BackgroundIdiopathic sudden sensorineural hearing loss (ISSNHL) is an otologic emergency with highly variable recovery despite prompt treatment. Because inflammatory chemotaxis and apoptosis are implicated in cochlear injury, serum biomarkers reflecting these pathways may improve early risk stratification.MethodsIn this single-center observational study, 209 consecutive adults with unilateral ISSNHL and 209 matched healthy controls were enrolled between July 2024 and February 2026. Patients were stratified by initial pure-tone average into mild, moderate, severe, or profound hearing loss. Fasting serum ENA-78, CXCL1, and caspase-3 were quantified at admission by enzyme-linked immunosorbent assay. After standardized 10-day treatment, complete, marked, or partial hearing improvement was classified as favorable early recovery and ineffective treatment as unfavorable. Analyses included multivariable logistic regression, receiver operating characteristic (ROC) analysis, 1,000-sample bootstrap internal validation, comparison with a clinical base model using net reclassification improvement (NRI) and integrated discrimination improvement (IDI), calibration, decision-curve analysis, a nomogram, and prespecified subgroup analyses.ResultsAll three biomarkers were higher in patients than in controls and increased progressively with severity (all p < 0.001). After treatment, 138 patients (66.0%) achieved favorable and 71 (34.0%) unfavorable early recovery. The unfavorable group had higher ENA-78 (147.69 ± 33.51 vs. 113.89 ± 21.89 pg/mL), CXCL1 (173.45 ± 36.44 vs. 137.24 ± 25.78 pg/mL), and caspase-3 (9.11 ± 2.43 vs. 6.63 ± 2.18 ng/mL) levels (all p < 0.001). Each biomarker remained independently associated with unfavorable recovery (ENA-78: odds ratio [OR], 1.019; CXCL1: OR, 1.024; caspase-3: OR, 1.106). The combined signature showed an apparent area under the curve (AUC) of 0.893 (95% CI, 0.843–0.932; 83.10% sensitivity, 81.88% specificity), which remained high after bootstrap correction (0.879). Adding the biomarkers to the clinical model improved discrimination (AUC, 0.709 to 0.912; NRI, 0.842; IDI, 0.187; all p < 0.001), with adequate calibration, positive net benefit across thresholds of 0.12 to 0.72, and consistent performance across subgroups.ConclusionAdmission serum ENA-78, CXCL1, and caspase-3 were associated with disease severity and early hearing recovery in ISSNHL. Their combined assessment outperformed any single marker and added value beyond routine clinical variables, although multicenter external validation and longer follow-up remain necessary before clinical implementation.