Article type
Year
Abstract
Background: Randomised trials provide an opportunity for collecting and reporting information on the harms of various interventions. Ideally systematic reviews should be able to collate and synthesize useful information on such harms on a large-scale. However, reporting of safety issues in randomised trials is known to be suboptimal and the extent to which systematic reviews can provide robust information on harms is debatable
Objective: To assess how frequently systematic reviews of randomised clinical trials provide large-scale evidence on specific, well-defined adverse events.
Methods: We screened the entire Cochrane Database of Systematic Reviews (issue 3, 2003) for reviews with quantitative data on specific, well-defined harms for at least 4000 randomised subjects (reflecting 80% power to detect at alpha=0.05 an adverse event that occurs in about 1% with an intervention but is extremely rare without it). Main outcome measures included the number of reviews with eligible large-scale safety data, number of reviews being ineligible for various reasons, and magnitude of recorded harms (absolute risk, relative risk) based on large-scale evidence.
Results: Across 1727 reviews, 138 included randomised evidence on >4000 subjects, but only 25 (18%) of these had eligible harms data, as defined above, while 77 had no data on any adverse events and 36 had data on harms that were either non-specific and/or pertained to fewer than 4000 subjects. Several systematic reviews were identified that had no eligible data on specific harms, while the largest randomized trials with over 4000 subjects that were included in these reviews had published data on specific harms. Of 66 specific harms with adequate data addressed in the 25 eligible reviews, 25 showed formally statistically significant differences between the compared arms. In 21 of those, the absolute risk difference between the compared arms was less than 4% and most pertained to serious or severe adverse events.
Conclusions: Availability of large-scale evidence on specific, well-defined harms is very uncommon is systematic reviews of randomized trials. Nevertheless, our survey provides proof of the concept that such reviews can provide important information in this regard. Efforts should be made to improve both the reporting of harms in randomised trials and the utilization of such data in systematic reviews.
Objective: To assess how frequently systematic reviews of randomised clinical trials provide large-scale evidence on specific, well-defined adverse events.
Methods: We screened the entire Cochrane Database of Systematic Reviews (issue 3, 2003) for reviews with quantitative data on specific, well-defined harms for at least 4000 randomised subjects (reflecting 80% power to detect at alpha=0.05 an adverse event that occurs in about 1% with an intervention but is extremely rare without it). Main outcome measures included the number of reviews with eligible large-scale safety data, number of reviews being ineligible for various reasons, and magnitude of recorded harms (absolute risk, relative risk) based on large-scale evidence.
Results: Across 1727 reviews, 138 included randomised evidence on >4000 subjects, but only 25 (18%) of these had eligible harms data, as defined above, while 77 had no data on any adverse events and 36 had data on harms that were either non-specific and/or pertained to fewer than 4000 subjects. Several systematic reviews were identified that had no eligible data on specific harms, while the largest randomized trials with over 4000 subjects that were included in these reviews had published data on specific harms. Of 66 specific harms with adequate data addressed in the 25 eligible reviews, 25 showed formally statistically significant differences between the compared arms. In 21 of those, the absolute risk difference between the compared arms was less than 4% and most pertained to serious or severe adverse events.
Conclusions: Availability of large-scale evidence on specific, well-defined harms is very uncommon is systematic reviews of randomized trials. Nevertheless, our survey provides proof of the concept that such reviews can provide important information in this regard. Efforts should be made to improve both the reporting of harms in randomised trials and the utilization of such data in systematic reviews.