Risks to predict blood loss and cranial nerve injury in carotid body paraganglioma resection

Abstract

Objective: To investigate clinical and imaging parameters to predict blood loss and cranial nerve injury (CNI) following carotid body paraganglioma (CBP) resection. Methods: A retrospective examination of clinical and imaging data was conducted on 63 patients who underwent CBP resection at Xiangya Hospital of Central South University from January 2016 to December 2022, including 23 males and 40 females, aged 26-87 years old. Three imaging parameters including tumor volume, the angle of contact with the internal carotid artery (ICA), and the distance to the base of skull (DTBOS) were gauged using the IMEDPACS software on CTA and MR imaging. The predictive efficacies of age, gender, Shamblin classification, and three imaging parameters for blood loss and CNI following surgery were analysed. Logistic composite parameter models were constructed and their predictive validity was assessed. Results: Multivariate logistic regression analysis underscored that only tumor volume (OR=1.381,95%CI:1.167-1.507,P=0.001) showed significant statistical correlations with blood loss following surgery. Area under curve (AUC) values of 0.910 for receiver operating characteristic (ROC) curves showed a sensitivity of 1.000 and a specificity of 0.694. Tumor volume (OR=1.126,95%CI:1.030-1.231, P=0.002) and DTBOS (OR=0.225,95%CI:0.081-0.630,P=0.005) were significantly associated with postoperative CNI. The analysis of logistic composite model showed AUC values for tumor volume, DTBOS and combination of the two parameters were 0.858, 0.788, and 0.872, respectively. The model for combination of tumor volume and DTBOS also proved superior in predicting postoperative CNI (Z=3.106, P<0.001), with a sensitivity of 0.833 and a specificity of 0.769. Conclusions: Tumor volume and DTBOS emerged as effective predictors for blood loss and/or CNI in patients with CBP resection. Moreover, the logistic composite parameter model outclassed single-parameter models in terms of their predictive clinical value.