Animal versus plant-based protein and risk of cardiovascular disease and type 2 diabetes: a systematic review of randomized controlled trials and prospective cohort studies

Objectives To systematically review the evidence on the effect of replacing the intake of animal protein with plant protein on cardiovascular disease (CVD) and type 2 diabetes (T2D) and their intermediate risk factors. Methods We searched MEDLINE, Embase, Cochrane Central Register of Controlled Trials, and Scopus up to 12th May 2022 for randomized controlled trials (RCTs) or prospective cohort studies that investigated replacement of animal protein with plant protein from foods. Outcomes were CVDs, T2D, and in RCTs also the effects on blood lipids, glycemic markers, and blood pressure. Risk of bias was evaluated with the Cochrane’s RoB2, ROBINS-I, and USDA’s RoB-NObS tools. Random-effects meta-analyses assessed the effects of plant vs. animal proteins on blood lipids in RCTs. The evidence was appraised according to the World Cancer Research Fund’s criteria. Results After screening 15,090 titles/abstracts, full text of 124 papers was scrutinized in detail, from which 13 RCTs and seven cohort studies were included. Eight of the RCTs had either some concern or high risk of bias, while the corresponding evaluation of cohort studies resulted in moderate risk of bias for all seven. Meta-analyses of RCTs suggested a protective effect on total cholesterol (mean difference -0.11 mmol/L; 95% CI -0.22, -0.01) and low-density lipoprotein cholesterol (-0.14 mmol/L; 95% CI -0.25, -0.02) by replacing animal protein with plant protein. The substitution of animal protein with plant protein (percentage of energy intake) in cohort studies was associated with lower CVD mortality (n = 4) and lower T2D incidence (n = 2). The evidence was considered limited-suggestive for both outcomes. Conclusion Evidence that the substitution of animal protein with plant protein reduces risk of both CVD mortality and T2D incidence is limited-suggestive. Replacing animal protein with plant protein for aspects of sustainability may also be a public health strategy to lower the risk of CVD mortality and T2D.


2
(page number not for citation purpose) Christel Lamberg-Allardt et al.
T he role of protein intake and its effect on health outcomes has been a long-standing research topic of interest and has been a high priority in nutrition research and disease prevention. In addition, efforts to combat climate change have identified protein intake as an important target, especially reducing protein of animal origin, since the production of animal protein generally is resource-intensive and environmentally impactful compared to plant protein sources (1). Compared to plant protein, animal protein sources are generally associated with larger carbon footprints, more land use, and larger blue water footprints (2).
Cardiovascular disease (CVD) and type 2 diabetes (T2D) are the major causes of morbidity and mortality worldwide and are associated with high societal costs (3). A recent systematic review (SR) and meta-analysis of observational studies indicated that habitual high intake of total and animal protein is associated with an increased risk of T2D (4). In contrast, Mousavi et al. (5) showed no association between dietary protein intake from different sources and risk of CVD in an SR of prospective studies. Likewise, in another recent SR, dietary protein intake from different sources showed no association with risk of coronary heart disease (CHD), but in subgroup analysis, there was a lower risk of CVD mortality with an increasing plant protein intake (6). The latter observation was further supported in an SR by Qi et al. (7) who demonstrated that higher plant protein intake was associated with a reduced risk of all-cause and CVD mortality. Equally, Chen et al. (8) presented evidence from prospective cohort studies that suggested that total protein intake was associated with an increased risk of all-cause mortality, mainly driven by an increased risk of CVD mortality by intake of animal protein. However, this SR showed that plant protein intake was inversely associated with all-cause and CVD mortality. The SR performed for the 2012 Nordic Nutrition Recommendations (NNR) on protein intake and several outcomes, including CVD, body weight, cancer, T2D, fractures, renal outcomes, physical training, muscular strength, and mortality concluded that many of the included studies found beneficial associations with plant protein intake (9).
In revising the NNR for the 2022 edition, the intake of animal protein vs. plant protein in adults was a prioritized subject by the NNR Committee for an SR. Criteria for shortlisting topics were published in 2021 (10). Briefly, it was deemed justified to perform a new SR if there were important new scientific data since NNR 2012 and no recent, relevant, and qualified SR available on the topic (11). A scoping review identified new data since 2011 that may be relevant. The aim of this SR was to examine the evidence for whether replacing animal protein with plant protein reduces the risk of CVD and T2D.

