Meat consumption has increased since World War II. While this increase has long been confined to the Western world, that is, North America, North and Western Europe, and Australia/New Zealand, meat consumption is now also on the rise in other countries, such as China, due to their economic development 1. From a physiological perspective, a diet rich in meat has several potential nutritional benefits but also some potential adverse effects. Meat is rich in protein, iron, zinc and B-vitamins, as well as vitamin A. The bioavailability of iron and folate from meat is higher than from plant products such as grains and leafy green vegetables. The drawback, however, is the high content of cholesterol and saturated fatty acids, both of which have been shown to be positively associated with plasma low density lipoprotein (LDL) concentrations and the risk of coronary heart disease 2. Although iron is essential for prevention of anemia, a high intake, especially of heme iron, is related to the endogenous formation of N-nitroso compounds in the gastro-intestinal tract 34 and, thus, may be a risk factor for some cancer entities, for example, colon cancer 5.

Some prospective studies have evaluated the association between meat intake and mortality 67891011121314151617, but several of them were studies comparing meat consumers with vegetarians 9111316. One of the most recent studies, conducted among EPIC-Oxford participants, revealed that vegetarians as well as non-vegetarians with a health-conscious lifestyle had a statistically significantly lower mortality compared with the British general population 9. This is similar to the results of a German cohort, in which both vegetarians and health-conscious non-vegetarians had a statistically significantly lower overall mortality compared with the general population 11. These results indicate that the decreased mortality in vegetarians compared with the general population is in large part due to a healthy lifestyle, that is, being non-smokers, being leaner and more physically active, and so on. However, large US cohorts have reported an increased risk for early mortality among individuals with a high red and processed meat consumption compared with low meat consumption independent of smoking, obesity and other potential confounders 68.

Within the European Prospective Investigation into Cancer and Nutrition (EPIC) including more than 500,000 participants from ten European countries and, thus, reflecting a very heterogeneous diet, we examined the association between meat consumption and the risk for overall and cause-specific mortality.

Men and women in the top categories of red or processed meat intake in general consumed fewer fruits and vegetables than those with low intake. They were more likely to be current smokers and less likely to have a university degree (Table 1). Men with high red meat consumption consumed more alcohol than men with a low consumption, which was not seen in women. Baseline characteristics by consumption of poultry differed somewhat from the pattern observed for red and processed meat; individuals consuming more than 80 g poultry per day had a higher consumption of fruits and vegetables than those with an intake of less than 5 g per day, but there was no difference in smoking habits at baseline.

Median follow-up time of our cohort was 12.7 years with a maximum of 17.8 years; median follow-up time was 8.5 years in cases and 12.9 years in non-cases. During the follow-up period, 26,344 study participants (11,563 men and 14,781 women) died. Of these, 5,556 died of cardiovascular diseases, 9,861 of cancer, 1,068 of respiratory diseases, 715 of digestive tract diseases, and 9,144 of other causes (including 976 who died from external causes). A high consumption of red meat was related to increased all-cause mortality (Table 2). Participants with an intake of 160+ g red meat/day had a HR = 1.37 (95% CI 1.23 to 1.54) compared with individuals with an intake of 10 to 19.9 g/day in the simple model taking into account age, study center, and sex. The association was attenuated in the multivariate model, but was still statistically significant (HR = 1.14, 95% CI 1.01 to 1.28). The association for processed meat was stronger than for red meat. In the multivariate model, the HR for high (160+ g/day) versus low intake was 1.44 (95% CI 1.24 to 1.66). Low consumption of poultry was associated with increased all-cause mortality compared with moderate consumption (Table 2), but there was no association of high poultry consumption compared with moderate consumption. Mutual adjustment for all meat groups did not appreciably change the observed associations for processed meat and poultry, but the elevated total mortality observed in the highest category of red meat consumption became statistically insignificant (Table 2). We estimated that 3.3% (95% CI 1.5 to 5.0%) of deaths could be prevented if all participants had a processed meat consumption of less than 20 g/day.

