Only a Small Fraction of Genetic Variants Ever Change Clinical Meaning, 13-Year Study Finds

Of the millions of genetic variants catalogued in ClinVar, the primary public archive for human variant classifications, only a small minority ever undergo a change that matters for patient care. That is the central finding of a 13-year temporal analysis published in Genome Medicine, in which researchers tracked 3.68 million variants across 148 monthly snapshots of the database from 2012 to 2025. While 9.3 percent of variants experienced at least one change in their aggregate classification category, just 0.68 percent — 25,105 variants — ever crossed the clinically actionable boundary separating pathogenic or likely pathogenic variants from variants of uncertain significance, benign, or likely benign ones.

Within that small group, 6,340 variants underwent the most drastic kind of reversal: a switch from pathogenic to benign, or the reverse. The study, led by Yilun Liang, Gongbo Zhang, and Chunhua Weng together with collaborators including Heidi Rehm and Wendy Chung, set out to disentangle two phenomena that earlier work had often conflated: genuine clinical reinterpretation of a variant's pathogenicity as evidence accumulates, and apparent reclassification driven by changes to the database's own structure. ClinVar's schema evolved through three distinct eras involving more than 120 structural changes that affect how the data can be interpreted.

The team built an era-stratified heuristic framework to separate these artifacts from evidence-based reinterpretation, analyzing both the older Reference ClinVar (RCV) and newer Variation ClinVar (VCV) formatted XML files. They found that 2.6 percent of all reclassification events were attributable to the RCV-to-VCV format transition alone — changes that reflect database engineering rather than new biology. Because variants enter the database at different times and have had different lengths of exposure to potential reclassification, the researchers also computed cumulative incidence figures that account for censoring.

On that basis, the cumulative incidence of a first reclassification was 2.5 percent at one year after a variant's first appearance, rising to 15.5 percent at five years. Reclassification was not distributed evenly across the genome's clinical landscape. Variants in cancer-related genes were reclassified at significantly higher rates than variants in other genes, at 16.6 percent versus 8.2 percent.

Similarly, variants linked to higher-prevalence Mendelian diseases were reclassified at 19.7 percent, compared with 9.2 percent for variants tied to lower-prevalence conditions. The authors argue that these patterns support domain-specific variant reassessment intervals — in other words, the appropriate frequency for re-examining a classification should depend on the disease area, with cancer genes warranting more frequent review. Disagreement among submitters emerged as another major theme.

Among variants assessed by multiple submitters, 72.1 percent carried concordant interpretations, while 17.6 percent maintained conflicting classifications. Of the 697,698 reclassification events recorded over the study period, 33.2 percent resulted from a submitter reinterpreting its own earlier assessment, and 18.3 percent arose when a new submitter joined and its classification shifted the aggregate category. These figures illustrate how ClinVar's aggregate labels are shaped not only by evolving scientific evidence but also by the composition of the contributing community.

The analysis also identified the evidence streams most associated with driving genuine reinterpretation. Functional evidence from high-throughput saturation mutagenesis experiments and targeted biochemical assays was likewise identified as an important contributor to reclassification. The authors conclude that their findings reinforce the importance of diverse population reference datasets and high-throughput functional assays in minimizing classification uncertainty.

They also emphasize the practical value of their methodological contribution: the team published an open-source pipeline and visualization framework designed to enable reproducible, era-aware temporal analysis of clinical variant reclassification, so that future studies can account for the database's structural history rather than treating thirteen years of data as a single homogeneous record. Yet the study also quantifies where vigilance is most needed: variants in cancer genes and variants linked to common Mendelian diseases were reclassified at roughly double the rate of their counterparts. As genomic testing continues to expand, the authors suggest, structured and domain-aware reassessment — supported by richer population data and functional experiments — will be essential to keeping variant classifications current and clinically trustworthy.

Subject of Research: Longitudinal analysis of clinical variant reclassification in the ClinVar database from 2012 to 2025

Article Title: Temporal analysis of clinical variant reclassification (2012–2025)

Article References: Liang, Y., Zhang, G., Babb, L., Rehm, H. L., Nestor, J. G., Kneifati-Hayek, J. Z., Luo, Y., Krishna Murthy, S. B., Shen, Y., Wang, K., Chung, W. K., & Weng, C. (2026). Temporal analysis of clinical variant reclassification (2012–2025). Genome Medicine. https://doi.org/10.1186/s13073-026-01805-9

Image Credits: AI Generated

DOI: 10.1186/s13073-026-01805-9

Keywords: ClinVar, variant reclassification, genomic medicine, pathogenicity, VUS, gnomAD, ExAC, ACMG/AMP criteria, cancer genes, Mendelian disease, database schema evolution, genetic testing