目的：通过分析美国监测、流行病学和最终结果（SEER）数据库资料，对比胰腺癌 患者行全胰腺切除术（TP）与胰十二指肠切除术（PD）的远期生存预后，并确定 TP 治疗胰腺癌的独立预后因素，建立 TP 治疗胰腺癌的列线图预后预测模型，为临床 预测胰腺癌患者行 TP 术后远期生存提供参考依据。 材料与方法：收集 SEER 数据库 2004-2015 年诊断为胰腺导管腺癌（PDAC）行全胰 腺切除术（TP 组）与胰十二指肠切除术（PD 组）患者的临床、病理资料。KaplanMeier 法和 log-rank 检验分析比较两种术式患者术后总生存期（OS）及肿瘤特异性 生存期（CSS），并采用倾向性评分匹配（PSM）平衡组间基线差异所致的混杂偏倚， 对两组匹配前及匹配后的 OS, CSS 分别进行比较，卡钳值设为 0.01。X-tile 软件确 定连续性变量的最佳截断值。单因素及多因素 Cox 回归分析确定 TP 组的独立预后 因素，并构建成生存列线图模型。通过模型的一致性指数（C-index）、校准曲线 （Calibration curve）验证模型的准确性，决策曲线分析（DCA）评价模型的临床应 用价值，并与临床广泛应用的 TNM 分期系统进行对比分析。X-tile 软件确定列线图 模型最佳截断值进行风险分层，将 TP 术后分为低、中、高风险组，并用 KaplanMeier 法和 log-rank 检验评估高、中、低风险组间 OS 的差异。 结果： 经筛选共纳入 TP 组 1248 例，PD 组 5619 例；经 1:1 匹配后，TP 组 1248 例，PD组1248例。TP、PD两组OS及CSS在PSM前、后均无显著性差异 (P>0.05)。 单因素 Cox 回归分析显示，年龄、肿瘤大小、肿瘤分化程度、T 分期、N 分期、放 疗、化疗、淋巴结比率（LNR）与 TP 组 OS 相关，多因素 Cox 回归分析显示，年 龄、肿瘤分化程度、T 分期、放疗、化疗、LNR 是影响 TP 组 OS 的独立预后因素， 纳入建立列线图模型。列线图模型训练集、验证集的 C-index 分别为 0.67, 0.69，TNM 分期系统训练集、验证集的 C-index 分别为 0.59, 0.60。Calibration curves 显示列线 图预测值与实际观测值具有较好的一致性。DCA 决策曲线分析显示列线图模型对 TP 术后具有良好的净获益，临床实用价值优于 TNM 分期系统。采用列线图风险分 层系统，以 123, 217 分为截断值，将 TP 组分为低 (<123 分)、中 (123-217 分)、高 (>217 分)风险组，三组间 OS 有显著性差异 (P<0.05)，可以较好区分 TP 术后生存 风险。 结论：（1）全胰腺切除术治疗胰腺导管腺癌患者的总生存期、肿瘤特异性生存期与胰十二指肠切除术无显著性差异；（2）患者年龄、肿瘤分化程度、T 分期、放疗、 化疗、淋巴结比率是全胰腺切除术治疗胰腺导管腺癌生存预后的独立影响因素；（3） 本研究构建的列线图模型具有较好的区分度及校准度，能较好地预测胰腺导管腺癌 行全胰腺切除术个体患者的总生存率，且较 TNM 分期系统可能具有更好的预测效 能和临床实用价值。 关键词：全胰腺切除术，胰腺导管腺癌，SEER 数据库，列线图，预后

Objective: By analyzing the Surveillance, Epidemiology, and End Results (SEER) 
database, to compare the long-term prognosis of pancreatic ductal adenocarcinoma (PDAC) 
undergoing total pancreatectomy (TP) versus pancreaticoduodenectomy (PD), and identify 
independent prognostic factors of TP for PDAC, then formulate a nomogram to predict 
overall survival (OS) for PDAC individuals following TP, which can provide a reference 
for clinical prediction of long-term survival of patients with PDAC following TP.
Materials and methods: Clinical and pathological data of PDAC patients who underwent 
TP and PD from 2004 to 2015 were collected from the SEER database. Kaplan-Meier 
method and log-rank test were used to compare OS and cancer-specific survival (CSS) of 
the two groups. A propensity score matching (PSM) was applied to balance the 
confounding due to baseline differences between groups, and the caliper value was set to 
0.01. OS and CSS of the two groups were compared both before and after PSM, 
respectively. X-tile software was used to determine the optimal cut-off values for 
continuous variables. Univariate and multivariate Cox regression were applied to identify 
the independent factors affecting OS in the TP group to construct the nomogram. The 
accuracy of the nomogram was measured according to the concordance index (C-index) 
and calibration plots. Decision curve analysis (DCA) was performed to evaluate the 
clinical utility of the model, and was compared with the widely-used American Joint 
Committee on Cancer (AJCC) tumor-node-metastasis (TNM) staging system. X-tile 
software was used to determine the optimal cut-off value of the nomogram for risk 
stratification, which divided TP patients into low risk, intermediate risk and high risk 
groups. Kaplan-Meier method and log-rank test were used to evaluate OS differences 
among low risk, intermediate risk and high risk groups.
Results: A total of 1248 patients were included in the TP group and 5619 patients in the 
PD group after screening. After a 1:1 PSM, there were 1248 cases in the TP group and 
1248 cases in the PD group. There were no significant differences in OS and CSS between TP and PD groups neither before nor after PSM (P>0.05). Univariate Cox regression 
analysis showed that age, tumor size, tumor differentiation, AJCC T stage, N stage, 
radiotherapy, chemotherapy, and lymph node ratio (LNR) were correlated with OS of 
patients in the TP group. Multivariate Cox regression analysis showed that age, tumor 
differentiation, T stage, radiotherapy, chemotherapy and LNR were independent 
prognostic indicators affecting OS in TP group, which were selected to construct the 
nomogram. The C-index of the nomogram model was 0.67 in the training set and 0.69 in 
the validation set, while C-index of the TNM staging system was 0.59 in the training set 
and 0.60 in the validation set. The calibration curves showed good uniformity between the 
nomogram prediction and actual observation. DCA curves indicated that the nomogram 
had a decent net benefit, which showed more preferable clinical practice than the TNM 
staging system. The risk stratification of the TP group was divided into low risk group 
(total score < 123), intermediate risk group (total score 123-217) and high risk group (total 
score > 217). There were significant differences in OS among the three groups (P<0.05), 
which could distinguish the survival risk of individuals following TP.
Conclusions: (1) For PDAC patients, there were no significant differences in overall 
survival and cancer-specific survival between total pancreatectomy and 
pancreaticoduodenectomy; (2) Age, tumor differentiation, T stage, radiotherapy, 
chemotherapy and LNR were independent prognostic indicators affecting OS of PDAC 
patients treated with TP. (3) The nomogram can well predict the prognosis of PDAC 
patients treated with TP, which demonstrates good discrimination and calibration, and may 
have better prediction efficiency and clinical practical value than the TNM staging system.
Keywords: total pancreatectomy, pancreatic ductal adenocarcinoma, SEER, nomogram, 
prognosis