Abstract
AIM: To investigate the reproducibility of the updated Standards for the Reporting of Diagnostic Accuracy Studies tool (STARD 2015) in a set of 106 studies included in a Cochrane diagnostic test accuracy (DTA) systematic review of imaging tests for diagnosing manifest glaucoma. METHODS: One senior rater with DTA methodological and clinical expertise used STARD 2015 on all studies, and each of three raters with different training profiles assessed about a third of the studies. RESULTS: Raw agreement was very good or almost perfect between the senior rater and an ophthalmology resident with DTA methods training, acceptable with a clinical rater with little DTA methods training, and only moderate with a pharmacology researcher with general, but not DTA, systematic review training and no clinical expertise. The relationship between adherence with STARD 2015 and methodological quality with QUADAS 2 was only partial and difficult to investigate, suggesting that raters used substantial context knowledge in risk of bias assessment. CONCLUSIONS: STARD 2015 proved to be reproducible in this specific research field, provided that both clinical and DTA methodological expertise are achieved through training of its users.