Article type
Year
Abstract
Background: Due to suspected problems with MetaView 4.0, all updated reviews for The Cochrane Library Issue 1, 2000 were checked for possible numerical errors.
Objective: To compare the numerical results reported in the text of Cochrane Reviews with those produced by the "gold standard" results of MetaView 3.1.
Methods: All updated reviews in all modules for The Cochrane Library Issue 1, 2000, were manually checked. All numerical results reported in the results section only were compared with results produced by MetaView 3.1. For some reviews, abstracts were also checked. Discrepancies were compared with both MetaView 4.0 and hand calculations.
Results: Ninety reviews were checked. Of these, 36 [40.0%] had errors or omissions that were not due to software problems. One review of the ninety had an error due to problems in MetaView 4.0. Types of errors included: mistranscription of digits, failure to identify excluded and included studies correctly, confusion between Peto Odds Ratio and Odds Ratio, omission of comparisons from comparison tables, duplications, reporting no significance when significant, confusion of Odds Ratio, Risk Ratio, and Weighted Mean Difference, and rounding errors.
Conclusion: The results sections and abstracts of systematic reviews often disagreed with the "gold standard" results generated by MetaView 3.1. More than a third of Cochrane reviews may have errors; it may be useful to develop a checklist or guidelines to improve the quality of the results sections of Cochrane reviews.
Objective: To compare the numerical results reported in the text of Cochrane Reviews with those produced by the "gold standard" results of MetaView 3.1.
Methods: All updated reviews in all modules for The Cochrane Library Issue 1, 2000, were manually checked. All numerical results reported in the results section only were compared with results produced by MetaView 3.1. For some reviews, abstracts were also checked. Discrepancies were compared with both MetaView 4.0 and hand calculations.
Results: Ninety reviews were checked. Of these, 36 [40.0%] had errors or omissions that were not due to software problems. One review of the ninety had an error due to problems in MetaView 4.0. Types of errors included: mistranscription of digits, failure to identify excluded and included studies correctly, confusion between Peto Odds Ratio and Odds Ratio, omission of comparisons from comparison tables, duplications, reporting no significance when significant, confusion of Odds Ratio, Risk Ratio, and Weighted Mean Difference, and rounding errors.
Conclusion: The results sections and abstracts of systematic reviews often disagreed with the "gold standard" results generated by MetaView 3.1. More than a third of Cochrane reviews may have errors; it may be useful to develop a checklist or guidelines to improve the quality of the results sections of Cochrane reviews.