metatrend

metatrend detects temporal trends in effect sizes in cumulative meta-analyses using regression-based methods to evaluate how individual studies influence evolving combined effect estimates.

Key Features:

• Regression-Based Trend Detection: Uses weighted linear regression where the cumulative combined effect size of studies published up to each point is the dependent variable and the publication rank/order is the independent variable.
• Autocorrelation Adjustment with Generalized Least Squares: Incorporates a first-order autoregressive (AR(1)) coefficient within Generalized Least Squares to account for correlation among successive cumulative effect sizes.
• Formal Statistical Testing: Provides formal statistical tests for the presence of temporal trends in cumulative meta-analytic effect sizes.
• Cumulative Aggregation: Aggregates studies cumulatively, updating combined effect sizes as new studies are added.
• STATA Implementation: Implemented as a STATA program.
• Applicability Across Research Areas: Applicable to any domain using cumulative meta-analysis, with demonstrated use in genetic association studies.

Scientific Applications:

• Genetic association meta-analyses: Detects and characterizes temporal changes and potential diminishing of effect sizes across published genetic association studies.
• Temporal evaluation of cumulative evidence: Identifies trends in effect sizes over time to assess how evolving evidence alters combined effect estimates.
• Interpretation of evolving results: Informs assessment of methodological or publication-driven changes in observed effects across successive studies.

Methodology:

Aggregates studies cumulatively updating combined effect sizes, applies weighted linear regression with the cumulative combined effect size as the dependent variable and publication rank as the independent variable, and adjusts for correlation between successive cumulative effect sizes using Generalized Least Squares with a first-order autoregressive (AR(1)) coefficient.