Updated: 8/26/2020

Systemic Review and Meta-Analysis

Review Topic
  • Systemic reviews synthesize previous studies to produce more precise and generalizable data
    • synthesized studies can be interventional or observational
  • Meta-analysis is the statistical procedure for combining numerical results from synthesized studies
Systemic Review
  • Systematic and procedural method of collecting, evaluating, and synthesizing qualitative or quantitative literature to answer a question
    • methods and selection criteria must be specified in advance
  • May or may not include meta-analysis
  • Statistical method to combine data from multiple studies 
    • weighted average
  • Advantages
    • better precision than individual studies
    • improves generalizability of study findings
    • considered to be the highest level of clinical evidence   
  • Limitations
    • quality of individual studies
    • bias of individual studies
      • must be assessed and accounted for
    • variability in study methods
      • studies must be similar enough to be meaningfully combined
      • e.g., meta-analysis may not be helpful for a study looking to compile evidence about the effects of reading on depression scores if the primary studies all focused on reading different genres 
        • would be more helpful to answer a specific question such as "How does reading fiction influence depression scores?"
    • study heterogeneity
      • when variation in effect size of included studies is greater than would be expected by pure chance
        • e.g. one study demonstrates a massive treatment effect and another demonstrates a negligible one
        • could be due to differences in participant characteristics
          • treatment could have larger impact in certain populations and pooling the results without examining that possibility will obfuscate that
        • statistical concerns
          • high heterogeneity could suggest absence of "true" effect, rendering pooled data meaningless
    • subject to publication bias
      • outcome and statistical significance of a study or experiment influences the likelihood of its publication
      • more papers with positive results are published despite similar quality
      • when present, sampled publications are not a true representation of gathered evidence
      • may lead to increase in false conclusions 
Funnel Plot
  • Evaluates for the presence of publication bias
  • Used in systemic reviews and meta-analyses
  • Plots precision versus results measure (e.g. odds ratio, mean, and relative risk) of included studies
  • Should take on a funnel (triangle) shape 
    • studies with high precision (high value on y-axis) should fall near the average result measure (center of x-axis) 
    • studies with low precision should fall scattered on either side of the average result
  • Deviation from funnel shape indicates publication bias 
Forest Plot
  • Used in systemic reviews and meta-analyses to visually display results from individual studies
  • Vertical "line of null effect" at point on horizontal axis representing no association between exposure and outcome
  • Positive and negative results measures (e.g., odds ratio, mean, and relative risk) of included studies plotted as boxes or shapes on either side of the line 
    • relative size of the shape indicates relative size of the study
    • 95% confidence interval (CI) plotted as a line extending horizontally from each point
      • any CI crossing the vertical line indicates the results are statistically insignificant
  • Combined point estimate (vertical line) and confidence intervals represented by a diamond 

Please rate topic.

Average 5.0 of 1 Ratings

Private Note