Predicting prolonged mechanical ventilation after endovascular treatment for acute vertebrobasilar artery occlusion: AIRFLOW score

Qiankun Cai^1*Minying Hong²Yingjie Xu³Shuai Zhang⁴Zhixin Huang⁵Pengfei Xu³Chunnuan Chen¹Jixing Chen¹Ye Lichao¹Wen Sun^3*

• ¹Department of Neurology, The Second Affiliated Hospital of Fujian Medical University, Quanzhou, China
• ²Department of General practice, The Second Affiliated Hospital of Fujian Medical University, Quanzhou, China
• ³Department of Neurology, The First Affiliated Hospital of University of Science and Technology of China, Hefei, China
• ⁴Department of Neurology, Affiliated Hospital of Yangzhou University, Yangzhou, China
• ⁵Department of Neurology, Guangdong Second Provincial General Hospital, Guangzhou, China

Background and purpose Vertebrobasilar artery occlusion (VBAO) is a rare yet severe type of ischemic stroke, often leading to respiratory failure that necessitates invasive mechanical ventilation and is associated with high mortality rates. While endovascular treatment (EVT) has improved outcomes for VBAO, many patients still require prolonged mechanical ventilation (PMV) post-EVT, further increasing mortality and posing challenging decisions for families. Currently, no predictive model exists to identify VBAO patients at risk of needing PMV after EVT. This study aims to develop and validate a predictive score for PMV in this patient population following EVT. Materials and methods The derivation cohort prospectively recruited VBAO patients undergoing EVT from four comprehensive stroke centers (CSCs) in China. PMV was defined as continuous mechanical ventilation lasting for ≥ 7 days. Multivariable logistic regression was conducted to develop a scoring system. The performance of the model was evaluated for discrimination, calibration, and clinical utility. 414 patients from acute Posterior circulation ischemic Stroke registry were enrolled to externally validate the model. Sensitivity analysis redefined PMV as using mechanical ventilation last for ≥ 14 days to further validate the model. Results The derivation cohort consisted of 419 patients from four CSCs, among whom 113 (27.0%) required PMV. The presence of malignant cerebellar edema, posterior circulation collateral status, symptomatic intracranial hemorrhage post-EVT, atrial fibrillation, intravenous thrombolysis, vasopressor therapy and Glasgow coma score classification are found to be independent predictors of PMV in logistic regression, then 'AIRFLOW' scoring system was created. The AIRFLOW score demonstrated good discrimination in derivation cohort (C-index, 0.85, 95% CI 0.81 to 0.89), as well as the validation cohort (C-index, 0.82, 95% CI 0.77 to 0.86). Calibration plots and decision curve analysis for AIRFLOW score indicated that the model accurately predicted the risk of PMV and had satisfactory net benefit across various thresholds. Similar results were found in sensitivity analysis. Conclusions The AIRFLOW score may help predict PMV in VBAO patients after EVT.