Reliable extraction of clinical-trial evidence with provenance, verification, and targeted review
- 1Arizona State University
Abstract
Reliable extraction of trial characteristics and outcomes is essential for systematic reviews, yet incorrect values can alter evidence synthesis and are difficult to detect without clear source evidence. We evaluate EviSearch, a clinical evidence-curation system that links extracted values to source passages, independently verifies cited evidence, and prioritizes higher-risk cells for review. Using ten trial publications, a 133-column schema, reference annotations, and consistent scoring across model families, we assess answer quality, inter-reading agreement, evidence reproduction, targeted review, and evidence modality. In the Qwen evaluation, EviSearch achieved 91.70% accuracy, cited 100.0% of admitted values, and reproduced 98.9% independently. Reviewing 15.6% of cells increased accuracy to 95.22%.
Related
- Macrophages in non-infectious uveitis: cell origin, disease stage, evidence provenance, and therapeutic maturity
- Toward Reliable Patient-Specific Aortic Strain Mapping from 4D CTA: Validation, Spectral Structure, and Clinical Potential
- Long-term salinity is associated with deteriorated soil properties, rhizosphere microbiome restructuring, and elevated wheat crown rot: evidence from a 19-year gradient experiment and a 3-year remediation verification
- LLM-Based Extraction of Clinical Practice Guidelines into Structured Arguments
- The consensus-based CINEX guideline for reporting clinical information extraction studies
BibTeX
@misc{najpande2026reliable,
title = {{Reliable extraction of clinical-trial evidence with provenance, verification, and targeted review}},
author = {G. Najpande and P. Titiya and M. A. Khan and K. Z. R. Khakwani and M. B. Sonbol and I. B. Riaz and V. Gupta},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.01.756091},
url = {https://www.biorxiv.org/content/10.64898/2026.10.01.756091v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Reliable extraction of clinical-trial evidence with provenance, verification, and targeted review
AU - G. Najpande
AU - P. Titiya
AU - M. A. Khan
AU - K. Z. R. Khakwani
AU - M. B. Sonbol
AU - I. B. Riaz
AU - V. Gupta
PY - 2026
DA - 2026/10/01
PB - bioRxiv
DO - 10.64898/2026.10.01.756091
UR - https://www.biorxiv.org/content/10.64898/2026.10.01.756091v1
AB - Reliable extraction of trial characteristics and outcomes is essential for systematic reviews, yet incorrect values can alter evidence synthesis and are difficult to detect without clear source evidence. We evaluate EviSearch, a clinical evidence-curation system that links extracted values to source passages, independently verifies cited evidence, and prioritizes higher-risk cells for review. Using ten trial publications, a 133-column schema, reference annotations, and consistent scoring across model families, we assess answer quality, inter-reading agreement, evidence reproduction, targeted review, and evidence modality. In the Qwen evaluation, EviSearch achieved 91.70% accuracy, cited 100.0% of admitted values, and reproduced 98.9% independently. Reviewing 15.6% of cells increased accuracy to 95.22%.
ER -