Red meat consumption in higher healthy eating index diets is associated with brain health critical nutritional adequacy, and fecal microbial diversity

Abstract

We evaluated whether red meat consumption, within diets of both high and low healthy eating index (HEI) quality, was associated with differences in brain health-critical micronutrient adequacy, mental health, diet quality, or fecal microbiota composition. Using data from 3643 adults in the American gut project, participants were stratified into four groups: high-HEI (≥ 80) with red meat (HH-R), high-HEI without red meat (HH-NR), low-HEI (< 80) with red meat (LH-R), and low-HEI without red meat (LH-NR). HH-R had higher protein intake and lower carbohydrate intake than HH-NR, with saturated fat levels within recommended limits. Brain health-critical micronutrient adequacies (selenium, vitamin B12, zinc, calcium, vitamin D3, choline) were significantly higher in HH-R (p < 0.001) than HH-NR. Higher HEI scores, irrespective of red meat consumption, were associated with reduced odds of depression (logOR = −2.22), PTSD (logOR = −3.80), and bipolar disorder (logOR = −5.90). Fecal microbiota diversity and richness were highest in HH-R, with higher Bacteroides caccae (p_adj = 0.003) and Clostridium hathewayi (p_adj < 0.001), while HH-NR showed higher Bifidobacterium adolescentis and Bacteroides eggerthii (p_adj < 0.001). Therefore, these findings suggest that the inclusion of red meat in a high-HEI diet improves brain health-supporting micronutrient adequacy without adverse effects on mental health or microbial diversity.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-18907-w.

Introduction

Diet plays an important yet underrecognized role in mental health research. A large body of evidence shows how dietary patterns influence our psychological well-being through nutrient availability, glycemia, immune activation, inflammation, and gut-brain axis signaling^1,2. Balanced diet with optimal essential micronutrients—such as iron (oxygen transport, myelination)³, zinc (neurotransmitter synthesis)⁴, folate (neurotransmitter synthesis, homocysteine metabolism)⁵, vitamin B12 (neuronal DNA methylation)⁶, and vitamin D (neurotrophic factor regulation, anti-inflammatory signaling)⁷—are consistently linked to reduced risk of depression, anxiety, and cognitive decline in observational studies^8–10. Deficiencies in these micronutrients may disrupt metabolic pathways, neurogenesis, and signaling that regulate mood cognition. Red meat, a culturally significant source of these micronutrients, especially in the U.S., has faced scrutiny due to its associations with some chronic diseases like cardiovascular disease and cancer^11,12. However, epidemiological studies have often failed to distinguish between unprocessed and processed meat or to account for overall dietary quality, despite evidence that diet quality rather than isolated nutrients drive health outcomes^11,12. Healthy eating index (HEI), a validated measure of adherence to U.S. dietary guidelines, provides a framework to resolve this ambiguity^13,14. High-HEI diets, characterized by dietary diversity, nutrient density, and moderation, are associated with a lower risk of mental health disorders compared to low-HEI diets^15,16. However, no studies have examined whether lean red meat within HEI-matched diets enhances micronutrient adequacy while supporting mental health or fecal microbiota—a gap this study addresses.

Furthermore, gut microbiota is shaped by diet and has been involved in linking nutrition and mental health via the production of neuroactive metabolites and modulation of systemic inflammation^17–20. High-HEI diets are associated with greater microbial diversity, without evidence of dysbiosis. and and have also been independently linked to improved psychological outcomes^15,16,21,22. Yet, the role of red meat in this process is poorly understood. While some studies suggest red meat may alter microbial composition, these findings are often confounded by low overall diet quality or processed meat intake^23–25. Diets rich in processed foods and saturated fats—common in low-HEI patterns—are linked to dysbiotic shifts associated with adverse outcomes (e.g., intestinal permeability, neuroinflammation)^26,27; while our study does not directly assess functional metrics, observed compositional changes in microbiota (e.g., taxa abundance) may signal metabolic or inflammatory pathways relevant to mental health^28–31. While most micronutrients are absorbed in the small intestine, other red meat components, such as carnitine, choline, and some bioactive peptides, can reach the colon, where they are metabolized by gut bacteria into metabolites with systemic effects, therefore, red meat’s impact on microbiota likely depends on the overall dietary context^24,32.

The American gut project (AGP), the largest open-source microbiome initiative to date, provides microbiome researchers a resource to address these questions, with dietary, microbial, and health data from over 10,000 participants across diverse demographics³³. By leveraging its validated food frequency questionnaire (FFQ) and 16S rRNA sequencing data, we stratified participants into four groups (details under methods) for this secondary analysis of this cross-sectional study. We hypothesized that when the diet quality is controlled, red meat consumption is not associated with adverse mental health outcome at the population level; and we also hypothesized that the incorporation of red meat within a higher-quality diet would not compromise nutrient intakes and instead will improve some brain health-critical micronutrients adequacy (selenium, vitamin B12, zinc, calcium, vitamin D3, choline); and the fecal microbiota diversity and composition would not be adversely altered in individuals who consumed red meat within a higher quality diet.

