| dc.description.abstract | Meta-analysis has traditionally occupied the highest position in evidence-based medicine because it combines data from multiple studies and appears to provide the most precise summary of treatment effects. Yet precision alone does not guarantee causal relevance. Conventional meta-analysis often synthesizes results across studies that differ in populations, interventions, outcome definitions, adherence patterns, and follow-up structures, while weighting studies primarily by statistical precision rather than relevance to a real-world target population. As a result, pooled estimates may be internally consistent yet poorly aligned with the causal questions. Causally interpreted meta-analysis (CIMA) offers a more rigorous alternative. Anchored in the potential outcomes framework, it shifts evidence synthesis from associational averaging toward estimation of treatment effects in explicitly defined target populations. This approach reframes heterogeneity as effect modification, incorporates transportability and target trial emulation, and supports robust estimation through methods such as the g-formula, inverse probability weighting, augmented inverse probability weighting, and targeted maximum likelihood estimation. It also provides a principled basis for integrating randomized trials, real-world data, and causal machine learning. This viewpoint argues that CIMA should be regarded as a higher form of evidence synthesis than conventional meta-analysis because it prioritizes causal interpretability, external validity, and decision relevance. As medicine moves toward precision care, real-world evidence, and automated analytics, evidence hierarchies must evolve accordingly. The future value of meta-analysis will depend not merely on how efficiently it pools evidence, but on how credibly it informs action in the populations where decisions are made. | |
