Development and Validation of Salivary Exosomal Tri-RNA Liquid Biopsy in Esophageal Carcinoma: A Multicenter Study

PURPOSE: Exosomal RNAs are emerging as cancer signatures, and saliva is a noninvasive biospecimen. Given the high mortality of patients with esophageal squamous cell carcinoma (ESCC) and limited early detection tools, we investigated a salivary exosome‑based Tri-signature for its diagnostic and prognostic potential. MATERIALS AND METHODS: The salivary exosome‑based signature (ie, a chimeric RNA seG-NchiRNA, a tRNA fragment GlyGCC-5, and a novel sRESE RNA) was quantified by qRT-PCR in a multicenter observational study across two ESCC-endemic regions. Model development and validation were performed in the training (n = 359) and validation (n = 225) cohorts using logistic regression, survival analyses, and Shapley Additive exPlanations-based feature interpretation. RESULTS: The Tri-signature showed excellent diagnostic accuracy (training cohort: AUC, 0.987; validation cohort: AUC, 0.964) and robust prognostic value (training cohort: overall survival [OS] hazard ratio [HR], 5.52, progression-free survival [PFS] HR, 4.46; validation cohort: OS HR, 4.76, PFS HR, 2.79). In the high Combined Risk Score for Prognosis (CRSP) subgroup, patients with relatively lower CRSP derived significant benefit from adjuvant therapy (training cohort: OS HR, 0.54, PFS HR, 0.47; validation cohort: OS HR, 0.38, PFS HR = 0.32), whereas no such benefit was observed in low Tri-signature patients. The Tri-signature exhibited strong early diagnostic performance, distinguishing early-stage ESCC without lymph node metastasis from healthy controls (training cohort: AUC = 0.975; validation cohort: AUC = 0.950). Patients with early-stage ESCC and high CRSP had significantly worse outcomes (training cohort: OS HR, 5.09, RFS HR, 3.80; validation cohort: OS HR, 8.79, RFS HR, 4.55). CONCLUSION: The salivary exosome-based tri-RNA signature showed robust multicenter reproducibility and strong diagnostic, prognostic, and treatment response-predictive performance, supporting its translational potential as a noninvasive biomarker panel for ESCC management.