Methods

Study dataset

We utilized data from the AGP, a large-scale, cross-sectional study citizen science initiative aimed at understanding the human microbiome and its relationship with diet and lifestyle factors³³. AGP recruited self-selected participants from the U.S. as well as internationally through online platforms and social media. Participants provided self-collected fecal samples and completed detailed metadata questionnaires, FFQ, medical history, mental health diagnoses, and lifestyle factors. The project data is available for use to researchers as a deidentified dataset and is accessible through the open-source microbiome analysis platform QIITA (Study ID 10317; https://qiita.ucsd.edu). Informed consent was obtained from all participants, and all study procedures are described previously in the original research³³. All study protocols were done in accordance to Declaration of Helsinki and were approved by the Institutional Review Board of the University of Colorado (12-0582) and the Human Research Protection Program at the University of California San Diego (141853)³³. This is a secondary analysis of publicly available data; therefore, we did not have any role in the original study protocols.

Dietary information

Dietary intake data were collected via the VioScreen® (VioCare) graphical Food Frequency Questionnaire (FFQ), a validated, web-based dietary assessment tool designed to capture habitual intake over the preceding 90 days³⁴. Participants independently completed the FFQ online, which included portion-size estimation aids (e.g., images, household measures) to improve accuracy. No additional manual review beyond the parent study was conducted. This VioScreen® tool estimates nutrient intake using the Food and Nutrient Database for Dietary Studies (FNDDS) from the University of Minnesota’s Nutrition Coordinating Center to calculate nutrient intake; dietary supplements were not included in the analysis³⁵. Diet quality was assessed using the HEI, a 100-point measure aligned with Dietary Guidelines for Americans. The HEI-2015 and HEI-2020 are identical and consist of 13 components: Adequacy components (e.g. total fruits, whole grains, total protein foods, seafood/plant proteins) scored 0 to 5 or 0 to 10 meaning higher intake of beneficial foods; Moderation components (e.g. refined grains, sodium, added sugars) scored inversely 0 to 10, penalizing excessive intake. The total HEI score, calculated as the sum of these components, ranges from 0 to 100, with higher scores representing superior diet quality, as it means greater adherence to dietary guidelines^33,36. Additionally, participants reported their red meat consumption through as a questionnaire as part of the parent study (https://qiita.ucsd.edu, project: 10317)., which was used to classify them into either red meat consumers or non-consumers.

Data processing and participant stratification

The initial dataset extracted from QIITA (study id 10317) had 43,813 raw sequences. For this present study, we only included 16S rRNA gene sequences. We then filtered the samples with non-unique entries (duplicate samples per participant ID) and participants aged < 18 years, yielding 4,915 unique sequences. We then removed participants lacking dietary information, resulting in 3,690 samples. Sample-level filtering, as discussed previously³³, was rarefied to 1,250 reads per sample. While this approach does not select a threshold across multiple rarefactions, this threshold is consistent with the American Gut Project’s published protocols. Furthermore, taxon-level filtering was performed independently, we followed MicrobiomeAnalyst’s recommended protocols³⁷. Filtering criteria included: 1) A minimum read count threshold of ≥ 4 (averaged across samples), and 2) Prevalence in at least 10% of samples for downstream differential abundance testing to reduce noise from low-abundance rare taxa. After applying these criteria, 3,643 samples remained for final analysis. The remaining samples were then stratified into four groups based on their HEI score and red meat consumption (Fig. 1): 1) High HEI with red meat (HH-R, n = 319); 2) High HEI without red meat (HH-NR, n = 325); 3) Low HEI with red meat (LH-R, n = 2,121); and 4) Low HEI without red meat (LH-NR, n = 878).

Fig. 1.

Participant inclusion flowchart. Data were derived from the American Gut Project, comprising 43,813 initial sequences. Participants were excluded if aged < 18 years, lacked complete dietary/metadata, or had duplicate entries, yielding 4915 unique samples. After removing individuals without dietary records (n = 1225), 3690 participants remained. Microbiome quality control reduced the cohort to 3,643 participants. These were stratified into four groups: high-HEI (≥ 80) with red meat (HH-R), high-HEI without red meat (HH-NR), low-HEI (< 80) with red meat (LH-R), and low-HEI without red meat (LH-NR).

Fecal microbiota processing

Fecal microbiota processing followed protocols established by the American Gut Project³³. Briefly, participants collected fecal samples at home and mailed them to American Gut Project laboratory at the University of California San Diego School of Medicine, where DNA was extracted, and the V4 region of the 16S ribosomal RNA gene was amplified and sequenced as previously described³³. The available raw sequences were downloaded from (PRJEB11419: The American Gut project) in May 2024. The obtained raw sequences were then filtered to select only human gut metagenome single end sequences. When filtering, we kept only one raw sequence per sample, keeping the highest byte size sequences to protect the most information, and the highest byte size correlated with the latest sequence. The raw sequences were processed using QIIME 2³⁸. Single-end FASTQ files were imported into QIIME2 using the Casava 1.8 format, generating a ‘.qza’ artifact for demultiplexed sequences. Quality filtering was conducted using ‘qiime quality-filter q-score’, removing low-quality reads. Denoising and error correction were performed using the Deblur pipeline with a trim length of 125 base pairs as recommended by the parent study³³. A feature table and representative sequences were generated, and sample statistics were done. For taxonomic classification, we used pre-trained Greengenes 13_8 Naïve Bayes classifier trained on the 515F/806R region.

