Gut microbiome signatures of vegan, vegetarian and omnivore diets and associated health outcomes across 21,561 individuals
By the diet they said they follow: 19,817 omnivores, 1,088 vegetarians and 656 vegans.
No summary yet. A plain-language summary, written by probiome.ai from the study's facts and checked before it's published, will appear here.
How well diet shows in the gut
- Diet told apart from gut microbes (mean AUC across five cohorts)
- 0.85
- Vegans told apart from omnivores (mean cross-cohort AUC)
- 0.90
- Vegetarians told apart from vegans
- 0.84
- Vegetarians told apart from omnivores
- 0.82
AUC: how well a model told the diets apart from gut microbes alone, from 0.5 (no better than chance) to 1 (every time).
Higher with each diet
Omnivores vs vegetarians 20,905 people
Vegetarians vs vegans 1,744 people
Only unnamed lab strains: see the lists.
“Higher” is the study’s measure: on average, a bigger share of gut bacteria across everyone in the group, carriers or not. It isn’t how many people carry it. An association the authors report, measured once per person. It doesn’t mean a diet changes this microbe, or that the microbe affects health. The authors also linked many of these microbes to cardiometabolic markers using a separate ranking; those per-microbe health links aren't in this study's tables, so we don't show them.
5 cohorts, 3 countries
| Cohort | Country | People |
|---|---|---|
| ZOE PREDICT 1 | UK | 1,062 |
| ZOE PREDICT 3 UK22A | UK | 12,353 |
| ZOE PREDICT 3 US22A | US | 7,931 |
| Tarallo et al. 2022 | Italy | 118 |
| De Filippis et al. 2019 | Italy | 97 |
Why it's a landmark study
21,561 people in five cohorts from three countries, all shotgun-sequenced; per-microbe results for every diet pair in open supplementary tables (CC BY 4.0); relative citation ratio 29.3 (NIH iCite, provisional, 105 citations, checked 2026-10-04).
We load the per-microbe results from 41564_2024_1870_MOESM4_ESM.xlsx (Supplementary Tables 1–23) (Supplementary Table 10, Supplementary Table 11, Supplementary Table 12) exactly as reported: standardized mean difference (adjusted Cohen's d) of arcsine-square-root relative abundance, adjusted for sex, age and BMI, pooled with a random-effects meta-analysis. A result is clear when its q-value is under 0.1, the authors' threshold. We show it when it's clear and most of its cohorts agree, at least 2. A difference under the usual size for a small one (under 0.2 for a standardized difference, under 0.1 for a correlation) is folded under its pair, labeled "a very small difference", and isn't counted in the lists. It's kept apart from our own GMrepo comparisons and never mixed into them.
1,294 of the 1,557 microbes with a species name in its tables are microbes we list.
| Pair | Microbes in the table | Clear in the table | The paper's text says |
|---|---|---|---|
| Omnivores vs vegetarians | 3,480 | 493 · 164 | 488 · 112 |
| Omnivores vs vegans | 3,469 | 641 · 137 | 626 · 98 |
| Vegetarians vs vegans | 2,673 | 30 · 13 | 30 · 11 |
Clear counts are omnivores · vegetarians and so on, by the table's own q-values. We count the table's own q-values; the paper's text gives slightly different counts.
Fackelmann G et al., Nature Microbiology 2025, CC BY 4.0 · doi:10.1038/s41564-024-01870-z · Data sources