key: cord-0795018-36tc61ig
authors: Kaiser, Jeenan; van Daalen, Kim R; Thayyil, Arjun; Cocco, Mafalda Tasso de Almeida Ribeiro Reis; Caputo, Daniela; Oliver-Williams, Clare
title: A Systematic Review of the Association Between Vegan Diets and Risk of Cardiovascular Disease
date: 2021-04-08
journal: J Nutr
DOI: 10.1093/jn/nxab037
sha: eb641c39da8a2a5faae65945667678b5e23e7f49
doc_id: 795018
cord_uid: 36tc61ig

BACKGROUND: Plant-based diets are gaining attention globally due to their environmental benefits and perceived health-protective role. A vegan diet may have cardiovascular benefits; however, evidence remains conflicting and insufficiently assessed. OBJECTIVES: We evaluated the utility of the vegan diet in cardiovascular disease (CVD) prevention. METHODS: We conducted a systematic review of studies evaluating the association between vegan diets and cardiovascular outcomes. We searched 5 databases (Ovid MEDLINE, EMBASE, Web of Science, Scopus, and OpenGrey) through 31 October 2020. Four investigators independently screened the full texts for inclusion, assessed quality, and extracted data from published reports. RESULTS: Out of the 5729 identified records, 7 were included, comprising over 73,000 participants, of whom at least 7661 were vegans. Three studies, with at least 73,426 individuals (including at least 7380 vegans), examined risks of primary cardiovascular events (total CVD, coronary heart disease, acute myocardial infarction, total stroke, hemorrhagic stroke, and ischemic stroke) in individuals who followed a vegan diet compared to those who did not. None of the studies reported a significantly increased or decreased risk of any cardiovascular outcome. One study suggested that vegans were at greater risk of ischemic stroke compared to individuals who consumed animal products (HR, 1.54; 95% CI, 0.95–2.48). Yet in another study, vegans showed lower common carotid artery intima-media thickness (0.56 ± 0.1 mm vs. 0.74 ± 0.1 mm in controls; P < 0.001), and in 3 studies of recurrent CVD events, vegans had 0–52% lower rates. Furthermore, endothelial function did not differ between vegans and nonvegans. Using the Grading of Recommendations Assessment, Development and Evaluation approach, evidence was deemed to be of low to very low strength/quality. CONCLUSIONS: Among the Western populations studied, evidence weakly demonstrates associations between vegan diets and risk of CVDs, with the direction of associations varying with the specific CVD outcome tested. However, more high-quality research on this topic is needed. This study was registered at PROSPERO as CRD42019146835.

Plant-based diets have increased in popularity due to concerns for the environment and animal welfare and due to perceived health benefits (1) (2) (3) . The Food and Agriculture Organization (FAO) indicates that adopting sustainable diets is essential to addressing the degradation of environmental resources and climate change (4) . As global livestock is considered to be responsible for 18% of anthropogenic greenhouse gas (GHG) emissions, this includes diets that reduce animal product consumption and increase plant consumption (5) . Such plantbased diets are associated with reduced diet-related GHG emissions, land use associated with food production, and cumulative energy demand indicators, regardless of the level of plant-based food consumption (2, 3, 6) . Furthermore, accumulating evidence on the potential health benefits of plant-C The Author(s) 2021. Published by Oxford University Press on behalf of the American Society for Nutrition. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited Manuscript received September 29, 2020. Initial review completed November 11, 2020. Revision accepted February 1, 2021. First published online April 8, 2021; doi: https://doi.org/10.1093/jn/nxab037. based diets has made them of interest to researchers, healthcare professionals, and public health practitioners, and may strengthen the case for the continued push towards sustainable diets to promote not only environmental sustainability but also population health (1, 7, 8) . Moreover, during the coronavirus disease pandemic, plant-based diets have been proposed as a potential means to prevent and mitigate future transmission of viruses between different species and humans (9) . However, the merits of plant-based diets relative to other diets in managing and preventing cardiovascular disease (CVD) has been the subject of intense debate (10) .

