A simple method to estimate prediction intervals and predictive distributions: Summarizing meta-analyses beyond means...

Area of Science:

• Biostatistics
• Epidemiology
• Evidence Synthesis

Background:

• Meta-analyses are crucial for evidence synthesis, typically presenting means via random-effects models.
• Confidence intervals offer incomplete heterogeneity summaries; prediction intervals are recommended for future study effect ranges.
• Conventional prediction intervals often assume normality of heterogeneity, which may not be accurate.

Purpose of the Study:

• To introduce a simple, nonparametric method for estimating prediction intervals and predictive distributions.
• To provide a ready-to-use spreadsheet tool for implementing the nonparametric method.
• To demonstrate the advantages of this novel approach over conventional methods.

Main Methods:

• Developed a nonparametric approach for calculating prediction intervals and predictive distributions.
• Utilized simulation studies to compare the new method with conventional techniques.
• Applied the method to a real-world meta-analysis on tuberculosis vaccination efficacy.

Main Results:

• The nonparametric method provides approximately unbiased estimates compared to conventional approaches.
• Simulation studies validate the accuracy and reliability of the new method.
• The nonparametric predictive distribution offers richer insights into the underlying distribution's shape.

Conclusions:

• The proposed nonparametric method enhances prediction interval estimation in meta-analysis.
• This approach offers a more informative summary of heterogeneity than standard methods.
• The tool is valuable for researchers conducting meta-analyses, improving the interpretation of evidence.