Skip to main content

Completeness of outcome specification across Cochrane systematic reviews of three common eye conditions: Time to be more explicit!

Date and Location




Sunday 22 September 2013 - 10:30 - 12:00


Presenting author and contact person

Presenting author

Ian Saldanha

Contact person

Ian Saldanha
Abstract text
Background: Complete specification of outcomes in a systematic review (SR) is important for promoting reproducibility, minimizing bias in the SR process (meta-bias), and evaluating time-dependent treatment effects. A completely-specified outcome has four domains: outcome title (e.g., visual acuity), technique/instrument used, outcome measure (e.g., continuous variable such as mean visual acuity), and time-point(s) for outcome assessment. Objectives: To evaluate the degree of completeness of outcomes specification across SRs addressing three common eye conditions: age-related macular degeneration (AMD), cataract, and diabetic retinopathy (DR). Methods: Using the four domains, we evaluated completeness of specification for the most frequently-assessed primary and non-primary outcomes for all Cochrane SRs addressing the three conditions. Results: We identified 36 completed Cochrane SRs and protocols addressing AMD (n=16), cataract (n=16), and DR (n=4). All reviews provided titles for the outcomes they assessed. Time-point(s) was the next most frequently specified domain; for example 13/16 (82%) SRs on AMD assessing visual acuity specified time-points at which the outcome was assessed (Table). Outcome measure was the least frequently specified domain; for example only 1/12 (8%) cataract SRs assessing quality-of-life specified the outcome measure. For the 5 most frequent outcomes across conditions, 16/35 (46%) SRs assessing visual acuity, 2/13 (15%) SRs assessing contrast sensitivity, and none of the 32, 7, and 13 SRs assessing quality-of-life, disease onset, and disease progression specified all 4 domains. Primary outcomes were more likely than non-primary outcomes to specify all four domains (46% vs. 5%, p<0.0001), and specified a higher mean number of domains (3.24 vs. 1.92, p<0.0001). Conclusions: Specification of outcomes was largely incomplete across SRs addressing the three conditions. To facilitate reproducibility and reduce bias, systematic reviewers should completely specify the primary and additional outcomes they propose to examine.