Mean and Mean difference are the two key statistical measures used in the statistical analysis. Both are essential for meta-analysis as well. Mean and Mean difference are used for the interpretation of a large set of values into a single number which explains the heterogeneity and variation among the individual values. However, one of a common challenge in meta-analysis is the unavailability of this data (mean and standard deviation).

Q & A Forum

Meta Analysis

Q: What is the Minimal sample required for Meta-analysis?

Introduction

Meta-analysis is one of the methods of systematic review. It is the quantitative methods of performing the literature review that helps to minimize the data not justified by subjective interpretation. It considers studies with the same outcome to draw conclusive evidence summarizes statistics of individual studies and combines them to derive an answer for hypothetical questions. The number of studies and a number of samples within the study is crucial for a confident answer. The meta-analysis is usually regarded as a true effect when the number of studies is more. Reducing the size can significantly reduce the time needed for review.[1]

The requirement of sample size is based upon many crucial factors such as heterogeneity, population demographics, and design of the study.

Number of studies:

Currently, there is no rule describing the minimal studies required to do a meta-analysis. But the fewer studies are not enough to derive conclusive evidence for the hypothesis.

Cochrane Library suggests that two studies are enough to perform a meta-analysis. [4]. A study was conducted on the threshold number of studies required for meta-analysis from the Cochrane Library dataset. Based on the study it was concluded that 9 studies are sufficient to conclude meta-analysis. Studies of more than 10 are considered more reliable. More than 20 studies provide high statistical power and subgroup analysis.[2][3]

Sample size:

While the requirement of sample size is greatly dependent on population heterogeneity and study bias, A study conducted by Dechartres et al [6] states that sample size greatly influences the treatment outcome. In the study, 93 meta-analyses were assessed. It concludes studies with 50 samples are 48% larger in efficacy outcome than 1000 sample studies. Studies with 500-999 patients are 10% larger than 1000 patients in efficacy outcome.

Conclusion:

From the above statement, we can conclude that the minimum number of studies required for meta-analysis is two. Studies of more than nine are considered more reasonable to draw conclusions. The sample size within each study influences the weightage of the effect of each study. Higher other the sample size higher the weightage, lower the sample size smaller the weightage.

Refernce:

Guzzo RA, Jackson SE, Katzell RA. Meta-analysis analysis. In: Research in Organizational Behavior. Vol 9. JAI Press Inc; 1987:407-442. ISBN: 0-89232-636-0. https://cccrg.cochrane.org/sites/cccrg.cochrane.org/files/uploads/meta-analysis_revised_december_1st_1_2016.pdf
McLellan, J., & Perera, R. (2018). Restricted meta-analyses versus full meta-analyses: T4hreshold number of studies based on study sample size. University of Oxford. BMJ Evidence-Based Medicine, 23(Suppl 2), A10. DOI: 10.1136/bmjebm-2018-111024.21
Dechartres, A., Trinquart, L., Boutron, I., & Ravaud, P. (2013). Influence of trial sample size on treatment effect estimates: Meta-epidemiological study. BMJ, 346, f2304. https://doi.org/10.1136/bmj.f2304