Article type
Year
Abstract
Background: The ’Summary of findings' (SoF) table is a key tool to present the main findings of Cochrane reviews in a transparent and simple format. The SoF table presents an absolute measure of treatment effect, using a measure of baseline risk to convert relative effect sizes. This is desirable as an absolute measure provides an estimate of how likely an intervention is to work in a specific population, or in an individual patient.
Objectives: We explored the relevant methodological literature, compared and identified strengths and weaknesses of different approaches to estimate baseline risk toward the goals of valid, feasible and understandable presentation of absolute measures of effect.
Methods: Baseline risk estimates for health populations were extracted from randomized controlled trials, prognostic systematic reviews, and individual prognostic studies. The table presents data sources and estimate measures compared. Simulation modeling was used to assess the potential impact of more or less extreme estimates.
Results: A summary of the methodological literature, results of exploratory analyses and simulation studies have been conducted. These analyses have demonstrated that the best estimate of baseline risk depends on the review question, including the size of the treatment effect, how well the trial control groups represent the population of interest, and the heterogeneity in the baseline risk values. It is appropriate to estimate the baseline risk from the median RCT control group rate in many situations; in some cases estimation of baseline risk from prognostic evidence is more appropriate.
Conclusions: We have developed several resources for reviewer authors and Cochrane Review Groups in making decisions about baseline risk estimates including: i) an algorithm to guide decision-making about appropriate choice for baseline risk estimate; ii) recommendations for identifying baseline risk estimates from prognosis studies; and iii) a tool to explore the impact of changes in baseline risk estimates on important absolute treatment effect.
Objectives: We explored the relevant methodological literature, compared and identified strengths and weaknesses of different approaches to estimate baseline risk toward the goals of valid, feasible and understandable presentation of absolute measures of effect.
Methods: Baseline risk estimates for health populations were extracted from randomized controlled trials, prognostic systematic reviews, and individual prognostic studies. The table presents data sources and estimate measures compared. Simulation modeling was used to assess the potential impact of more or less extreme estimates.
Results: A summary of the methodological literature, results of exploratory analyses and simulation studies have been conducted. These analyses have demonstrated that the best estimate of baseline risk depends on the review question, including the size of the treatment effect, how well the trial control groups represent the population of interest, and the heterogeneity in the baseline risk values. It is appropriate to estimate the baseline risk from the median RCT control group rate in many situations; in some cases estimation of baseline risk from prognostic evidence is more appropriate.
Conclusions: We have developed several resources for reviewer authors and Cochrane Review Groups in making decisions about baseline risk estimates including: i) an algorithm to guide decision-making about appropriate choice for baseline risk estimate; ii) recommendations for identifying baseline risk estimates from prognosis studies; and iii) a tool to explore the impact of changes in baseline risk estimates on important absolute treatment effect.