Statistical analyses

All statistical analyses were performed in R (version 4.4.1; R Core Team, 2023) using our previously published protocol^{22,29,39–41} and MicrobiomeAnalyst³⁷. Continuous variables are reported as mean ± standard deviation (SD) or standard error of the mean (SEM). Differences between dietary groups were assessed using t-test for parametric data or Mann–Whitney-U test for non-parametric data and log odds were calculated using logistic regression. Data visualizations were done using R studio (packages: ggplot2, corrplot, ggcorrplot, lattice, reshape2, ggpubr, vegan, and tidyverse). Diversity analyses were performed using the vegan package, with Shannon diversity index used to assess α-diversity and Jaccard index and Bray–Curtis for β-diversity, analyzed via permutational multivariate analysis of variance (PERMANOVA). To address concerns about variability introduced by single rarefying, we performed a sensitivity analysis implementing multiple rarefactions using the vegan::avgdist() function in R. This function calculates beta diversity metrics by averaging distance matrices computed from 100 repeated rarefactions at a minimum sequencing depth of 1,250 reads per sample. The results were consistent with those obtained from the original unrarefied data, supporting the robustness and reproducibility of our beta diversity findings (see Supplementary Figure S1). Differential taxonomic abundance was evaluated using linear discriminant analysis effect size (LEfSe) and EdgeR⁴² LEfSe was used for identifying taxa with significant differential abundance considering both statistical significance and effect size, while EdgeR provided a model-based approach for count data allowing for normalization and dispersion estimation. Both methods were applied with their default settings. Specifically, LEfSe may overestimate effect sizes and lacks multiple testing correction, which can reduce reproducibility. Similarly, EdgeR, while effective in RNA-seq contexts, shows lower agreement with microbiome-specific methods and may not handle compositional data robustly. These limitations may impact the interpretation of differential taxa and should be considered when comparing results across studies. Statistical significance was considered at p ≤ 0.05. P-values from edgeR analyses (n = 973 comparisons) were adjusted for false discovery rate using the Benjamini–Hochberg correction. We considered taxa significant at an FDR of less than 0.05, which corresponded to a raw p-value significance threshold of 0.027.

Result

Individuals consuming high HEI-diet, regardless of red meat intake, maintained healthy BMI

Participants in the HH-R group were older with a mean age of 55.47 ± 13.92 years, marginally higher than those in the HH-NR group, who had a mean age of 53.10 ± 15.14 years (p < 0.05). In the low-HEI groups, LH-R participants averaged 52.88 ± 13.95 years, compared to 50.59 ± 14.07 years in LH-NR (p < 0.05, Fig. 2). Participants in the HH-R group had a higher mean BMI (23.76 ± 3.73 kg/m²) compared to those in the HH-NR group (22.98 ± 4.08 kg/m², p = 0.001). In the low-HEI groups, LH-R group also had a higher mean BMI (25.17 ± 4.91) compared to LH-NR (23.80 ± 5.34, p < 0.001). High-HEI participants collectively had lower BMI than low-HEI groups (p < 0.001), with only the LH-R group exceeding recommended BMI thresholds. Additionally, while female participants outnumbered males in all diet groups, high-HEI groups showed a similar sex distribution regardless of red meat consumption (34.9% male in HH-NR vs 36.7 in HH-R, Fig. 2). Higher imbalances appeared between low groups.

Fig. 2.

Age, BMI, and weight by diet quality and red meat intake. The bar graph illustrates the mean (± SD) values for age, BMI, and weight in the following groups: high-HEI (≥ 80) with red meat (HH-R), high-HEI without red meat (HH-NR), low-HEI (< 80) with red meat (LH-R), and low-HEI without red meat (LH-NR). Statistical significance is based on t-tests, denoted by asterisks: *p < 0.05, **p < 0.01, ***p < 0.001.

High-HEI diets, with or without red meat, preserve healthier macronutrient profiles

