Introduction: As the growing number of clinical trials reported each year makes it increasingly difficult to keep up to date, the importance of quantitative overviews correspondingly increases. The quality of a trial overview depends on the quality of the trial summaries on which it is based. Although experienced readers do not always extract the same data from a given report we think that, with guidance, even inexperienced readers should be able to extract data reliably and reproducibly. Since summarizing trials is difficult, we decided to examine this process. The numerous possible trial designs, endpoints and treatments, made a paper-based questionnaire for data extraction impractical. An adaptable computer-based questionnaire has, therefore, been designed.

Discussion: The first step in the development of this Trial Summary Report Software (TSRS) was the elaboration of an abstract conceptual model representing the information characterizing a clinical trial, and the expression of this model as a relational database architecture, using the MERISE method. After definition, the conceptual model was converted into a representation of the physical structure of a relational database. During early development stages this database will be implanted on a local server, but will later migrate to a distributed architecture. The questionnaire was developed using a 4th generation database application development language and runs under MS Windows. The thesaurus incorporated is based on the Unified Medical Language System (UMLS). The next step will involve cyclic evaluation using a restricted group of reviewers. The TSRS's comprehensive structure, conceptual vocabulary and on-line help incorporate the essential elements of trial methodology. Thus inexperienced users, both students and practitioners, can be taught to critically assess clinical trial reports while extracting data. The integrated search and aggregation functions enables us to select related trials and extract the data necessary for meta-analysis.