JMIR Publications

ABSTRACT

Background:

Systematic reviews and meta-analyses are cornerstones of evidence-based medicine, yet the proliferation of overlapping and redundant secondary evidence syntheses has emerged as a critical meta-research concern. When multiple review teams independently address near-identical clinical questions using overlapping sets of primary studies, the resulting redundancy wastes finite research resources and risks distorting the perceived weight of clinical evidence. Three methodological challenges remain undertheorized: the reproducibility of AI-assisted evidence synthesis tools, the fuzzy boundary between warranted replication and wasteful duplication, and the scientific validity of resource-waste estimates routinely cited in the meta-research literature.

Objective:

This review develops a three-dimensional analytical framework, spanning methodological origin, execution-level primary study overlap, and systemic proliferation drivers, to characterize the sources, scale, and consequences of redundancy in evidence synthesis. It critically evaluates proposed solutions against their implementation barriers and identifies three unresolved conceptual tensions that undermine both empirical quantification and policy design.

Methods:

A narrative synthesis of meta-research studies, meta-epidemiological investigations, and policy analyses published between 2018 and 2026 was conducted, drawing on literature from the evidence synthesis methodology, research-on-research, and clinical epidemiology domains. The review integrates empirical findings on primary study overlap quantified through the corrected covered area metric, meta-research on PROSPERO registration duplication, and analyses of incentive structures driving redundant synthesis production.

Results:

Evidence indicates that methodological redundancy originates from inadequate pre-protocol scoping searches and the absence of automated duplicate detection in registration platforms such as PROSPERO. At the execution level, meta-epidemiological studies using the corrected covered area metric consistently document moderate to high primary study overlap across systematic reviews on the same clinical question, with overlap rates frequently exceeding 20%. The COVID-19 pandemic functioned as a natural experiment demonstrating that, without coordination infrastructure, surge research conditions amplify redundancy dramatically. The downstream consequences include artificial inflation of evidence certainty through pseudo-replication, direct resource waste with estimates that vary widely depending on costing methodology, and erosion of methodological trust in evidence synthesis as a scientific method.

Conclusions:

Redundancy in evidence synthesis is a structural pathology of the current evidence ecosystem rather than an individual failure, and requires coordinated interventions at methodological, infrastructural, and policy levels. Three critical unresolved challenges demand attention: establishing consensus on acceptable redundancy thresholds that distinguish warranted replication from wasteful duplication, developing validated and reproducible methods for AI-assisted overlap detection and evidence synthesis automation, and constructing resource-waste estimates grounded in transparent, empirically defensible costing frameworks rather than extrapolation from small convenience samples.