Among individuals adhering to high-HEI diets, macronutrient profiles were distinct between red meat consumers and non-consumers (Fig. 3). Energy intake did not differ significantly (p = 0.24) between the HH-R group (1838.69 ± 558.03 kcal) and the HH-NR group (1791.3 ± 630.52 kcal). Whereas energy intake was significantly higher in the LH-R group (1910.72 ± 769.58 kcal) compared to the LH-NR group (1688.45 ± 685.19 kcal, p < 0.001). Total carbohydrate intake was lower among red meat consumers (HH-R: 187.24 ± 62.63 g; LH-R: 162.57 ± 90.43 g) compared to non-consumers (HH-NR: 195.6 ± 74.22 g; LH-NR: 172.86 ± 87.7 g; both, p < 0.05), though all the values were within the acceptable macronutrient distribution ranges for carbohydrates. The protein intake on the other hand was higher (p < 0.001) in both HH-R (77.58 ± 26.84 g) and LH-R (80.74 ± 35.8 g) compared to HH-NR (67.35 ± 26.85 g) and LH-NR (62.99 ± 29.56 g), suggesting the contribution of red meat. Dietary fiber intake across all four groups was below adequate intake level of 25-38 g/day. Both red meat groups had lower fiber intake: HH-R (21.87 ± 8.99 g) and LH-R (17.6 ± 11.17 g) compared to HH-NR (24.78 ± 11.45 g) and LH-NR (19.24 ± 11.89 g); however, HH-R still exceeded both Low-HEI groups. Saturated fatty acid consumption was different between groups: in the high-HEI group, HH-R (19.88 ± 8.32 g) had significantly higher saturated fatty acid intake (p < 0.001) than HH-NR: 16.89 ± 8.84 g, corresponding to 9.73% vs. 8.49% of total calories—both remained below the DGA threshold of 10%. This suggests that red meat inclusion in overall high-quality diets does not inherently exceed SFA recommendations. By contrast, in the low-HEI group, LH-R (28.69 ± 14.76 g; 13.51% of calories) and LH-NR (20.67 ± 12.77 g; 11.02% of calories), both exceeded this 10% recommendation threshold. While LH-NR’s intake represents a modest improvement overall, both low-HEI groups remain at higher SFA consumption than DGA recommendations.

Fig. 3.

A lollipop plot showing differential nutrient intake associated with red meat consumption in High- and Low- HEI diets. Lollipop chart illustrating log₂ fold changes (FC) in macro- and micronutrient intakes between red meat consumers and non-consumers, stratified by diet quality (high-HEI [blue] vs. low-HEI [red]). Fold differences were capped at ± 0.5 for visualization clarity. Lollipops extending to the left denote reduced nutrient intake in consumers, while those to the right indicate increased intake relative to non-consumers. Dot size corresponds to the -log₁₀(p-value), with larger dots representing greater statistical significance (p < 0.05). HEI healthy eating index.

Inclusion of red meat in a diet improves micronutrient adequacy

In both the high- and low-HEI categories, participants consuming red meat consistently demonstrated higher intakes of some brain-health critical micronutrients than their non-red-meat counterparts (Fig. 3). Vitamin B12 levels among red meat consumers (5.04 ± 2.91 mcg in HH-R; 5.55 ± 4.35 mcg in LH-R) exceeded those of non-consumers (3.63 ± 2.96 mcg in HH-NR; 3.45 ± 3.24 mcg in LH-NR). The higher absolute intake also translated to a much greater proportion meeting the Estimated Average Requirement (EAR)—over 93% in both red meat groups (HH-R and LH-R) compared to just 65–71% of non-consumers (Figure S2). Zinc intake followed a similar pattern, with red meat groups (11.82 ± 5.06 mg in HH-R; 11.22 ± 5.36 mg in LH-R) exceeding non-consumers (10.65 ± 5.08 mg in HH-NR; 8.93 ± 4.35 mg in LH-NR). Higher percentage of participants in the red meat groups meeting the EAR (HH-R: 84.9%; LH-R: 72.0%) than in the non-red meat groups (HH-NR: 71.9%; LH-NR: 61.6%). Selenium intake ranged from 114.27 ± 40.28 mcg (HH-R) to 112.03 ± 52.79 mcg (LH-R), compared to 100.89 ± 44.39 mcg (HH-NR) and 90.48 ± 48.33 mcg (LH-NR) While selenium adequacy was high for all participants, red meat groups still showed the highest proportion meeting the EAR, with over 97% achieving the threshold. Calcium levels were below the recommended levels, but the HH-R group (994.74 ± 447.68 mg) came closest overall. Vitamin D intake was significantly higher among red meat consumers, but remained below the recommended level as well as the EAR in all groups. On the other hand, while over 79% of the high-HEI consumers (HH-R) and 64% of the low-HEI consumers (LH-R) meeting the EAR, folate intake was lower in both the high-HEI with red meat group (519.86 ± 244.14 mcg) and the low-HEI with red meat group (457.38 ± 218.98 mcg) compared to the high-HEI without red meat (549.01 ± 270.5 mcg) and low-HEI without red meat (488.75 ± 238.45 mcg) groups. Importantly, despite higher intakes of certain nutrients in the red meat groups, consumption remained safely below the Tolerable Upper Intake Level (UL) for virtually all participants. Less than 3% of any group exceeded the UL for calcium or zinc. Overall, the findings suggest that integrating red meat into a high-HEI diet elevates intake of some micronutrients essential for cognitive and mental health and increases the probability of meeting or exceeding the EAR level.

