2. META ANALYSIS
• Analysis of analyses
• Meta-Analysis is the process of using statistical methods to combine the
results of different studies.
• It involves the systematic, organized, and structured evaluation of a
problem of interest using information (commonly in the form of
statistical tables or other data) from a number of independent studies of
the problem.
• The motivation of a meta-analysis is to aggregate information in order to
achieve a higher statistical power for the measure of interest, as
opposed to a less precise measure derived from a single study.
• In performing a meta-analysis, an investigator must make choices many
of which can affect its results, including deciding how to search for
studies, selecting studies based on a set of objective criteria, dealing
with incomplete data, analyzing the data, and accounting for or choosing
not to account for publication bias.
3. Steps in a meta-analysis
1. Formulation of the problem
2. Identify studies with relevant data
3. Define inclusion criteria for studies
4. Statistical Analysis
4. Formulation of the problem
1. Specify the Research Objectives
2. Review the Environment or Context of the
Research Problem
3. Explore the Nature of the Problem
4. Define the Variable Relationships
5. Identify studies with relevant data
• Meta-analysis requires a comprehensive search
strategy which interrogates several electronic
databases (for example, MEDLINE, EMBASE,
Cochrane Central Register of Controlled Trials).
• Hand searching of key journals and checking of
the reference lists of papers obtained is also
recommended.
• The search strategy the key terms used to search
the database needs to be developed with care.
6. Define inclusion criteria for studies
Should be determined in advance, to reduce
investigator bias.
Inclusion criteria involves
• Types of studies included (case control, cohort, etc)
• Years of publication covered
• Languages
• Restrictions on sample size
• Definition of disease, exposures
• Confounders that must be measured
• Dose response categories similar
7. Statistical Analysis
Two types of models are used to produce
summary effect measures
• Fixed Effect Model
• Random Effects Model
8. Fixed Effects Model
• In a fixed effect model, we assume that the effect sizes in our
meta-analysis differ only because of sampling error and they all
share a common mean.
• Our effect sizes differ from each other because each study used a
different sample of participants – and that is the only reason for
the differences among our estimates
• The fixed effect model provides a weighted average of a series of
study estimates.
• The inverse of the estimate’s variance is commonly used as study
weight, such that larger studies tend to contribute more than
smaller studies to the weighted average.
• When studies within a meta-analysis are dominated by a very
large study, the findings from smaller studies are practically
ignored . Most importantly, the fixed effects model assumes that
all included studies investigate the same population, use the same
variable and outcome definitions, etc.
9.
10. Random effects model
• In a random effects model, we assume two components
of variation:
– Sampling variation as in our fixed-effect
model assumption.
– Random variation because the effect
sizes themselves are sampled from a
population of effect size.
• In a random effects model, we know that our effect sizes
will differ because they are sampled from an unknown
distribution.
• Our goal in the analysis will be to estimate the mean and
variance of the underlying population of effect sizes .
11.
12.
13.
14.
15. Advantages
• Results can be generalized to a larger population.
• The precision and accuracy of estimates can be improved
as more data is used. This, in turn may increase the
statistical power to detect an effect.
• Inconsistency of results across studies can be quantified
and analyzed. For instance, does inconsistency arise from
sampling error, or are study results (partially) influenced
by between study heterogeneity.
• Hypothesis testing can be applied on summary estimates.
• Moderators can be included to explain variation between
studies.
• The presence of publication bias can be investigated.
16. Limitation
• Meta-analysis has a high chance of being perceptable
to publication bias, this is where the researcher
collecting the data will pick specific studies that only
provide the outcome that the researcher is looking
for.
• Publication bias here can be done purposefully to
manipulate results or by accident through
unconsciously knowing that it is being done as they
are unaware that they are looking for a certain
outcome.
• Poor designs are also mixed with good ones which
can skew the statistical result.
