预测最大径≤2 cm胆囊良恶性息肉指标的筛选及模型构建

Index Screening and Model Construction for Predicting Benign and Malignant Gallbladder Polyps with Maximum Diameter ≤2 cm

目的:通过联合超声影像学表现、血液检验及患者一般情况等相关指标,探讨上述指标在影像学表现≤2 cm的胆囊息肉性质的诊断价值。方法:收集2012年~2022年在青岛大学附属医院386例≤2 cm的胆囊息肉样病变患者临床资料,包括术前超声、血液肿瘤标记物、性别、基础信息、既往史、代谢综合征等指标,分析胆囊癌的危险因素及多指标联合早期诊断的准确性。通过Logistic回归方程,计算出各个指标的个体预测概率值,比较其ROC曲线下面积,并建立预测模型Y,绘制联合指标ROC曲线下面积,同单独指标比较其预测及应用价值,并以术后病理结果检验预测模型的准确性。结果:通过多因素分析,得到与胆囊癌相关的危险因素:超声息肉大小、胆囊壁厚、CEA、CA199、性别、年龄。联合上述指标得到预测模型为Y = −12.761 − 1.873性别(女) + 0.116年龄 + 6.558胆囊壁厚 + 2.724息肉大小 + 0.198CEA + 0.029CA199。对预测模型进行评价:联合预测指标Y的ROC曲线下面积为0.933,95%置信区间为(0.887, 0.979)。当Y > 0.1047557时模型预测息肉为恶性可能的灵敏度为93.6%,特异度为86.4%。结论:基于上述相关指标构建的预测模型,经内部验证具有良好的预测效能,灵敏度和特异度较高,有助于预测术前≤2 cm的胆囊息肉的性质,减少术中意外胆囊癌的发生,具有一定的临床使用价值。

Objective: To explore the diagnostic value of the above indexes in gallbladder polyps with imaging manifestations ≤2 cm by combining ultrasound imaging findings, blood tests and general conditions of patients. Methods: The clinical data of 386 patients with gallbladder polypoid lesions ≤2 cm in the Affiliated Hospital of Qingdao University from 2012 to 2022 were collected, including preopera-tive ultrasound, blood tumor markers, gender, basic information, past history, metabolic syndrome and other indicators. The risk factors of gallbladder cancer and the accuracy of multiple indicators combined early diagnosis were analyzed. The area under the ROC curve of each index was compared and the prediction model Y was established. The area under the ROC curve of the combined index was plotted to compare its prediction and application value with that of the single index. The accu-racy of the prediction model was tested by postoperative pathological results. Results: Through multivariate analysis, the risk factors related to gallbladder cancer were obtained: ultrasonic polyp size, gallbladder wall thickness, CEA, CA199, gender and age. Combined with the above indicators, the prediction model was Y = −12.761 − 1.873 gender (female) + 0.116 age + 6.558 gallbladder wall thickness + 2.724 polyp size + 0.198CEA + 0.029CA199. The prediction model was evaluated: the area under the ROC curve of the combined predictor Y was 0.933, and the 95% confidence interval was (0.887, 0.979). When Y > 0.1047557, the sensitivity and specificity of the model to predict the possibility of malignant polyps were 93.6% and 86.4% respectively. Conclusions: The prediction model constructed based on the above related indicators has good prediction efficiency, high sensi-tivity and specificity, which is helpful to predict the nature of gallbladder polyps ≤2 cm before op-eration, reduce the occurrence of accidental gallbladder cancer during operation, and has certain clinical application value.