Systematic
Review &
Meta-analysis
Group B12:
Fatimah Mansour 171210009
Shaima Sharaf Aldeen 171210014
Marah Alrahal 171210016
Mayyas khader 171210017
Dana Hendie 171210018
Systematic
01 Review and process
Definition
TABLE Meta-analysis
OF 02 Defenition and forest
plot
CONTEN 03 HOMOGENEITY &
HETEROGENEITY
TS Publication bias
04 funnel polt and how to
avoid it.
Introduction
A literature search often
reveals many studies with
similar aims and hypotheses.
Examining one study in
isolation can mean that we
miss out on findings discovered
by other researchers. Ideally,
all the studies around one
subject area should be collated.
Overview
Attempts to access
and review Is the quantitative
systematically ALL assessment of a
the pertinent articles systematic review.
in the field
Why?
- Complex issue
- To increase precision of estimates
- Resolve discrepancies
Systematic reviews
are the gold-
standard source of
research evidence
in the hierarchy of
research
evidence
01
Systematic
Review
●It’s a process of identifying, assessing
and interpreting all of the pertinent
Systematic
articles and the available research
evidences, to provide answers for a
particular research question.
●It should effectively explain the
research question, search strategy, and
the design of the studies that were review
selected. Then, the results of these
studies are pooled.
●The evidence drawn can therefore
be very powerful and valuable.
definition
●The overall conclusions are
more accurate and reliable
than those of individual studies .
Process
Interpretation of the
data
Quality
assessment
04
Search strategy 03 Analyze and
combine the data
'Extraction' of
Research 02 relevant data
question
01 search for all
relevant data
Defining
a question
Advantages Disadvantages
• By increasing number of
participants, the power • Systematic review is
increases very comprehensive
• Able to see if the effect is and therefore it’s very
consistent across different time consuming.
clinical populations. • it’s a resource
• The well defined intensive. Because It
methology makes the typically requires a
results of the literature team of researchers
less likely to be biased . (either ‘in-house’ or
• Identifies all available contracted to carry out
evidence rather than the work).
‘cherry picking’ of • Findings are only
evidence. relevant to a single
question.
02
Meta-
analysis
What is Meta-analysis?
• Is a quantitative statistical analysis that is applied
to separate but similar experiments of different
and usually independent researchers and that
involves pooling the data and using the pooled
data to test the effectiveness of the results.
• The benefit of this approach is the aggregation of
information leading to a higher statistical power
with a larger sample size.
• This means that the overall estimate of effect is
more likely to be accurate and with a smaller
confidence interval.
• It also allows for smaller studies to be merged into
a larger study giving more comprehensive
analysis.
5- Review and
Key steps: interpretation of the
findings
1- Using statistical
techniques to
combine results
4- Check for
publication bias
2- Calculation of a
pooled estimate effect of
an intervention with its
P-value and confidence 3- Check for variations
interval between the studies
(heterogeneity)
Forest plot
• Is a diagram with a list of studies on the
vertical axis, often arranged in order of
effector chronologically, and the
common outcome measure on the
horizontal axis.
• The results of a meta-analysis are
presented as a forest plot of pooled
results.
• The overall outcome of the meta-
analysis is a diamond shape.
- The center = point estimate
- The width = confidence interval
03
HOMOGENEITY
AND
HETEROGENEI
TY
The aim of the meta-analysis is to summate
the results of similar studies (summary
effect size)
However, we must consider whether we should be
trying to combine them, or whether they differ too
much for
this to be a sensible thing to do
Heterogeneity
Did You Know This?
Clinical Statistical
Homoge Heterogene
heteroge heterogen
neity ity
neity eity
Occurs when the
When studies There is more Occurs when
results of the
have similar variation than the
studies (the true
and consistent would be individuals,
treatment effects
results and expected by interventions
or risk ratios)
any observed chance alone, or outcomes
differ from one
differences are even after chosen in
another in
due to random allowing for studies differ
magnitude or
variation random from one
direction
variation another
significantly
significantly
more than would
be expected by
chance
Testing
heterogene
ity:
mathematical equations
Using Cochran’s Q (chi2, x2) and I2
Statistic
visual representation
Forest plot Galbraith L’Abbé plot
plot
If there is heterogeneity there are a number of
possibilities:
- We could decide not to pool the study estimates
at all
- We could ignore the heterogeneity and analyze the data
using a fixed effect model
- We could explore the heterogeneity and try to
explain it and remove it
- We could allow for the heterogeneity in our
analysis using a random effects model
Investigating sources of heterogeneity
Heterogeneity comes about Meta-regression is a method that
because the effects in the can be used to try to adjust for
populations which the studies heterogeneity in a meta-analysis.
