Methodological review showed correct absolute effect size estimates for time-to-event outcomes in less than one-third of cancer-related systematic reviews

Ref ID 473
First Author N. Skoetz
Journal JOURNAL OF CLINICAL EPIDEMIOLOGY
Year Of Publishing 2019
URL https://www.jclinepi.com/article/S0895-4356(18)30530-4/fulltext
Keywords • Cochrane
• Statistical
• Error
• Oncology
Problem(s) • Errors in effect estimate calculations or data synthesis
• Weaknesses identified in some Cochrane reviews
Article Type Empirical
Article Subtype Cross-sectional survey/Methodological systematic review
First Author Country Germany
Aim To assess how many cancer-related Cochrane reviews hazard ratio-based absolute effects in summary of findings tables have been correctly calculated and reported published between January 2011 and December 2017 in the Cochrane Database of Systematic Reviews.
Level of Investigation Analytical
Summary of Findings In 29% of included Cochrane reviews, absolute effects in the summary of findings tables were calculated in a correct manner. In 24%, absolute effects had been correctly calculated (e.g., overall survival), but labelled incorrectly in summary of findings tables (e.g., death). 13% of reviews provided incorrect absolute effects in summary of findings tables. For 7%, it was unclear if absolute effects were correctly calculated. In 27% reviews, no absolute effects based on the given hazard ratio were calculated.
Number of systematic reviews included 96
Number of eligible systematic reviews assessed 483
Treatment impacted Yes
Treatment impacted description The authors examine whether data on the treatment effect are accurately calcuated and reported in the summary of findings tables of Cochrane reviews.
Interpretation impacted Yes
Interpretation impacted description In less than one-third of cancer-related Cochrane reviews, absolute effect size estimates were correctly calculated and reported indicating that erroneous interpretations about treatment effects could be made from two thirds of reviews examined.