Whether prospective registration can improve the overall reporting and methodological quality of systematic reviews: a comparative meta-epidemiological study

Tags: Oral
Ge L1, Tian J1, Song F2, Zhang J3, Yang K1
1Evidence-Based Medicine Center of Lanzhou University, China, 2Norwich Medical School, University of East Anglia, UK, 3School of Nursing, Gansu University of Chinese Medicine, China

Background: A key feature of high-quality systematic reviews (SRs) is the development of a protocol that sets out the main objectives, key design features and planned analyses for the review. A protocol written in advance can avoiding bias in the conduct and reporting of SRs. A large number of prospectively registered SRs have been published. It is not known yet whether the overall reporting and methodological quality of prospectively registered SRs is better than that of unregistered SRs.

Objective: To compare the quality of registered and not-registered SRs.

Methods: One reviewer searched PubMed to identify SRs/meta-analysis published in 2015 in English. Two reviewers independently selected full-text to identify eligible SRs and then divided them into a registered group and not-registered group. Registered SRs were defined as having a protocol in advanced of the review, whether a registration number was available or not. For each group, eligible SRs were randomly ordered, and the first 100 reviews were selected. If a selected SR was not eligible, a successive record was used to replace it until the total number of included SRs was 100 for each of the two groups. Data extracted from SRs included general characteristics, reporting of literature search, selective reporting bias, reporting quality based on the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement, and methodological quality based on the AMSTAR (A MeaSurement Tool to Assess systematic Reviews) checklist. The main characteristics and quality of registered SRs versus not-registered SRs were tabulated. The summary PRISMA and AMSTAR scores were ranked by quartiles for analysis. Ordered logistic regression analyses were conducted to compare the reporting and methodological quality of SRs between the registration and no registration groups after adjusting for multiple review characteristics. SPSS version 21.0 was used for statistical analyses.

Results and conclusion: This study is ongoing and available results will be presented at the Colloquium.