Meta-analysis

It is a statistical process by which the results of several studies, frequently randomized clinical trials, are combined to develop a single estimate of the effect of an intervention, treatment, or exposure on disease [1]. It is a specific type of systematic analysis that uses a quantitative method to combine data from two or more studies. It is a series of systematic methods for combining data from more than one investigation in order to draw conclusions that could not be drawn from a single investigation [2]. Therefore, every meta-analysis should be based on an underlying systematic review, but not every systematic review leads to a meta-analysis [3]. It gives a pretty good idea about the best trend to go with, which is supported by multiple studies, with a larger number of patients (by Dr. Nabil Mobarek, Medical Advisor Oncology at Eli Lilly and Company).

It «is a statistical method for combining the information residing in the summary statistics, say odds ratios and their standard errors, from a group of independent studies, e.g., clinical trials, of a given treatment or diagnostic procedure. In biochemical and epidemiological investigations it is usually undertaken to achieve one of two aims: (1) obtaining pooled point and interval estimates of an overall "true treatment effect". (2) "borrowing strength" across a group of studies to obtain improved point and interval estimates. There are four models of meta-analysis: fixed effects, random effects, mixed effects, and Stein effects. The first three models are used to achieve the first aim. The last two are used to achieve the second aim. The Stein, or empirical Bayes model, like most Bayes models, provides a useful answer to the extrapolation problem often described as Bernard's dilemma: The response of the "average" patient to therapy is not necessarily the response of the patient being treated [4].»

Bibliographic references:

[1] Balcer, L. (n.d.). Basic Epidemiologic and Biostatistics Terminology. [online] Available at: http://content.lib.utah.edu:81/cgi-bin/showfile.exe?CISOROOT=/ehsl-nam&CISOPTR=1175&filename=1176.pdf [Accessed 20 Jun. 2016].

[2] Riegelman, R. (2005). Most Frequently Used Terms in Biostatistics. In: Studying a study and testing a test. Philadelphia: Lippincott Williams & Wilkins, pp.1-38.
[3] Researchcore.org. (2016). ResearchCore.org - What is the difference between a "systematic review" and a "meta-analysis"?. [online] Available at: http://www.researchcore.org/faq/answers.php?recID=5 [Accessed 21 Jun. 2016].

[4] Deasy JO, et al.; AAPM/NIH. Methodological issues in radiation dose-volume outcome analyses: summary of a joint AAPM/NIH workshop. Med Phys. 2002 Sep;29(9):2109-27. Available at: https://doi.org/10.1118/1.1501473.