Although there is significant heterogeneity in the types and specific definitions of plant-based diets, veganism, specifically, is defined as the complete exclusion of animal products, including meat, fish, poultry, seafood, dairy, and eggs (7, 11) . The vegan diet has been found to be significantly associated with beneficial changes in cardiometabolic CVD risk factors, such as lower BMIs, serum total cholesterol levels, serum glucose levels, inflammation, and blood pressure, compared to omnivorous diets, which are typically lower in whole grains, fruits, nuts, and vegetables (7, 12, 13) . Such positive cardiovascular health effects may result from lower intakes of dietary cholesterol, saturated fat, and red and processed meat, as well as higher intakes of fiber, plant protein, and phytonutrients (14, 15) . These observations suggest that the vegan diet may have therapeutic potential in preventing or treating CVD.

Contrastingly, levels of nutrients such as EPA, DHA, selenium, zinc, iodine, and vitamin B 12 are significantly lower in vegans compared to nonvegans, which may similarly have adverse cardiovascular health effects (16, 17) . Nevertheless, data on vegan diets are limited. Very few studies have evaluated vegan diets separately from other no-meat diets, such as ovolacto-vegetarianism, making it difficult to explore the specific effects of the vegan diet on cardiovascular health.

To date, no studies have compiled evidence for the role of the vegan diet alone on the risk of primary, intermediate, and recurrent CVD. A previous review conducted in 2015 evaluated only primary CVD, and therefore did not give a complete account of the effect of vegan diets on CVD (7) . To address this knowledge gap, we performed a systematic review of studies assessing the association between vegan diets and risks of primary, intermediate, and recurrent CVD.

We systematically searched 5 databases, including a gray literature database (Ovid MEDLINE, EMBASE, Web of Science, Scopus and OpenGrey), through to 31 October 2020 using a controlled vocabulary. Search terms included those related to diet (e.g., veganism, plant-based), cardiovascular health (e.g., stroke, carotid-intima media thickness), and the relevant population (e.g., humans). We cross-referenced bibliographies of relevant publications identified in our search to capture any additional studies that fit our inclusion criteria (Supplemental Table 1 ).

The exposure of interest was adherence to a vegan diet, defined as no consumption of animal products or byproducts. By this definition, the vegetarian diet did not qualify as a vegan diet. Studies that met the inclusion criteria were written in any language, assessed humans, and examined associations between adherence to vegan diets and risks of CVD outcomes compared to a control group that followed a nonvegan diet. For studies that compared a vegan diet with several predefined diets (e.g., vegetarian diets and omnivorous diets), we considered the least restrictive diet as the comparison group (e.g., omnivorous diets).

Based on our exclusion criteria, nonhuman studies (e.g., animal or in vitro studies); studies that did not compare individuals who followed a vegan diet to individuals who followed a nonvegan diet; studies that did not evaluate a cardiovascular outcome; and studies that did not include any original analyses (e.g., reviews) were excluded. After removing duplicate records, the titles and abstracts of identified studies were screened according to the inclusion and exclusion criteria by 6 researchers (JK, KRvD, AT, MTdARRC, DC, and CO-W) using the software Abstrackr (18) . All full texts for studies that satisfied all selection criteria were retrieved and screened (JK, KRvD, AT, and MTdARRC). Any divergences between authors on study eligibility were discussed and adjudicated by CO-W.

Data from included studies were independently extracted by 4 authors (JK, KRvD, AT, and MTdARRC). Discrepancies between authors were considered by CO-W and discussed among JK, KRvD, and CO-W until consensus was reached. The following information was extracted from each study: study design, study population, participant demographics, baseline characteristics, method of recruitment, details of vegan and control diets (e.g., definition and duration), total number of participants, numbers of vegan and nonvegan participants, reported outcomes, outcome ascertainment, number of events, and association measures with their 95% CIs. Where raw data were provided but no association measure given, the association measure and 95% CI were calculated. An open field to record additional relevant information was also included.

Four authors (JK, KRvD, AT, and MTdARRC) independently assessed the quality of studies using the Cochrane Collaboration Tool for randomized controlled trials (RCTs) (19) and the Newcastle-Ottawa Scale for cohort studies and cross-sectional studies (20) . The final score was converted to Agency for Healthcare Research and Quality (AHRQ) standards. Discrepancies between authors were adjudicated by CO-W.

