Meta-analysis: What, Why, and How
Posted on 2nd December 2016 by Ammar Ismail
What is a meta-analysis?
Meta-analysis is a statistical technique for combining data from multiple studies on a particular topic.
Meta-analyses play a fundamental role in evidence-based healthcare. Compared to other study designs (such as randomized controlled trials or cohort studies), the meta-analysis comes in at the top of the ‘levels of evidence’ pyramid in evidence-based healthcare. This is a pyramid which enables us to weigh up the different levels of evidence available to us. As we go up the pyramid, each level of evidence is less subject to bias than the level below it. Therefore, meta-analyses can be seen as the pinnacle of healthcare evidence (1).
Meta-analyses began to appear as a leading part of research in the late 70s. Since then, they have become a common way for synthesizing evidence and summarizing the results of individual studies (2).
Why should we carry out and use meta-analyses?
To make a valid decision about using an intervention, ideally we should not rely on the results obtained from single studies. This is because results can vary from one study to another for various reasons, including confounding factors, and the different study samples used.
By combining individual studies, and thus using more data, the precision and accuracy of the estimates in the individual studies can be improved upon. Additionally, if the individual studies were underpowered, combining them in a meta-analysis can increase the overall statistical power to detect an effect.
**for further reading, please visit: (https://www.meta-analysis.com/pages/why_do.php)
How is a meta-analysis performed?
Below are the basic steps involved in a meta-analysis (3):
Identifying/formulating a problem: this is a question to be answered e.g. “to determine the effectiveness of exercise for depression compared with no treatment and comparator treatments”.
Doing a literature search: this will probably involve searching multiple databases that index reliable peer-reviewed articles, such as PubMed, Scopus, Web of science, Embase, etc.
Deciding on selection/inclusion criteria: you should use inclusion and exclusion criteria that will ensure that high-quality evidence, of direct relevance to your research question, is included. For this reason, we tend mostly to include randomized controlled trials (and may exclude observational studies). Ideally, we would also include unpublished studies in order to avoid publication bias. (If we fail to include all of the relevant studies, our conclusions may be erroneous. Specifically, we may overstate the benefit of a treatment (for example), because studies which fail to find a significant effect are less likely to be published than those which do not find a significant effect. See here for more information).
Data extraction: you ought to extract data for your outcomes of interest to be pooled (combined) in the final analysis set.
Doing the basic meta-analysis: there are a range of software for this purpose, such as Review Manager and Comprehensive Meta-Analysis Software.
1. Haidich AB. Meta-analysis in medical research. Hippokratia [Internet]. 2010 Dec [cited 2016 Sep 25];14(Suppl 1):29–37. Available from: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3049418&tool=pmcentrez&rendertype=abstract
2. Chalmers TC, Matta RJ, Smith H Jr, Kunzler AM. Evidence favoring the use of anticoagulants in the hospital phase of acute myocardial infarction. N Engl J Med. 1977; 297: 1091–1096. Available from: https://www.ncbi.nlm.nih.gov/pubmed/909566
3. Field AP, Gillett R. How to do a meta-analysis. Br J Math Stat Psychol [Internet]. 2010 Nov [cited 2016 Sep 25];63(Pt 3):665–94. Available from: http://www.ncbi.nlm.nih.gov/pubmed/20497626