2. USE OF MOBILE
PHONES LEAD TO
NEUROMASGERMANY
DRINKING RED WINE
IS BENEFICIAL TO
HEALTH-US STUDY
CHOCOLATES
ARE HARMFUL
SOFT DRINKS
ARE SAFE
SOFT DRINKS
CONTAIN HARMFUL
PESTICIDES
RED WINE IS
HARMFUL-US
CHOCOLATES ARE
GOOD FOR HEART
MOBILE PHONES
CAUSE NO HARMEUROPE
4. Meta-analysis
A meta-analysis is a quantitative study wherein a
set of statistical procedure is used to summarize
and synthesize results of a number of
independently conducted research studies. If
done well it can be very valuable to a researchers
, because it provides an extensive bibliography of
existing research on a topic ,while also providing
a combined analysis of a number of study.
Meta-analysis
is
a
very
time
consuming, understanding and usually conducted
by a team of researchers.
5. Thus, Meta Analysis is A Quantitative approach
For
systematically combining
the results of
previous researches
to
arrive at conclusions
about
the body of research.
6. What does it mean?
Quantitative
: numbers
Systematic
: methodical
Combining
: putting
together
Previous research
: already
done
Conclusions
: new
7. Important
1- MA falls under a broader classification of
reviews known as Systematic Reviews
2- There are two types of SRs –
(a) Qualitative
(b) Quantitative (meta-analysis)
8. Important
Both follow the same rigorous steps, EXCEPT that
a qualitative review does not combine the
endpoints for statistical analysis, usually
because it’s not appropriate to combine them
into any type of common Metric.
9. The Etymology
"Meta" implies something occurring
later, more comprehensive.
Alternative terms are less specific — for
example, "overview" is also used for
traditional reviews, & "pooling" incorrectly
implies that the source data are merged .
10. Historical notes
Karl Pearson (1904) - Use of formal
techniques to combine data from
different samples.
Glass (1976) coined the term metaanalysis
11. Rationale
“by combining the samples
of the individual studies,
the overall sample size is increased, thereby
improving the statistical power of the analysis
as well as the
precision of the estimates
of treatment effects”
12. Objectives
The benefits or hazards that might not
be detected in small studies can be
found .
Integrating the findings.
Identifying the treatment effect (or
effect size) when it is consistent from
one study to the next.
Identifying the reason for the variation
when the effect varies from one study
to the next .
13. Importance MA
Too much scientific information.
A Well-informed clinical decision is difficult to
reach, time consuming & cost-ineffective.
Decisions about the utility of an intervention or
the validity of a hypothesis can’t be based on
results of a single study.
Information from many studies with less effort
& hassle .
14. Advantages of MA
Saves effort and time
Increases sample size – Gain in
statistical power by reducing Random
errors.
Enhances reliability (precision) &
accuracy (validity).
Explores & Reduces bias.
15.
Resolves controversies &conflicting
reports
Identifies crucial areas & questions
that have not been adequately
addressed with past research.
Generalizes study results.
May explain heterogeneity & its
sources between the results of individual
studies.
16.
Answers questions about whether
an overall study result varies
among subgroups—for
example, among men and
women, older and younger
patients, or subjects with different
degrees of severity of disease.
Reproducible numerical values –
no place of unhelpful descriptors such
as "no relation," "some evidence of a
trend”, "a weak relation," and "a
17. Applications OF M A
Clinicians & applied researchers - to
determine which interventions
work, and which ones work best.
Basic research - to evaluate the
evidence in diverse areas.
Planning new studies.
Some funding agencies now require it
as part of the grant application to fund
new research.
18. CAUTION
Not used or meaningless, when –
1. Studies are different in terms of their
population, intervention or how outcomes were
measured.
2. Treatments, evaluated in the individual
are different.
studies
3. The findings of individual studies differ significantly
– because combining widely differing results to
produce an average effect would fail to represent the
great variation in the outcomes .