To assess the risk of bias across individual studies, 3 authors (CO-W, JK, and KRvD) applied the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach (21) . As individual studies were assessed using the Cochrane Collaboration's tool or the Newcastle-Ottawa Scale, the GRADE approach was only applied where 2 or more included studies evaluated the same outcome. Due to heterogeneity in the directions of the associations with different outcomes, we did not deem it appropriate to apply the GRADE approach to all of the studies as a single collective. Evidence from RCTs starts at high quality. In contrast, evidence from observational studies starts at low quality, due to residual confounding, among other issues. We downgraded the evidence by 1 level for each serious study limitation that was identified. Study limitations were risk of bias, indirectness of evidence, serious inconsistency, imprecision of effect estimates, or potential publication bias. Two or more studies were identified for the following outcomes and were therefore included in the GRADE assessment: primary coronary heart disease (CHD) (22, 23, 24) , primary total stroke (22, 23) and recurrent CHD (25, 26) .

To provide a comprehensive assessment of CVD risks, our outcomes of interest included primary, intermediate, and recurrent CVD end points (Supplemental Table 2 ).

A meta-analysis was deemed inappropriate due to high heterogeneity between studies.

The flow diagram in Figure 1 depicts the search results from all databases. Characteristics of the 7 studies included in our systematic review are summarized in the Tables 1 and 2.

Due to unspecified partially overlapping populations between 2 studies (22, 27) it was not possible to estimate the exact number of individuals included in the review, but there were at least 73,852 participants, of whom at least 7661 were vegan (12, (22) (23) (24) (25) (26) 28) . Studies were conducted in the United States, (n = 4) (12, 24, 25, 28) , the United Kingdom (n = 1) (23), and New Zealand (n = 1) (26), and 1 study combined data from the United States, Germany, and the United Kingdom (22) . There were 4 prospective cohort studies (12, (22) (23) (24) , 1 crosssectional study (28) , and 2 RCTs (25, 26) . Cardiovascular events were ascertained through health service records (22, 23) , the US National Death Index (22, 24) , church records (22), data gathered through telephone or from personal contacts (12, 22, 25) , and/or the registrar's office (22) . Dietary assessments were conducted via FFQs (22-24), 24-hour dietary recalls (25) , food diaries (25, 26, 28) , and self-reported elimination of animal products (12) .

Three studies, with at least 73,426 individuals (including at least 7380 vegans) (27) , examined the risk of primary cardiovascular events in individuals who followed a vegan diet compared to those who did not (27) . Risks of the following events were assessed: total CVD (24), CHD (22) (23) (24) , acute myocardial infarction (MI) (23), total stroke (22, 23) , hemorrhagic stroke, and ischemic stroke (23) . None of the studies reported a significantly increased or decreased risk of any cardiovascular outcome for individuals following a vegan diet. Orlich et al. (24) Outcome comparing the adherence to a vegan dietary pattern to a nonvegan dietary pattern. The least restrictive diet was chosen as a comparison group (e.g., omnivorous diet). 3 The 3 groups consisted of a low-calorie low-protein vegan diet group, an endurance runner group and a Western diet group. Key et al. (22) found no evidence for an increased risk of mortality from CHD or cerebrovascular disease [HR, 0.74 (95% CI, 0.46-1.21) and 0.70 (95% CI, 0.25-1.98), respectively] among 32,519 participants (2.3% vegan) who were followed for an average of 11.7 years.

A cross-sectional study (28) of 36 individuals and an RCT (25) of 100 individuals evaluated the association between vegan diets and the intermediate cardiovascular outcomes of carotid artery intima-media thickness (CIMT) and endothelial function, respectively. CIMT was measured using high-resolution realtime B-mode ultrasonography (29) and endothelial function was measured using an EndoPat Device (25) .

Common CIMT, a measure of atherosclerotic vascular disease, was reported to be significantly lower in the low-calorie, low-protein, vegan diet group compared to the Western diet group (0.56 ± 0.1 mm versus 0.74 ± 0.1 mm, respectively; P < 0.001) (28) . In contrast, the EVADE CAD (Effects of a Vegan versus American Heart Association-Recommend Diet in Coronary Artery Disease) trial of participants with a history of angiographically-defined CHD reported no significant change in endothelial function following 8 weeks of adherence to a vegan diet compared to the AHA diet (25) .

The risk associated with vegan diets was assessed by 2 RCTs (25, 26) and 1 prospective cohort study (12) using recurrent cardiovascularoutcomes, including CHD. These studies had a total of 361 participants, including 57 women and 304 men (aged 57-63 years) (12, 25, 26) . No cardiovascular events were observed in the BROAD trial for any participants, regardless of diet (26) . In the EVADE CAD trial, no participants in the vegan group had an MI or cerebrovascular event, underwent repeat coronary revascularization, or died during the 8-week study period (25) . In the AHA diet group of the same trial, 2 participants had a probable ischemic attack. In the cohort study, a lower risk of worsening cardiac symptoms was observed in individuals who adhered to a vegan diet compared to those who did not (unadjusted RR, 0.16; 95% CI, 0.09-0.29; P < 0.001 (12) .

