Understanding Meta-Analyses and Systematic Reviews

Abstract

E vidence-based practice is fundamentally about answering patient-centered questions. A pitfall in trying to answer such questions is finding 1 relevant study, appraising it, and making a clinical decision based on that single study. Sometimes a single study is all that exists, but in many cases there are multiple studies that address similar clinical questions. Systematic reviews (SR) and meta-analyses (MA) are frequently used to synthesize large amounts of information from different studies to get an overall assessment of the benefit of a specific treatment, diagnostic tool, or laboratory test, or the overall risk from a specific exposure. SR and MA are fundamentally different from review articles. A narrative review article is usually written by a content expert and can be an excellent way to get a broad overview of a topic, but does not necessarily answer a specific clinical question. The process the author used to select the articles for review is not usually transparent and could be biased, so a review article is impossible to replicate. A systematic review is a qualitative or quantitative synthesis of data and is typically narrow in scope, addressing a specific clinical question. Each study included in the SR is assessed for quality, and the process of study identification and selection is systematic and explicitly described. The risk of bias of each individual study is discussed. A meta-analysis quantitatively combines data from multiple studies (and potentially large numbers of patients) and therefore has increased statistical power. It is especially useful for finding statistically significant results from combining multiple small studies.While SR and MA do some of the work for you, they cannot be taken on faith. To appraise a SR or MA, look for:• A clear clinical question that is being answered.• An explicit description of the literature search, including whether unpublished papers, conference proceedings, and non-English articles were searched. Omitting these could lead to publication bias. • The eligibility criteria and quality of each included study.• The methods and results of each study succinctly presented, ideally in a table that includes interventions or exposures tested, number of study participants, and the outcomes in each group. • An overall pooled statistic on the main outcomes including data from studies included in the SR or MA. This is frequently presented graphically with the use of a "forest plot. " If the 95% confidence intervals around the pooled point estimate of the outcome do not include 0 (or 1, depending on the assessment), the overall effect is statistically significant. • A discussion of heterogeneity. There are 2 ways to detect this, a statistical method and a "common sense" method. The I2 statistic ranges from 0% to 100%, and it represents the proportion of variation among studies that is due to heterogeneity. An I2 of more than 50% is, by convention, unacceptable. The "common sense" method involves considering the specific patient and original clinical question and determi...

Economic impact of routine opt‐out antenatal human immune deficiency virus screening: A systematic review

“…Nevertheless, for the purposes of this study, we synthesised a wide range of high‐quality studies comprising of trial‐based studies and decision modelling studies. High‐quality systematic review of interventions is a panacea to identifying suitable alternative evidence‐based alternatives that replace the existing ones (Centre for Reviews Guidance, ; Cunningham, ; Nagappan, ; Sackett, ). Ideally, economic decisions should rely on data obtained from systematic review provided the studies are sufficiently homogenous to be comparable and the results generalisable to the population with different baseline risks, instead of a specific population group identified for an individual trial (Centre for Reviews Guidance, ).…”

“…Ideally, economic decisions should rely on data obtained from systematic review provided the studies are sufficiently homogenous to be comparable and the results generalisable to the population with different baseline risks, instead of a specific population group identified for an individual trial (Centre for Reviews Guidance, ). Systematic review is essential because it employs rigorous, orderly and transparent methods to reduce bias in results (Centre for Reviews Guidance, ; Nagappan, ).…”