结外鼻型NK/T细胞淋巴瘤患者中枢受累风险评分模型的建立与验证

Establishment and Verification of Prognostic Model for Central Nervous System Involvement in Patients with Extranodal Natural Killer/T-Cell Lymphoma,Nasal Type

目的构建结外鼻型NK/T细胞淋巴瘤(ENKTL)患者中枢神经系统(CNS)受累风险的评分模型并验证其效能。方法训练组纳入2010年1月至2018年12月郑州大学第一附属医院收治的306例ENKTL患者,验证组纳入2009年11月至2019年12月河南省肿瘤医院收治的223例ENKTL患者。回顾性分析所有患者的临床资料,通过单因素及多因素Cox分析筛选ENKTL患者CNS受累的独立危险因素并构建风险评分模型,以受试者工作特征(ROC)曲线下面积(AUC)评估模型区分度。结果训练组纳入306例患者,其中16例(5.2%)发生CNS受累,验证组纳入223例患者,其中10例(4.5%)发生CNS受累。Cox分析发现肾上腺受累(HR=9.225,95%CI:2.702~31.495)、非上呼吸消化道原发(NUAT)型(HR=4.044,95%CI:1.187~13.776)、Ann ArborⅣ期(HR=3.500,95%CI:1.103~11.107)以及CD56阴性(HR=0.229,95%CI:0.056~0.927)是CNS受累的独立危险因素。风险函数提示,训练组和验证组中高低危患者的CNS受累风险显著不同,将验证组患者的数据代入本模型及CNS-PINK模型,AUC值分别为74.0828和59.5305。结论以肾上腺受累、NUAT型、Ann ArborⅣ期及CD56阴性4个风险因素构建的评分模型有助于识别ENKTL CNS受累的高危患者,为早期预防和治疗提供决策依据,从而改善患者预后。

Objective To construct and verify a prognostic model of central nervous system(CNS)involvement risk in patients with extranodal natural killer/T-cell lymphoma,nasal type(ENKTL).Methods The training set included 306 patients with ENKTL at the First Affiliated Hospital of Zhengzhou University from January 2010 to December 2018,and the validation set included 223 patients with ENKTL at Henan Cancer Hospital from November 2009 to December 2019.The clinical date of all patients were analyzed retrospectively,and univariate and multivariate Cox analysis was used to screen the independent risk factors of CNS involvement in ENKTL patients.The area under the receiver operating characteristic(ROC)curve(AUC)was used to verify the accuracy of the model.Results Of the 306 patients in the training set,16(5.2%)developed CNS involvement,and of the 223 patients in the validation set,10(4.5%)developed CNS involvement.The Cox analysis revealed that adrenal involvement(HR=9.225,95%CI:2.702-31.495),non-upper aerodigestive tract(NUAT)type(HR=4.044,95%CI:1.187-13.776),Ann Arbor stageⅣ(HR=3.500,95%CI:1.103-11.107),and CD56 negative(HR=0.229,95%CI:0.056-0.927)were independent risk factors for CNS involvement.The hazard function suggested that the cumulative incidence of CNS involvement was significant different between high and low risk patients in training and validation sets.The date of the validation set were brought into this model and CNS-PINK model,and the AUC values was 74.0828 and 59.5305,respectively.Conclusion The prognostic model,which included factors of adrenal involvement,NUAT type,Ann Arbor stageⅣand CD56 negative are helpful to identify high-risk patients with ENKTL CNS involvement,and provides a decision-making for early prevention and treatment,thereby improving the prognosis of patients.