According to the AHRQ standards, 3 studies (12, 25, 26) were of poor quality, 4 studies (22) (23) (24) 28) were of fair quality, and no studies were of good quality (Supplemental Tables  3-5) . Shortfalls in study quality in prospective cohort and cross-sectional studies commonly stemmed from low external validity and poor-quality exposure ascertainment methods (i.e., self-reported diet and lack of repeated measurements, leading to increased susceptibility to measurement errors and misclassification) (12, (22) (23) (24) 28) . In RCTs, study quality was impaired by a high risk of performance bias due to the lack of blinding. Furthermore, only 2 studies performed a power or sample size calculation (25, 26) . The GRADE approach was used to determine the risk of bias across studies where 2 or more studies assessing a particular outcome were identified. Details of the grading based on the current review are indicated in Supplemental Table 6 . Low to very low evidence was found due to observational studies and downgrading of RCT evidence due to the risk of bias.

To our knowledge, our review provides the most comprehensive assessment of the effect of a vegan diet on incident, recurrent, and intermediate CVD, and adds to the growing body of evidence on the health impacts of vegan dietary patterns (1). There was no significant evidence of an association between adherence to a vegan diet and risks of primary CVD or a CHD event. Although there was some evidence that a vegan diet may prevent recurrent events, the studies were of poor quality as assessed by the Cochrane Collaboration Tool. Findings from 2 studies exploring intermediate outcomes were inconsistent. However, this may be due to the short duration of followup (8 weeks) in the EVADE CAD trial, which may have been insufficient to observe a change in endothelial function. The overall evidence for the role of the vegan diet in CVD development or prevention remains weak due to a limited number of high-quality studies.

Another systematic review and meta-analysis indicated that there was low-quality evidence for the association between vegetarian diets and reductions in CHD mortality and incidence, but no evidence for reductions in CVD and stroke mortality (7) . This review found inconsistent associations between a vegan diet and different CVD health outcomes. This includes a decreased risk of recurrent CVD events and weak evidence of an increased risk of ischemic stroke with a vegan diet. This inconsistency may reflect a potential protective effect of plantbased foods against CHD and the lower intakes of certain nutrients in vegans versus nonvegans, which may negatively impact cardiovascular health in other ways.

The protective effect may arise because both vegan and vegetarian diets have been associated with significant reductions in established CVD risk factors (7, (29) (30) (31) (32) (33) , including type 2 diabetes (34), as well as emerging risk factors, such as highsensitivity C-reactive protein and IL-6 (25, 35) . Any favorable effects of the vegan diet on CVD risk factors and cardiovascular functions relative to the omnivore diet are likely to derive from its more optimal macro-and micronutrient contents, including reduced dietary cholesterol and saturated fats (22, 36) and increased plant sterols and fiber (37) , as well as its reduced overall energy intake. The vegan diet is also characterized by reduced carnitine and choline consumption, and subsequently lower levels of trimethylamine-N-oxide (TMAO), which is a proatherogenic metabolite produced by intestinal microbiota (38, 39) . TMAO, associated with CVD morbidity and mortality, is produced in lower amounts in vegans than in omnivorous individuals (39) . However, improvements in CVD risk factors can be made through diets other than veganism, particularly if they emphasize plant food consumption. Importantly, discrepant findings between the 2 studies of intermediate outcomes may have resulted from the EVADE CAD trial using the omnivorous AHA diet as their comparison diet; the AHA diet, which encourages plant foods, is designed to promote cardiovascular health (25) .

Conversely, adverse cardiovascular effects associated with vegan diets, such as the suggested increased risk of ischemic stroke, may be partly explained by lower intakes in certain nutrients in vegans compared to nonvegans. These include lower intakes in essential amino acids (40) , vitamin B 12 (41) , and long chain n-3 PUFAs (25, 28, (42) (43) (44) . Additionally, some evidence suggests a protective role of meat-derived factors against stroke (25, 28, 43, 44) . Previous studies have reported an increased risk of total stroke (particularly hemorrhagic stroke) among individuals with a low intake of animal products, including animal fat and protein (25, 28, 43, 44) , as well as among individuals with low serum cholesterol levels (45, 46) . However, food fortification and supplement use may ameliorate these possible adverse effects in vegans.

