天津医药 ›› 2024, Vol. 52 ›› Issue (11): 1177-1182.doi: 10.11958/20240546

• 临床研究 • 上一篇    下一篇

卵巢癌化疗耐药预测模型的建立及效果评价

喻萍1,2(), 周敏2, 苏丹1,2,()   

  1. 1 川北医学院(邮编637100)
    2 四川省医学科学院?四川省人民医院妇产科(邮编610072)
  • 收稿日期:2024-05-07 修回日期:2024-07-26 出版日期:2024-11-15 发布日期:2024-11-12
  • 通讯作者: △E-mail:303483765@qq.com
  • 作者简介:喻萍(1981),女,主治医师,主要从事妇科肿瘤方面研究。E-mail:hellenyp@163.com
  • 基金资助:
    四川省科技厅重点研发项目(2022YFS0088);四川省自然科学基金项目(2022NSFSC1499);四川省干部保健科研项目(川干研2022-212)

Construction and validation of chemotherapy resistance prediction model for ovarian cancer

YU Ping1,2(), ZHOU Min2, SU Dan1,2,()   

  1. 1 North Sichuan Medical College, Nanchong 637100, China
    2 Department of Gynecology and Obstetrics, Sichuan Academy of Medical Sciences and Sichuan Provincial People’sHospital
  • Received:2024-05-07 Revised:2024-07-26 Published:2024-11-15 Online:2024-11-12
  • Contact: △E-mail:303483765@qq.com

摘要:

目的 探讨卵巢癌患者术后化疗发生耐药的影响因素,构建预测模型并评价模型效能。方法 收集经肿瘤细胞减灭术及化疗的407例卵巢癌患者的临床资料,至随访终点根据是否复发分为复发组363例和未复发组44例,其中复发组根据化疗耐药将其分为耐药组59例和敏感组304例。使用单因素分析和Lasso回归筛选变量,建立Logistic模型,用R软件建立列线图并进行评价。结果 与未复发组比较,复发组年龄偏低,低分化比例及FIGO分期Ⅲ—Ⅳ期比例较高(P<0.05)。与敏感组比较,耐药组淋巴结增大、病理类型为非浆液性、FIGO分期Ⅲ—Ⅳ期比例、肿瘤组织免疫组化重组蛋白Ki-67(Ki-67)、蛋白53(P53)、血管内皮生长因子(VEGF)及肾母细胞瘤基因1(WT-1)阳性率较高,手术前后糖类抗原125(CA125)变化率、化疗前后罗马指数(绝经前)变化率及免疫组化蛋白16(P16)阳性率较低(P<0.05)。以Lasso回归筛选出的8个自变量进行Logistic回归,结果显示:术前全腹增强CT有淋巴结增大、病理类型为非浆液性、FIGO分期Ⅲ—Ⅳ期、免疫组化WT1、VEGF阳性,P16阴性是卵巢癌患者发生化疗耐药的独立危险因素。据此建立的列线图模型受试者工作特征曲线下面积为0.837(0.783~0.880),Hosmer-Lemeshow检验结果示模型拟合优度较好,校准曲线及临床决策曲线提示模型有较高的校准度及临床使用度。结论 根据临床数据成功构建了卵巢癌化疗耐药Logistic模型,据此建立的列线图预测模型可有效评估卵巢癌患者发生化疗耐药的风险。

关键词: 卵巢肿瘤, 化放疗, 抗药性, 肿瘤, Logistic模型, 列线图

Abstract:

Objective To investigate the influencing factors on the occurrence of chemo-resistance after postoperative chemotherapy in ovarian cancer patients, and construct a prediction model and evaluate the model efficacy. Methods The clinical data of 407 ovarian cancer patients who underwent tumor cytoreduction and chemotherapy were collected. At the endpoint of follow-up, patients were divided into the recurrence group (n=363) and the non-recurrence group (n=44). Patients in the recurrence group were re-divided into the resistant group (n=59) and the sensitive group (n=304) according to the chemotherapy resistance. Variables were screened using univariate analysis and Lasso regression. Logistic model was established. R software was used to build a nomogram and evaluate it. Results Compared with the non-recurrence group, the age of the recurrence group was lower, and the proportion of low differentiation and the proportion of FIGO stage Ⅲ-Ⅳwere higher (P<0.05). Compared with the sensitive group, in the resistant group, lymph node enlargement, non-serous pathological type, the proportion of FIGO stage Ⅲ-Ⅳ, positive rate of immunohistochemical recombinant proteins Ki-67, protein 53 (P53), vascular endothelial growth factor (VEGF) and nephroblastoma gene 1 (WT-1) were higher. The change rate of glycan antigen 125 (CA125) before and after surgery, the change rates of Rome index (ROMA) (premenopausal) before and after chemotherapy and the positive rate of immunohistochemical protein 16 (P16) were lower (P<0.05). The eight variables screened by Lasso regression were used as independent variables for Logistic regression. Results showed that there were enlarged lymph nodes in preoperativeCT imaging, the pathological type was non-serous, the FIGO stages were Ⅲ-Ⅳ, and immunohistochemistry results of WT1 and VEGF were positive. P16 negative was an independent risk factor for chemo-resistance in ovarian cancer patients. Accordingly, the area under the receiver operating characteristic curve of the nomogram model established was 0.837 (0.783-0.880), and the result of Hosmer-Lemeshow test indicated a good model fit. The calibration curve and the clinical decision curve (DCA) suggested a high calibration and clinical use of the model. Conclusion We have successfully constructed a Logistic model of chemotherapy resistance in ovarian cancer based on clinical data, and the nomogram prediction model can effectively assess the risk of chemo-resistance in ovarian cancer.

Key words: ovarian neoplasms, chemoradiotherapy, drug resistance, neoplasm, Logistic models, nomograms

中图分类号: