Background:
Systematic review (SR) authors should define when outcomes are measured. If not pre-specified, this may introduce bias into the selection of reported results. However, there has been no investigation into how often authors define the timing of outcome measurement and whether or not this is associated with the reporting of statistical significance.
Objectives:
We aimed to describe the definition of the timing of outcome measurement and to examine whether pre-specification is associated with the reporting of statistical significance in SRs registered in PROSPERO.
Methods:
All SR protocols for interventional randomized controlled trials (RCTs) that were registered in PROSPERO by 31 December 2014 were eligible. We excluded Cochrane protocols, SRs with network meta-analyses, individual patient data meta-analyses and SRs without meta-analysis. We identified publications using the authors' names and keywords for participants or interventions in PROSPERO via PubMed and Google Scholar. We regarded the timing of outcome measurement to be pre-specified when the authors defined this for all primary outcomes. We described the proportion of SR protocols defining the timing of outcome measurement. We further examined the association between defining the timing of outcome measurement and the reporting of statistical significance using a multivariable logistic regression model to adjust for year of registration, number of included RCTs, number of primary outcomes, multinational collaboration and funding.
Results:
We included a total of 379 SR protocols. Of these, 209 full publications and nine conference abstracts were published before February 2018. Eighty-five protocols (22%) defined the timing of outcome measurement. The proportion did not change from 2011 to 2014 (P for trend = 0.51). Defining the timing of outcome measurement in the protocol was not associated with reporting of statistical significance (adjusted odds ratio 1.48, 95% confidence interval 0.61 to 2.54).
Conclusions:
The timing of outcome measurement was not usually defined among SRs registered in PROSPERO. The present study found no evidence to suggest that defining the timing of outcome measurement reduced the reporting of statistical significance in the publication.
Patient or healthcare consumer involvement:
None.