This is a chapter from "Methods Guide for Effectiveness and Comparative Effectiveness Reviews."
People using assistive technology may not be able to fully access information in these files. For additional assistance, please contact us.
In quantitative synthesis of randomized clinical trials (RCTs) for a comparative effectiveness review, continuous outcomes are usually less straightforward to analyze than binary outcomes. Continuous outcomes are often measured at both baseline and followup time points. Results of continuous data can be reported as means, mean differences, or differences in change score from baseline, and measures of precision are reported as standard deviation (SD), standard error (SE), or confidence intervals. The distribution of the data is not always symmetric, and journal publications may not report all of the information required for meta-analysis.
The original quantitative synthesis chapter of the “Methods Guide for Effectiveness and Comparative Effectiveness Reviews” has a very brief continuous outcomes section that provides limited guidance on using mean difference versus standardized mean difference, but the section does not provide guidance on a number of other issues relating to meta-analysis of continuous outcomes. To fill this gap, this report updates the guidance on quantitative synthesis of continuous outcomes measured in RCTs.
Accordingly, we address the following topics applicable to quantitative synthesis of continuous outcomes measured in RCTs: choice of effect measures of continuous outcomes, choice of estimates for mean difference and baseline imbalance; calculation of SD and SE, how to handle missing data and skewed data, use and interpretation of the standardized mean difference (SMD) and of the ratio of means (RoM) as an alternative measure, and dichotomization of continuous outcomes in meta-analyses.
For each of the topics related to quantitative synthesis of continuous outcomes, we searched for relevant methodological or applied methodological papers in the Effective Health Care Program Methods Library and in Ovid Medline, Current Index to Statistics, and Scopus databases (Appendix A, full report). Recommendations for each topic were then developed based on current knowledge of the literature along with group discussion and consensus. A draft report of the workgroup’s key conclusions and recommendations was circulated for comment to peer reviewers and Agency for Healthcare Research and Quality officers, and those comments were considered by the team in preparing this report. The summary of final key points and recommendations are presented in the full report.