Preoperative Radiological Features Predict Facial and Cochlear Nerve Function after Vestibular Schwannoma Surgery: A Single-Center Retrospective Study.

Zhang, Ding; Guo, Zining; Wang, Qun; Shi, Xudong; Liu, Yuyang; Zhang, Jun · World Neurosurg · 2026

retrospective_cohort · Level III

Where this comes from

Abstract

Vestibular schwannoma (VS) surgery carries risks of facial nerve (FN) dysfunction and hearing loss. This study aimed to identify preoperative radiological predictors of postoperative FN function and develop a predictive nomogram. We retrospectively analyzed 262 patients who underwent retrosigmoid VS resection at a single institution (December 2017 -December 2025). The dataset was divided into training (n=185) and validation (n=77) sets. Univariate and multivariate logistic regression identified independent predictors of poor FN function (House-Brackmann grade ≥III), and a nomogram was constructed. Hearing outcomes were explored in 81 patients with preoperative serviceable hearing. Multivariate analysis identified tumor laterality (OR = 2.69, p = 0.027), maximum tumor diameter (OR = 1.07, p = 0.002), fundal fluid cap sign (OR = 0.35, p = 0.020), and cerebrospinal fluid cleft sign (OR = 0.31, p = 0.009) as independent predictors of postoperative FN dysfunction. The nomogram demonstrated good discrimination, with an area under the ROC curve of 0.834 (95%CI: 0.753-0.906) in the training set and 0.880 (95%CI: 0.772-0.965) in the validation set. Calibration and decision curve analyses confirmed clinical applicability. For hearing outcomes, tumor diameter (OR = 1.11, p = 0.007) and brainstem compression (OR = 4.50, p = 0.020) were associated with postoperative hearing loss. In this single-center cohort, tumor laterality, maximum diameter, FFC sign, and CSFC sign independently predicted postoperative facial nerve function; the nomogram demonstrated satisfactory discrimination and calibration. These findings have refined our surgical approach for left-sided tumors and cleft-absent cases, pending multicenter validation.