Development and application of a graded evidence framework for the association between gut microbiota-derived metabolites and diabetic retinopathy

Objective The gut–retina axis hypothesis suggests that microbial metabolites contribute to diabetic retinopathy (DR); however, the existing evidence remains heterogeneous, and no systematic grading framework has been established to date. We developed a four-tier (I–IV) evidence-grading framework specifically tailored to the gut–retina axis. Methods The framework was adapted from established methodologies, including the Systematic Review and Integrated Assessment (SYRINA), the Evidence Integration Framework (EIF), and the International Agency for Research on Cancer (IARC) Preamble, with prespecified criteria for assessing metabolite–DR associations. Systematic searches in PubMed, Scopus, Embase, and Web of Science (2001–2026) identified 29 original studies. Two reviewers independently graded the evidence, with substantial inter-rater agreement (Cohen’s κ = 0.89). Results None of the 17 metabolites met Level I criteria. Valine attained Level II-B, supported by Mendelian randomization (MR) evidence for a composite DR endpoint (OR = 1.51, 95% CI: 1.31–1.76; statistically significant after Bonferroni correction (P < 3.78 × 10 -5 ); exact P not reported). Level III (suggestive) comprised three metabolites: SCFA deficiency, TMAO, and LPS; TUDCA was classified as III-conflicting. Level IV (exploratory) included 12 metabolites: ImP, colibactin, virome, fungal microbiota, PCS, H 2 S, UDCA, BCAAs, IPA, PAGln, IS, and MGO. Conclusion MR studies are needed for most substances to establish causal evidence. The evidence map provides a proof-of-concept demonstration (version 1.0). Grades are hypothesis-generating; external validation is required before clinical use. Systematic review registration https://doi.org/10.17605/OSF.IO/D48Z2 , identifier 10.17605/OSF.IO/D48Z2