A thorough look into meta-analysis
Posted on 10th October 2017 by Ammar Ismail
This blog follows on from Ammar’s previous post “Meta-analysis: What, Why and How” (you will find the link to that blog at the bottom of this page, together with other handy resources).
A historical perspective
Meta-analysis is a type of research introduced in the 17th century in astronomy. The first clinical study to be published in this manner was published by the statistician Karl Pearson in the British Medical Journal (1).
The word “Meta-analysis” was first given to this specific model by the statistician Gene V. Glass (2). Some of his precious words were “my major interest currently is in what we have come to call …the meta-analysis of research. The term is a bit grand, but it is precise and apt … Meta-analysis refers to the analysis of analyses”.
What is the definition of meta-analysis?
Meta-analysis is a quantitative, rigorous, and precise type of research analysis. It intends to summarize and consolidate the results of an intervention in a specific cohort of people.
The value of using this statistical approach
The value of using Meta-analysis can be summarised in multiple points:
Firstly; it combines data from many studies – all studies available about a single specific topic – then gives a net effect estimate from all these studies. So, its sample size and power will be greater than any single study within the literature on this subject.
Secondly; we use Meta-analysis when there is a great degree of conflict between studies which can not be resolved without using meta-analysis. So, meta-analysis combines data from all these different studies and then gives us a precise result which can be easily understood.
Thirdly; it helps in establishing guidelines and straightening the clinical practice, as data from meta-analyses are derived from a wide scope of population so the meta-analysis will be more generalizable and powerful.
Level of evidence supplied by meta-analyses
Meta-analysis provides class I evidence according to the evidence pyramid. In this pyramid, meta-analysis of randomized controlled trials occupies the top part, the “apex of the pyramid”. After this comes systematic reviews, randomized controlled trials, cohort studies, case-control studies, case reports/ case series, then expert opinion which is not followed anymore in clinical sciences.
Pharmaceutical companies can depend on meta-analyses to gain approval of their drugs. In clinical practice, physicians use the meta-analysis approach for choosing the suitable course and effective doses for their patients. Also, researchers use this analytic approach to apply for grants to do research about what is called “gap of knowledge”, for example, to start testing a new regime.
How can we implement the steps of meta-analysis?
Firstly; we specify a topic for our project,
Secondly; we search relevant databases using keywords,
Thirdly; we screen the records retrieved from our search,
Fourthly; we extract data that can be analyzed (data available from two studies at least),
Fifthly; extracted data should be analyzed in software such as Review Manager or Stata,
Finally; retrieved data should be documented.
For further reading, please visit the websites below:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3049418/
https://www.meta-analysis.com/pages/why_do.php
https://en.wikipedia.org/wiki/Meta-analysis
https://s4be.cochrane.org/meta-analysis-what-why-and-how/
https://s4be.cochrane.org/the-evidence-based-medicine-pyramid/
References:
1. Pearson K. Report on Certain Enteric Fever Inoculation Statistics. Br Med J [Internet]. 1904 [cited 2017 Jun 19];2(2288):1243–6. Available from: http://www.bmj.com/content/2/2288/1243
2. Glass GV. Primary, Secondary, and Meta-Analysis of Research. Educ Res [Internet]. 1976 Nov 1 [cited 2017 Jun 19];5(10):3–8. Available from: http://journals.sagepub.com/doi/abs/10.3102/0013189X005010003