19. Types of MA
1. Literature-based MA (LBMA)
- most frequent type of MA
- may be misleading as Data extraction
&
analysis may be less accurate
2. Individual patient data MA (IPDMA)
- Gold-standard, but has problems
- inability of investigators to supply data
- the increased costs
20. Principles of MA
1-The need to consider the totality of evidence.
2-Requirement for Reproducibility
Transparent, explicit & systematic approach.
3-Principles of reliable detection of the effects of
health care interventions
21. The Process
The process simply involves:
1. Calculation of the treatment effect i.e. OR/RR.
2. Calculation of the 95% Confidence interval
around the individual OR/RR.
3. Giving a weight to the individual OR/RR (shown
as the size of the box in the forest plot). The
weight is calculated as the inverse of the square
of the standard error of each OR/RR (1/SE2).
22. How do we conduct MA
1.
2.
3.
4.
5.
6.
7.
Write a protocol – the blueprint
Locating relevant studies
Selecting & appraising studies for inclusion
Data extraction from selected studies
Statistical methods to combine the effect
measures extracted from primary studies
Addressing biases and limitations
Results in graphical form – the Forest Plot
23. How do we conduct MA
1.
2.
3.
4.
5.
6.
7.
Write a protocol – the blueprint
Locating relevant studies
Selecting & appraising studies for inclusion
Data extraction from selected studies
Statistical methods to combine the effect
measures extracted from primary studies
Addressing biases and limitations
Results in graphical form – the Forest Plot
24. Recipe of a good protocol
1.
2.
3.
4.
5.
6.
7.
8.
Purpose of meta-analysis.
Design a research question.
Search for studies.
Specify study selection (inclusion & exclusion) &
appraisal criteria.
Decide data extraction procedures (including statistical
reanalysis).
Select an analytical strategy (use of models & sensitivity
analysis).
Anticipate systematic errors (biases)/limitations.
Present & disseminate results .
25. How do we conduct MA
1.
2.
3.
4.
5.
6.
7.
Write a protocol – the blueprint
Locating relevant studies
Selecting & appraising studies for inclusion
Data extraction from selected studies
Statistical methods to combine the effect
measures extracted from primary studies
Addressing biases and limitations
Results in graphical form – the Forest Plot
26. Locating relevant studies
Systematic approach.
Primary objective – Strategically locate as
much of the completed research on the topic
as possible.
Document strategy in sufficient detail to allow
others to critique it’s quality.
Usually include e-databases
(MEDLINE, CINAHL, Psyclit, Embase, Cochrane
Library) and others
27. How do we conduct MA
1.
2.
3.
4.
5.
6.
7.
Write a protocol – the blueprint.
Locating relevant studies.
Selecting & appraising studies for inclusion .
Data extraction from selected studies.
Statistical methods to combine the effect
measures extracted from primary studies
Addressing biases and limitations.
Results in graphical form – the Forest Plot.
28. Selecting & appraising studies for
inclusion
Selecting - Judge the relevance of the studies to the
review question.
Appraising - Judge numerous features of design &
analysis .
Methodical, impartial and reliable strategies are
necessary as MA are retrospective exercises & are
therefore susceptible to both random & systematic
sampling errors .
Rationale - by excluding lesser quality studies
the risk of error/bias will be lessened.
29. How do we conduct MA
1.
2.
3.
4.
5.
6.
7.
Write a protocol – the blueprint.
Locating relevant studies.
Selecting & appraising studies for inclusion.
Data extraction from selected studies.
Statistical methods to combine the effect
measures extracted from primary studies.
Addressing biases and limitations.
Results in graphical form – the Forest Plot.
30. Data extraction
Eligibility criteria for inclusion of data .
Data collection in standardized record form.
2 independent observers extract the data, to
avoid errors.
Blinding observers to the names of the
authors, their institutions, the names of the
journals, sources of funding, and
acknowledgments leads to more consistent
scores.
