Summary
An analysis of 6,803 participants in Japan found two contrasting gut-microbiome patterns statistically mediating links between lifestyle factors, general health and 15 chronic diseases. The patterns were associated with microbial metabolism, faecal short-chain fatty acids and inflammatory markers, but the study design cannot establish that the microbiome causes the health outcomes.
A study of 6,803 participants in Japan identified two contrasting gut-microbiome patterns statistically linking lifestyle factors with general health and 15 chronic diseases. The analysis, published in Nature Medicine, associated plant-based foods, fermented foods and physical activity with a health-oriented microbial pattern, while refined-carbohydrate intake and prolonged sitting were associated with a disease-oriented pattern. These are mediation-analysis findings from observational data, not evidence that changing a particular microbiome feature will prevent disease.
Researchers used faecal shotgun metagenomic data to characterize microbial species and microbial genes, then applied statistical mediation analysis to examine whether those features lay along associations between 28 lifestyle factors and health outcomes. Diagnoses were physician-confirmed and checked against clinical records. The models adjusted for age, sex, body-mass index and medication use. The study also assessed faecal short-chain fatty acids and blood inflammatory markers in subsets of more than 1,700 participants.
Two contrasting microbial patterns
The health-oriented group of microbial features was associated with higher faecal short-chain fatty acids—particularly butyrate—and lower levels of several inflammatory cytokines. The disease-oriented group showed the opposite pattern. Short-chain fatty acids are compounds produced when gut microbes ferment some dietary components; butyrate is one such compound. The researchers found statistical links between plant foods, including vegetables, fruit and mushrooms, and microbial functions involved in breaking down plant cell-wall carbohydrates and pathways associated with butyrate formation.
Refined-carbohydrate intake, including foods such as rice, bread and soft drinks, was associated with microbial genes involved in starch breakdown, sugar uptake and lactate production. The disease-oriented group also contained more oral-derived bacteria and gut pathobionts—microbes that can be associated with disease in certain circumstances. These results describe patterns across the cohort; they do not mean that a named food or bacterial species has the same effect in every person.
The findings for other lifestyle factors were less one-directional. Alcohol intake was associated with both health- and disease-oriented microbial features, including pathways related to the use of host-derived glycans. Dairy intake also showed mixed patterns: some microbial features statistically mediated associations with general health, while others were linked to disease outcomes. The study found that microorganisms used as starter cultures in fermented foods could be detected in participants’ gut microbiomes, but they were not statistically significant mediators of the observed fermented-food–health associations.
A framework tested across cohorts
The researchers developed MADE scores—microbiome-based assessments of disease and health—from species-level profiles. In the Japanese cohort, these scores distinguished participants with general health from those without, and separated disease groups from comparison groups across 15 conditions. They then assessed the scores in 13 independent, non-Japanese cohorts comprising 3,095 participants. Discriminatory patterns were consistent for most chronic conditions across 11 of the 13 cohorts. Combining the microbiome scores with conventional host information improved discrimination for nine of 16 outcomes in the study, including general health, inflammatory bowel disease, type 2 diabetes and depression.
A longitudinal analysis of 402 participants, each sampled at two time points separated by a median of two years, found that key lifestyle-associated microbial patterns were also observed over time. This supports the consistency of the reported associations, while the main mediation analysis remains cross-sectional. In such analyses, statistical mediation describes a modelled pathway; it cannot by itself establish that a lifestyle factor changes the microbiome and thereby causes a health outcome.
The work offers a way to compare lifestyle, microbial functions and multiple diseases within one analytical framework. Establishing whether interventions that alter these patterns can reduce disease risk will require experimental and clinical studies. The authors also note that the discovery cohort was Japanese and that the clinical usefulness and cost-effectiveness of microbiome profiling remain to be established.