Unit 4
SYSTEMATIC REVIEW
METHODS
1. Identify and determine appropriate methods for data synthesis
2. Differentiate between clinical and statistical heterogeneity
3. Identify the components assessed in applying GRADE to systematic reviews 4. Identify strategies for communicating
the results of a systematic review
OBJECTIVES: UNIT 4
Descriptive summary
Synthesis
Narrative synthesis
Meta-analysis
SYNTHESIS
Clinical heterogeneity
Population
Intervention
Outcome
Statistical heterogeneity
Other
SYNTHESIS DECISION
Degree to which there is variation between studies beyond chance
Informs
synthesis decision
statistical model for meta-analysis
Tests
χ 2 (chi-squared)
I 2
STATISTICAL HETEROGENEIT Y
Practice
Heterogeneity Assessment
IN-CLASS ACTIVIT Y
Intervention Review Question
What are the effects of mobile phone messaging reminders for attendance at healthcare
appointments?
Clinical Heterogeneity
p. 10-11
p. 25-35
Statistical Heterogeneity
p. 13-15
IN-CLASS ACTIVIT Y: HETEROGENEIT Y
Avoid reporting study by study
Direction, size, and consistency of the effect
Overall assessment of the strength of the evidence
NARRATIVE SYNTHESIS
1. Summary statistic to describe intervention effect is calculated for each study
Description of effect
Dichotomous
Continuous
2. Summary (pooled) intervention effect
estimate across studies is calculated as a weighted average of the intervention effects estimated in the individual studies
Statistical method
Fixed effects
Random effects
META - ANALYSIS
Dichotomous
1. Consistency
2. Ease of understanding 3. Other
Continuous
Measurement of outcome
DESCRIPTION OF EFFECT
Fixed-Effects Random-Effects
Conceptual Considerations
• Assumes effects are the same in all studies
• e.g. Among
these X studies and no others, what is the
impact of the intervention(s) on the outcome?
• Assumes effects differ across studies and the pooled estimate is the mean effect
• e.g. Among all studies, of which these X studies are a random sample, what is the impact of the
intervention(s) on the outcome?
Statistical
Considerations
• within-study variance
• within-study and between- study variances
Practical
Considerations
• Narrow CI
• Large studies
have much more weight than
small studies
• Wider CI
• Large studies have more weight than small studies, but the gradient is smaller than in fixed-effects model
Guyatt, G., Rennie, D., Mead, M. O., Cook, D. J. Users' Guides to the Medical Literature: A Manual for Evidence-Based Clinical Practice, 3rd ed.
Different approaches, no one knows the best
approach and (fortunately, rarely) the method can yield noticeable differences in results [ M u r a d et a l . ( 2 01 5 ) Us er s G ui d e to t h e M ed i c a l L i ter at u r e 3
r de d . , p . 5 0 9 ]
Rationale for Fixed
One study is much larger or more trustworthy than one or more smaller studies and yield quite different results
Number of studies is very small (< 5)
Rationale for Random
Uncertainty about the accuracy and applicability of a particular point estimate increasing with increasing variability in results across studies
Interested in not just available studies, but applying them to a wider population
RANDOM VS FIXED MODELS
Dichotomous
Fixed
Inverse variance
If studies are small or low event rates, consider Mantel Hanzel/Peto
Random
DerSimonian and Laird is the most common
Mantel Hanzel
Continuous
Inverse-variance (either fixed or random)
METHODS
Purpose
explore the certainty of pooled results by repeating the analyses having made some changes to the data or methods
if pooled results are not affected by these changes -> higher degree of certainty about the pooled results
Examples
study design
statistical methods
SENSITIVIT Y ANALYSIS
Differ from sensitivity analysis
Produce “new” estimates of effect
Identify in advance, with rationale
Require an adequate number of studies
Magnitude of the difference
SUBGROUP ANALYSIS
Studies with positive effects have a better chance of being published
Always a risk in systematic reviews
Funnel Plot
plot of the intervention effect estimates from individual studies against some
measure of each study’s size or precision
PUBLICATION BIAS
Center for Reviews and Dissemination.(2008). Systematic reviews: CRD’s guidance for
undertaking reviews in healthcare . https://www.york.ac.uk/media/crd/Systematic_Reviews.pdf
System for rating the quality of a body of evidence in systematic reviews and
guidelines
Different “end” points
Systematic reviews “end” after rating the
quality of evidence for outcomes and clearly presenting the results in evidence tables
Guidelines continue to formulating recommendations
GRADE
Risk of bias/Study limitations
Inconsistency
Indirectness
Imprecision
Publication bias
FACTORS FOR DETERMINING
THE QUALIT Y OF EVIDENCE
Focus is on the outcome, not the individual study
Study limitations for RCTs
Lack of allocation concealment
Lack of blinding
Incomplete accounting for patients and outcome events
Selective outcome reporting bias
Other bias
RISK OF BIAS/STUDY LIMITATIONS
Risk of bias
Across studies
Interpretation Considerations GRADE
Assessment of Study
Limitations
Low risk of bias.
Most
information is from studies at low risk of
bias.
Plausible bias unlikely to
seriously alter the results.
No apparent limitations. No serious
limitations, do not downgrade
Unclear risk of bias.
Most
information is from studies at low or unclear risk of bias.
Plausible bias that raises some doubt about the results.
Potential limitations are
unlikely to lower confidence in the estimate of effect.
No serious
limitations, do not downgrade.
Potential limitations are likely to lower confidence in the estimate of effect
Serious limitations, downgrade one level.
High risk of bias.
The proportion of information from studies at high risk of bias is
sufficient to affect the interpretation of results.
Plausible bias that seriously weakens confidence in the results.
Crucial limitation for one criterion, or some limitations for multiple criteria, sufficient to lower confidence in the estimate of effect.
Serious limitations, downgrade one level
Crucial limitation for one or more criteria sufficient to
substantially lower confidence in the estimate of effect
Very serious limitations, downgrade two levels
Cochrane Handbook
Unexplained heterogeneity of results
Never rate “up”
Judge related to relative measures of effect
Consider rating down if
Point estimates vary widely across studies
Confidence intervals (CIs) show minimal or no overlap;
Statistical test for heterogeneity
low P-value
Large I
2INCONSISTENCY
Guyatt et al., (2011).
Journal of Clinical Epidemiology, 64, 1294-1302.