We also evaluated the association between meat consumption and all-cause mortality in two continuous models, that is, obtaining uncalibrated and calibrated risk estimates (Table 3). Similarly to the observation of no association between red meat consumption and all-cause mortality in the multivariable categorical model, we observed no statistically significant association in the continuous models either. While the associations were similar in the uncalibrated and calibrated models for red meat and poultry, the association between processed meat consumption and all-cause mortality was stronger in the calibrated model: per 50 g increase in daily processed meat consumption, the HR for all-cause mortality was 1.18 (95% CI 1.11 to 1.25). In the calibrated spline models, we observed significantly higher all-cause mortality with higher consumption of processed meat and no statistically significant association with red meat or poultry intake (Figure 1). However, all-cause mortality was higher among participants with very low or no red meat consumption.

We also explored the association of meat intake with mortality in models without adjusting for total energy intake. However, the results were identical for models not including (data not shown) and including total energy intake. Results were also similar for models including total energy and fruit and vegetable intake. The associations between red or processed meat or poultry intake and all-cause mortality were also similar for the first two years or after the first two years of follow-up (data not shown).

For processed meat, for which we observed statistically significant associations with overall mortality, we examined whether this effect differed by sub-groups of our population. We did not observe statistically significant effect modification by sex (Table 4), with similarly increased all-cause mortality in both sexes, although the association was statistically significant only among men (HR = 1.35, 95% CI 1.16 to 1.58, 160+ versus 10 to 19.9 g/day), but not among women (HR = 1.38, 95% CI 0.95 to 2.00; P-interaction 0.88). This may be due to the relatively small number of deaths among women in the highest processed meat consumption category (29 women; 194 men). There was also a statistically significant interaction between smoking and processed meat consumption (P-interaction 0.01), with mortality being significantly increased among former (HR = 1.68, 95% CI 1.29 to 2.18) and current smokers (HR = 1.47, 95% CI 1.18 to 1.83), but there was no association among never smokers (HR = 1.24, 95% CI 0.89 to 1.72). However, the small number of deaths among never smokers has to be taken into account (n = 44; 72 among former and 107 among current smokers in the top consumption category). We observed a statistically significant interaction with body mass index, such that the association between processed meat consumption and all-cause mortality was stronger in lean than in overweight and obese participants (P-interaction 0.04). Those with a lower fruit and vegetable intake (below median intake) had a higher overall mortality in the highest consumption category of processed meat (160+ g/day) as compared to subjects with a fruit and vegetable intake above the median intake (P-interaction 0.001).

No statistically significant heterogeneity between countries was observed for the associations of processed meat and poultry consumption with mortality (P-values >0.05), but was observed for the association of red meat intake with mortality (P-value 0.006). This heterogeneity was not driven by risk estimates from a particular country (data not shown).

Very high consumption of red meat was associated with a non-significantly increased cancer mortality, but not with deaths due to cardiovascular diseases, respiratory diseases, diseases of the digestive tract, or any other cause of deaths (Table 5). However, the increase in risk was not observed in the continuous model. In contrast, participants who consumed 160+ g of processed meat per day had an increased risk of dying of cardiovascular diseases compared with those who consumed only moderate amounts (10 to 19.9 g/day; HR = 1.72, 95% CI 1.29 to 2.30); this association was also observed in the continuous model, even after correcting for measurement error (HR = 1.30, 95% CI 1.17 to 1.45 per 50 g/day). There was also a significant positive association between processed meat consumption and risk of dying from cancer (HR = 1.11, 95% CI 1.03 to 1.21 per 50 g/day) or other causes of death (HR = 1.22, 95% CI 1.11 to 1.34 per 50 g/day). Since the results of the categorical model deviated from the continuous model, we reexamined the association between processed meat intake and cancer risk using the lowest consumption category (0 to 9.9 g/d) as reference and observed a statistically significantly increased risk for cancer mortality for those who consumed 80 to 159.9 g/day (HR = 1.12, 95% CI 1.01 to 1.24) and a non-significantly increased risk in the highest consumption category (HR = 1.19, 95% CI 0.93 to 1.51). Although for some categories of processed meat intake a positive association with mortality from respiratory diseases and digestive tract diseases was obtained, the continuous models failed to reach statistical significance. Poultry consumption was not consistently associated with deaths from any of these causes.

