Dataset for the study: Evidence Triangulator: A Large Language Model Approach to Extracting and Synthesizing Causal Evidence across Study Designs

Health strategies increasingly emphasize both behavioral and biomedical interventions, yet the complex and often contradictory guidance on diet, behavior, and health outcomes complicates evidence-based decision-making. Evidence triangulation across diverse study designs is essential for balancing biases and establishing causality, but scalable, automated methods for achieving this are lacking. In this study, we introduce Evidence Triangulator, a framework leveraging large language models to automate evidence triangulation through ontological and methodological extraction from scientific literature. A two-step extraction approach—focusing on exposure-outcome concepts first, followed by relation extraction—outperforms a one-step method, particularly in identifying the direction of effect (F1=0.86) and statistical significance (F1=0.96). Using salt consumption related health outcome as a case study, we calculate the Convergency of Evidence and Level of Convergency, finding a strong excitatory effect of salt on blood pressure (942 studies), and weak excitatory effect on cardiovascular diseases and mortality (124 studies). This approach complements traditional meta-analyses by integrating evidence across study designs, and enabling rapid, dynamic assessment of scientific controversies.健康战略日益重视行为干预与生物医学干预并重，但针对饮食、行为与健康结局的相关指南往往复杂且存在矛盾，这使得循证决策变得更为复杂。针对不同研究设计开展证据三角验证（evidence triangulation）是平衡偏倚、确立因果关系的必要手段，但目前仍缺乏可规模化的自动化实现方法。本研究提出了证据三角验证器（Evidence Triangulator），这是一种借助大语言模型（Large Language Model，LLM），通过从科学文献中提取本体论与方法论信息来实现证据三角验证自动化的框架。该框架采用两步提取法：先聚焦于暴露-结局概念提取，再开展关系提取，其性能优于单步提取方法，尤其在识别效应方向（F1=0.86）与统计学显著性（F1=0.96）方面表现突出。本研究以食盐摄入相关健康结局为案例，计算了证据收敛性（Convergency of Evidence）与收敛性等级（Level of Convergency），结果发现食盐对血压存在显著兴奋性影响（纳入942项研究），而对心血管疾病与死亡仅存在微弱兴奋性影响（纳入124项研究）。该方法通过整合不同研究设计的证据，弥补了传统元分析的不足，同时能够快速、动态地评估科学争议。