Unpublished systematic reviews and financial support: a meta-epidemiological study

Abstract

Objective: PROSPERO, an international prospective register of systematic reviews, was launched in February 2011 to reduce publication bias of systematic reviews (SRs). A questionnaire survey of SR researchers conducted in 2005 indicated the existence of unpublished SRs and the potential influence of lack of funding as a reason for non-publication. Here, we investigated the publication status of registered SRs in the 1st year that PROSPERO launched and assessed the association between publication and the existence of funding or conflicts of interest (COIs).

Results: We identified 326 SRs registered in PROSPERO from February 2011 through February 2012. Among them, 85 SRs (26%) remained unpublished at least 65 months after registration. We found 241 published reports, including four conference abstracts and one poster presentation. Median time to publication from protocol registration was 16.3 months. Funding for SRs was associated with publication [odds ratio (OR) = 2.10; 95% confidence interval (CI) = 1.26 to 3.50]. We found no significant association of author-reported COIs with publication (OR = 2.35; 95% CI = 0.67 to 8.20). Twenty SRs were not published despite the authors reporting completion of the reviews in PROSPERO.

Keywords: Dissemination bias; Meta-epidemiology; PRISMA; PROSPERO; Publication bias; Registry; Systematic review; Unpublished results.

Fig. 1

We first identified published SRs with published or updated status in PROSPERO records. One of them was only conference abstract publication. For 211 SRs with non-published status in PROSPREO, We investigated dissemination of the result in the publicly available space or published in any journals searched by Google or Google Scholar

Fig. 2

The minimum and maximum time to publication were 0.87 and 56 months. The median was 16.3 months. The average was 19.9 months. *The time (in months) from protocol registration to publication of SRs was defined as the number of days from registration to publication divided by 30. The date of publication was selected for each review according to the following hierarchy: (1) the acceptance date, if available; (2) the online publication date, if available; (3) the earlier of the journal publication date or the date of author-reported “published” status in PROSPERO; (4) the earlier of the journal publication date of a conference abstract or the documented date of public poster dissemination detected by searching Google or Google Scholar. If only the publication month was reported, the midpoint of that month (day 15) was set as the publication day