Our study has several strengths, including a detailed comprehensive search strategy of all available evidence and the assessment of multiple primary, intermediate, and recurrent cardiovascular end points to explore the overall utility of the vegan diet in preventing and managing CVD.

However, it also has some major limitations. First, data insufficiency prevented us from 1) performing meta-analyses; 2) assessing publication bias; and 3) performing subgroup analyses to assess the effects of relevant characteristics (e.g., duration of adherence) on the relationship between the vegan diet and CVD. Low proportions of participants adhered to a vegan diet (a minimum of 2.3% of study participants were vegans) and few participants were included in studies of recurrent outcomes, which limits the power to detect associations and draw conclusions. Second, the definitions of vegan diets and their qualities varied between studies (e.g., some excluded excess salt or added oils). As consumption of unhealthy plant foods like refined grains and starchy vegetables is associated with a greater CHD risk than consumption of healthy plant foods like fruits and nuts, the comparability of diet quality between vegans is an important consideration (47) . Third, it should be noted that the studies were conducted in high-income populations, and the studies of recurrent events had varied inclusion criteria, particularly with respect to disease characteristics. This may limit the generalizability of our findings, particularly to low-or middle-income countries, where diet composition and quality may differ and individuals may face unique nutritional challenges and requirements for good health (48) . Finally, poor quality diet ascertainment methods using self-report or single dietary measurements may have led to measurement errors, misclassifications, and inaccuracy, and the lack of participant and/or personnel blinding in the included RCTs also increased their susceptibility to performance bias (12, (22) (23) (24) (25) (26) 28) . Low strength/quality of evidence was also found using the GRADE assessment as a result of observational studies and downgrading of RCT evidence due to the risk of bias.

Whilst society is slowly transitioning to sustainable plantbased diets due to their purported health and environmental benefits, it is important to assess the risks of primary, intermediate, and recurrent cardiovascular outcomes in individuals who follow a vegan diet in order to elucidate its implications on cardiovascular health. Greater knowledge of the cardiovascular impacts of vegan diets may advance global efforts to promote sustainable dietary patterns that are beneficial for both the environment and public health. Future studies should elucidate whether vegan diet quality, demographics, or clinical characteristics modify a potential association between the vegan diet and CVD. Ideally, this should include dose-response analyses of foods, characterizing the various diets and analyses across different CVD outcomes.

In conclusion, to our knowledge, this is the most comprehensive systematic evaluation of vegan diets and cardiovascular health. Transitioning to vegan diets provides significant individual health (e.g., diabetes mellitus) (34) and environmental benefits (1) (2) (3) . However, our review highlights the need for more studies on the role of vegan diets in cardiovascular health. Further experimental evidence and research in large, ethno-geographically diverse cohorts is warranted to better understand the clinical relevance and public health implications of the vegan diet.

Health and nutritional aspects of sustainable diet strategies and their association with environmental impacts: a global modelling analysis with country-level detail

Reducing food's environmental impacts through producers and consumers

The relative greenhouse gas impacts of realistic dietary choices

Plates, pyramids and planets: developments in national healthy and sustainable dietary guidelines: a state of play assessment. FAO; 2016, Food and Agriculture Organization of the United Nationsand The Food Climate Research Network

Healthy and sustainable diets for European countries 2

Environmental impacts of plant-based diets: how does organic food consumption contribute to environmental sustainability?

Vegetarian, vegan diets and multiple health outcomes: a systematic review with meta-analysis of observational studies

Position of the Academy of Nutrition and Dietetics: vegetarian diets

Is there a guarantee that the crisis of Covid-19 will not be repeated

Optimal diets for prevention of coronary heart disease

Can lifestyle changes reverse coronary heart disease?

