In the previous chapter (Chapter 22, The Process of a Systematic Review and Meta-analysis), we provided guidance on how to evaluate the credibility of the process of a systematic review with or without a meta-analysis. In this chapter, we address how—if the systematic review is sufficiently credible—to decide on the degree of confidence in the estimates that the evidence warrants. As you will see, systematic review authors may have conducted a credible review and analysis and one may still have little confidence in the estimates of effect. We will return to the clinical scenario discussed in the previous chapter and obtain the relative and absolute effects of the intervention from a credible systematic review and meta-analysis1 and determine the confidence in these estimates (quality of evidence). The general framework for judging confidence in estimates is based on the approach offered by the GRADE (Grading of Recommendations Assessment, Development and Evaluation) Working Group.2 This chapter focuses on questions of therapy or harm. This framework can, however, be adapted for other types of questions, such as issues of prognosis3 or diagnosis.4
We continue with the scenario of a 66-year-old male smoker with type 2 diabetes and hypertension undergoing noncardiac surgery for whom we are considering prescribing perioperative β-blockers to prevent the cardiovascular complications of nonfatal infarction, death, and nonfatal stroke.
Understanding the Summary Estimate of a Meta-Analysis
If the systematic review authors decide that combining results to generate a single estimate of effect is inappropriate, a systematic review will likely end with a table or tables describing results of individual primary studies. Often, however, systematic reviews include a meta-analysis with a best estimate of effect (often called a summary or pooled estimate) from the weighted averages of the results of the individual studies. The weighting process depends on sample size or number of events (see Chapter 12.3, What Determines the Width of the Confidence Interval?) or, more specifically, study precision. Studies that are more precise have narrower confidence intervals (CIs) and larger weight in meta-analysis.
In a meta-analysis of a therapeutic question looking at dichotomous outcomes (yes/no) for estimates of the magnitude of the benefits or risks, you should look for the relative risk (RR) and relative risk reduction (RRR) or the odds ratio (OR) and relative odds reduction (see Chapter 9, Does Treatment Lower Risk? Understanding the Results). When the outcome is analyzed using time-to-event methods (eg, survival analysis), the results could be presented as a hazard ratio. In a meta-analysis addressing diagnosis, you should look for summary estimates of likelihood ratios or diagnostic ORs (see Chapter 18, Diagnostic Tests).
In the setting of continuous variables rather than dichotomous outcomes, meta-analysts typically use 1 of 2 options to aggregate data across studies. If the outcome is ...