Vegan diet in young children remodels metabolism and challenges the statuses of essential nutrients

Abstract

Vegan diets are gaining popularity, also in families with young children. However, the effects of strict plant-based diets on metabolism and micronutrient status of children are unknown. We recruited 40 Finnish children with a median age 3.5 years-vegans, vegetarians, or omnivores from same daycare centers-for a cross-sectional study. They enjoyed nutritionist-planned vegan or omnivore meals in daycare, and the full diets were analyzed with questionnaires and food records. Detailed analysis of serum metabolomics and biomarkers indicated vitamin A insufficiency and border-line sufficient vitamin D in all vegan participants. Their serum total, HDL and LDL cholesterol, essential amino acid, and docosahexaenoic n-3 fatty acid (DHA) levels were markedly low and primary bile acid biosynthesis, and phospholipid balance was distinct from omnivores. Possible combination of low vitamin A and DHA status raise concern for their visual health. Our evidence indicates that (i) vitamin A and D status of vegan children requires special attention; (ii) dietary recommendations for children cannot be extrapolated from adult vegan studies; and (iii) longitudinal studies on infant-onset vegan diets are warranted.

Keywords: development; metabolism; nutrition; vegan; vitamin.

Figure 1. Anthropometric measurements and dietary intake of omnivores, vegetarians, and vegans

1. A, B

   Height (A) and body mass index (B) of the participants by age in the diet groups compared to 2 SD intervals (shaded area) and median in the Finnish reference population.

2. C

   Dietary intake of selected macro‐ and micronutrients of participants.

Data information: The summary bar is presented as mean ± SEM. Differences between diet groups were evaluated with age‐ and sex‐adjusted permutation tests with Benjamini–Hochberg correction for multiple testing. Only significances P < 0.05 are displayed. n = 24 omnivores, 10 vegetarians, six vegans. *P < 0.05, **P < 0.01, ***P < 0.001. Exact P‐values are provided in Table EV1. Source data are available online for this figure.

Figure 2. Biomarkers of micronutrient statuses and cholesterol metabolism in omnivores, vegetarians, and vegans

1. Serum, plasma, or urine levels of vitamin D, vitamin A, folate, zinc, iron, and iodine status biomarkers in diet groups. Solid lines indicate cut‐offs for deficiency, and dashed lines indicate cut‐offs for insufficiency. The RBP limits for vitamin A insufficiency and deficiency are validated for the method at 1.17 and 0.83 µmol/l, respectively. Vitamin A status was calculated based on RBP, transthyretin, and CRP (Talsma et al, 2015).

2. Cholesterol metabolism biomarkers in the diet groups. The subfigure summarizes cholesterol metabolism in humans and highlights cholesterol‐related metabolites measured in this study with black font (Risley, 2002; Nelson & Cox, 2012).

Data information: The summary bar is presented as mean ± SEM. Differences between diet groups were evaluated with age‐ and sex‐adjusted permutation tests with Benjamini–Hochberg correction for multiple testing. Only significances P < 0.05 are displayed. n = 24 omnivores, 10 vegetarians, six vegans except for panel (A) iodine, where n = 13 omnivores, nine vegetarians and six vegans. *P < 0.05, **P < 0.01, ***P < 0.001. Exact P‐values are provided in Table EV6. DMAPP, dimethylallyl pyrophosphate; HDL, high‐density lipoprotein; IPP, isopentenyl pyrophosphate; LDL, low‐density lipoprotein; OMN, omnivore; RBP, retinol‐binding protein; VGN, vegan; VGTR, vegetarian; VLDL, very low‐density lipoprotein. Source data are available online for this figure.

Figure 3. Analysis of untargeted flow injection TOF‐MS metabolomics data and further targeted analysis of serum bile acid concentrations

1. Pathway analysis from 872 detected untargeted metabolites between omnivores and vegans.

2. Targeted analysis of bile acid concentrations in serum of omnivores, vegetarians and vegans.

3. Amino acid levels in untargeted metabolomics. Numeric means and standard deviations in this figure can be found in Appendix Table S4. ^#Valine and betaine are two major components under the same m/z peak, causing uncertainty in the interpretation of valine levels.

Data information: Pathway analysis in subfigure (A) was performed with a gene‐set enrichment analysis (GSEA)–based method. The summary bar of subfigures (B, C) is presented as mean ± SEM. Differences between diet groups in subfigure (B) were evaluated with age‐ and sex‐adjusted permutation tests with Benjamini–Hochberg correction for multiple testing. Only significances P < 0.05 are displayed. n = 24 omnivores, 10 vegetarians, six vegans in panel (B). *P < 0.05. Exact P‐values for panel (A) are provided in Appendix Table S2 and for panel (B) in Table EV6. ALA, alpha‐linolenic acid; BCAA, branched‐chain amino acid; BA, bile acid; LA, linoleic acid; OMN, omnivore; VGTR, vegetarian; VGN, vegan, VLCFA, very long‐chain fatty acid. Source data are available online for this figure.

Figure 4. Fatty acid analysis from untargeted MS metabolomics in omnivores, vegetarians, and vegans

1. Fatty acid analysis from untargeted MS metabolomics including free alpha‐linolenic acid (ALA) and docosahexaenoic acid (DHA).

2. Carnitine‐bound fatty acids commonly found in serum that were detected in mass spectrometry.

3. Lysophosphatidylcholine and lysophosphatidylethanolamine with C16:0 (palmitic acid) and C18:1 (oleic acid).

4. Combined analysis of all found triglycerides corresponding to medium‐chain fatty acid lengths and long or very long‐chain fatty acid lengths.

Data information: Box plot center represents group median, hinge covers 25^th to 75^th percentile and whiskers cover interval from minimum to maximum. Zero‐level represents omnivore mean to which each individual is compared in log₂ scale. Differences between diet groups in subfigures (A–C) were evaluated with Student's t‐test (unequal variance, two‐sided) with Benjamini–Hochberg correction for multiple testing. Only significances P < 0.05 are displayed. Exact P‐values are provided in Appendix Table S8. In subfigures (A–C), n = 24, 10, 6 for omnivores, vegetarians, and vegans, respectively. In subfigure (D), multiple‐related metabolites (three MCFAs and six LCFAs/VLCFAs) are combined to same box plot yielding n = 72, 30, and 18 in MCFA subfigure and n = 144, 60, and 36 in (V)LCFA subfigure for omnivores, vegetarians, and vegans, respectively. *P < 0.05, ***P < 0.001, NA, not applicable. Cxx:y, fatty acid of xx carbon atoms and y double bonds; FC, fold change; LysoPC, lysophosphatidylcholine; LysoPE, lysophosphatidylethanolamine; MCFA, medium‐chain fatty acid; LCFA, long‐chain fatty acid; VLCFA, very long‐chain fatty acid. (1) Student's t‐test is not suitable for testing of multiple dependent metabolites in a single test. Of single metabolites included in subfigure (D), Benjamini–Hochberg adjusted t‐test was significant (α = 0.05) for C50:0‐o, C53:4, and C58:0‐o. Source data are available online for this figure.

Figure EV1. Hierarchical clustering of participants based on untargeted biomarkers in blood

Hierarchical clustering of participants based on the detected 872 metabolites in untargeted metabolomics from serum. Hierarchical clustering was performed using Ward's method and Euclidean distance as the measure of dissimilarity. n = 24 omnivores, 10 vegetarians, six vegans. Source data are available online for this figure.