31. How to Extract Data
Create a spreadsheet (Excel)
For each study, create the following columns:
name
of the study
name of the author, year published
number of participants who received intervention
number of participants who were in control arm
number who developed outcomes in intervention
number who developed outcomes in control arm
32. We got like 22 studies to do our meta analysis, after all
33. How do we conduct MA
1.
2.
3.
4.
5.
6.
7.
Write a protocol – the blueprint
Locating relevant studies
Selecting & appraising studies for inclusion
Data extraction from selected studies
Statistical methods to combine the effect
measures extracted from primary studies
Addressing biases and limitations
Results in graphical form – the Forest Plot
34. Statistical Methods
They attempt to answer basic questions:
(a) Are results of different studies similar?
-Check for heterogeneity.
(b) To what extent that they are similar?
-Calculate the amount of heterogeneity.
(c) What is the best overall estimate?
- Combine the effect measures using suitable model
& calculate the summary effect size & its CI.
(d) How precise & robust is this estimate?
- Do Sensitivity Analysis.
(e) Finally, can dissimilarities be explained?
35.
Heterogeneity
if
present, should not simply be
ignored after a statistical test is
applied; rather, it should be
scrutinized and explained.
More weight is given to –
(a) larger trials
(b) Studies with narrow CI
36. • Assess the heterogeneity of effect size
across the studies
• Decide the type of model for combining
the effect size of all studies.
• 2 models to adjust the potential
confounding effects of study –
(1) Fixed Effect model .
(When the combined trials are a homogeneous set)
(2) Random Effect model.
(When heterogeneity is detected)
37. How do we conduct MA
1.
2.
3.
4.
5.
6.
7.
Write a protocol – the blueprint.
Locating relevant studies.
Selecting & appraising studies for inclusion.
Data extraction from selected studies.
Statistical methods to combine the effect
measures extracted from primary studies.
Addressing biases and limitations .
Results in graphical form – the Forest Plot.
39. Publication Bias
Synthesis of published data can yield an
exaggerated effect as studies that yield relatively
large/beneficial treatment are more likely to
publish .
English language bias, citation bias, & multiple
publication bias- In English ,studies are more likely
to be cited, and more likely to be published
repeatedly.
40.
High likelihood of publishing –
- Studies sponsored by government or NGO.
- Multi-centric studies.
Many authors may not submit studies with
negative findings because they anticipate
rejection.
41. Other Biases
Selection/Inclusion Bias-Manipulation of the
inclusion criteria could lead to selective inclusion
of studies with positive findings.
Database bias – Selective inclusion of studies
from developed countries.
Citation bias – Ease of locating and contacting
authors from reference lists.
Data provision Bias – due to willingness of
investigators to make their data available.
42. Testing for bias – Funnel Plot
The presence of bias should be examined in
sensitivity analyses and funnel plots.
Funnel plot – It is graphical test for any type of
bias that is associated with sample size.
Results from small studies will scatter widely at the
bottom of the graph. The spread will narrow as
precision increases among larger studies.
In the absence of bias, the plot should thus
resemble a symmetrical inverted funnel.
If the plot shows an asymmetrical & skewed
shape, bias may be present .
43. Funnel Plot: what & how to read
To study a funnel plot, look at
its LOWER LEFT
corner, that’s where negative
or null studies are located
If EMPTY, this indicates
“PUBLICATION BIAS”
Note that here, the plot fits in
a funnel, and that the left
corner is not all that
empty, but we cannot rule out
publication bias
44.
45. How do we conduct MA
1.
2.
3.
4.
5.
6.
7.
Write a protocol – the blueprint.
Locating relevant studies.
Selecting & appraising studies for inclusion.
Data extraction from selected studies.
Statistical methods to combine the effect
measures extracted from primary studies.
Addressing biases and limitations
Results in graphical form – the Forest Plot.