Methods
The methodology for the present SR followed the guidelines developed for the NNR 2022 (12,13) and the Preferred Reporting Items for SRs and Meta-Analyses (14,15). A protocol was pre-registered online on PROSPERO (https://www.crd.york.ac.uk/prospero, CRD42021240630). A focused research question was developed by the NNR 2022 Committee, defining the population/participants, intervention/exposure, control, outcome, timeframe, study design, and setting (PI/ ECOTSS), in an iterative process with the SR authors. The funding source for NNR 2022 was the Nordic Council of Ministers and governmental food and health authorities of Norway, Finland, Sweden, Denmark, and Iceland (10).

Eligibility criteria
The inclusion and exclusion criteria are outlined in the PI/ ECOTSS statement (Table 1). Briefly, prospective cohort studies and non-randomized and randomized controlled trials (RCTs) conducted in healthy adult populations (>18 years) were eligible for inclusion. Studies including subjects with mild hypercholesterolemia (as reported by the authors), who were not treated with cholesterol-lowering medication, were included in the analyses of RCTs. We excluded prospective cohort studies that did not report on substitution of animal protein with plant protein in relation to the outcomes, and those that were from settings otherwise not relevant for the Nordic/Baltic population. In this case, studies that evaluated a parallel comparison between the intake of animal and plant protein were excluded as no substitution was performed in such studies. For RCTs using soy protein as plant protein source, we included only RCTs intervening soy with zero or low isoflavone content and excluded those with moderate or high isoflavone content. For interventions using soy protein with different levels of isoflavones, only the group with 3 (page number not for citation purpose) Animal versus plant-based protein and risk of CVD and T2D the lowest isoflavone content was included to discount effects of isoflavones and focus on those of the protein (16). Outcomes included CVD (mortality and incidence), T2D, and related cardiometabolic risk factors. Selection and data collection process Two investigators (JB and BN) independently reviewed titles, abstracts, and full-text articles for inclusions according to the PI/ECOTSS statement (Table 1), first in a pilot test of 10% of the papers, using the web tool Rayyan (https://rayyan.qcri.org) in blinded mode. Potentially eligible papers were retrieved and read in full text by the same two reviewers. Disagreements about inclusion were resolved by a third reviewer (AÅ).

Information sources and search strategy
Another four authors (JD, EA, AR, and FS), in pairs, independently extracted data from the included studies into pre-specified Excel forms. Disagreements were solved by discussion. Among the variables extracted were study design, information on recruitment, dietary intake, interventions and controls, assessment of outcomes, follow-up, drop-out, confounders, etc.

Study risk of bias assessment
Risk of bias in each included study was assessed by two authors (CLA and BT), working independently. The assessment tools used were Cochrane's Risk of Bias 2.0 (17) and Risk of Bias In Non-randomized Studies of Interventions (18,19) for intervention studies, while 'Risk of Bias for Nutrition Observational Studies' (RoB-NObS) (20) was used for prospective observational studies. The risk of bias in each individual study was classified as 'low', 'some concerns', or 'high'. Risk of bias was visualized by using the web app Risk-of-bias VISualization (robvis) (21).

Synthesis methods
We performed a qualitative synthesis of the included studies by describing the main characteristics. Following the recommendations of the Healthcare Research and Quality (AHRQ), the Cochrane Handbook, and the NNR 2022 Handbook, a meta-analysis was performed if >3 independent RCTs or >5 cohort studies were available (12,(22)(23)(24).
Consequently, quantitative syntheses were performed of RCTs reporting effects on total cholesterol, LDL-cholesterol, HDL-cholesterol, and triglycerides. Measures expressed in mg/dl were converted to mmol/l by dividing mg/dl by 38.67 for cholesterol and 88.57 mg/dl for triglycerides. We used the random-effects meta-analyses with variance (τ 2 ) estimated by the restricted maximum likelihood method. For most parallel-group and crossover trials, we used pooled differences in means and standard deviations (SD) of follow-up values, while if post-intervention outcomes were not reported, we included change from baseline scores. The SDs were imputed from standard errors if not reported. Homogeneity was assessed by the Cochran Q test, and we used the I 2 statistic to quantify variability explained by between-study heterogeneity. I 2 of ≥50% was considered 'substantial', and ≥75% 'considerable'. Potential small study bias was assessed by Egger's test (significance level P > 0.1) and visual inspection of funnel plots. For studies using soy with different amounts of isoflavones, we included only the intervention arm using the lowest isoflavone dose. Differences between plant protein sources were evaluated by subgroup analyses of soy vs. non-soy interventions, with between-group heterogeneity assessed by Cochran's Q. The meta-analyses were performed with Stata/SE version 17.0 (StataCorp LLC, College Station, Texas, USA).

