中国实用外科杂志

中国实用外科杂志 ›› 2025, Vol. 45 ›› Issue (06): 670-676.DOI: 10.19538/j.cjps.issn1005-2208.2025.06.13

• 论著 • 上一篇    下一篇

不同营养评分预测模型评估胰腺癌行胰十二指肠切除术病人预后研究

李宏涛，黄子健，苏    铁，李南南，张文彬，王    刚   

  1. 哈尔滨医科大学附属第一医院肿瘤腔镜外科，黑龙江哈尔滨 150001
  • 出版日期:2025-06-01 发布日期:2025-07-01

  • Online:2025-06-01 Published:2025-07-01

摘要： 目的    比较不同营养评分对胰腺癌行胰十二指肠切除术（PD）后总生存期（OS）的预测效能，并构建个性化预测模型。方法    回顾性分析2013年12月至2022年12月于哈尔滨医科大学附属第一医院接受PD的188例胰腺癌病人的临床资料，通过单因素和多因素Cox回归分析确定影响OS的独立预后因素。将各营养评分［营养风险指数（NRI）、那不勒斯预后评分（NPS）、预后营养指数（PNI）、控制营养状态评分（CONUT）］分别与独立影响因素结合建立预测模型，采用一致性指数（C-index）、校正曲线、决策曲线分析（DCA）及时间依赖性受试者工作特征（ROC）曲线比较各模型的预测效能与稳定性。最终基于最佳营养评分构建列线图（Nomogram）模型。结果    多因素Cox回归分析结果显示，淋巴结转移（HR=1.299，95%CI 0.926~1.824，P=0.047）、肿瘤最大径≥2.5 cm（HR=1.411，95%CI 1.010~1.972，P=0.043）及肿瘤分化程度低（HR=1.494，95%CI 1.069~2.089，P=0.019）是PD术后病人预后不良的独立危险因素。随后，联合4种营养评分指标，建立相应的预测模型进行横向对比分析。DCA和校正曲线结果表明，NRI和NPS模型预测性能优于PNI和CONUT模型。C指数显示NRI模型（0.687，95%CI 0.663~0.712）与NPS模型（0.684，95%CI 0.660~0.707）预测效能相近，时间依赖性ROC曲线结果显示2种评分具有相似的准确性和稳定性，但NPS评分的长期预测能力更优。基于NPS和预后独立影响因素构建列线图模型，该模型的截断值为80.6分。结论    基于NPS评分构建的预测模型可准确评估胰腺癌行PD术后病人的生存风险，对营养不良风险分层及个体化干预具有重要参考价值。

关键词: 营养评分, 胰腺癌, 胰十二指肠切除术, 预后模型, 列线图, 总生存期 

Abstract: To evaluate the predictive performance of distinct nutritional scores for overall survival (OS) in pancreatic cancer patients after pancreaticoduodenectomy (PD) and establish a personalized prognostic model. Methods  A retrospective cohort study was performed on 188 pancreatic cancer patients who underwent PD at the First Affiliated Hospital of Harbin Medical University, between December 2013 and December 2022. Independent prognostic factors for OS were identified through univariate and multivariate Cox regression analyses. Four predictive models were constructed by integrating each nutritional score (NRI, NPS, PNI, and CONUT) with identified independent predictors. Model performance was compared using the concordance index (C-index), calibration curves, decision curve analysis (DCA), and time-dependent ROC curves. A nomogram was ultimately developed based on the optimal nutritional score. Results    Multivariate analysis confirmed lymph node metastasis (HR=1.299, 95%CI 0.926~1.824, P=0.047), tumor diameter ≥2.5 cm (HR=1.411, 95%CI 1.010~1.972, P=0.043), and poor tumor differentiation (HR=1.494, 95%CI 1.069~2.089, P=0.019) as independent risk factors for poor prognosis after PD. Comparative assessment of the four models revealed superior predictive capability for NRI and NPS models compared to PNI and CONUT models, as evidenced by DCA and calibration curves. While the NRI (C-index=0.687, 95%CI 0.663~0.712) and NPS (C-index=0.684, 95%CI 0.660~0.707) models demonstrated equivalent discriminatory power, time-dependent ROC analysis indicated comparable accuracy and stability, with NPS exhibiting enhanced long-term predictive performance. A nomogram incorporating NPS and independent prognostic risk factors was established, with a cutoff score of 80.6. Conclusion    The NPS-derived predictive model enables accurate survival risk stratification for pancreatic cancer patients after PD, providing critical insights for nutritional risk-adapted management and tailored clinical interventions.

Key words: nutritional scores, pancreatic cancer, pancreaticoduodenectomy, prognostic model, nomogram, overall survival