Introduction: Reviewers of clinical trial reports develop implicit ad hoc summaries of the characteristics of the reported clinical trial. The "Validata" project demonstrated serious inter-reviewer variability between such summaries. We are developing an explicit and generic representation of these ad hoc summaries in the form of a relational database and interface program.

Objective:

This research has four major

Objectives:

The standardisation of trial report interpretation and the limitation of inter-reviewer variability.

The provision of mechanisms for the blinded and reproducible application of meta-analysis protocols using a collection of clinical trial summaries.

The promotion of team work, information reuse and the development of persistent representations of clinical trials.

The development of a contents addressable database of clinical research.

Methods:

Entity relationship modelling has been used to develop the database structure, and Gupta SQLWindows has been used to implement the interface. The system is being developed in an evolutive fashion: various sources were used to define an initial model, which is being developed with reference to problems arising whilst extracting summaries from an evaluation set of trial reports.

Results:

A simplified model has been implemented as a database and data entry program, which are undergoing alpha testing and evolution. This model defines trial interventions, outcomes and results, the baseline population characteristics, trial selection criteria and miscellaneous general trial information. Our priority is to accurately define these core elements of the model, and then to build around them. In the near future, indicators of trial validity and strength of evidence will be incorporated. Three difficulties have been met with: the identification of the essential nature of a clinical trial; the representation of the complex entity, the clinical trial, in a structure as rigid as a relational database; and adequate information coding methods.

Discussion:

Trial summaries might be of interest to the Cochrane Collaboration. Trial summaries provide a model for co-working and reuse. A register of trial summaries could help to eliminate the duplication of effort and enable the recombination of evidence in cross-sectional fields, further reviews or cumulative reviews. A register of trial summaries would provide a contents addressable persistent index to clinical research. The results of this project will need to be rigorously evaluated, to be seen as acceptable to the Collaboration.