Meta Analysis
[Link]
Study sample
Study title
(20-40)
10
50
70
Definition
Meta-analysis is a quantitative approach
for systematically combining results of
previous research to arrive at conclusions
about the body of research
Karl Pearson is the first
medical researcher to use
formal techniques to combine
data from different studies
(1904)
He synthesised data from
several studies on efficacy of
typhoid vaccination
Traditional Systemic
Meta analysis
review review
Author Individual Team Team
Search
Individual Protocol based Protocol based
strategy
Summary
Authors Can be
Results statistical
judgement qualitative
technique
Terms you should
know
Publication bias - When only Published
data is taken which is 5% of the total
actual research
Heterogeneity - Variation between study
results because of any one of the PICO
elements or the difference in the quality
of the studies
Assessing quality of the studies
Jadad scale- Oxford quality scoring system, is a procedure to
independently assess the methodological quality of a clinical
trial
What is the weighting studies
and how it is done
Assessing which studies are more important
How - which give us more information
More participants
More events
Lower variance
Steps in meta analysis
Formulation of the research question
PICO model (Population, Intervention, Comparison,
Outcome)
Search of literature
Selection of studies
Decide which dependent variables or summary measures are
allowed.
Selection of a meta-analysis model
Examine sources of between-study heterogeneity
More deaths in
Adding up the studies gave - the steroids
For example : in the studies with steroids group
adding up the results of the studies - 34.3% in steroids group and 32.5%
301/925 in the control group with no in non steroids group
steroids died and in steroids group Thus the relative risk comes out to
410/1194 died
be - 1.06
Break
s dow
n
rando the powe Why follow so
miza r of
ti o n many steps and
not just add up the
result of previous
studies?
b a la n c e s the trials
Im
a b ias
u c es
d Lower deaths in
Intro the steroids
group
But if the meta analytic methods
implemented - Relative risk was
0.96
Models
Direct evidence: Models incorporating
study effects only
Fixed effects model
Random effects model
IVhet model
Models
Direct evidence: Models incorporating
additional information
Quality effects model
Models
Indirect evidence: Network meta-analysis
methods
Bayesian framework
Frequentist multivariate framework
Generalized pairwise modelling framework
Tailored meta-analysis
Aggregating IPD and AD
Fixed effect models
Fixed effects model assumes that the true
effect of treatment is the same value in each
study (fixed); the differences between studies
is solely due to random error (same
population, use the same variable and
outcome definitions, etc)
Provides a weighted average of a series of
study estimates
Mantel- Haenszel method (Relative risk
and odd’s ratio)
Peto‟s method (combining odds ratios )
Maximum-Likelihood techniques
Exact methods of interval estimation
Random effect models
In random effects models, the treatment
effects for the individual studies are
assumed to vary around some overall
average treatment effect
Weighted least squares
Un-weighted least squares
Maximum likelihood
Restricted Maximum likelihood
Exact approach to random effects of
binary data.