中国实用外科杂志

中国实用外科杂志

• 论蓍 • 上一篇    下一篇

胰腺囊性肿瘤恶性风险预测模型研究

赵    练，胡鹏飞，邓    尧，刘    猛，王    巍   

  1. 复旦大学附属华东医院胆胰中心，上海 200040
  • 出版日期:2020-09-01 发布日期:2020-09-21

  • Online:2020-09-01 Published:2020-09-21

摘要： 目的    明确与恶性胰腺囊性肿瘤（PCN）相关的术前危险因素，建立准确的预测模型，并予以验证。方法    纳入2013年1月至2020年5月复旦大学附属华东医院经术后病理检查证实的114例PCN病例，分为模型组（n=80）和验证组（n=34）。回顾性分析模型组术前的临床资料并探索与恶性PCN相关的影响因素，建立PCN恶性风险预测模型，绘制受试者工作特征（ROC）曲线和校正曲线评价模型，最后基于验证组数据对模型进行临床验证。结果    单因素回归分析提示临床症状、CA19-9水平升高、中性粒细胞淋巴细胞比值（NLR）、淋巴细胞单核细胞比值（LMR）、肿瘤最大直径、胰管扩张和实性成分与恶性PCN显著相关，进一步行多因素回归分析确定了NLR≥2.146、CA19-9水平升高、胰管扩张是恶性PCN的独立预测因素。基于多因素回归分析结果建立恶性PCN预测模型，绘制模型的ROC曲线，计算AUC为0.921（95%CI 0.863~0.979），Youden指数最大时取得最佳临界预测值为0.203，此时相对应的特异度为83.3%，敏感度为92.9%，准确率为85%。同时校正曲线显示模型具有较好的拟合度，最后代入验证组数据显示模型预测准确率为82.4%，特异度81.2%，敏感度100%。结论    CA19-9水平升高、NLR升高以及胰管扩张是恶性PCN的高危因素，基于此建立的恶性胰腺囊性肿瘤的预测模型具有较好的准确率，可为今后的临床诊疗提供辅助参考。

关键词: 胰腺囊性肿瘤, 统计模型, 炎性指标, 风险因素, 中性粒细胞/淋巴细胞比值

Abstract: Preoperative system predicting malignant potential in patients with pancreatic cystic neoplasm and its clinical validation        ZHAO Lian, HU Peng-fei, DENG Yao, et al. Department of General Surgery, Huadong Hospital Affiliated to Fudan University, Shanghai 200040, China
Corresponding author：WANG Wei，E-mail：wangw2003cn@126.com
Abstract    Objective    To clarify the preoperative risk factors differentiating malignant pancreatic cystic neoplasm(PCN) from benign PCN and establish an accurate clinical predicting system followed by evaluations. Methods    A total of 114 pathologically confirmed PCN patients who had surgeries from January 2013 to May 2020 in Huadong Hospital Affiliated to Fudan University were adopted in the study and were divided chronologically into model group（n=80）and validation group（n=34）. The clinical data of model group were evaluated retrospectively and the significant factors associated with malignancy were assessed by univariate and multivariate logistic regression analyses. Based on the significant variables in the multivariate analysis，a system to predict malignant potential in patients with PCN was developed. The area under the receiver operating characteristic curve（AUC）and calibration curve were used to assess diagnostic value. Validation group data were used to verify the predicting system. Results    Univariate analysis revealed that symptoms，elevated CA19-9，neutrophil-to-lymphocyte ratio（NLR），lymphocyte-to-monocyte ratio（LMR），cyst size，main pancreatic duct dilation，presence of solid component were significantly associated with malignant PCN. Among them，NLR≥2.146，elevated CA19-9，and main pancreatic duct dilation were identified as independent significant parameters to predict malignant PCN. Based on the parameters，a predicting system for malignant were established. The AUC for the system was 0.921（95%CI 0.863~0.979）and the optimal cut-off value was 0.203 in reference to Youden index. The specificity, sensitivity and accuracy were 83.3%，92.9%, 85.0% respectively based on the optimal cut-off value. At the same time，the model had a good fitting degree evaluated by calibration curve and showed an accuracy of 82.4%, specificity of 81.2%, sensitivity of 100% when applied into validation group. Conclusion  Elevated CA19-9，NLR≥2.146 and main pancreatic duct dilation are of great values in differentiating malignant PCN. The predicting system based on the factors shows great accuracy and may aid in the diagnosis of malignant PCN in the future.

Key words: pancreatic cystic neoplasm, statistical model, systemic inflammatory markers, risk factors, lymphocyte-to-monocyte ratio