Page 25 - 南京医科大学自然版

第44卷第7期 南京医科大学学报（自然科学版）
2024年7月 Journal of Nanjing Medical University（Natural Sciences） ·909 ·

·临床研究·

基于SEER数据库的胰腺胶样癌生存预测模型的构建与验证

李晨晨 ，陈重发 ，尹凌帝 ，彭云鹏 ，苗 毅 1，2，3*
1，2
1，2
2，3
2，3
南京医科大学附属明基医院胰腺中心，江苏 南京 210019；南京医科大学胰腺研究所，江苏 南京 210029；南京医科大
1 2 3
学第一附属医院胰腺中心，江苏 南京 210029

［摘 要］ 目的：明确影响胰腺胶样癌（colloid carcinoma，CC）患者预后的危险因素，构建临床预后模型，对患者生存预后进行
评估。方法：选取美国国家癌症研究所监测、流行病学和最终结果（Surveillance，Epidemiology，and End Results，SEER）数据库
中510例患者，将其分为训练集和验证集，并根据训练集的临床变量构建预后模型，以预测CC患者在1、3、5年的癌症特异性生
存率的概率值，而后用验证集验证训练集筛选出的变量及校准预测模型。结果：利用SEER数据库筛选CC患者预后的影响因
素，包括肿瘤分级、检出淋巴结数量、是否手术、T分期、是否转移，并根据以上影响因素构建了临床预后模型。受试者特征曲
线及曲线下面积值表明本预测模型具有较高的准确性。校正曲线显示模型预测的生存率与实际生存率相近。结论：本研究构
建出CC患者的临床预后模型，并利用此模型对患者的生存预后进行了有效评估，填补了这种罕见肿瘤预后研究的空白，为临
床医生评估CC患者预后提供了一项有力依据。
［关键词］ 胰腺导管腺癌；胶样癌；SEER数据库；预后模型
［中图分类号］ R735.9 ［文献标志码］ A ［文章编号］ 1007⁃4368（2024）07⁃909⁃07
doi：10.7655/NYDXBNSN231036

Construction and validation of a survival prediction model for pancreatic colloid carcinoma
based on the SEER database
1，2
2，3
1，2
2，3
LI Chenchen ，CHEN Chongfa ，YIN Lingdi ，PENG Yunpeng ，MIAO Yi 1，2，3*
2
1 Pancreas Certer，the Affiliated BenQ Hospital of Nanjing Medical University，Nanjing 210019；Pancreas Institute of
Nanjing Medical University，Nanjing 210029；Pancreas Center，the First Affiliated Hospital of Nanjing Medical
3
University，Nanjing 210029，China
［Abstract］ Objective：To identify the risk factors that affect the prognosis of patients with pancreatic colloid carcinoma（CC），
construct a clinical prognostic model，and evaluate the survival prognosis of patients. Methods：A total of 510 patients from the
Surveillance，Epidemiology，and End Results（SEER）database of the National Cancer Institute were selected and divided into the
training and validation sets. A prognostic model was constructed based on the clinical variables of the training set to predict the
probability of cancer⁃specific survival rates of CC patients at 1，3，and 5 years Subsequently，the selected variabes and the calibration
predictive model were validated using the validation set. Results：In this study，the factors affecting the prognosis of CC patients were
screened by using the SEER database，including tumor grade，the number of detected lymph nodes，whether surgery was performed，T
stage，and whether metastasis was present. A clinical prognostic model was constructed based on the above⁃mentioned risk factors. The
final receiver operator characteristic curve and area under the curve value showed that the prediction of this model had a high accuracy.
The calibration curve showed that the survival rate predicted by the model was similar to the actual survival rate. Conclusion：In this
study，the clinical prognostic model was constructed，and the survival prognosis of patients was effectively evaluated by using this
model，which fills the gap in the prognostic studies of this rare tumor，providing a powerful basis for clinicians to assess the prognosis of
patients.
［Key words］ pancreatic ductal adenocarcinoma；colloid carcinoma；SEER database；prognostic model
［J Nanjing Med Univ，2024，44（07）：909⁃914，940］

［基金项目］ 国家自然科学基金（82173206，822003690）
∗
通信作者（Corresponding author），E⁃mail：miaoyi@njmu.edu.cn