Certainty assessment
We categorized the strength of evidence according to the World Cancer Research Fund's grading: 'Convincing', 'Probable', 'Limited -suggestive', 'Limited -no conclusion', and 'Substantial effects unlikely' (9). The quality (risk of bias), quantity, consistency, and precision in the body of evidence were considered in categorizing the strength of evidence. Figure 1 shows the literature search, screening, and the number of papers/studies excluded (including the reasons) as well as the studies retrieved and included in the SR. The potentially eligible studies excluded after the full-text assessment is listed together with reasons in the online supplement (Supplementary file 2).
There were seven reports (38)(39)(40)(41)(42)(43)(44) from seven cohort studies, including between 2,332 and 416,104 subjects (total, n = 720,663 for CVD mortality; n = 5,873 for CHD incidence; n = 281,341 for T2D incidence) with endpoint data ( Table 3). The cohorts included subjects from USA, Japan, Finland, and the Netherlands. Animal versus plant-based protein and risk of CVD and T2D Christel Lamberg-Allardt et al. Animal versus plant-based protein and risk of CVD and T2D  Animal versus plant-based protein and risk of CVD and T2D

Types of intervention/exposures
Eight RCTs compared the effect of low-isoflavone soy protein supplementation to casein or milk protein supplementation on different outcomes (27,28,30,31,(33)(34)(35)(36) ( Table 2). Three RCTs (25,26,37) compared the effect of lupin protein supplementation to milk protein or casein supplementation, and one (27) compared, in addition, the effect of lupin protein supplementation to milk protein + arginine supplementation on different outcomes. One RCT investigated the effect of barley protein supplementation in comparison to casein supplementation in bread (32), and one compared the effect of cowpea protein supplementation to casein supplementation (29) on different outcomes. The protein supplementation amount ranged between 25 and 30 mg/d for all studied protein sources. The outcomes in all studies were related to lipid metabolism. In some RCTs, the effects on glucose metabolism or blood pressure were studied.
Four reports from five prospective cohorts investigated the association between plant protein as E% substitution of animal protein and risk of CVD mortality (38,39,41,42) (Table 3) and one on CHD incidence (44). Two reports from four prospective cohorts examined the association with plant protein intake as E% substitution of animal protein and the incidence of T2D (40, 43) ( Table 3).

Outcome assessment
The duration of the interventions in the RCTs ranged from 4 weeks to 24 weeks, all reporting on serum/plasma total cholesterol concentrations (total cholesterol), serum/plasma LDL (low-density lipoprotein)-cholesterol concentrations (LDL-cholesterol), serum/plasma HDL (high-density lipoprotein)-cholesterol concentrations (HDL-cholesterol), andserum/plasma triacylglycerol concentrations (triacylglycerol, TG). In addition, three studies (25,26,32) reported on effects on blood pressure, one on fasting serum/plasma insulin concentration (30), and four on blood glucose concentration (29,30,35,37). If blood was drawn at several time points, only the results from the baseline and latest time point were considered. In the cohort studies, the follow-up time between assessment of diet and outcome ranged from 16 to 19.3 years (median or average in those where it was reported).

Risk of bias in included studies
The risk of bias assessment per domain in RCT studies is outlined in Figs. 2 and 3. Five RCTs had overall low concerns for risk of bias (25-27, 30, 31). Four RCTs had overall some concerns, due to the lack of information on the randomization process (28,29,34,37). Four RCTs had overall high concern of bias, mostly due to non-adherence to the study intervention (32,33,35,36). The risk of bias for all prospective cohort studies was moderate overall (Figs. 4 and 5).
Both crossover and parallel studies were pooled in the meta-analyses. The summary effect sizes showed significantly decreased total cholesterol ( Fig. 6; -0.11 mmol/L, 95% CI, -0.22, -0.01, I 2 = 8.3%) and LDL-cholesterol ( Fig. 7; -0.14 mmol/L, 95% CI, -0.25, -0.02, I 2 = 43.8%), with plant protein interventions compared to animal protein, a borderline significantly increased HDLcholesterol ( Fig. 8; 0.04 mmol/L, 95% CI, 0.00, 0.07, I 2 = 0.01%), but unsignificant effects on TG ( Fig. 9; -0.00 mmol/L, 95% CI, -010, 0.09, I 2 = 0.00%). It should be       Based on inspection of funnel plots (not shown), and Egger's test for all meta-analyses including all intervention studies, we did not find evidence of publication bias in the form of small study-effects bias.  Plant protein, blood pressure, blood glucose, and insulin concentration Two studies (25,26) investigated the impact of lupin protein or barley (32) and observed no effect on blood pressure compared to the animal protein (Tables 2 and 4). Three papers studied the effect of plant protein in comparison with animal protein on blood glucose (30,35,37) and one on fasting insulin (30), with no differences between the treatment groups. No meta-analyses were conducted for these outcomes, as the number of studies were insufficient.

