Background: the Living Overview of Evidence (LOVE, powered by Epistemonikos) is a database that compiles systematic reviews and is kept up to date through voluntary collaborators and computational algorithms. Thanks to a tool that allows linking systematic reviews of similar health issues, it offers the possibility of identifying with less effort and in less time all the potentially relevant evidence for a recommendation. Currently, the Chilean Ministry of Health uses this database to inform recommendation development. However, is it uncertain to what extent this pragmatic method may miss relevant evidence.
Objective: the aim of this study was to compare the evidence retrieved on the LOVE database with a traditional search of relevant systematic reviews and primary studies.
Methods: we compared the evidence obtained by a traditional search and by LOVE database in the recommendations of the clinical guideline about the management of HIV/AIDS. In the traditional approach, we conducted a search of systematic reviews in the Cochrane Library databases, Embase and MEDLINE from the date of inception to the date of guideline development (2017). In addition, when the latest systematic reviews identified was prior to 2015, we conducted a search of randomized controlled trials (RCTs). In the LOVE approach, we searched the database for systematic reviews about HIV treatment and categorized them into those that were relevant to the questions included. Different teams, unaware of the results of the other method (blinding) made the selection of relevant studies, first by evaluating the title and abstract, and then by evaluating the full text of the potentially relevant articles. Disagreements were resolved by a third investigator.
Results: the sensitivity and specificity of the LOVE database were 76% and 100%; while for the traditional method they were 69% and 99%, respectively.
In the qualitative analysis of the evidence potentially missed by searching in the LOVE database, we observed that only in one question could the use of the traditional search method have changed the evidence presented to the panel. In this case, the latest systematic review identified was outdated and the update of the search identified seven new RCTs, one of them a landmark trial. For the rest of the five questions, in one question the traditional method identified only one small additional trial, which did not substantially change the evidence, and in four questions no additional trial was found.
Conclusion: when recent systematic reviews exist, the use of LOVE database saves time and resources and provides almost all the relevant evidence for decision making. When there are no recent systematic reviews, the LOVE database has to be complemented by a search of primary studies.