Background: The growing volume of circulating RNA biomarker studies has generated thousands of candidate biomarkers across human diseases; however, evidence supporting individual candidates often remains fragmented across discovery and validation studies that differ in biological, technical, and analytical design, making robust biomarker qualification difficult. Although systematic reviews commonly catalogue reported biomarkers, they provide limited ability to distinguish robust candidates from isolated observations. We developed an evidence-qualification framework to systematically prioritize circulating RNA biomarkers by integrating four complementary evidence domains: evidence depth, statistical support, directional consistency, and concordance between discovery and targeted validation studies. Age-related macular degeneration (AMD) was used as a representative case study to demonstrate the framework.
Methods: Original studies investigating circulating RNAs in individuals with AMD were systematically reviewed. For each RNA, evidence was synthesized sequentially across the four evidence domains to evaluate the extent of independent investigation, consistency of statistical support, reproducibility of the reported direction of change, and agreement between unbiased large-scale profiling and targeted validation studies. Biological and technical study characteristics were evaluated alongside the framework to identify factors limiting biomarker qualification.
Results: Thirty studies met the inclusion criteria. Large-scale profiling studies identified 185 circulating RNA candidates, while targeted studies evaluated 118 candidates. Application of the framework demonstrated that most reported biomarkers lacked sufficient evidence for qualification because they had limited independent evaluation, inconsistent statistical support, conflicting directions of change, or no supporting evidence from complementary discovery and validation approaches. Sequential evidence qualification reduced the large pool of reported candidates to only a small number supported across multiple evidence domains. The framework further identified methodological and biological heterogeneity, inconsistent statistical practices, fragmented miRNA nomenclature, and limited replication under comparable biological–technical conditions as major barriers to biomarker qualification.
Conclusions: We present an evidence-qualification framework that moves beyond cataloguing reported biomarkers to systematically distinguish well-supported candidates from isolated findings. By integrating evidence depth, statistical support, directional consistency, and discovery–validation concordance within a single qualification workflow, the framework provides a standardized approach for evidence synthesis, biomarker prioritization, and identification of methodological sources of irreproducibility. Although demonstrated using AMD, the framework is designed to support qualification of circulating RNA biomarkers across diverse diseases and to facilitate the development of more reproducible biomarker discovery pipelines.