Substitution of animal protein with plant protein and CVD
Only one prospective cohort study (44) was retrieved that focused on the incidence of CHD using substitution model design (Tables 3 and 5). Although non-significant, a higher plant protein intake tended to be associated with a lower risk of CHD when consumed at the expense of animal protein. All four prospective studies (38,39,41,42) with an isocaloric substitution of animal protein with plant protein showed lower risk of CVD mortality (Tables 3 and 5). Of these, Song et al. (41) found that substituting animal protein from processed or unprocessed red meat, fish, or dairy with plant protein was associated with lower CVD mortality. Budhathoki et al. (38) found that replacing animal protein from red meat (not from processed meat, chicken, egg, dairy, or fish) with plant protein reduced CVD mortality. Huang et al. (39) found that replacing total animal protein with plant protein was associated with lower mortality from CVD, heart disease, and stroke in both men and women. When separating on sources of animal protein, results remained for red meat, dairy, and egg, but replacing white meat protein with plant protein was only significantly associated with lower stroke mortality in men.

Substitution of animal protein with plant protein and T2D incidence
Only two papers (40,43) fulfilled our inclusion criteria for T2D incidence, and both showed associations with reduced T2D incidence with isocaloric substitution of animal protein with plant protein (Tables 3 and 5).
Virtanen et al. (43) also showed that replacing any animal protein except for protein from eggs with energy from plant protein was associated with a 14-20% decreased risk of T2D, although not all associations reached statistical significance.

Certainty in the evidence
The evidence for a favorable association between plant protein intake in comparison to animal protein and CVD mortality was considered limited-suggestive based on consistent results from cohort studies with moderate risk of bias, supported by evidence of biological plausibility from the RCTs. The corresponding evidence for T2D incidence was considered limited, suggestive, while the few available RCT studies on blood glucose and insulin did not support an effect.

Summary of key findings
This SR summarizes both RCTs and cohort studies for whether substituting plant protein for animal protein is associated with lower risk of CVD and T2D or lower levels of cardiometabolic risk factors. While the cohort studies reported associations with decreased risks of CVD and T2D in substitution models of animal protein with plant protein, the biological plausibility based on the RCTs was supported for CVD alone. Evidence was considered limited-suggestive for reduced CVD mortality and T2D, when replacing animal protein with plant protein.  Animal versus plant-based protein and risk of CVD and T2D

Strengths and limitation of review
A strength of this review is that we followed established processes for undertaking robust SRs. The NNR 2022 Committee established criteria for the prioritization and selection of a SR topic (10). We developed and registered a detailed protocol before undertaking the review, which improved transparency of the review process. We searched four foremost electronic databases, which cover most of the literature in medicine and public health, why it is unlikely that we may have missed any relevant literature. Moreover, the review processes were thoroughly implemented, with independent assessments taken at each stage of the process, including literature screening and data extraction.
One-third of the RCTs was graded as having a high risk of bias, especially due to deviations from the intended intervention, another third was graded having some concerns regarding risk of bias, mainly arising from the randomization process. Additional limitations include the habitual diets in the RCTs, which may have affected the ability to detect effects of the intervention. Moreover, the animal protein in the RCTs was milk protein or casein, which may not be totally representative for animal protein sources. Among the RCTs, eight investigated soy protein (27)(28)(29)(30)(31)(33)(34)(35)(36) and five other plant proteins, including other legumes (25,26,29,37) and grains (32). Although overall results were not different for the different sources of plant protein, it could be worth in future studies to focus on other legumes and grains instead of soy. We did not find RCTs comparing other sources of plant protein intake, than those above mentioned, to animal protein intake in our search.
All included cohort studies were graded as having a moderate risk of bias, which may constitute a limitation of the underlying evidence. We extracted studies that reported on plant protein intake in relation to animal protein intake, but this may, however, not cover all possible sources of plant protein. Most of the studies were prone to limitations inherent in many observational epidemiologic studies -the starting time of the exposure, method of assessment of dietary intake as it was based on self-reported data (which, in addition, is usually done once at baseline), and inadequate adjustment for confounding factors during the long follow-up, thereby given a possibility for residual/unmeasured confounding across the reported estimates in the studies.

