Dietary intake of animal and plant proteins and risk of all cause and cause-specific mortality: The Epic-Italy cohort

1Introduction

An overwhelming body of data indicate that lower intake of dietary protein, methionine, tryptophan and branched chain amino acids, independently of energy intake, improve health and extend lifespan in multiple model organisms including rodents [1–4]. Results from human epidemiological studies on the association between total and animal protein consumption and mortality are still contentious, with some reporting an inverse association, but many no or even a positive relationship [5, 6]. In contrast, most studies show a protective effect of plant proteins that typically have lower content of methionine, tryptophan and branched chain amino acids [5–9].

Several residual or unmeasured dietary and metabolic confounding factors may explain these discrepancies. For example, animal protein consumption (and associated iron and saturated fat content) is considerably higher in the large North American study populations compared to those living in Southern Europe countries like Italy [7]. Moreover, as it has been shown for the BMI-mortality relationship, the protein-mortality associations may be influenced by other differences in behavioral factors, including degree of physical activity, smoking, alcohol consumption and diet quality, which deeply affect metabolic health independently of calorie and protein intake [10].

Here, we used data from a large Italian cohort study (EPIC-Italy, 1993–1998) of 47,749 men and women, and up to 15 years of follow-up, to prospectively examine the associations of total, animal and plant protein intake with the risk for all-cause, cardiovascular and cancer mortality. We also performed isocaloric substitution associations of different protein sources with mortality, and analyzed the joint association of protein intake and lifestyle factors (i.e., smoking status, body mass index, physical activity, alcohol consumption, and adherence to a Mediterranean dietary pattern) with disease-specific mortality. Finally, in a subset of study participants, we collected detailed data on a wide spectrum of metabolic risk factors that allowed for differential subgroup analysis of the relationship between animal and plant protein intake with key metabolic determinants of cardiovascular risk.

2Methods

3Results

After a median time of observation of 15.2 years with 644,254 person-years of follow-up, we documented 2,449 deaths, of which 1414 were due to cancer, and 441 were due to cardio- and cerebrovascular diseases. Table 1 shows the baseline characteristics of participants according to quintiles of animal and vegetable protein intake expressed as percent of energy. Mean energy from animal protein varied from 6.4 to 14% across quintiles that it is markedly lower than the amounts consumed by participants enrolled in large cohort studies of North America [Supplementary Table 1); in contrast, mean energy from vegetable protein was substantially similar ranging from 3.4% (lowest quintile) to 7.1% (highest quintile). Compared with participants in the lowest energy from animal protein quintile, those in the highest quintile consumed more fat, and less vegetable protein, starch, sugar, alcohol and energy; they had higher BMI and lower waist-to-hip ratio, and were less educated and physically active. Participants in the highest quintile of energy from animal proteins consumed more protein from processed and unprocessed red meat, poultry, egg and dairy but less pasta, bread and pizza. In contrast, those with higher plant protein intake consumed less fat, sugar and alcohol, and more starch and energy, than those in lowest quintile. Similarly to the participants in the highest quintile of energy from animal, they had higher BMI and were less educated. Participants in the highest quintile of energy from vegetable proteins were more often smokers, physically inactive, and consumed more protein from fish, and calories from pasta, bread, pizza, and legumes. Vegetable protein consumption was higher in southern Italy (Napoli and Ragusa), whereas animal protein intake was greater in northern and central Italy (Turin, Varese and Florence).

Participants with higher intake of energy from animal protein had significantly higher total, LDL and non-HDL cholesterol (p = 0.0001), and fasting glucose (p = 0.008), despite consuming significantly less calories. In contrast, those with the highest intake of energy from vegetable protein had lower total cholesterol (p = 0.001), LDLc (p = 0.029), HDLc (p = 0.001), fasting insulin (p = 0.01) but higher triglycerides (p = 0.003), triglyceride/HDL ratio [p = 0.001), remnant-cholesterol (p = 0.003), and C-reactive protein (p = 0.021) than participants in the lowest quintile (Table 1 and supplementary Table 2). Adjustment for saturated intake did not change the results for total cholesterol and LDL cholesterol (data not shown).

