gms | German Medical Science

We argue that the basis for a more efficient process for creating and updating (living) systematic reviews is the semantic publication of clinical trial results. We present the C-TrO ontology, which has been developed to support the publication of clinical trials results using the Resource Description Framework (RDF). The ontology has been designed to describe the results of a clinical trial with respect to primary and secondary endpoints and has the pooling of evidence across studies as its main use case.

We further present an implemented tool that supports the automatic aggregation and synthesis of evidence as required by (living) systematic reviews, given that clinical trial results are described in RDF following the C-TrO ontology [1]. The proposed methodology supports the definition of inclusion and exclusion criteria and filtering the evidence according to quality criteria (e.g., considering only randomized trials), and it thus makes searching within the evidence, extracting results, and aggregating evidence more efficient than with current methodologies. It further supports the automatic update of a systematic review as, given the availability of a new clinical trial available in RDF, the systematic review is automatically updated. The methodology has been successfully used to reproduce the main results of different systematic reviews in the area of Glaucoma and Diabetes Type 2. We will present a live demo of the tool and methodology during the conference.

During the talk, we also discuss how the large-scale semantic publication of clinical trial results can become a reality and which stakeholders need to be involved in this endeavor. Most notably, we think that publishers should introduce policies to require authors of published studies to describe their results semantically. This leads to an economic approach in which the results of studies are described once at publication time, allowing reuse of the data by any party, and thus reducing the duplicate effort of extracting the data by each party performing a systematic review. During recent experiments with medical students we showed that the effort involved in modelling the results of a clinical trial via a tool developed by us called the C-TrO editor is modest, ranging between 1 and 2 hours. This shows that the semantic publication of clinical trials is feasible at a larger scale. Via such an approach, we reduce the effort required by every single party in extracting and combining evidence from multiple clinical trials substantially, and we come closer to the vision of (living) systematic reviews that are automatically updated.

Keywords: ontologies, semantic publication, clinical trials, systematic reviews