17. Overweight, Obesity, and Incident Asthma
A Meta-analysis of Prospective Epidemiologic Studies
• AMERICAN JOURNAL OF RESPIRATORY AND
CRITICAL CARE MEDICINE VOL 175 2007
• Beuther and Sutherland: Obesity and Asthma
Incidence
18. Objective of study
• To quantify the relationship between
categories of body mass index (BMI) and
incident asthma in adults and to evaluate the
impact of sex on this relationship.
19. Studies with relevant data
• Targeted studies were those in which the relationship
between BMI and incident asthma was evaluated.
• MEDLINE
• Cumulative Index to Nursing and Allied Health
Literature
• International Pharmaceutical Abstracts
• all Evidence-Based Medicine Reviews (EBMR)
(Cochrane Database of Systematic Reviews, ACP
Journal Club, Database of Abstracts of Reviews of
Effects, and Cochrane Central Register of Controlled
Trials)
20. • A date range of 1966 to May 2006, crossing
keywords
• overweight and asthma
• Obesity and asthma
• body mass index and asthma
• body weight and asthma
• anthropometry and asthma
• The systematic search yielded 2,006 total
references of which 1,569 were unique.
21. Inclusion criteria
Predetermined inclusion criteria included
• (1) adult subjects,
• (2) primary outcome of incident asthma,
• (3) use of BMI as a measure of overweight or
obesity,
• (4) minimum 1-year follow-up,
• (5) follow-up of at least 70%, and
• (6) data that could be categorized by standard
ranges of BMI obtained at study inception.
22.
23. Statistical Analysis
• Stata 7.0 was used to generate summary ORs
using inverse variance-weighted randomeffects
meta-analysis.
• Random effects methodology was chosen to
account for within-study and between-study
variation.
• Heterogeneity of data was evaluated using the Q
statistic .
• Summary ORs were represented as a point
estimate and 95% confidence intervals in
aggregate and stratified by sex.
24. Result
Total Men Women
Comparison OR (95% CI ) OR (95% CI ) OR (95% CI )
Overweight vs.
normal BMI
1.38 (1.17–1.62) 1.44 (1.01–2.04) 1.42 (1.18–1.72)
Obese vs. normal
BMI
1.92 (1.43–2.59) 1.63 (0.92–2.89) 2.30 (1.88–2.82)
Overweight and
obese (BMI 25) vs.
normal BMI
1.51 (1.27–1.80) 1.46 (1.05–2.02) 1.68 (1.45–1.94)
Obese vs.
overweight
1.49 (1.20–1.85) 1.17 (0.66–2.07) 1.58 (1.25–1.99)
25. Limitation
• Many of the studies included in this meta-analysis relied
on self reported, these reports cast some doubt on the
validity of self-reported asthma in large epidemiologic
studies. It is reasonable to believe that some of these
patients with “asthma” may have respiratory symptoms
due to obesity but may not meet rigorous objective
physiologic criteria for asthma .
• It is possible that asthma may be over diagnosed in an
obese population.
• The calculated ORs may have been underestimated due to
the grouping together of underweight and normal weight
subjects.
MEDLINE (Medical Literature Analysis and Retrieval System Online, or MEDLARS Online) is a bibliographic database of life sciences and biomedical information.
Embase (often styled EMBASE for Excerpta Medica dataBASE) is a biomedical and pharmacological database of published literature designed to support information managers and pharmacovigilance in complying with the regulatory requirements of a licensed drug.
The Cochrane Central Register of Controlled Trials (CENTRAL) is a bibliographic database that provides a highly concentrated source of reports of randomized controlled trials.
In statistics, an effect size is a quantitative measure of the strength of a phenomenon. Examples of effect sizes are the correlation between two variables, the regression coefficient, the mean difference, or even the risk with which something happens, such as how many people survive after a heart attack for every one person that does not survive.
The q statistic or studentized range statistic is a statistic used for multiple significance testing across a number of means.