Relatively absolute (one size may not fit all)

Article type
Authors
Cates C
Abstract
Background: There is a proposal to include Summary of Findings tables in Cochrane reviews to allow readers to see the relative and absolute effects of interventions. If the relative effects of treatment are stable across different levels of baseline risk, the absolute effect of treatment can be calculated for a specified level of baseline risk. How should this baseline risk be generated?

Objectives: To assess the impact of using different methods of generating a baseline risk to calculate the absolute effects of treatment in Cochrane reviews.

Methods: Using data from existing reviews the variability in baseline risks across trials will be measured using a Chi-square test and the reasons for heterogeneity explored.

Results: For understandable clinical reasons (severity of patients and trial duration, for example) the trials in many reviews are heterogeneous in terms of baseline risk. The average baseline risk is therefore not clinically meaningful, and if used to generate absolute treatment effects can be misleading, especially if clinicians use the absolute effects to compare treatment options.

Conclusions: Using an average baseline risk to generate absolute treatment effect can be misleading. Instead the trial duration and patient severity should be specified for baseline risk. Ideally, the potential for the end user to adjust the baseline risk should be available.