Higher HEI scores associated with reduced mental and neurodevelopmental disorders, independent of red meat intake

Higher HEI scores showed inverse association with the prevalence of nearly all mental health and neurodevelopmental disorders examined (Table 1). Higher HEI scores were significantly associated with reduced log odds of depression (log odds = −2.22, p < 0.001), bipolar disorder (log odds = −5.903, p < 0.001), PTSD (log odds = −3.80, p < 0.001), and migraines (log odds = -1.47, p < 0.001). Within individuals with high HEI scores, depression prevalence was 9.8% (36/367) in HH-NR vs 10.8% (42/388) in HH-R and the difference was not statistically significant. Similarly, red meat consumption was not associated with statistically significant differences in the prevalence of bipolar disorders, PTSD, non-specified mental illnesses, and migraines. These findings suggest that adherence to a high-quality diet is important for mental health disorders, but the impact may be largely independent of red meat consumption. In the analysis of neurodevelopmental conditions, a higher HEI score was significantly associated with lower odds of attention deficit disorder/attention-deficit/hyperactivity disorder (log odds = −3.4042, p < 0.001 and autism spectrum disorder diagnoses (log odds = −4.7185, p =  < 0.001). Red meat consumption showed a borderline significant increase in attention deficit disorder/attention-deficit/hyperactivity disorder odds (log odds = 0.6648, p = 0.07584), it did not significantly affect autism spectrum disorder risk (p = 0.89). These findings raise the possibility of reverse causality, which means neurodevelopmental conditions may influence dietary patterns, complicating the ability to maintain high quality diets. The cross-sectional nature of the data limits us from making temporal interpretations and observed associations could reflect the adaptations that is secondary to the mentioned conditions.

Table 1.

Frequency of select mental health conditions in study cohort by HEI level and red meat consumption.

Condition	Response	HH-NR	LH-NR	HH-R	LH-R
Migraine	Yes	65 (18.7%)	221 (22.0%)	69 (18.3%)	467 (20.1%)
	No	283 (81.3%)	751 (74%)	308 (81. 7%)	1859 (79.9%)
Depression	Yes	36 (9.8%)	103 (10.3%)	42 (10.8%)	238 (9.7%)
	No	331 (90.2%)	931 (89.7%)	346 (89.2%)	2214 (90.3%)
PTSD	Yes	8 (2.2%)	12 (1.2%)	3 (0.8%)	39 (1.6%)
	No	359 (97.8%)	1022 (98.8%)	385 (99.2%)	2412 (98.4%)
Schizophrenia	Yes	0 (0%)	1 (0.1%)	0 (0%)	6 (0.2%)
	No	367 (100%)	1033 (99.9%)	388 (100%)	2446 (99.8%)
Non specified mental illness	Yes	235 (66.0%)	704 (70.1%)	250 (65.8%)	1612 (69.3%)
	No	121 (34.0%)	300 (29.9%)	130 (34.2%)	714 (30.7%)

Red meat with a high-HEI diet preserves favorable microbial phyla profile

At the phylum level, HH-R (6.3) had the highest phylum diversity among the four groups and LH-NR (5.83) had the lowest. HH-NR, HH-R, and LH-R (6.2) all had a significantly higher mean diversity compared to LH-NR group (5.83, p < 0.01) (Fig. 4a). Similar outcome was also observed when using Shannon diversity measures (Fig. 4a). Among major phyla, Firmicutes showed the highest abundance across all groups, with a mean abundance of 36.59%. Proteobacteria was the second most abundant phylum overall (31.48%), with the highest levels observed in the HH-NR group (36.59%) and the lowest in the LH-R group (27.61%). Bacteroidetes followed as the third most abundant phylum, with relatively stable levels across groups. Regarding differences, Actinobacteria showed significantly lower relative abundance in the HH-R compared with the HH-NR group (p = 0.02); Bacteroidetes was significantly higher in the LH-R relative to the HH-R (p = 0.016); Firmicutes was lower in the LH-NR than the HH-NR (p = 0.038) and higher in LH-R than LH-NR (p = 0.0018) (Fig. 4b). At the OTU level, No difference was observed within the High-HEI groups as HH-R and HH-NR were not different for both alpha diversity and Shannon index measures. However, at the Low-HEI levels, both metrics of alpha diversity were significantly higher in LH-R compared to LH-NR (Fig. 4c).

Fig. 4.

Alpha diversity and relative abundance of microbial communities across dietary groups. (A) Phylum-level alpha diversity: Boxplots of phylum richness (left) and Shannon diversity index (right) across dietary groups. Significant differences were assessed using Kruskal–Wallis tests (B) Alpha diversity at the OTU level, represented by OTU richness (left) and Shannon diversity index (right), highlighting additional diversity patterns among groups. (C) Stacked bar plots of dominant bacterial phyla, showing group-specific variations in the composition. High-HEI (≥ 80) with red meat (HH-R), high-HEI without red meat (HH-NR), low-HEI (< 80) with red meat (LH-R), and low-HEI without red meat (LH-NR); Statistical significance is denoted where applicable, with group-specific comparisons annotated within the plots.

