METAANALYSIS

1
META-ANALYSIS

• ACHMAD SURJONO
• Child Health Dept. Faculty of Medicine
• Gadjah Mada University, Yogyakarta

2
Integrative primary studies

• Narrative /literature review
• Systematic review/ Overview
• Meta-analysis

3
Narrative review

• Non-systematic
• in gathering relevant studies
• no sufficient appraisals
• Prone severe bias error
• cite only studies support (L Pauling- vit C)
• vote prones and cons study
• Still valuable (pathophysiology, basic sciences)

4
Meta-analysis

• Quantitative approach, systematically combining
  results using statistical method
• Consistentgeneralized across population,
  settings, Tx variations, subset
• Explicit method to limit bias
• A research study analytic-retrospective

5
Tabel 1. Defferences between Narrative Reviews
and Systematic Reviews
6
Why or rationale MA

• Quantities information to be reduced
• Integrate the critical information
• Efficient scientific technique
• Establish generalisability of findings
• Assess consistency of relationship
• Explain inconsistency and conflict data

7
Cont-rationale

• Increase power
• Increase precision effect estimation
• Improve accuracy (vs narrative)
• -replication
• -detect small important effect
• -continuous update - cumulative

8
(No Transcript)
9
(No Transcript)
10
Who need MA

• Clinicians
• Researchers
• Policy makers
• Consumers
• Reviewers/editorials

11
Scope systematic review/MA

• Therapy/ RCT
• Observational studies
• Diagnostic - screening
• Prognostic
• Others

12
Strength of evidence

• Tx
• Systematic review/MA
• RCT
• Non randomized, cohort, case series, case-control
• Descriptive, peer expert opinion
• Personal experience

13
Anatomy of MA

• Introduction
• explicit clinical problem
• justification
• objectives/hypothesis

14
Cont methods

• Method
• Eligible criteria inclusion exclusion
  (population, Tx, outcome, design)
• Searching resources (database, agencies, files,
  expert, hand) and any restrictions (year publ.,
  status publ., language)
• Validity assess criteriaprocess used (masked,
  quality assess, and the findings)
• Data abstraction process used (complete
  independent, duplicate)

15
Cont methods

• Study characteristic type study design,
  participant charact., detail Tx, outcome
  definition etc assess clin heterogeneity
• Synthesis effect measures (OR,RR), combining
  data (statisticsCI), handling missing data, how
  stat heterogeneity, a priory sensitivity,
  sub-group analysis, assess publication bias

16

• Results
• Trial flow profile summarizing flow
• Study charact data each trial (age, sample
  size, intervention, dose, duration f-up)
• Synthesis agreement on selection validity,
• summary results (each trial for each
  outcome), data needed to calculate effect CI
  in intention-to-treat)

17

• Discussion
• Summarize key finding, clinical inferences based
  int ext validity, interpret in available
  evidence, potential bias, practical
  implication,future research agenda

18
Steps in MA

• 1. Definition research question
• 2. Methods identifying articles
• 3. Selection articles for inclusion
• 4. Quality appraisal of included studies
• 5. Extraction of data
• 6. Synthesis of data interpret

19
Steps in conducting MA

• 1. Formulate review questions (PICO)
• 2. Define inclusion exclusion criteria
• (study design, method quality)
• 3. Locate studies data base, on-line, reference
  list, hand searching, expert
• 4. Select studies eligibility by gt1 person,
  solving disagreement, keep the excluded

20
Jadad validated quality scale

• 1. Described as randomized?
• 2. Described as double blind?
• 3. Description of withdrawals and drop outs?
• a.Yes1 , No0
• b.Add 1 if R / blinding appropriate
• c.Deduct 1 if R/ blinding inappropriate
• Scoring range 0 5 Poor quality lt3

21

• 5. Appraise study quality
• assess gt1 person use simple checklist blinding
  to authors, institution, journals
• 6. Extract data designpilot form by gt1 person
  blinding observers
• 7. Analyse present results tabulation results
  forest plot source heterogeneity combined all
  trials subgroups perform sensitivity analysis,
  funnel plot list excluded studies
  

22
Cont Steps

• 8. Interpret results
• limitations, incl publication other bias
• strength of evidence
• applicability
• NTT to benefit/harm
• economic implications
• implication for future research

23
Effect assessment

• Risk, Odds
• Risk Ratio, Odds Ratio
• Absolute Risk Reduction, Relative RR
• NNT(B), NN(T)H
• Mean difference
• Precision 95 confidence interval

24
Assess heterogeneity

• Real difference between studies
• Test
• statistics Q/Chi square
• graphical forest plot/ blobbogram,
• z-score, radial, LAbbe plot

25
Possible causes

• -due to chance
• -spurious due scale used
• -tx characteristic
• -patient level covariate
• -design conduct
• -unexplainable
• adjust the analysis

26
Identification publication bias

• Funnel plot
• Rank correlation test
• Linear regression test

27
(No Transcript)
28
Combining study

• Many techniques, still controversial
• Focus on invariability
• A. Fixed effect models homogeneity
• general inverse variance-weighted
• specific MH, Peto, maximum-likelihood

29
Combining studies, cont

• B. Random effect models consider intra and
  intervariability heterogeneity
• Different studies estimating different but
  related effects
• DerSimonian and Laird method
• C. Bayesian method (less popular)

30
Sensitivity test was the results robust?

• The more confident of the results
• Similar result as analysis by changing inclusion
  criteria, lower quality articles, unpublished
  studies, publication bias, missing values

31
Evaluating meta-analysis

• A. Validity
• B. Importance
• C. Applicability

32
A. Validity

• Asking focused clinical question
• Appropriate criteria in articles selection
• Likely important studies missed
• Appraised the validity of included studies
• Assessment of studies reproducible
• The results similar from study to study

33
B. Importance

• Was the overall results clinically important
• NNT can be determined by formulas or tables
  (precision estimates)

34
C. Applicability

• Similarity your patients and reviewed patients
• Your local setting
• Patients preference

35

• THANK YOU
• MATUR NUWUN

36
The QUORUM

• Title Identify MA of RCT
• Abstract Use structure format
• Objective Clinical question explicitly
• Data source List databases and others
• Review method Selection criteria (population,
  Tx, outcome, design), methods validity assess,
  data abstraction, study characteristics, data
  synthesis
• Results Characteristic RCT included-excluded,
  quali-quanti- tative findings (point estimates
  CI)
• Conclusion The main results

37
(No Transcript)
38
(No Transcript)
39
(No Transcript)