Constructing a Finer-Grained Representation of Clinical Trial Results from ClinicalTrials.gov

Abstract

AbstractRandomized controlled trials are essential for evaluating clinical interventions. ClinicalTrials.gov serves as a primary repository for such data, yet extracting and synthesizing information from it remains challenging. This study introduces a novel methodology for constructing a detailed arm-centered representation of clinical trial results, moving beyond the traditional PICO (Patient, Intervention, Comparison, Outcome) framework. The representation attentively uncovers both efficacy outcomes and adverse drug events in safety outcomes, promoting a dual-faceted understanding of intervention effects. Through a structured acquisition, extraction, and initialization process, we present a knowledge graph incorporating arm-level efficacy with safety results, categorizing outcomes into three distinct groups: biomarkers, patient-reported outcomes, and clinical endpoints. The goal is to bridge the gap between the generally described searchable design information and the specifically detailed reported results. This approach aims to offer a structured dataset towards better utilization and interpretation of ClinicalTrials.gov data, facilitating a more feasible and complete evidence synthesis practice to include both positive and negative results hidden in clinical trials registries.