MASLD-FN: Data, code, and reproducibility materials for a systematic evidence-structure analysis of microbiome-targeted randomized trials
Description
This data and code release accompanies the study “From Class-Level Benefit to Formulation-Level Uncertainty in MASLD: A Systematic Evidence-Structure Analysis of Microbiome-Targeted Randomized Trials.” It provides versioned, machine-readable evidence authorities and reproducibility materials for a systematic evidence-structure analysis of microbiome-targeted randomized trials in adults with metabolic dysfunction-associated steatotic liver disease (MASLD), including studies historically reported under NAFLD/NASH/MAFLD terminology. The final evidence set comprises 136 reports mapped to 78 adult eligible randomized trials. Of these, 65 were ALT candidates and 56 trials contributed 121 selected arms and 65 harmonized ALT mean-difference contrasts. The release includes report-to-trial reconciliation and study-flow authorities; trial- and arm-level ALT and formulation data; ALT harmonization and contrast provenance; formulation/component structural-identifiability outputs; result-specific Cochrane RoB 2 authorities; public-safe title/abstract screening decisions; exact archived database search queries; the frozen six-trial recurrent seven-species-plus-fructooligosaccharide frequentist and Bayesian prior-sensitivity/posterior-predictive analyses; R scripts; reproducibility run outputs; session information; and SHA-256 checksum manifests. The package is designed to reproduce and audit the reported evidence structure rather than redistribute source publications. Primary article PDFs, raw licensed bibliographic-database exports, and verbatim article abstracts are therefore intentionally excluded. The associated Open Science Framework record (DOI: 10.17605/OSF.IO/CT32D) is a retrospective, time-stamped record of protocol evolution, methodological amendments, decision gates, stopping rules, provenance rules, and the frozen analytical framework; it is not a prospective preregistration.
Files
Steps to reproduce
1. Download and extract the repository archive. Verify file integrity using 00_README_AND_MANIFEST/CHECKSUMS_SHA256.txt and the release SHA-256 sidecar. 2. Read README.md and DATA_DICTIONARY_AND_EVIDENCE_UNIT_RULES.md. The analytical structure is 78 eligible trials → 65 ALT candidates → 56 Gate E trials → 26 recurrent-admissible trials → 6 Gate G/H trials; this is post-inclusion analytical resolution, not additional PRISMA exclusion. 3. Use PACKAGE_BUILD_SESSIONINFO.txt to reproduce the R environment and install packages required by individual scripts. Use the deposited machine-readable authorities as inputs. 4. Reproduce Gate H by running 08H_MASLD_Biotics_Bayesian_Prior_Predictive_Gate_v1.0_2026-08-29.R with the frozen Gate G authority. 5. Independently reproduce Gate H using 08H_GATEH_independent_reproducibility_rerun_v1.1_2026-08-29.R. The archived reproducibility run is 20260829T090410. 6. Run 09_FINAL_Evidence_Structure_Reproducibility_Audit_v1.1_2026-08-29.R. The archived final audit is 20260829T092444 and verifies frozen denominators, structural results, Gate G/H outputs, and checksums. 7. Recreate Figures 2–5 using scripts in 02_CODE/03_FIGURE_CODE/. Figure 1 uses the final report-flow and analytical-resolution authorities and the deposited two-panel blueprint. 8. Exact database queries and search documentation are in 04_SEARCH_AND_SCREENING/01_SEARCH_QUERIES_AND_AUDIT/. Embase, Web of Science, and Scopus reruns may require subscriptions and may yield different counts as indexing changes. 9. Public screening files intentionally exclude verbatim abstracts. Primary PDFs and raw licensed database exports are also not redistributed; these omissions do not affect reproduction of the analytical results. 10. Successful reproduction should recover the frozen Gate G result (ALT MD −14.64 U/L; mKH 95% CI −38.05 to 8.78; I² 95.37%; PI −73.39 to 44.11) and Gate H findings (P(μ<0)=93.25%–99.54%, with all future-study 95% posterior predictive intervals crossing zero).