A way to reverse CAD

Nutrient profiles of vegetarian and nonvegetarian dietary patterns

Consumption of a defined, plant-based diet reduces lipoprotein(a), inflammation, and other atherogenic lipoproteins and particles within 4 weeks

Omega-3 polyunsaturated fatty acids and vegetarian diets

Food and nutrient intake and nutritional status of Finnish vegans and non-vegetarians

Preferred Reporting Items for Systematic Reviews and Meta-Analyses: the PRISMA statement

Deploying an interactive machine learning system in an evidence-based practice center: Abstrackr

Cochrane handbook for systematic reviews of interventions

Critical evaluation of the Newcastle-Ottawa scale for the assessment of the quality of nonrandomized studies in meta-analyses

GRADE guidelines: a new series of articles in the Journal of Clinical Epidemiology

Mortality in vegetarians and nonvegetarians: detailed findings from a collaborative analysis of 5 prospective studies

Risks of ischaemic heart disease and stroke in meat eaters, fish eaters, and vegetarians over 18 years of follow-up: results from the prospective EPIC-Oxford study

Vegetarian dietary patterns and mortality in Adventist Health Study 2

Anti-inflammatory effects of a vegan diet versus the American Heart Association-recommended diet in coronary artery disease trial

The BROAD study: a randomised controlled trial using a whole food plant-based diet in the community for obesity, ischaemic heart disease or diabetes

EPIC-Oxford:lifestyle characteristics and nutrient intakes in a cohort of 33 883 meat-eaters and 31 546 non meat-eaters in the UK

Long-term low-calorie low-protein vegan diet and endurance exercise are associated with low cardiometabolic risk

Vegetarian diets and blood pressure ameta-analysis

Association between plant-based diets and plasma lipids: a systematic review and meta-analysis

Effect of vegetarian dietary patterns on cardiometabolic risk factors in diabetes: a systematic review and meta-analysis of randomized controlled trials

Effect of plant protein on blood lipids: a systematic review and meta-analysis of randomized controlled trials

A systematic review and metaanalysis of changes in body weight in clinical trials of vegetarian diets

Association between vegan diet and risk of CVD 1551

Association between plant-based dietary patterns and risk of type 2 diabetes: a systematic review and meta-analysis

Effect of plant-based diets on obesity-related inflammatory profiles: a systematic review and meta-analysis of intervention trials

Vegetarian diets: what do we know of their effects on common chronic diseases?

Cardio-metabolic benefits of plantbased diets

The contributory role of gut microbiota in cardiovascular disease

Intestinal microbiota metabolism of l-carnitine, a nutrient in red meat, promotes atherosclerosis

Plasma concentrations and intakes of amino acids in male meat-eaters, fish-eaters, vegetarians and vegans: a cross-sectional analysis in the EPIC-Oxford cohort

Serum concentrations of vitamin B12 and folate in British male omnivores, vegetarians and vegans: results from a crosssectional analysis of the EPIC-Oxford cohort study

Long-chain n-3 polyunsaturated fatty acids in plasma in British meat-eating, vegetarian, and vegan men

Epidemiologic studies of coronary heart disease and stroke in Japanese men living in Japan, Hawaii and California. Incidence of myocardial infarction and death from coronary heart disease

Animal protein, animal fat, and cholesterol intakes and risk of cerebral infarction mortality in the adult health study

Serum cholesterol and hemorrhagic stroke in the Honolulu Heart Program

Prospective study of fat and protein intake and risk of intraparenchymal hemorrhage in women

Healthful and unhealthful plant-based diets and the risk of coronary heart disease in U.S. adults

Associations of fats and carbohydrate intake with cardiovascular disease and mortality in 18 countries from five continents (PURE): a prospective cohort study

This work was supported by core funding from: the UK Medical Research Council (MR/L003120/1), the British Heart Foundation (RG/13/13/30194; RG/18/13/33946), and the National Institute for Health Research (Cambridge Biomedical Research Centre at the Cambridge University Hospitals NHS Foundation Trust). Author disclosures: CO-W received funding from Homerton College, University of Cambridge. KRvD received funding by the Gates Cambridge Scholarship (OPP1144) and received funding for publication from the Gates Foundation. All other authors report no conflicts of interest. The funders had no role in the planning, analysis, write-up, or decision to submit this article. Supplemental Tables 1-7 are available from the "Supplementary data" link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/jn/. JK and KRvD contributed equally to this work.

We would like to thank Dr Marinka Steur for her suggestions on how to improve our manuscript and Isla Kuhn for her assistance in the development of the search strategy.The authors' responsibilities were as follows -CO-W: designed the research; JK, KRvD: supported the design of research; CO-W, JK, KRvD: assessed the study quality, wrote the manuscript, and had primary responsibility for the final content; and all authors: conducted screening and data extraction and read and approved the final manuscript.