2026-10-12T15:24:57Z https://api.figshare.com/v2/oai

oai:figshare.com:article/33974467 2026-09-23T16:32:54Z category_8 portal_316 item_type_3 month_year_09_2026

Supplementary file 2_Noninvasive microbiota-based diagnostic models for pancreatic ductal adenocarcinoma: a systematic review and diagnostic test accuracy meta-analysis.docx Liang Zhang (41392) Jian Liu (33711) Mengyao Pang (25098778) Chenfei Liang (727824) Da Li (194457) Yali Wang (332169) Microbiology diagnostic test accuracy meta-analysis microbiota pancreatic ductal adenocarcinoma saliva stool systematic review tongue coat Background<p>Due to delays in diagnosis, many patients with pancreatic ductal adenocarcinoma (PDAC) miss the window for curative interventions. The oral and gut microbiomes are attractive as noninvasive triage instruments; however, while the literature provides AUROC values, it has not established clinically useful thresholds.</p>Objective<p>To systematically assess the diagnostic performance and clinical maturity of noninvasive microbiota-derived tools for identifying PDAC.</p>Methods<p>Following PRISMA 2020 and PRISMA-DTA, we searched eight databases up to April 19, 2026, and conducted a targeted supplementary search on September 3, 2026. Risk of bias was assessed using PROBAST and QUADAS-2. The nine-row meta-analysis, which included approximations and correlated estimates, was exploratory.</p>Results<p>The initial search and targeted supplementary search yielded 16 studies covering 21 sample-domain items; 20 model-based records constituted the units of analysis for the risk-of-bias assessment. Fifty-one AUROC estimates formed a separate list of discrimination estimates. 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>https://www.crd.york.ac.uk/PROSPERO/view/CRD420261370926, Identifier, CRD420261370926.</p> 2026-09-23T16:32:54Z Dataset Dataset 10.3389/fmicb.2026.1866794.s002 https://figshare.com/articles/dataset/Supplementary_file_2_Noninvasive_microbiota-based_diagnostic_models_for_pancreatic_ductal_adenocarcinoma_a_systematic_review_and_diagnostic_test_accuracy_meta-analysis_docx/33974467 CC BY 4.0