Background:

Reviewers should take into account the methodological quality of the studies in any meta-analysis. Several quality assessment tools have been developed for evaluating diagnostic accuracy studies. Different ways exist to incorporate quality in meta-analyses: ignore differences, select only high-quality studies, or incorporate quality differences in the analysis. It is not clear whether different incorporation strategies result in different estimates of test performance and if they do, to what extent.

Objectives:

To study the effect of different strategies for incorporating methodological quality into the meta-analysis on the outcome measures of a meta-analysis of diagnostic accuracy studies.

Methods:

We repeated the analyses in published systematic reviews, using different strategies comparing the resulting summary estimates of diagnostic accuracy. We used an existing set of 28 systematic reviews, containing 31 meta-analyses in different clinical fields.
Methodological quality of all included studies was assessed using the QUADAS checklist. We applied the following analysis

strategies:

analysis of high quality papers only, sensitivity analysis, and sequential incorporation of results based on quality rankings. The diagnostic odds ratio was chosen as a summary measure for diagnostic accuracy, calculated using the model of Moses and Littenberg. To express the impact of study quality on the result of the meta-analysis, a relative diagnostic odds ratio was calculated for each strategy, comparing each high quality set of studies with the complete set or low-quality set.

Results and Conclusion:

Different strategies for incorporating quality can produce very different summary estimates of diagnostic accuracy. Analyses are in progress. Final results will be presented at the Cochrane Colloquium in October 2005.