Epigenetic Signatures as Predictors of Fracture Non-union: A Systematic Review

Non-union is a recognized complication of fracture healing that imposes prolonged disability, pain, and repeated surgical intervention. Clinical prediction currently depends on patient and injury characteristics that discriminate poorly at the level of the individual patient, and a molecular marker capable of identifying high-risk patients early would permit intensified surveillance or biological augmentation while intervention remains feasible. Epigenetic mechanisms, comprising DNA methylation, histone modifications, and non-coding RNAs, regulate skeletal repair and have been proposed for this purpose. This systematic review, reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 and Synthesis Without Meta-analysis (SWiM) guidelines, synthesized evidence published between 2014 and 2025, distinguishing throughout between markers of established non-union and validated predictors of future non-union. A structured search of MEDLINE supplemented by Google Scholar and semantic literature retrieval yielded 460 records, of which 159 unique records were screened, 120 full texts were assessed, and 30 studies met the inclusion criteria. Six studies formed the core evidence base: four primary studies and two systematic reviews. Serum lnc_GABARAPL2 was elevated in 45 patients with long bone non-union relative to 45 matched controls, with an area under the receiver operating characteristic curve (AUC) of 0.922 and an adjusted odds ratio of 11.661 (95% CI, 3.794 to 35.840), acting through sequestration of miR-302a-3p to suppress osteogenic differentiation. Long non-coding RNA ENST00000563492 was downregulated in non-union tissue from 24 patients and promoted osteogenesis and angiogenesis through CDH11 and vascular endothelial growth factor as a competing endogenous RNA for miR-205-5p. Five microRNAs, miR-31a-3p, miR-31a-5p, miR-146a-5p, miR-146b-5p, and miR-223-3p, were upregulated in a rat atrophic non-union model, with peak differential expression at day 14. Systemic inflammation reduced Dnmt3b expression in skeletal progenitors, producing hypomethylation of the Rbpjk promoter and consequent Rbpjk upregulation with impaired differentiation, such that promoter methylation at this locus is protective, and restoration of Dnmt3b rather than demethylation rescued repair. No primary study examined histone modifications or non-canonical non-coding RNA classes in non-union. A single discrimination estimate exists, derived from a case-control design without external validation, in which serum was obtained from patients already undergoing non-union surgery, and no study sampled human material before the outcome was known. Certainty of evidence was low to very low across all marker classes. The literature therefore identifies biologically plausible candidates and one promising circulating marker but does not establish prospective predictive validity; prospective cohorts with early post-injury sampling, standardized outcome definitions, and stratification by fixation strategy are required.