A 13-part ‘gut microbiota’ diet score tracked with 19% lower bowel-disease risk. In 198,204 UK adults, coffee was the only ingredient that stood alone

Researchers gave 198,204 UK Biobank participants a score built around foods that earlier literature linked with gut microbes. Avocado, broccoli, chickpeas, coffee, cranberries, fermented dairy, fibre, green tea, soybeans and whole grains counted favourably; a high-fat pattern, red or processed meat and refined grains counted unfavourably. The study did not collect stool samples, so the score is a dietary label, not a reading of anyone's microbiome.

During a median 9.46 years of follow-up, 1,170 people received a new inflammatory bowel disease diagnosis. Each 1.68-point increase in the score — one standard deviation in this sample — was associated with an 8.6% lower relative hazard after adjustment. Comparing the high group, seven points or more, with the low group, zero to four, gave a hazard ratio of 0.806: about 19% lower.

The model adjusted for age, sex, ethnicity, body mass index, education, deprivation, employment, smoking, alcohol, physical activity, medicines, diabetes, high blood pressure and supplement use. Results were directionally similar when early diagnoses were removed and when death was treated as a competing event.

But a useful stress test softened the claim. Among the 120,589 people who had at least two food questionnaires, the estimated hazard ratio per standard deviation was 0.939, with a confidence interval from 0.866 to 1.017. That interval includes no association. Nearly four in ten participants had only one 24-hour recall, which may be a poor portrait of their usual diet.

Coffee was the only single score component that remained significant after correction for testing many ingredients, with a hazard ratio of 0.828. That does not make coffee a treatment or show it caused the overall pattern. People with early bowel symptoms may already change what they eat, healthier diets cluster with other advantages, and UK Biobank volunteers are not a random slice of the population. The score is best read as a clue for trials and better repeated dietary studies, not as a 13-item prescription.