As shown in Table 2, no association was observed between total, animal and vegetable protein intake and all-cause, cancer or cardiovascular mortality when energy from protein was substituted with carbohydrates. These results were similar in both sexes with a p for interaction of 0.577 for animal protein intake and of 0.612 for vegetable protein. Table 3 shows HRs for all-cause and cause-specific mortality associated with iso-proteic replacement of 3% of energy from animal protein and animal protein sources with plant protein. A reduction in all-cause of mortality risk was observed when substituting processed meat protein for vegetable protein in the crude (HR 0.77, 95% CI 0.61–0.97) but not in the fully adjusted model (HR 0.89, 95% CI 0.71–1.12). Similarly, a risk reduction for overall and CVD mortality was observed in men in the crude model when substituting processed meat protein for vegetable protein (HR 0.65, 95% CI 0.45–0.93; HR 0.42, 95% CI 0.19–0.91, respectively), but not in the fully adjusted model. A significant reduction in CVD mortality was observed in men only, even in the fully adjusted model, when fish protein was substituted for vegetable protein (HR 0.43, 95% CI 0.20–0.91). In women, CVD mortality was significantly reduced by substituting egg protein for vegetable protein (HR 0.25, 95% CI 0.06–0.99 in the fully adjusted model).

When we analyzed associations between animal and vegetable protein and mortality risk by subgroups of BMI category (< 25 and > 25 kg/m²), waist-to-hip ratio, or age, no significant interaction was found (data not shown). However, when we categorized participants into healthy- and unhealthy-lifestyle groups according to a priori–defined criteria (Table 4), we found a protective association with CVD mortality of substituting vegetable protein for carbohydrate that was restricted only to participants with at least 1 unhealthy lifestyle risk factor and poor adherence to the Mediterranean diet (HR 0.45, 95% CI 0.22–0.94). Similarly, in the unhealthy-lifestyle group but not in the healthy lifestyle group, the multivariable HR per 3% replacement of animal with vegetable protein was 0.47 (95% CI, 0.24–0.92).

4Discussion

In this large cohort study of 45,009 Italian men and women who were followed for a mean of 15.2 years, we observed no clear significant association between total or animal protein intake and total or cause-specific mortality. However, a higher plant protein consumption was associated with a ∼50% lower risk of cardiovascular disease mortality but only in participants with at least 1 unhealthy lifestyle risk factor and poor adherence to the Mediterranean diet. High vegetable protein intake was associated with significantly lower plasma cholesterol and insulin levels, while high animal protein was associated with elevated plasma cholesterol and fasting glucose, despite consuming significantly less calories and having similar BMIs, suggesting that protein source plays crucial roles in modulating cardiometabolic health independently of body weight.

Adherence to high protein diets, especially animal protein rich in branched-chain amino acids and sulfur amino acids, has become very popular because of its hypothetical role in preserving muscle mass, and in promoting fat loss and associated improvements in cardiometabolic health [19–22]. However, conclusive evidence that consuming high protein diets has meaningful effects on body composition or metabolic health is controversial, and accumulating evidence suggest it might have adverse effects on insulin sensitivity, and it may increase the risk of developing obesity, type 2 diabetes, and therefore cardiovascular disease [23].

Branched-chain amino acids are known activators of the mammalian target of rapamycin complex 1 [mTORC1) pro-aging signaling pathway that is deeply implicated in the regulation of lipid/glucose metabolism, protein turnover, autophagy, and nucleotide and ribosomal biogenesis [24, 25]. On the other end, non-essential amino acids, which predominate in plant proteins, stimulate glucagon secretion [26]. Glucagon can promote a cAMP-dependent down-regulation of lipogenic enzymes and cholesterol synthesis, while up-regulating hepatic LDL receptors and IGFBP-1 production in the liver [27]. A lower sulfur amino acid intake, especially of methionine, increases longevity in rodents, and has been associated with reduced risk for cardiometabolic diseases in a recent analysis from the NHANES III Study [28]. As methionine is specified by the AUG start codon, it is essential for translation initiation of most proteins. Methionine restriction thus has a dramatic effect in downregulating protein synthesis and activating FGF21, a key regulator of insulin sensitivity and energy expenditure [25, 29].