In the EPIC cohort, a high consumption of processed meat was related to moderately higher all-cause mortality. After correction for measurement error, red meat intake was no longer associated with mortality, and there was no association with the consumption of poultry. Processed meat consumption was associated with increased risk of death from cardiovascular diseases and cancer.

The largest study so far, the National Institutes of Health-American Association of Retired Persons (NIH-AARP) cohort in the US, reported positive associations of both red and processed meat consumption with risk for all-cause mortality 8. In that cohort, the association was stronger for red meat than for processed meat intake, which might be due to the fact that red meat in that US cohort also included processed meat. Similarly, in the Nurses' Health Study and the Health Professionals Follow-up Study (HPFS), high red meat intake was related to higher all-cause mortality 6. The effect was similar for unprocessed and processed red meat. Similar associations were reported in other 101417, but not all, studies 15. Also, several vegetarian studies did not find increased all-cause mortality among non-vegetarians compared with vegetarians 9111316. The EPIC results do not show the lowest relative risks (RRs) for subjects in the lowest meat intake category, but a slight J-shaped association with the lowest risk among subjects with low-to-moderate meat consumption. This was observed for red meat and poultry. Also, taking into account the results from the studies that evaluated vegetarian and low-meat diets, it appears that a low - but not a zero - consumption of meat might be beneficial for health. This is understandable as meat is an important source of nutrients, such as protein, iron, zinc, several B-vitamins as well as vitamin A and essential fatty acids (linoleic acid and to a minor extent eicosapentaenoic and docosahexaenoic acids also). A sub-optimal supply of some of these nutrients due to an unbalanced type of vegetarian diet seems possible and might be associated with an increased risk for morbidity and mortality. However, support for this hypothesis from the literature is not strong, especially when looking at the population level. Alternatively, subjects with very moderate meat consumption may be the group with the highest proportion of health-conscious subjects who also try to optimize their diet (as part of a healthy lifestyle).

In contrast to the US results, we observed a consistent association between processed meat consumption and total mortality but not between red meat consumption and total mortality. Processed meats such as sausages, salami and bacon have a higher content of saturated fatty acids and cholesterol than fresh red meat; the latter is often consumed after removing the visible fat tissue, whereas the proportion of fat in sausages often reaches 50% of the weight or even more. Both high saturated fat and cholesterol intake have been found to be related to the risk of coronary heart disease 2. Also, processed meat is treated by salting, curing, or smoking in order to improve the durability of the food and/or to improve color and taste. These processes, however, lead to an increased intake of carcinogens or their precursors (polycyclic aromatic hydrocarbons, heterocyclic aromatic amines, nitrosamines) or to a high intake of specific compounds possibly enhancing the development of carcinogenic processes (for example, nitrite).

We estimated that 3.3% of all deaths could be prevented if processed meat consumption were below 20 g/day. In the AARP cohort, the preventable fraction was estimated to be much higher, that is, 20% if women decreased their processed meat consumption to less than 1.6 g/1,000 kcal/day (the authors did not state the preventable fraction for men 8). The preventable fraction was estimated to be 9.3% in the HPFS and 7.6% in the Nurses' Health Study if the participants lowered their red meat (processed and unprocessed) consumption to less than 0.5 servings per day. The difference between the US studies and our result is likely due to the stronger risk estimates observed in the US cohorts compared with our cohort, but may also be explained by higher meat consumption in the US than in Europe.

