Sulfur-linked diet associated with modestly higher ischemic stroke risk

In a prospective study of 208,415 middle-aged and older adults, people whose diets most closely matched a pattern linked to sulfur-metabolizing gut bacteria had a 13% higher hazard of ischemic stroke than those with the lowest adherence. Over a mean follow-up of 11.45 years, researchers recorded 3,273 ischemic strokes and found that each one-standard-deviation increase in the sulfur microbial diet score was associated with a 4% higher stroke hazard. The findings were modest, the researchers reported, and the estimate became statistically uncertain when the analysis was restricted to participants with at least two dietary recalls, underscoring the need for independent replication.

The population-based cohort enrolled more than 500,000 people aged 40 to 69 years across England, Scotland, and Wales between 2006 and 2010. For the present analysis, investigators included participants with at least one valid 24-hour dietary recall, the information needed to calculate the dietary score, no previous cerebrovascular disease and available follow-up. Individuals were excluded if their records were unreliable, major components of the score were missing or follow-up was incomplete. Stroke outcomes were identified through linked hospital admissions and death registries, while follow-up ended at the first ischemic stroke, death from any cause, loss to follow-up or the close of the study.

At the center of the analysis was a dietary index previously derived and validated against metagenomic data. The sulfur microbial diet, abbreviated SMD, is not a direct measurement of microbes or sulfur metabolites. Instead, it estimates a pattern of foods associated with the functional capacity of sulfur-metabolizing gut bacteria. Its eight components were processed meat, low-calorie beverages and liquor or spirits, which contributed positively to the score, and beer or cider, fruit juice, legumes, other vegetables and sweets or desserts, which contributed negatively. Each intake measure was standardized, weighted with the published coefficients and combined, so a higher score represented greater adherence to the overall pattern rather than the simple consumption of a single food or nutrient.

Dietary information came from repeated Oxford WebQ 24-hour recall questionnaires, and investigators used the mean of each participant’s valid recalls to estimate average food-group intake. The score was divided into four quartiles and also examined as a continuous exposure. Cox proportional hazards models estimated hazard ratios and 95% confidence intervals while progressively accounting for demographic, medical, socioeconomic and lifestyle factors. The most fully adjusted model included age, sex, ethnicity, body mass index, deprivation, education, income, employment, smoking, alcohol use, physical activity, diabetes, coronary heart disease, hypertension and total energy intake. Missing covariates were handled with multiple imputation, and restricted cubic splines tested whether the association curved or changed across the exposure range.

After the full adjustment, participants in the highest quartile of the SMD had a hazard ratio of 1.13 for ischemic stroke compared with those in the lowest quartile, with a 95% confidence interval of 1.02 to 1.25 and a p-value of 0.019. The continuous estimate was similarly modest: a one-standard-deviation increase corresponded to a hazard ratio of 1.04, with a 95% confidence interval of 1.01 to 1.08 and a p-value of 0.018. The trend across quartiles was significant, and spline analysis found no statistical evidence that the relationship departed from an approximately linear pattern. These values describe an association in a cohort, not proof that the dietary pattern caused strokes.

Most subgroup estimates pointed in the same direction, and statistical tests did not indicate significant differences by age, sex, smoking or drinking frequency. The association appeared stronger among women and participants aged 60 or older, but the relevant interaction tests were not significant. Ethnicity was an exception. Among White participants, the highest versus lowest SMD quartile was associated with a hazard ratio of 1.16, whereas among non-White participants the estimate was 0.48. The interaction between ethnicity and dietary score was significant at p = 0.003. Because this was a post-hoc comparison and the non-White group was comparatively small, the researchers treated the apparent heterogeneity as exploratory and said it requires confirmation in cohorts specifically designed to examine ethnic differences.

Sensitivity analyses produced estimates that were generally directionally consistent, including after exclusion of strokes occurring within the first two years, when the fully adjusted hazard ratio was 1.12. The association also remained above 1 after excluding people with baseline diabetes or coronary heart disease, with a hazard ratio of 1.12. However, among participants with at least two valid dietary recalls, the estimate was 1.08 with a 95% confidence interval of 0.93 to 1.24 and a p-value of 0.310. That result did not reverse the finding, but it showed that improved dietary measurement reduced statistical precision, a limitation the authors said could reflect the fact that a single 24-hour recall may not adequately characterize long-term eating habits.

Exploratory food-substitution models did not provide statistically significant support for any specific replacement at the conventional 0.05 threshold. Replacing low-calorie beverages with an equal energy intake of other vegetables produced a hazard ratio of 0.90, with a 95% confidence interval of 0.81 to 1.01 and a p-value of 0.063. Replacing sweets or desserts with other vegetables yielded a hazard ratio of 0.99, while replacing sweets or desserts with low-calorie beverages yielded 1.10 with a 95% confidence interval of 1.00 to 1.21 and a p-value of 0.078. Because these estimates did not reach statistical significance, the researchers did not present them as evidence for preventive dietary prescriptions.

Several biological pathways could connect the pattern with cerebrovascular disease, but none was directly tested in this cohort. Greater intake of processed meat and lower intake of plant foods may encourage an intestinal environment favorable to sulfur-metabolizing microbes. Previous mechanistic work cited by the authors suggests that increased sulfur metabolism, including production of hydrogen sulfide, may affect intestinal barrier function, low-grade inflammation, oxidative stress, glucose and lipid metabolism, and vascular endothelial function. These processes have been linked in prior research to obesity, insulin resistance, dyslipidemia, atherosclerosis and thrombosis. However, the present study measured neither microbiome composition nor circulating sulfur metabolites and did not formally evaluate adiposity, glycemia or inflammation as mediators, leaving the proposed microbial pathway hypothesis-generating rather than demonstrated.

The study’s scale, long follow-up, standardized outcome ascertainment and use of repeated dietary recalls were strengths, but its observational design left room for residual confounding and reverse explanations that adjustment could not eliminate. The participants were predominantly middle-aged and older White adults, which limits generalization to younger people and more diverse populations. The relatively modest estimate, the loss of statistical significance in the better-measured dietary subset, the exploratory subgroup findings and the absence of direct microbial or metabolite measurements all argue against treating the score as a proven risk factor. The investigators concluded that replication in independent cohorts with more repeated dietary assessments is warranted. Future studies would also need shotgun metagenomic sequencing, targeted metabolomics and appropriately designed analyses of mediation to determine whether sulfur-related microbial activity helps explain the observed dietary association with ischemic stroke.

Subject of Research: Prospective analysis of a sulfur-microbe-linked dietary pattern and incident ischemic stroke in 208,415 adults

Article Title: Sulfur microbial diet and ischemic stroke risk

Article References: Wang, K., Guo, Z., Han, L., & Wang, B. (2026). Sulfur microbial diet and ischemic stroke risk. Frontiers in Nutrition, 13, 1885612. https://doi.org/10.3389/fnut.2026.1885612

Image Credits: AI Generated

DOI: 10.3389/fnut.2026.1885612

Keywords: ischemic stroke, gut microbiome, sulfur metabolism, dietary patterns, prospective cohort, cardiovascular risk, diet and stroke, microbial diet score, Sulfur, microbial, diet, ischemic