Health Outcomes Associated with Vegetarian Diets: An Umbrella Review of Systematic Reviews and Meta-Analyses

Abstract

Background: Several meta-analyses evaluated the association between vegetarian diets and health outcomes. To integrate the large amount of the available evidence, we performed an umbrella review of published meta-analyses that investigated the association between vegetarian diets and health outcomes. Methods: We performed an umbrella review of the evidence across metaanalyses of observational and interventional studies. PubMed, Embase, Cochrane Database of Systematic Reviews, and ISI Web of Knowledge. Additional articles were retrieved from primary search references. Metaanalyses of observational or interventional studies that assessed at least one health outcome in association with vegetarian diets. We estimated pooled effect sizes (ESs) using four different random-effect models: DerSimonian and Laird, maximum likelihood, empirical Bayes, and restricted maximum likelihood. We assessed heterogeneity using I² statistics and publication bias using funnel plots, radial plots, normal Q-Q plots, and the Rosenthal's fail-safe N test. Findings: The umbrella review identified 20 meta-analyses of observational and interventional research with 34 health outcomes. The majority of the meta-analyses (80%) were classified as moderate or highquality reviews, based on the AMSTAR2 criteria. By comparison with omnivorous diets, vegetarian diets were associated with a significantly lower concentration of blood total cholesterol (pooled ES = -0·549 mmol/L; 95% CI: -0·773 to -0·325; P < 0·001), LDL-cholesterol (pooled ES = -0·467 mmol/L; 95% CI: -0·600 to -0·335); P < 0·001), and HDLcholesterol (pooled ES = -0·082 mmol/L; 95% CI: -0·095 to -0·069; P < 0·001). In comparison to omnivorous diets, vegetarian diets were associated with a reduced risk of negative health outcomes with a pooled ES of 0·886 (95% CI: 0·848 to 0·926; P < 0·001). In comparison to omnivores, Seventh-day Adventists (SDA) vegetarians had a significantly reduced risk of negative health outcomes with a pooled ES of 0·721 (95% CI: 0·625 to 0·832; P < 0·001). Non-SDA vegetarians had no significant reduction of negative health outcomes when compared to omnivores (pooled ES = 0·973; 95% CI: 0·873 to 1·083; P = 0·51). Vegetarian diets were associated with harmful outcomes on one-carbon metabolism markers (lower concentrations of vitamin B12 and higher concentrations of homocysteine), in comparison to omnivorous diets. Interpretation: Vegetarian diets are associated with beneficial effects on the blood lipid profile and a reduced risk of negative health outcomes, including diabetes, ischemic heart disease, and cancer risk. Among vegetarians, SDA vegetarians could represent a subgroup with a further reduced risk of negative health outcomes. Vegetarian diets have adverse outcomes on one-carbon metabolism. The effect of vegetarian diets among pregnant and lactating women requires specific attention. Welldesigned prospective studies are warranted to evaluate the consequences of the increased prevalence of vitamin B12 deficiency during pregnancy and infancy on later life and of trace element deficits on cancer risks. Funding Statement: French National Institute for Health and Medical Research (INSERM) (UMR_S 1256). Declaration of Interests: Authors declare no conflict of interest. Ethical Approval Statement: This review is registered with PROSPERO, number CRD42018092470.