hill criteria further conclude that causality between meat consumption and health outcomes cannot be reliably established, with many studies affected by bias, confounding, and selective reporting. Most of this evidence comes from observational studies with low relative risk (RR) values, often below the threshold required for meaningful causal inference, particularly in the likely presence of confounding factors. In contrast, other components of the Western diet, such as visceral fat and metabolic syndrome, demonstrate much stronger associations and plausible biological mechanisms. Moreover, when absolute risk (AR) is reported, it is typically minimal for red and processed meat consumption. Despite frequent recommendations to reduce red meat intake, the current evidence base lacks the strength and reliability needed to support strong causal claims or broad dietary guidelines for the general population.

Epidemiological research often relies on self-reported data from surveys and food frequency questionnaires, which are prone to significant errors, including memory-based inaccuracies, social desirability bias, and survivorship bias. Categorization issues further compound these problems, as pre-defined food groups often fail to match regulatory definitions, leading to heterogeneous and misleading classifications, especially for ‘red’, ‘white’, and ‘processed’ meats. Similarly, plant-based diets range from healthy to unhealthy, making generalizations problematic. Social desirability bias affects data reliability, as respondents often report intake that aligns with perceived health norms rather than actual consumption. For example, obese individuals underreport calorie intake, and self-reported fruit and vegetable consumption often correlates poorly with actual biomarkers. Self-identified vegetarians may underreport occasional meat consumption, and many studies group vegetarians with semi-vegetarians, which skews results. Survivorship bias also distorts findings, as a majority of individuals revert to eating animal-source foods within a few months of adopting a vegetarian or vegan diet. Studies on plant-based diets may therefore inadvertently focus on a minority who thrive on these diets, potentially due to genetic factors influencing lipid metabolism and brain function. Given these methodological issues, relying solely on self-reported dietary data for health outcomes leads to questionable conclusions, necessitating biomarker-based validation for greater accuracy.

In observational research, associations between animal-source foods and health risks are influenced by broader dietary patterns and lifestyle factors, making it challenging to isolate specific dietary effects. Diet quality, socio-demographic characteristics, and lifestyle habits obscure the relationship between food intake and health outcomes. For example, both healthier vegetarians and healthier meat eaters tend to have lower health risks than less healthy meat eaters, underscoring the importance of overall context rather than focusing on individual food groups. Healthy user bias adds another layer of complexity, as individuals who follow dietary advice, such as reducing red meat intake, are often socioeconomically advantaged and more health-conscious, which skews results. In many cases, associations between meat consumption and health risks become non-significant after adjusting for confounders like obesity, smoking, and physical activity. Meta-analyses that exclude studies on Seventh-Day Adventist communities, which adhere to health-focused lifestyles, show diminished benefits of vegetarian diets on cardiovascular health. Furthermore, even after statistical adjustments, residual confounding cannot be ruled out. Cultural and regional differences further complicate findings, with the health impacts of meat and egg consumption varying across populations. For instance, global analyses, such as those from the PURE cohort, show a positive correlation between red meat intake and life expectancy, while studies in the US reveal differing and harmful associations. Similarly, the Southern European Atlantic diet, which includes red and processed meats, has demonstrated protective effects, challenging traditional assumptions.

Developing a case against animal source foods is often based on cherry-picked evidence from observational studies. In many studies there is no unambiguous association between their avoidance and improved health. The evidence on unprocessed red meat, for instance, is conflicting, some large studies showing no association. Poultry, eggs, and dairy display mostly neutral or even protective associations with health. Butter's associations with mortality, cardiovascular disease, and diabetes are relatively small or neutral. The overall picture is complex, and context plays a crucial role in interpreting the results from observational studies on animal source foods.

Embracing observational data to criticize omnivore diets would undermine the notion that vegan diets are always superior for avoiding chronic diseases and death. Those who rely on nutritional epidemiology to argue that we need more ‘plant-based’ eating need to acknowledge that some animal source foods also come with protective associations – implying that their underconsumption may be suboptimal for certain aspects of health. The global PURE study, for instance, has associated the intake of fish, eggs, full-fat dairy, and white meat, along with plant foods, to protective health outcomes. In the PURE cohort, red meat is mostly considered neutral in terms of health outcomes. In some studies, red meat consumption has even been associated with benefits, including lower morbidity, increased telomere length, and decreased risk of depression. Also, some studies have linked vegetarian and vegan diets to poorer health outcomes and lower quality of life. However, it is essential to recognize that both pro-meat and pro-plant studies may have limitations in establishing causal relationships, and they should be considered together in the overall debate about dietary choices.