Meta analysis is defined as "The statistical analysis of large collection of analysis results from individual studies for the purpose of integrating the findings."



Origin

The predecessor of Meta analysis originated from Fisher's idea of "merging P-values" in 1920;

The initial concept was first proposed by Beecher in 1955;

In 1976, the psychologist Glass further developed into "consolidated statistics" according to his thoughts, which is called Meta analysis;

In 1979, British clinical epidemiologist Archie Cochrane proposed the concept of Systematic Review (SR) and published the "Systematic Review of Randomized Controlled Trials of Hormone Therapy for Premature Pregnant Women to Reduce Neonatal Mortality."



Scope of application

Meta in a broad sense refers to a scientific clinical research activity, which refers to the entire process of comprehensively collecting all relevant studies and conducting rigorous evaluation and analysis one by one, and then statistically processing the data using quantitative synthesis methods to reach a comprehensive conclusion;

Meta in the narrow sense refers to a purely quantitative and synthetic statistical method.



Analyze the advantages

1) Able to evaluate the consistency of multiple research results on the same subject;

2) Systematic evaluation and summary of multiple research results on the same subject;

3) Put forward some new research questions to point out the direction for further research;

4) When subject to certain conditions, such as time or research object constraints, meta-analysis may be an option;

5) From the perspective of methodology, evaluate the research design of a certain topic at this stage;

6) Found some unexplained problems in a single study;

7) For clinical experimental studies with small samples, meta-analysis can statistic the accuracy of power and effect size estimation. Therefore, a well-designed and rigorous meta-analysis article can evaluate evidence more objectively (compared with traditional descriptive reviews), evaluate effect indicators more accurately and objectively, and explain the differences between the results of different studies. Heterogeneity. Meta analysis conforms to the process of people's understanding of objective laws, is completely consistent with the thinking of evidence-based medicine, and is a huge improvement.



Step method

The main steps

Clearly and concisely put forward the problem to be solved

Develop search strategies and collect randomized controlled trials comprehensively and extensively

Determine inclusion and exclusion criteria, and eliminate documents that do not meet the requirements

Data selection and extraction, including original result data, charts, etc.

Quality assessment and characterization of each test

Statistical processing

Interpretation of results, conclusions and evaluations

Maintain and update information.

Approach

Heterogeneity test (homogeneity test, that is, through the most commonly used Q test, I2, and H value, etc.).
Calculate the combined effect size (weighted combination, calculate the effect scale and 95% confidence interval) and perform statistical inference.
Generally, a forest diagram is used to represent the results of a single experiment and the combined results.
Sensitivity analysis
Through the Egger’s method, the calculation of the “failure number” and the “inverted funnel chart” to understand the potential publication bias.


Applicability

Increase statistical power: Because individual clinical trials often have small samples, it is difficult to determine certain effects, and these effects may be important to clinicians

Resolve the inconsistency of the research results

Seeking new hypotheses



Limitation

Not all relevant studies included

Unable to extract all relevant data

(Publication bias) Only studies with positive findings published / accepted

Unclear definition of clinical endpoints used for pooled statistics.



About us

Located in the USA, CD BioSciences provides a full range of clinical trial consulting, design, conduction, management, reporting and analysis services to advance your clinical trial development. CD BioSciences continuously provides professional support to local, national and international clients. Our clinical trial services focus on the entire clinical trial lifecycle, including: Variance Component and Linear Models Analysis, Analytical Writing, Non-parametric Methods Analysis, Statistical Significance Tests, etc.