As in the US cohorts, EPIC participants with a high processed meat intake had an increased risk of cardiovascular and cancer mortality. We have previously reported an increased risk of colorectal 29 and gastric 30 cancer with high meat, in particular processed meat, consumption. However, in contrast to the US cohorts 68, there was no statistically significant association of red meat consumption with risk of cancer or cardiovascular mortality. Also, in the Japan Collaborative Cohort Study, meat consumption up to 100 g/day was not related to increased mortality from cardiovascular disease 7.

The EPIC study has several strengths including its prospective design, the large sample size and the assessment of diet using two different methods, that is, dietary questionnaires and a 24-hour dietary recall in a representative sub-sample of the cohort. In a series of validation studies, correlation coefficients for meat intake between 12 24-hour recalls and food questionnaires ranged between 0.4 and 0.7 31. The single 24-hour recalls in a representative sample of the cohorts allow for partly correcting for systematic over- and underestimation of dietary intakes 3233. It is, therefore, important to note that the impact of the calibration method in our study was such that the risk estimates from the calibrated data are usually stronger than the non-calibrated results. Nevertheless, measurement error may still have an effect on calibrated RRs to a certain extent because the error structure in the reference method is not entirely independent of that in the FFQ 3435. A further methodological strength of the EPIC cohort is the inclusion of individuals from 10 European countries with distinctly diverging meat consumption habits 36. A high between-person variation in diet decreases the impact of measurement error and enables the detection of only modest diet-disease relationships. We did explore meat intake in models with and without adjusting for total energy intake. In models adjusting for energy intake, meat intake must substitute the intake of other non-specified energy-providing foods. However, the results were identical for models including and not including total energy intake and also for models including total energy and fruit and vegetable intake, which have also been considered important in the development of chronic diseases. The results observed in this study were, thus, robust in a number of different models with different interpretation. Lastly, loss to follow-up is negligible as vital status is known for 98% of the cohort.

We cannot exclude residual confounding, in particular due to incomplete adjustment for active and passive smoking. The sub-group analysis for processed meat showed heterogeneity according to smoking, with significant associations only in former and current smokers and no significant associations in never smokers, which is compatible with residual confounding by smoking. Although EPIC includes ten European countries with a wide range of dietary behaviors, we observed relatively little heterogeneity in the association between meat consumption and overall mortality.

We relied on mortality information from death certificates but cause of death as coded on death certificates is not perfect. Deaths due to cancer are most correctly diagnosed, whereas deaths due to coronary heart disease tend to be overrepresented and respiratory diseases might be underrepresented 373839.

The results of our analyses suggest that men and women with a high consumption of processed meat are at increased risk of early death, in particular due to cardiovascular diseases but also to cancer. In this population, reduction of processed meat consumption to less than 20 g/day would prevent more than 3% of all deaths. As processed meat consumption is a modifiable risk factor, health promotion activities should include specific advice on lowering processed meat consumption.

AARP: American Association of Retired Persons; CI: confidence interval; EPIC: European Prospective Investigation into Cancer and Nutrition; FFQ: food frequency questionnaire; HPFS: Health Professionals Follow-up Study; HR: hazard rate; ICD-10: 10^th Revision of the International Classification of Diseases; LDL: low density lipoprotein; NIH: National Institutes of Health; PAR: population attributable risk; RR: relative risk.

The authors declare that they have no competing interests.

ER is the overall coordinator of the EPIC study, which he conceptualized, designed, and implemented in collaboration with the main investigators in the collaborating centers: Denmark: ATjønneland, KO; France: MCBR, FCC; Germany: RK, HB; Greece: ATrichopoulou, DT; Italy: VK, DP, SP, RT; Netherlands: HBBdM, PHMP; Spain: MJS, AB; Sweden: GH; UK: KTK, NW, TJK; IARC: IR. All authors contributed to recruitment, data collection/acquisition and/or biological sample collection, and are responsible for the ongoing follow-up and management of the EPIC cohort. This analysis was initiated and supervised by JL. The article was written by SR with assistance from JL, KO, and HBBdM, taking into account the comments and suggestions of the coauthors. All coauthors had the opportunity to comment on the analysis and interpretation of the findings and approved the final version for publication.