Several key bacterial species were different between the groups

No significant differences in beta-diversity composition were detected via principal coordinate analysis (PCoA) using the Jaccard dissimilarity metric (PERMANOVA, p > 0.05; Fig. 5a,b) and Bray–Curtis distance (Figure S1)When looking at the individual species, we observed several differences in the relative abundance of specific bacterial species Fig. 5c). B. adolescentis showed a significantly lower abundance in HH-R compared to HH-NR (log₂FC = −1.3957, p_adj < 0.001) as well as LH-R compared to LH-NR (log₂FC = −0.5895, p_adj < 0.001). Within the genus Bacteroides, B. caccae was significantly higher in HH-R compared to HH-NR (log₂FC = 0.7457, p_adj = 0.003), with no significant differences observed between low-HEI groups. B. eggerthii showed a consistent and significantly lower abundance across both: HH-R (log₂FC = −1.0163, p_adj < 0.001), LH-R (log₂FC = −0.5337, p_adj < 0.001), and overall (log₂FC = -0.6174, p_adj < 0.001). Similarly, B. fragilis was less abundant in HH-R (log₂FC = −0.8234, p_adj = 0.003) compared to HH-NR but showed no significant changes in low groups. And B. ovatus showed a significantly lower abundance in both HH-R (log₂FC = −0.5306, p_adj = 0.009) compared to HH-NR and LH-R (log₂FC = −0.3134, p_adj = 0.001) compared to LH-NR. Furthermore, within the genus Blautia, both B. obeum and B. producta were significantly less abundant among the red meat consumers across all comparisons. B. obeum was lower in HH-R (log₂FC = −0.6163, p_adj < 0.001) and LH-R (log₂FC = −0.1856, p_adj = 0.017). Similarly, B. producta was lower in HH-R (log₂FC = −0.4692, p_adj = 0.021), LH-R (log₂FC = −0.8112, p_adj < 0.001). Butyricoccus pullicaecorum was another species that showed significantly less abundance across all comparisons: HH-R (log₂FC = −0.6036, p_adj = 0.001) and LH-R (log₂FC = −0.3479, p_adj < 0.001).

Fig. 5.

Microbial community composition and differential abundance across dietary groups. (A) Principal Coordinate Analysis (PCoA) plot at the genus level based on the Jaccard index, illustrating microbial community clustering across dietary groups. (B) PCoA plot at the OTU level using the Jaccard index, showing distinct microbial community structures among groups. (C) Log₂ fold change (log₂FC) plot depicting differential abundance of microbial species, comparing HH-R vs. HH-NR and LH-R vs. LH-NR groups. Positive values indicate higher abundance in the red meat group, while negative values represent higher abundance in the no-red-meat group. Significant differences are annotated within the plot; high-HEI (≥ 80) with red meat (HH-R), high-HEI without red meat (HH-NR), low-HEI (< 80) with red meat (LH-R), and low-HEI without red meat (LH-NR).

Among the species from Clostridium genus, C. clostridioforme also showed a consistent lower abundance in red meat consumers: HH-R (log₂FC = -0.3864, p_adj = 0.018) and LH-R (log₂FC = −0.2290, p_adj = 0.004). Whereas C. colinum and C. ramosum both were higher significantly in HH-R (C. colinum: log₂FC = 0.5799, p_adj = 0.003; C. ramosum: log₂FC = 0.7271, p_adj = 0.001), but was lower significantly in LH-R (C. colinum: log₂FC = −0.8000, p_adj < 0.001; C. ramosum: log₂FC = -0.6078, p_adj < 0.001). C. hathewayi was higher in HH-R (log₂FC = 2.9310, p_adj < 0.001) but did not show difference between low-HEI groups. C. saccharogumia was significantly less abundant in LH-R (log₂FC = −1.1328, p_adj < 0.001) but was not statistically significant between high-HEI groups. Furthermore, C. spiroforme was lower significantly in HH-R (log₂FC = −0.4873, p_adj = 0.004), while C. symbiosum was lower in LH-R (log₂FC = −0.2484, p_adj = 0.005).

Within the genus Coprococcus, C. catus showed a significant lower abundance in HH-R compared to HH-NR (log₂FC = -0.4532, p_adj = 0.001), with no significant difference between low-HEI groups. Conversely, C. eutactus showed a significant lower abundance in LH-R (log₂FC = −0.5739, p_adj < 0.001) but not in HH-R. Among others, Dorea formicigenerans showed a significant lower abundance in HH-R (log₂FC = −0.4146, p_adj = 0.002) and not in low-HEI groups. Eggerthella lenta was consistently less abundant across all red meat consumer groups: HH-R (log₂FC = −0.6128, p_adj = 0.001), LH-R (log₂FC = -0.1946, p_adj = 0.017), and overall (log₂FC = −0.2108, p_adj = 0.003). Collinsella aerofaciens was significantly higher in red meat consumers compared to non-consumers in the overall analysis (log₂FC = 0.1489, p_adj = 0.043). Eubacterium biforme was less abundant in HH-R compared to HH-NR (log₂FC = –0.6855, p_adj = 0.003) but was significantly higher in LH-R compared to LH-NR (log₂FC = 0.2512, p_adj = 0.030). Eubacterium dolichum, was significantly less abundant among LH-R (log₂FC = −0.5451, p_adj < 0.001), while no significant difference was observed in HH-R.

