Should all inclusion criteria be used for inclusion?

Article type
Authors
Verbeek J1, Ijaz S2, Tikka C2
1Finnish Institute of Occupational Health
2Cochrane Work Review Group
Abstract
Background: Cochrane reviews state that there are four (PICO) criteria for inclusion of studies but the guidance is usually not to use the outcomes (O) for inclusion of studies to prevent outcome reporting bias.

Objectives: To survey how outcomes in recent Cochrane Reviews were defined and used for inclusion of studies and how current practice compares with existing guidance on preventing outcome reporting bias.

Methods:We performed a survey of the latest systematic reviews published by each Cochrane review group as of August 2016. We extracted data on the names and number of outcomes and which were used in summary of findings (SoF) tables and abstract, and recorded if and how the outcomes were used for inclusion of studies in the review. We compared this with guidance documents.

Results:We included 52 reviews with a mean of 8.4 (SD 4.3) outcomes. All but two reviews contained SoF tables. Of all reviews 47 (90%) used primary and secondary outcomes as the names for their review’s outcomes, but without further definition. None reported using a core outcome set. Forty reviews (77%) did not explain if and how they used the outcomes for inclusion of studies, 8 (15%) stated that studies were included if they reported either primary or secondary outcomes, 1 (2%) reported that outcomes were not used for inclusion and for 3 (6%) it was unclear how outcomes were used. In 15 (29%) reviews, the outcomes were used for inclusion because the type of participants/patients did not sufficiently define the outcome such as when the intervention is preventive. In 8 (17%) reviews, secondary outcomes were not reported in SoF table or in the abstract.

Conclusions: In a sample of recent Cochrane Reviews, it remained unclear if and how outcomes were used for inclusion of studies and for reporting in the SoF tables in most reviews. Better explanation of inclusion decisions at review level is needed to be able to understand the risk of outcome reporting bias in a review. Consistent guidance in names and definitions for different types of outcomes used in systematic reviews is needed.