Rindler Rima S, Robertson Henry, De Yampert LaShondra, Khatri Vivek, Texakalidis Pavlos, Eshraghi Sheila, Grey Scott, Schobel Seth, Elster Eric A, Boulis Nicholas, Grossberg Jonathan A
Sierra Neurosurgery Group, Reno, Nevada, USA.
Department of Neurosurgery, Emory University, Atlanta, Georgia, USA.
Neurosurgery. 2024 Oct 11. doi: 10.1227/neu.0000000000003224.
Prediction of patient outcomes after severe traumatic brain injury (sTBI) is limited with current clinical tools. This study aimed to improve such prognostication by combining clinical data and serum inflammatory and neuronal proteins in patients with sTBI to develop predictive models for post-traumatic vasospasm (PTV) and mortality.
Fifty-three adult civilian patients were prospectively enrolled in the sTBI arm of the Surgical Critical Care Initiative (SC2i). Clinical, serum inflammatory, and neuronal protein data were combined using the parsimonious machine learning methods of least absolute shrinkage and selection operator (LASSO) and classification and regression trees (CART) to construct parsimonious models for predicting development of PTV and mortality.
Thirty-six (67.9%) patients developed vasospasm and 10 (18.9%) died. The mean age was 39.2 years; 22.6% were women. CART identified lower IL9, lower presentation pulse rate, and higher eotaxin as predictors of vasospasm development (full data area under curve (AUC) = 0.89, mean cross-validated AUC = 0.47). LASSO identified higher Rotterdam computed tomography score and lower age as risk factors for vasospasm development (full data AUC 0.94, sensitivity 0.86, and specificity 0.94; cross-validation AUC 0.87, sensitivity 0.79, and specificity 0.93). CART identified high levels of eotaxin as most predictive of mortality (AUC 0.74, cross-validation AUC 0.57). LASSO identified higher serum IL6, lower IL12, and higher glucose as predictive of mortality (full data AUC 0.9, sensitivity 1.0, and specificity 0.72; cross-validation AUC 0.8, sensitivity 0.85, and specificity 0.79).
Inflammatory cytokine levels after sTBI may have predictive value that exceeds conventional clinical variables for certain outcomes. IL-9, pulse rate, and eotaxin as well as Rotterdam score and age predict development of PTV. Eotaxin, IL-6, IL-12, and glucose were predictive of mortality. These results warrant validation in a prospective cohort.
目前的临床工具对重型颅脑损伤（sTBI）患者预后的预测能力有限。本研究旨在通过整合sTBI患者的临床数据以及血清炎症蛋白和神经元蛋白，以改善此类预后评估，从而建立创伤后血管痉挛（PTV）和死亡率的预测模型。
53例成年平民患者前瞻性纳入外科重症监护计划（SC2i）的sTBI组。使用最小绝对收缩和选择算子（LASSO）和分类与回归树（CART）等简约机器学习方法，将临床、血清炎症和神经元蛋白数据相结合，构建预测PTV发生和死亡率的简约模型。
36例（67.9%）患者发生血管痉挛，10例（18.9%）死亡。平均年龄为39.2岁；22.6%为女性。CART确定较低的IL9、较低的就诊脉搏率和较高的嗜酸性粒细胞趋化因子是血管痉挛发生的预测因素（全数据曲线下面积（AUC）=0.89，平均交叉验证AUC =0.47）。LASSO确定较高的鹿特丹计算机断层扫描评分和较低的年龄是血管痉挛发生的危险因素（全数据AUC 0.94，敏感性0.86，特异性0.94；交叉验证AUC 0.87，敏感性0.79，特异性0.93）。CART确定高水平的嗜酸性粒细胞趋化因子对死亡率的预测性最强（AUC 0.74，交叉验证AUC 0.57）。LASSO确定较高的血清IL6、较低的IL12和较高的血糖可预测死亡率（全数据AUC 0.9，敏感性1.0，特异性0.72；交叉验证AUC 0.8，敏感性0.85，特异性0.79）。
sTBI后炎症细胞因子水平对于某些结局可能具有超过传统临床变量的预测价值。IL-9、脉搏率和嗜酸性粒细胞趋化因子以及鹿特丹评分和年龄可预测PTV的发生。嗜酸性粒细胞趋化因子、IL-6、IL-12和血糖可预测死亡率。这些结果有待在前瞻性队列中进行验证。
