This document provides an overview of a meta-analysis workshop. It will introduce descriptive and inferential statistics, the concept of meta-analysis, and meta-analysis software and models. The workshop covers new topics like quality effects meta-analysis, heterogeneity models, and assessment of publication bias. It explains that simply averaging study results is incorrect, and meta-analysis statistically combines studies while weighting them by size and power to provide a single pooled effect estimate. Meta-analysis has advantages like larger power but must address heterogeneity and differences between studies.
1. Meta-analysis Workshop
From ZERO to HERO
Ahmed Negida
MBBCh candidate, Fifth Year Student
School of Medicine, Zagazig University, Egypt
2. Learning Objectives
• Introduction to descriptive and inferential statistics.
• Concept of meta-analysis
• Meta-analysis software
• Meta-analysis models
• Assessment of heterogeneity
• Assessment of publication bias
3. What is new in this workshop?
• Overview of different meta-analysis softwares
• Meta-analysis functions on MS Excel
• Quality effects meta-analysis model
• Q index as a safe-guards score
• Inverse variance heterogeneity model
• DOI plot and LFK index for assessment of publication bias
• Overview of the concept of indirect comparison of three-
treatment network using the “single closed loop” method.
4. Important Terms
• Effect estimate
• Effect size
• Mean difference
• Standardized mean difference
• Prospective study
• Retrospective study
• Clinical Trial
• Randomized Controlled trial
5. Combining Effect Estimates
The drug (x) is being investigated for the treatment of insomnia.
Multiple clinical trials have evaluated the efficacy of drug (x).
Sleeping hours (h): Improvement = Post - Pre
There are 10 studies reporting the improvement in sleeping
duration with drug (x) in patients with insomnia. How can you
combine the results of these studies in one single combined
effect estimate?
7. Using the mean average to summarize effect
estimates is not correct!
The mean average assumes that all studies have the same power
(weight) to influence the final effect estimate. This is not correct
because studies do not have the same power and are not equal in
sample size or population distribution.
Why?
8. Concept ofmeta-analysis
A meta-analysis is a statistical
analysis that combines the
results of multiple scientific
studies.
The idea of meta-analysis is to
pool the weighted average of
studies/subgroups to give one
single effect estimate. Pooled Effect Estimate
Study
3
Study
1
Study
2
10. Why can not werely onsingle high-quality studies?
Under the most rigorous study design conditions, a well-planned single
study, even if prospective and randomized, rarely provides definitive
results, and primary studies tend to overestimate treatment effects.
A recent study reported that 32% of a set of studies with at least 1,000
citations were either contradicted by or reported stronger effects than
subsequent studies (Ioannidis, 2005).
Ioannidis JPA. Why Most Published Research Findings Are False. PLoS Medicine.
2005;2(8):e124. doi:10.1371/journal.pmed.0020124.
11. Advantages of meta-analysis
- Larger power (sample size)
- Reproducible (systematic review protocol)
- A small but clinically significant effect size might be
apparent in meta-analysis
12. Disadvantages of Meta-analysis
- Complicated statistical analysis
- Heterogeneity among studies
- Difference in outcome reporting