| Objectives: This study aimed to explore the risk factors for platinum-resistant recurrence in patients with advanced epithelial ovarian cancer(EOC)after first-line treatment,as well as the prognostic factors for Overall survival(OS)in platinum-sensitive and platinumresistant patients based on the multi-center data of National Union of Real-world Gynecological Oncology Research and Patient Management(NUWA)platform.The prediction models were established and evaluated in the above patient groups to provide real-world evidence for personalized management of risk factors.Methods: The data sources were electronic medical records and follow-up data from 7 hospitals in NUWA Platform.The research objects were International Federation of Gynecology and Obstetrics(FIGO)stage Ⅲ-Ⅳ EOC patients who were diagnosed between January 2006 and June 2020.Detailed methods were as follows:(1)To compare the clinical characteristics,first-line treatment and survival between platinum-sensitive and platinumresistant patients.(2)Patients from 4 centers were allocated as the training set,and those from the other 3 centers were included as the external validation set.Univariate and multivariate Logistic regression analyses were performed in the training set to establish a nomogram for predicting platinum-resistant recurrence.The performance of the nomogram was evaluated by the following indicators: correct classification rate,Receiver operator characteristic(ROC)curves,Area under the curve(AUC),Hosmer-Lemeshow goodnessof-fit test,calibration curves,as well as Net reclassification improvement(NRI),Integrated discrimination improvement(IDI)and Decision curve analyses(DCA)compared with the basic histological classification and staging model.(3)Univariate and multivariate Cox regression analyses were performed in the training sets of platinum-sensitive and platinumresistant patients respectively to construct nomograms for predicting OS,and the following indicators were used to evaluate the performance: survival curves of different risk groups,ROC curves,AUC,calibration curves,as well as NRI,IDI,and DCA compared with basic models.Results: A total of 2307 FIGO stage Ⅲ-Ⅳ EOC patients were included in the study,including 1693 platinum-sensitive patients and 614 platinum-resistant patients.Detailed results were as follows.(1)Compared with platinum-sensitive patients,the proportions of cases with comorbidities,mucinous or clear cell histotype,FIGO stage IIIC& IV,severe ascites,positive pathology of ascites,lymphovascular tumor emboli,metastatic or unresected lymph nodes,R2 cytoreduction,Clavien-Dindo grade III or above postoperative complications,more than 4 cycles of neoadjuvant chemotherapy(NACT),less than 6 cycles of first-line chemotherapy and intravenous chemotherapy alone were higher in platinumresistant patients.The median progression free survival(PFS)of platinum-sensitive patients was 29.4 months,with the median OS being 67.8 months.The median PFS and OS of platinum-resistant patients were 8.4 months and 25.2 months.(2)The variables applied to construct the nomogram for the risk of platinum-resistant recurrence were comorbidities,histological type,FIGO stage,ascites volume,lymphovascular tumor emboli,residual disease,number of NACT cycles,number of first-line chemotherapy cycles and chemotherapy ways.In the training set,the correct classification rate was 75.8%,and the AUC was 0.759.Compared with the basic model,the NRI was 0.425 and IDI was 0.115.In the external validation set,the correct classification rate was 80.8%,and the AUC was 0.732.Compared with the base model,the NRI and IDI were 0.494 and 0.126,respectively.The calibration curves of the nomogram displayed high consistency with reference lines in the both datasets,with the Hosmer-Lemeshow test showing no significant difference.The nomogram’s decision curve was higher than reference lines and that of the basic model at each threshold probability in the both datasets except for the probability of 0.75-0.85 in the external validation set.(3)The variables included in the nomogram for predicting OS of platinum-sensitive patients were age at diagnosis,histological type,FIGO stage,ascites volume,lymph node metastasis,residual disease,number of NACT cycles and number of first-line chemotherapy cycles.The AUCs of the nomogram for predicting 3-year and 5-year OS were 0.777,0.807 in the training cohort and 0.685,0.761 in the external validation cohort,respectively.There were significant differences in OS among patients with different risk levels in both datasets.Compared with the basic model,the nomogram had an NRI of 0.416 and an IDI of 0.140 for 3-year OS,along with an NRI of 0.433 and an IDI of 0.197 for 5-year OS in the training cohort.In the external validation set,the NRI and IDI for 3-year OS were 0.342 and 0.090,and the NRI and IDI for 5-year OS were 0.464 and 0.151,respectively.The calibration curves of the nomogram for predicting 3-year and 5-year OS were highly consistent with the reference lines.With regards to the DCA,the net clinical benefit of the nomogram at each threshold probability was higher than that of the reference lines and basic model except for the probability of 0.66-0.89 in the external validation set.(4)The following variables were included in the nomogram for OS of platinum-resistant patients: age at diagnosis,histological type,FIGO stage,pathology of ascites,lymph node metastasis,residual disease and number of first-line chemotherapy cycles.The AUCs for predicting 1-year,3-year and 5-year OS were 0.785,0.727,0.823 in the training set,and 0.694,0.655,0.711 in the external validation set,respectively.In the both data sets,there were significant differences in OS among patients with different risk levels.Compared with the basic model,the NRIs of the nomogram in the training set for 1-year,3-year and 5-year OS were 0.403,0.246 and 0.335,and the IDIs were 0.095,0.086 and 0.080.In the external validation set,the NRIs for 1-year,3-ear and 5-year OS were 0.183,0.136 and 0.136,and the IDIs were 0.006,0.025 and 0.030.In the training set,calibration curves were highly consistent with the reference lines.In the external validation set,the calibration curves for predicting 1-year and 5-year OS were highly consistent with the reference lines,but there was some deviation in predicting 3-year OS.In regard to DCA,the net clinical benefit of the nomogram at each threshold probability was higher than that of the reference lines and basic model except for the probability of 0.81-0.93 in the external validation set.Conclusions:(1)There were significant differences in clinical characteristics,first-line treatment and survival between platinum-sensitive and platinum-resistant patients with advanced EOC.(2)The nomogram for predicting platinum-resistant recurrence of advanced EOC exhibited good discrimination and accuracy in both training set and external validation set.Compared with the basic histological classification and staging models,the nomogram had improved predictive ability and clinical benefit,as well as compatibility in multi-center patients.(3)The nomogram for predicting OS of platinum-sensitive patients showed good discrimination,accuracy and net clinical benefit in the both data sets.The nomogram for predicting OS of platinum-resistant population had good performance in the training set,but its applicability required improvement in a larger external validation set. |
PDF Full Text Request