Comparison with other SRs
We retrieved three previous SRs and meta-analyses related to the comparison of animal protein intake to plant protein intake or other diets with blood lipids as outcomes in RCTs settings (45)(46)(47). Guasch-Ferré et al. (45) included 36 RCTs, comparing diets with red meat to diets that replaced red meat with a variety of foods. They concluded that substituting red meat with high-quality plant protein sources, but not with fish or low-quality carbohydrates, leads to more favorable changes in blood lipids and lipoproteins. Li et al. (46) included 104 RCTs, also including individuals with, e.g., T2D and renal disease, comparing the effect of plant protein in substitution for animal protein on blood lipids. They concluded that substitution of plant protein for animal protein decreases LDL-cholesterol and non-HDL cholesterol. Zhao et al. (47) focused on effects of plant protein and animal protein on lipid profile as well as body weight and body mass index in patients with confirmed hypercholesterolemia. They concluded that compared with animal protein, the consumption of plant protein could improve lipid profile in patients with hypercholesterolemia. Our results support the results from previous SRs, even though we only included soy intake with low concentrations of isoflavones and subjects with normal serum cholesterol concentrations or mild hypercholesterolemia, which was reflected in the low number of studies included.
We found two previous SRs focused on protein intake, including plant protein intake and risk of CVD mortality (6, 7). Naghshi et al. (6) concluded that higher intake of plant protein was associated with a lower risk of CVD mortality, whereas there was no association of total protein or animal protein with the risk of CVD mortality. Qi et al. reported (7) that higher plant protein intake (but not total protein) was associated with a reduced risk of CVD related-and all-cause mortality. In conclusion, our results seem to be in line with these two SRs.
A previous SR and meta-analysis showed that total protein and animal protein intake was associated with a higher risk of T2D in both males and females, and that plant protein decreased the risk of T2D in females. These associations were also dependent on the food source, as e.g. red meat and processed meat were risk factors of T2D, while soy, dairy, and dairy products were protective against T2D (48). Our results point in the same direction, but we included fewer cohort studies, as the exposure was defined as substitution of animal protein with plant protein.
Altogether we found six recent SRs that could be considered comparable to the current paper (5,6,(45)(46)(47)(48). The inclusion criteria were overall not exactly the same as ours as we did not include interventions with soy containing high or medium levels of isoflavones, in contrast to the previous reviews. In addition, we included only prospective cohorts, which compared substitution of animal protein with plant protein, i.e. substitution analyses. These differences in inclusion criteria lead to a lower number of included studies in comparison to previous SRs.

Interpretation and implications of findings
The intervention studies showed significantly, albeit only small lowering of total cholesterol and LDL-cholesterol along with higher HDL-cholesterol as a result of plant protein intake in comparison with animal protein intake. Soy, which has been studied extensively, may have a favorable effect on blood lipids, since it contains or can be fortified with high amount of isoflavones, which are known to have these effects (16). Although the magnitudes of the differences in cholesterol levels were small, they may be relevant in a life-course population perspective. The results of the cohort studies indicated an association between substitution of animal protein with plant protein on the risk of CVD and T2D. In comparison with most animal protein sources, plant protein sources contain less saturated fat and no cholesterol and more monounsaturated and polyunsaturated fat, fiber, antioxidants, polyphenols, and other bioactive compounds (49). Other mechanisms have also been suggested, i.e., related to amino acid metabolism. Lysine, which is more prevalent in animal proteins, has been shown to increase cholesterol levels in animal models, whereas arginine, which is found more in plant proteins, has been found to have the opposite effect (47).

Conclusion
We found limited-suggestive evidence that substitution of animal protein with plant protein may decrease the risk of CVD mortality and T2D incidence. Protective effects seen in RCTs on established risk factors for CVD supported the evidence from observational studies. Replacement of animal protein with plant protein for sustainability may also be considered as a public health strategy to lower the risk of CVD and T2D.