46. Forest plots
Effective way of presenting results:
Studies, effect sizes, confidence intervals
Provides an overview of consistency of
effects
Summarizes an overall effect (with
confidence interval)
Useful visual model of a meta-analysis
47. Anatomy of a forest plot…
Study effect size (with C.I.)
N of study
Line of no
effect
C.I
Studies
Weighting
of study
in metaanalysis
Study
effect
size
Pooled
effect
size
Pooled effect size
48.
When individual studies are inconclusive
deficient or its not possible to do Multicentric
RCT , Money problem ,Time nahi mila.
49.
50.
51.
52.
53.
54. Weighting studies
56
More weight to the studies which give us
more information
More participants
More events
Lower variance
Weight is closely related to the width of the
study confidence interval: wider confidence
interval = less weight
55. EFFECT OF β-BLOCKADE AFTER MI
1-
3+
SG1 = MIX POP
5+
SG3 =
GERMAN POP
2+
SG2 = MIX POP
4+
SG5 = MIX POP
SG4 = MIX POP
META ANALYSIS
BENEFICIAL EFFECT OF β-BLOCKADE
13+
56. Limitations of MA
Can’t improve the quality or reporting of the
original studies.
Limitations arising from mis-applications:
- when study diversity is ignored or
mishandled , and
- when variability of patient populations’,
quality of data & potential for underlying
biases are not addressed.
Publication bias is a major limitation.
57.
Some clinicians consider it as "a tool that has
become a weapon” & which represents "the
unacceptable face of statisticism" & "should be
stifled at birth”
At the other end of the spectrum, the application
has been hailed “Newtonian”.
Some reject and see it as "mega-silliness”
58. MA continues to be controversial
technique ?
The mixed reception is not surprising
The pooling of results from a particular set of
studies may be inappropriate from a clinical
point of view, producing a population "
average" effect.
Meta-analyses of the same issue may reach
opposite conclusions.
59. Still, Meta Analyses hold promise….
If original studies of the effects of clot busters
after heart attacks had been systematically
reviewed, the benefits of therapy would have
been apparent as early as the mid-1970s.
Traditional approaches were inadequate in
summarizing the current state of knowledge &
omitted mention of effective therapies.
60. Popularity of Meta Analyses
3000
2500
Number of Publications
2000
1500
1000
500
0
93-94
94-95
95-96
96-97
97-98
98-99
99-00
Year of Publications
2000-1
2001-2
2002-3
2003-4
61.
62. Refrences
Lipsey, M.W., Wilson, D. B. Practical meta-analysis. Thousand Oaks, CA Sage; (Applied Social
Research Methods Series; 49), 2001.
Petitti, D. B. Meta-analysis, decision analysis, and cost-effectiveness analysis: Methods for
quantitative synthesis in medicine (2nd ed.). New York Oxford University Press; 2000.
Sutton AJ, Abrams KR, Jones DR, Sheldon TA, Song F. Methods of meta-analysis in
medical research. West Sussex, England Wiley, 2000. The Handbook of Research SynthesisHarris Cooper & L Hedges
Statistical Methods for Meta-Analysis – L Hedges & I OlkinLike
Practical Meta-Analysis - Mark Lipsey and David
Systematic Reviews in Health Care: Meta-Analysis in context - M Egger, G Davey-Smith, D G
Altman, Foreword by Iain
Methods for Meta-Analysis in Medical Research- AJ Sutton, K R Abrams, DR Jones, TA Sheldon, F
SongLike
Meta-Analysis in Medicine and Health Policy - DK Stangl, DBerry
Publication Bias in Meta-Analysis - H Rothstein, A Sutton, M Borenstein
How Science Takes Stock: The Story of Meta-Analysis - Morton
Methods of Meta-Analysis: Correcting Error and Bias in Research Findings - John E. Hunter and
Frank L. Schmidt
Synthesizing Research - A Guide for Literature Reviews - Harris
Meta-Analysis of Controlled Trials - Anne Whitehead