Our current findings together with the results of other epidemiological studies suggest that replacement of foods high in animal protein with plant protein sources, but not total protein intake per se, is associated with lower cardiovascular mortality [5, 7, 30–32]. However, unlike in other North American studies of people with substantially higher intake of animal food, we did not find an association between animal protein and cardiovascular mortality. Moreover, our data confirm that even in this cohort of Italian consuming less animal protein, the protective effect of vegetable protein is confined only to participants with unhealthy diets, and one or more negative lifestyle risk factors.

There are multiple explanations for these findings. Indeed, residual confounding from metabolic factors may contribute to the observed protein-mortality associations even after adjustment for age, sex, BMI, energy intake, smoking, physical activity, fiber, alcohol, and total fat intake. At least eighty per cent of cardiovascular disease risk and mortality can be explained by the additive effects of classical cardiovascular risk factors [33–35]. Therefore, when evaluating the relation between protein intake and mortality, it might be important to consider their interactive and potentially counteractive actions. Interestingly, in our study both higher animal and plant protein intakes were associated with similar increase in BMI. However, only animal protein consumption was strongly associated with increased fasting glucose and LDLc, whereas high vegetable protein was associated with significantly lower LDLc and fasting insulin, and a trend for lower fasting glucose as well. This is in agreement with recent data from a randomized clinical trial demonstrating that increasing protein intake from 0.8 to 1.2 g/kg per day by adding whey protein rich in branched-chain amino acids completely prevented the beneficial effect of visceral and liver fat loss induced by calorie restriction on insulin-mediated glucose disposal [36].

Our findings suggest that other components in vegetable protein-rich foods [e.g., fiber, starch), in addition to protein per se, may have a critical cardiometabolic modulatory effects. The elevated fiber content of foods rich in plant protein might explain their LDLc lowering effect because this association persisted even after adjustment for saturated fatty acid content [37, 38]. In contrast, the high content of starch from white bread, pizza, and pasta might explain the detrimental effects of vegetable protein-rich foods on circulating remnant-cholesterol, cholesterol:HDL ratio, triglycerides and C-reactive protein levels, which may mitigate the protective effects of lower LDLc and higher insulin sensitivity [39–41]. It is possible that other phytochemicals that predominate in vegetable protein-rich foods, such as sterols, phenolics, flavonoids, saponins, sulfides and organosulfur compounds, could also affect lipid and cholesterol metabolism through multiple mechanisms, including inhibition of cholesterol biosynthesis and LDL oxidation, and reduction of cholesterol absorption by disrupting the solubility of micelles [42].

This study has strengths and weaknesses. Our analysis was performed on a large sample size of Italian men and women who were followed for 15.2 years and consumed substantially less animal protein than other cohorts of North America. This is important because our findings can be more easily translated to populations living in developing countries consuming carbohydrate-rich diets low in animal food products. The availability of a wide range of lifestyle factors allowed for rigorous confounding adjustment and subgroup analysis, whereas the cardiometabolic biomarker phenotyping provided a unique mechanistic insight into these associations. Calculation of protein intake according to food sources, and the substitution effect for protein of various origins is another strength of our study. One limitation is that dietary exposure was assessed only once at baseline. Some participants may have changed their diet during the follow-up, giving rise to potential misclassification of exposure, which could have weakened diet-disease associations. Moreover, although we have adjusted our results for several dietary and lifestyle factors, residual confounding remains a possibility because of unmeasured or unknown factors.

5Conclusions

In conclusion, we found that substitution of plant protein for animal protein is associated with lower cardiovascular mortality, which is consistent with its beneficial effects on plasma cholesterol levels and glucose metabolism, especially in people with unhealthy lifestyle factors. These results have important public health implications especially when the cardiometabolic detrimental effects of consuming starch-rich plant foods can be mitigated by regular exercise training and higher intake of plant foods rich in protein and low in refined carbohydrates such as legumes.