Supplementary file 2_Noninvasive microbiota-based diagnostic models for pancreatic ductal adenocarcinoma: a systematic review and diagnostic test accuracy meta-analysis.docx

Six-row fixed-threshold data from five studies constituted the primary descriptive core. Sensitivity ranged from 0.682 to 0.964, and specificity ranged from 0.667 to 0.963. The exploratory nine-row bivariate model yielded pooled sensitivity of 0.86 (95% CI 0.80–0.90) and specificity of 0.80 (95% CI 0.73–0.86).

These nine rows represented seven studies; Li 2025 and Chen 2025b each provided a correlated row, so the pooled estimates were hypothesis-generating. The certainty of evidence was rated very low using an adapted, GRADE-informed appraisal. All model-based rows (20/20) were at high risk of bias in the PROBAST analysis domain and high overall risk on QUADAS-2.

Although stool-based testing provided the clearest signal of external validation, benign pancreatic mimics remained underrepresented.</p>Conclusion<p>Detecting PDAC via noninvasive microbiota classifiers carries a biologically plausible yet very low-certainty diagnostic signal. Current findings do not support stand-alone clinical deployment and support only exploratory adjunctive triage. Validating prospectively against benign pancreatic mimics is mandatory before routine clinical adoption.</p>Systematic review registration<p>crd.york.ac.uk/PROSPERO/view/CRD420261370926, Identifier, CRD420261370926.</p>