Editorials

Reporting diagnostic tests

Complying with STARD is likely to improve the quality of reporting

Education and debate p 41

As a clinician, I need high quality evidence about the usefulness, precision, and accuracy of diagnostic tests, and I need it now. Such evidence is rare even for the clinical examination, the most critical component of the diagnostic process. ^{1 2} The situation is getting worse with the exponential increase of diagnostic tests, most of which have never been evaluated properly and can mislead the diagnostic process. Although rigorous methodological standards in research about diagnostic tests have been applied more rigorously in the past decade, their reporting and methodological quality remain inadequate.^2-5 Against this background, the proposal in this issue from the authors of Standards for Reporting of Diagnostic Accuracy (STARD) for reporting diagnostic research should be applauded (p 41).³

There is a precedent for a favourable effect of such standards in the reporting of randomised trials. Since the development of the Consolidated Standards of Reporting Trials (CONSORT)⁶ and their adoption by the International Committee of Medical Journal Editors, the Council of Science Editors, and the World Association of Medical Editors, the reporting of randomised trials has improved. Although some of this may be due to a growing sophistication among trialists in general, the quality of reports in journals that promoted CONSORT (BMJ, JAMA, and Lancet) showed greater improvement than in a journal that did not advocate its use (New England Journal of Medicine).⁷ Similarly, although Devereaux and colleagues found that six of 11 methodological factors outlined in the CONSORT statement were still reported less than 50% of the time in 105 recently reported randomised trials published in 29 journals,⁸ journals that promoted the CONSORT statement did better than those that did not. Although these are encouraging results, referees and editors of journals need to do a better job in ensuring that their authors implement the CONSORT recommendations.

Given that precedent, the STARD criteria, if applied by investigators and required by editors, may lead to more high quality evidence about diagnostic tests. But if this evidence is to be used by busy clinicians, it must be available quickly and in an easily understandable form. The promise of the "more informative abstract" has to be fulfilled by authors and journals.⁹ The STARD methodological criteria provide the basis for more informative abstracts to accompany diagnostic articles.

Will the publication and recognition of the STARD statement have a favourable impact on our clinical practice? Although I agree with the authors of STARD that complete and informative reporting about diagnostic tests can only lead to better health care, I am confident that they would agree that the mere introduction of STARD is unlikely significantly to improve the quality and reporting of diagnostic research. Just as with the CONSORT statement, more intensive efforts to apply the guidelines will have to be made by those who develop diagnostic tests, fund and execute the studies that determine their clinical usefulness, and report and disseminate their results. This will not happen without substantial increases in the support of diagnostic research and in the translation and presentation of its results to the front lines of clinical care. Alliance between high quality diagnostic studies that observe the STARD recommendations and programmes of systematic reviews of diagnostic studies such as the one initiated by Matthias Egger and Daniel Pewsner are welcome. Egger and Pewsner have created the Bayes Library of Diagnostic Studies and Reviews, an international consortium conducting rigorous systematic reviews of studies of diagnostic accuracy. The alliance should provide journal editors and translational services with the raw materials that could give high quality evidence about the usefulness, precision, and accuracy of diagnostic tests to clinicians, and give it to them now.

Sharon E Straus, assistant professor. 

Toronto General Hospital, 200 Elizabeth Street, Toronto M5G 2C4, Canada (sstraus{at}mtsinai.on.ca)

Footnotes

Competing interests: SS attended the first STARD consensus conference but did not receive any honorariums.