represent are not the same. We It can determine if there is
can look for possible evidence of different effects in
explanations of this in variations different subgroups of trials.
in study characteristics. For example, the use of statins to
For example: different lower cholesterol levels may be
categories of participants, such investigated by a series of trials.
as hospital-based and A meta regression analysis will
community-based patients. We provide information on the role of
investigate these differences statin dosage or duration of
and remove them treatment
Fixed-effects Random effects
model model
This assumes that there is no This allows for between-study
heterogeneity between the studies, variations, ie the effect is not the
ie that the trials are all comparable same in all the studies
and the effect is the same in all the (heterogeneity)
studies HOWEVER, if a random effect
(homogeneity) model is not valid model may result
HOWEVER, if using a fixed effect in a confidence interval which is
model when it is not valid can have unnecessarily wide and a P value
misleading results. which is unnecessarily large.
How do we decide between these
two methods of meta-analysis?
Irrespective of the numerical data, decide whether the
assumption of a fixed effect model is plausible. Could
the studies all be estimating the same effect? If not,
consider a random effects model
If fixed effect assumption is plausible, are the data
compatible with it? If the fixed effect assumption looks
compatible with the data, use a fixed effect model,
otherwise consider random effects
If we consider a random effects model, do the studies
represent a population where the average effect is
interesting? Do we want to pool them? If yes, then use
a random effects model. If no, then do a
narrative review and do not try to get a numerical
estimate
To sum up the difference
between meta-analysis
and Systematic review ?
Systematic Review Meta-analysis
• It is the process of • It is the statistical
undertaking the process synthesizes
review Findings from many
studies which all have
similar issue
• Systematic review • Meta analysis
include meta preferably uses
analysis, but not systematic
always but can not always.
04
Publicatio
n bias
Publicational bias
The tendency to submit or accept studies for publication based on the
direction or the strength of the study findings.
((Positive studies tend to get published more than negative studies))
the exclusion of negative results means that the overall results are
misleading. The over-representation of positive studies in systematic
reviews may mean that the results are biased toward a positive result.
(( there is an overestimation of benefit and underestimation of harm))
There are several methods available to identify publication bias,
including funnel plots, the Galbraith plot and tests such as the Egger’s
test, Begg’s rank correlation test and Rosenthal’s fail-safe N.
Funnel plots
is a graph designed to
check for the existence
of publication bias.
are commonly used
in systematic
reviews and meta-analyses.
In which the scatter plot of
effect size (x axis) against Publication bias
some
In the measure
absence of study size
of publication due to absent of
(y bias,
axis).it assumes that negative result
studies with high
precision will be plotted
near the average, and
studies with low precision
will be spread evenly on
both sides of the average,
creating a roughly funnel-
Avoiding publication bias
The risk of publication bias can be
reduced by identifying and including
unpublished studies, redefining the
inclusion and exclusion criteria
The ‘trim and fill’ method can be used
with funnel plots to correct for
publication bias
Conclusion
• The evidence drawn from systematic reviews can
therefore be very powerful and valuable.
• A forest plot is a graphical display of estimated results
from a number of scientific studies addressing the
same question, along with the overall results.
• Homogeneity: Where Observed differences are due to
random variation (Chance); Heterogeneity: Differences
between studies are not due to Chance.
• We use Visual and Mathematical means for testing
Heterogeneity and
Meta-Regression Analysis for adjusting It.
• Publication bias is one of the reporting biases it results
because positive studies tend to get published more
than negative studies.
THANK
Do you have any questions?
[email protected] you
+91 620 421 838
yourcompany.com
Please keep this slide for attribution
CREDITS: This presentation
template was created by
Slidesgo, including icons by
Flaticon, infographics & images
by Freepik
Refrances
●Gosall, N. and Gosall, G., n.d. The Doctor's Guide To Critical
Appraisal. 4th ed.
●Bland, M., 2015. An Introduction To Medical Statistics. 4th ed.
Oxford: Oxford University Press.