Ruminococcus, Ruminococcus callidus was significantly less abundant in HH-R (log₂FC = −0.5619, p_adj = 0.010), but significantly more abundant in LH-R (log₂FC = 1.3757, p_adj < 0.001) compared to respective non-consumers. R. flavefaciens was significantly less abundant in LH-R (log₂FC =  −0.5331, p_adj < 0.001), but not in HH-R. R. gnavus was significantly less abundant in LH-R (log₂FC = −0.4056, p_adj < 0.001), but no significant difference was seen in HH-R. R. lactaris was more abundant in LH-R (log₂FC = 0.1868, p_adj = 0.046), but no significant changes were observed in HH-R, and R. torques was higher in LH-R (log₂FC = 0.8395, p_adj < 0.001), but not in HH-R.

Several other species showed differential responses to diet quality and red meat consumption, which are shown in (Fig. 5c). A few more important differences include: Roseburia faecis showed a consistent and significantly lower abundance among red meat consumers across all comparisons (p_adj < 0.001 in all comparisons). Gemmiger formicilis was significantly less abundant in HH-R (log₂FC = -0.4911, p_adj = 0.007), but no significant difference was seen in LH-R. Haemophilus parainfluenzae showed a divergent pattern: it was significantly less abundant in HH-R (log₂FC = −1.1420, p_adj < 0.001) compared to HH-NR, but significantly more abundant in LH-R (log₂FC = 0.6370, p_adj < 0.001) compared to LH-NR. Parabacteroides distasonis showed a significantly higher abundance among HH-R (log₂FC = 0.4329, p_adj = 0.018), but no significant difference was seen in LH-R. Prevotella copri was significantly less abundant in LH-R (log₂FC = -0.6686, p_adj < 0.001), but no difference was observed in HH-R. Veillonella dispar, both HH-R (log₂FC = −0.7551, p_adj = 0.007) and LH-R (log₂FC = −0.5314, p_adj < 0.001) had significantly lower abundance. Conversely, Veillonella parvula was significantly more abundant in HH-R (log₂FC = 0.5388, p_adj = 0.029) and LH-R (log₂FC = 1.6065, p_adj < 0.001).

Discussion

Higher-quality dietary patterns favorably influence human health by modulating nutrient intake, metabolic pathways, and gut microbiota composition^{20,24,32,43–45}. HEI serves as a benchmark for assessing diet quality, with higher scores linked to improved health outcomes^46,47. The role of red meat in human health remains contentious, with some studies associating its consumption with adverse metabolic and gut health outcomes, while others underscore its nutritional benefits^11,12,48,49. However, the role of red meat within high-HEI diets remains underexplored, particularly in relation to microbiome diversity and mental health. Our findings suggest that when red meat is consumed within a high-HEI diet, it is not associated with a higher BMI or altered fecal microbes but is linked to improved adequacy of micronutrients associated with mental health. Specifically, we show that: both high-HEI groups showed favorable macronutrient profiles and inclusion of red meat contributed to improved micronutrient adequacy particularly for essential micronutrients such as iron, B vitamins, zinc, and selenium; higher HEI scores were associated with a reduced prevalence of mental and neurodevelopmental disorders, independent of red meat intake and high-HEI diet preserved a favorable fecal microbial phyla profile, irrespective of red meat intake, with distinct differences in key bacterial species between groups.

HEI is designed to promote healthier dietary patterns with appropriate macronutrient ratio within acceptable macronutrient distribution ranges¹⁴. As expected, both high-HEI groups exhibited favorable macronutrient profiles, aligning with previous research^48,50. Energy intake did not differ between HH-R and HH-NR, however, LH-R had a significantly higher energy intake compared to LH-NR. This may reflect that red meat consumption within a low-quality diet could lead to higher energy intake, potentially due to the inclusion of lower-quality cuts (such as fatty or processed red meat) or pairing with energy-dense foods like refined carbohydrates and ultra-processed snacks. Total carbohydrate intake was higher among high-HEI participants, likely reflecting increased inclusion of fruits, vegetables, and whole grains; however, fiber intake remained below adequate intake levels across all groups, including high-HEI participants. This aligns with broader national trends⁵¹, suggesting that even individuals adhering to dietary guidelines may struggle to meet fiber recommendations, indicating a need for further emphasis on fiber-rich food sources in the guidelines. Furthermore, high-HEI groups consumed significantly less total fat than low-HEI groups. Interestingly, saturated fat intake among high-HEI red meat consumers remained below the threshold of 10% of total calories. This implies that when red meat is integrated into a nutrient-dense dietary pattern, it may not inherently lead to excessive saturated fat intake. While we did not look at individual consumption data on red meat types (e.g. lean vs processed), the observed macronutrient balance in HH-R suggests participants may have prioritized leaner cuts, consistent with recent studies showing that lean red meat can align with DGA recommendations^11,12,24,49. The HEI score does not explicitly differentiate between lean and processed red meats but are indirectly penalized through the moderation components (sodium and saturated fats).

