Assessing the magnitude of reporting bias in a sample of recent systematic reviews - A comparison of a Prospero and Cochrane Sample
Objective: To explore differences in published reviews and their respective protocols in a sample of 97 non-Cochrane systematic reviews (non-CSRs) and 97 Cochrane systematic reviews (CSRs) in terms of PICOS elements and to which extent they were reported.Study Design and Setting: We searched PubMed and Cochrane database to identify non-CSRs and CSRs that were published in 2018. Afterwards we searched for their corresponding Cochrane or PROSPERO protocol. The reviews were compared to their protocols. The primary outcome was the change from protocol to review in terms of PICOS elements.Results: More than the half of each sample (54.6% of CSRs & 67% of non-CSRs) (ARR 12.4% [12% ; 12.8%]) presented changes in PICOS elements. A total of 227 changes were identified, distributed as 108 (47.5%) in CSRs and 119 (52.5%) in non-CSRs. For both samples an approximate third of total changes corresponded to changes related to primary outcomes. Marked differences were found between samples with regards to declaration of changes. Only 4.2% of changes in PICOS items have been declared in non-CSRs compared to 57.4% in CSRs (ARR 53.2% [52.9% ; 53.5%]).Conclusion: CSRs showed better results than non-CSRs in terms of reporting changes and showed higher quality. Nevertheless, reporting of changes from protocol should be advanced and requires improvement in general. Limitation of this study lies in its cross-sectional nature.