Meta-analyses and systematic reviews

Systematic Review & Meta-Analysis - Evidence Synthesizers

• Systematic Review: A qualitative synthesis of all high-quality evidence on a focused question using rigorous, predefined methods to minimize bias.
• Meta-Analysis: A quantitative technique that statistically combines results from multiple studies to produce a single pooled estimate, increasing precision and power.

⭐ Heterogeneity (variability among study outcomes) is key. It's measured by the $I^2$ statistic. An $I^2$ value > 50% indicates substantial heterogeneity, questioning the validity of pooling results.

The PRISMA Method - Blueprint for a Review

• PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) provides a standardized, evidence-based checklist and flow diagram for reporting.
• Its goal is to ensure clarity, transparency, and completeness in systematic reviews and meta-analyses, enhancing reproducibility and critical appraisal.

⭐ Following the PRISMA checklist is crucial for reducing the risk of reporting bias and is often a submission requirement for high-impact journals.

Forest Plots - Seeing the Big Picture

• A forest plot graphically summarizes individual studies in a meta-analysis, showing both individual and pooled results.

• Key Components:

  • Squares: Represent the point estimate (e.g., RR, OR) of each study. The size of the square is proportional to the study's weight.
  • Horizontal Lines: Show the 95% Confidence Interval (CI) for each study.
  • Vertical Line: The line of no effect (e.g., at OR=1). If a CI line crosses it, the study's result is not statistically significant.
  • Diamond: Represents the pooled result of all studies. Its width is the pooled CI.

⭐ If the diamond (the pooled result) does not touch or cross the vertical line of no effect, the overall result of the meta-analysis is statistically significant.

• Heterogeneity: Assessed by the I² statistic; visually, significant overlap of CIs suggests low heterogeneity.

Bias & Bumps - Gauging Review Quality

• Heterogeneity: Are studies too different to combine?
  • Assess with:
    • Forest Plot: Visually check for non-overlapping confidence intervals.
    • Statistics: Cochran's Q test & the $I^2$ statistic.
  • $I^2$ Interpretation:
    • <25%: Low
    • 25-75%: Moderate

    • 75%: High heterogeneity → Use a random-effects model.

• Publication Bias: Are negative or small studies missing?
  • Assess with: Funnel Plot.
  • Interpretation:
    • Symmetrical plot (inverted funnel) → Low bias risk.
    • Asymmetrical plot → High bias risk.

⭐ Asymmetry in a funnel plot is most famously due to publication bias, but can also arise from true heterogeneity where effect size differs by study size (e.g., smaller studies show larger effects).

• Systematic reviews provide a qualitative summary, whereas meta-analyses use quantitative methods to pool data and ↑ statistical power.
• Forest plots visually summarize results, with a diamond representing the pooled effect estimate and its confidence interval.
• A major limitation is publication bias, the tendency to publish only studies with positive findings; a funnel plot can help detect this.
• Heterogeneity (inter-study variability) is measured by the I² statistic.