The primary observed benefit of red meat consumption in this study was its association with higher micronutrient intake. Across both high- and low-HEI categories, red meat consumers had higher intakes of key micronutrients, including vitamin b12, zinc, calcium, selenium, and choline—nutrients critical in metabolic function, cognitive performance, and mental well-being. Previous studies have associated deficiencies in these micronutrients with increased risks of depression, cognitive decline, and neurodevelopmental disorders^1,2,8–10. Red meat is naturally abundant in these micronutrients, therefore its inclusion within a high-quality dietary pattern may support optimal nutrient intake, particularly for mental and cognitive health⁵². However, our study did not assess clinical deficiency status; thus, the clinical relevance of these intake differences remains unclear. Additionally, in our findings, higher HEI scores were associated with a lower prevalence of mental and neurodevelopmental disorders—including depression, bipolar disorder, PTSD, migraines, attention deficit disorder/attention-deficit/hyperactivity disorder, and autism spectrum disorder—consistent with the hypothesis that overall diet quality plays a role in brain health^2,8,9,47. However, we did not see a relationship (positive or negative) between red meat consumption and these conditions. Red meat consumption was linked to improved micronutrients implicated in mental health; therefore, we speculate that red meat inclusion in a high-quality diet may contribute to better mental health outcomes over time. However, the relationship between diet and mental health is complex, with the observational nature of the study, the potential for reverse causation cannot be ignored, as individuals with these conditions may adopt different dietary patterns due to factors such as appetite changes, dietary restrictions, or socioeconomic influences.

Our microbiome analysis further supports the role of diet quality in shaping microbiota composition. High-HEI diets were associated with greater microbial diversity—a marker of gut health—regardless of red meat consumption. Both the red meat-consuming groups had higher diversity compared to their non-consumer counterparts, suggesting red meat’s potential impact on microbial diversity. Among taxa-specific changes, Bifidobacterium adolescentis levels were lower in HH-R individuals. B. adolescentis, a well-documented probiotic species involved in carbohydrate fermentation and short-chain fatty acid production, has been previously linked to improved sleep, stress regulation, and mental health outcomes^53–55. The reduction in these species in red meat consumers may be attributed to reduced intake of carbohydrates. Conversely, Bacteroides caccae—a taxon reduced in anxiety-related conditions⁵⁶—was significantly higher in HH-R, aligning with the protective role of high-quality diets against mental health risks. While HH-R showed potentially beneficial shifts e.g., higher in B. caccae, reductions in taxa like B. eggerthii (↓ in HH-R), which is reported to increase in PTSD⁵⁷, and Veillonella dispar (↓ in red meat consumers), previously reported to increase in schizophrenia⁵⁸. On the other hand, HH-R also showed a greater abundance in Collinsella aerofaciens linked to bipolar disorder⁵⁹, and Veillonella parvula, associated with schizophrenia⁵⁸. Therefore, gut microbiota alterations are complex and influenced by multiple dietary and environmental factors.

This study benefits from a large, well-characterized cohort with detailed dietary, anthropometric, and microbiota data. While we have highlighted several implications of the data, some limitations must be considered. Participants were self-selected, they joined online and paid fee for microbial analysis themselves. This introduces ingherent bias within the cohort. This group likely had higher disposable income, greater health literacy, and healthier behaviors than the general population as reflected by their anthropometrics as well as higher HEI than average population. These factors may limit the generalizability of our findings and could influence the microbiota composition as well. Additionally, micronutrient and macronutrient data were not formally adjusted for differences in age, sex, or total energy intake, as these variables did not vary substantially between the comparison groups. However, it is important to consider this when interpreting the results, as we report raw means and standard deviations without adjustment. Furthermore, given the exploratory use of p-value thresholds, findings risk overinterpretation; future studies should emphasize effect sizes and pre-registered hypotheses to confirm associations and mechanistic studies are needed to understand causal pathways. Our findings highlight red meat’s role in supporting micronutrients important for brain health, dietary guidance must contextualize these benefits. Processed red meats are independently linked to increased colorectal cancer risk and gut microbial metabolism of red meat-derived compounds like carnitine and choline may elevate trimethylamine N-oxide, a metabolite associated with cardiovascular risk. In conclusion, our results show that high-HEI diets, whether including or excluding red meat, are associated with favorable health outcomes. The inclusion of red meat within high-HEI instead contributed to improved micronutrient adequacy. We emphasize the necessity of longitudinal and mechanistic studies to further clarify the relationships observed.

Author contributions

Conceptualization, manuscript writing, supervision of researchers, project administration, funding acquisition: SD; Methodology, manuscript review and final approval: All co-authors; Formal data analyses: SD, MH, SP.

Data availability

The raw data is available open-access from the parent study. All sequence data and deidentified participant responses can be found in EBI(European Bioinformatics Institute) under project PRJEB11419 and Qiita study ID 10317. Our group performed the secondary analyses of the existing data available open-access. All the downstream data and R scripts are available upon reasonable request from the corresponding author.

Declarations

Competing interests

The authors declare no competing interests.