The investigators evaluated how eating meat related to total and cause-specific mortality in the large European Prospective Investigation into Cancer (EPIC) cohort, including 448,568 participants in 23 participating centers across 10 European countries. Importantly, that study took care to separately evaluate unprocessed red meat, unprocessed poultry, and processed meats (including processed red meat and processed poultry). During an average follow-up of 12.7 years, 26,344 deaths occurred, comprising 5,556 due to CVD, 9,861 to cancer, 1,068 to respiratory disease, 715 to digestive tract diseases, and 9,144 to other causes. Notably, the authors appropriately accounted for potential effects of residual confounding (which would cause, in this case, overestimation of harm of meat intake) and random errors in diet assessment (which would cause underestimation of associations).
In calibrated and adjusted models for various lifestyle and dietary factors, consumption of unprocessed red meat was not significantly associated with CVD mortality (per 100 g/day, relative risk (RR) = 1.09, 95% confidence interval (CI) = 1.00 to 1.18); consumption of unprocessed poultry was associated with a non-significant trend toward lower risk (per 50 g/day, RR = 0.84, 95% CI = 0.69 to 1.03); and consumption of processed meat was associated with a 30% higher risk (per 50 g/day, RR = 1.30, 95% CI = 1.17 to 1.45). Matching the serving sizes, each 100 g/day of processed meats was associated with an approximately 70% higher risk (RR = 1.69, 95% CI = 1.37 to 2.10). Translated to weekly servings, each 100 g/week of unprocessed red meats had no significant association with CVD mortality (RR = 1.01, 95% CI = 1.00, 1.02), and each 100 g/week of processed meats was associated with 8% higher risk (RR = 1.08, 95% CI = 1.05, 1.11).
Do these findings suggest cause and effect? Observational studies can be limited by residual confounding, that is, the observed associations being due to other unmeasured or poorly measured factors. However, when considering such effects, it is crucial to consider plausible directions of confounding. As seen in previous studies, unprocessed and processed meat consumption in EPIC were each associated with higher-risk demographics and worse lifestyles, including older age, higher body mass index lower fruit intake, greater current smoking, and lower education; conversely, many of these associations were attenuated or reversed for poultry consumption. Although the authors adjusted for these factors, residual confounding could still be present as a result of imperfect covariate measurement. In addition, the authors did not adjust for other key dietary confounders such as fiber, whole grains, nuts, legumes, fish, and trans fats. Based on the associations of meat intake with these risk factors, residual confounding could overestimate the harmful associations of processed meat consumption and the protective associations of poultry consumption. However, residual confounding could not plausibly explain the absence of a link between unprocessed red meats and CVD, as the direction of residual bias in this case would be toward showing more harm, not less.
A second method to evaluate potential confounding is use of a ‘negative control’, that is, a health outcome on which the risk factor of interest would have little plausible effect [11]. In the EPIC investigation, when other causes of death were evaluated, intake of unprocessed red meat was not associated with cancer, digestive, respiratory, or other deaths, whereas intake of processed meat was associated with higher rates of cancer and other deaths (with a smaller magnitude than for CVD deaths) and was not associated with respiratory or digestive deaths. The absence of associations of processed meat intake with biologically unrelated causes of death supports a low likelihood of confounding as an explanation for the observed higher risks of CVD and cancer deaths.
What are the implications of these findings? In 2010, we performed a meta-analysis of observational studies that showed no significant association between intake of unprocessed red meat and coronary heart disease (CHD) (per 100 g/day, RR = 1.00, 95% CI = 0.81 to 1.23), and significant positive associations between processed meat intake and CHD (per 50 g/day, RR = 1.42, 95% CI = 1.07 to 1.89) [5]. However, whereas the findings for processed meat were based on 21,308 incident CHD events, the studies available for our meta-analysis of unprocessed red meats and CHD covered less than 1,000 cases. Subsequent analyses from large prospective cohorts in the USA supported stronger associations of processed meat intake with CVD, but also suggested statistically significant, although modest, associations of unprocessed red meats [12, 13]. This investigation in EPIC, including nearly half a million participants across 10 European countries and more than 5,000 cardiovascular events, confirms that consumption of processed meat is strongly associated with CVD risk, and that consumption of unprocessed red meat has little to no association.
These findings, taken together with previous studies, have important implications for understanding how meat consumption influences cardiovascular health. In previous analyses, we found that average contents of saturated fat, cholesterol, and heme iron are similar between unprocessed red meats and processed meats (indeed, average cholesterol and heme iron contents are lower in processed meats) [5]. The strong association of processed meats with CVD, compared with the weak or absent association of unprocessed red meats with CVD, suggests that none of these ingredients have major effects on CVD risk. This is supported by evidence for no overall association of saturated-fat consumption with incident CHD [2–4], and little overall association of dietary cholesterol with CHD [14].
These findings also inform the extent to which other meat ingredients might be relevant for risk. Experimental evidence suggests that trimethylamine N-oxide, a metabolite of L-carnitine formed by intestinal microbiota, is pro-atherogenic [9], yet, unprocessed red meats, which have the highest L-carnitine content, have little association with CHD, whereas processed meats, which are commonly made from pork or even poultry that contains much lower L-carnitine levels, are associated with higher CHD risk. In sum, these results suggest that trimethylamine N-oxide may not mediate the observed associations with risk.
Preservatives are the most notable difference between unprocessed and processed meats. In the USA, processed meats contain an average of 400% more sodium and 50% more nitrates than unprocessed red meats [5]. The predicted blood-pressure effects of the high sodium content alone can account for more than 2/3 of the observed relationship between processed meats and CHD risk [15].
Comments
View archived comments (1)