Systematic review

Systematic review

1. Highest hierarchy of the clinical evidence

   • Systematic reviews
   • RCTs
   • Observational
   • Consensus
   • Case report
   • Expert opinion

2. Meaning

   • Meta-analysis
     • Biostatistics technique that pools all results (> 2 studies)
       • Reported as Forest plot
   • Scoping review
     • Identify types of literature, key concepts, gaps of knowledge
       • Precursor of systematic review
   • Narrative review
     • Explain mechanisms of disease/development of technique
       • Risks of selection bias, specific use
   • Systematic review
     • Methods including analysis and summary of evidence

3. Characteristics

   • Several steps to decrease risks of bias
     • Bias = systematic errors
       • Define risks of bias to prevent misinterpretation
         • Cochrane RoB
           • Review author judgement
             • Low
             • High
             • Unclear
               • Not enogh information
           • 7 domains
             • Random sequence generation
               • Decrease selection bias
                 • Low risk: unpredictable
                 • High risk: predictatble
             • Allocation concealment
             • Blinding participants/personnel
               • Decrease performance bias
               • Fail: high risk
             • Blinding outcome assessment
               • Decrease detection bias
               • Some objective outcomes are not affected: low risk
               • Subjective outcomes
             • Incomplete outcome data
               • Decrease attrition bias
               • Missing data
                 • How many?
                 • From what?
                 • How to analyze?
                   • Intention-to-treat
                     • Prefered for efficacy
                   • Per-protocol
                     • Preferred for adverse events
                   • As treated
                     • High risk of bias
                   • Imputation
                     • Need some assumption
                       • Consult statistician
             • Selective reporting
               • Decrease reporting bias
                 • Different between protocol/method and reported results
             • Others
               • Include
                 • Non-parallel RCT, baseline imbalance, block without blind, DDx
               • Not include
                 • Sample size, diversity, EC, funding
           • Overall: aggregated effect
             • Selection/subgroup/meta-regression: prespecified
         • Summarized by outcomes, by studies
   • Specific PICOT question
     • Intervention
     • Diagnosis
     • Prognosis
   • Use PRISMA guideline
   • Define inclusion criteria for screening
   • PROSPERO registration
   • Searching strategies
     • At least 2 electronic databases
     • References of searched studies
     • Grey literature
     • Communication to researchers/sponsors
     • No language restriction
   • Duplicate process: at least 2 assessors
     • Kappa > 0.6
     • At least for screening for inclusion, extraction, RoB
   • Results
     • Qualitative (narrative)
     • Meta-analysis
       • Increase effect size, precision
       • Risks of overestimation/contaminated from poor study
       • Heterogeneity: I2 fixed/random effect
   • Summary of findings
     • Confidence of evidence
       • Veryl low/low/moderate/high
       • GRADE system
         • Decreased if...
           • Risks of bias
           • Inprecision
           • Inconsistency
           • Indirectness
           • Publication bias
             • Funnel plot
         • Increase if...
           • Large effect
           • Dose-response effect
           • Demonstrating plausible confounding (opposite direction)
       • Starting GRADE
         • RCT: high, moderate
         • Observational: low/very low

4. Applicability

   • Patients' values and preferences
   • Before giving recommendation
   • Benefits and harms
   • Implementation (health care rescource)
   • Feasibility
   • Acceptablity

5. Conclusion

   • PICO: critical steps
   • Teamworks
   • Meta-analysis if applicable only
   • Confidence: GRADE system