| Objectives Brain metastasis is the important complication and cause of death in patients with cancer.About 10% to 30% of lung cancer patients meet brain metastases.Stereotactic radiosurgery(SRS)is widely used for patients with brain metastases.With the progress of molecular targeted therapy and immunotherapy,the survival of patients with lung cancer has been improved.This trend and the widespread use of magnetic resonance imaging(MRI)leads to more patients with brain metastases.To predict survival in patients with brain metastases,data from 1200 patients in three clinical trials based on the Radiation Therapy Oncology Group(RTOG)were used to develop Recursive Partitioning Analysis(RPA).Another scoring system based on data from 1960 patients in five RTOG clinical trials(RTOG7916,8528,8905,9104,and 9508)was Graded Prognostic Assessment(GPA).The two prognostic systems were designed to divide patients into different groups for survival prediction and have been widely used in practice.However,the value of the two scoring systems to predict the specific and individual survival probability following Cyberknife radiosurgery in patients with brain metastases needs to be validated.In our study,we evaluated the predictive value for early death(≤3months)and long-term(≥12months)survival probability in patients with intracranial metastases underwent Cyberknife radiosurgery.We also developed two nomograms to predict the short(≤6months)and long-term(≥12months)survival in patients with metastases of lung cancer.The nomograms were validated and compared with the prognostic scoring systems,and can be used to predict the individualized survival probability.In order to comprehensively evaluate the prognostic factors for overall survival of patients with brain metastases treated with Cyberknife,the present study analyzed patients treated with second SRS for intracranial recurrences and metastases following the initial radiosurgery treatment.In addition,analysis of factors influencing survival after intracranial progression provides a basis for guiding the choice of treatment options.Methods The research was divided into four parts.The first part analyzed 492 patients who had undergone Cyberknife radiosurgery for intracranial metastases in Tianjin medical university cancer institute and hospital.Inclusion criteria were patients with intracranial metastases from whom the date of death was known,or patients with intracranial metastases who had a follow-up of at least 1 year.The estimated early death(≤3months)and long-term(≥12months)survival probabilities were predicted using RPA and GPA.Discrimination,calibration and clinical usefulness were assessed to evaluate models.In the second part of the study,patients with brain metastases of lung cancer were systematically reviewed from September 2006 to July 2016.All patients were treated with Cyberknife in Tianjin medical university cancer institute and hospital.Inclusion criteria were patients with intracranial metastases from whom the date of death was known,or patients with intracranial metastases who had a follow-up of at least 1 year.All patients were randomly divided into training cohort and validation cohort.In the training cohort,two nomograms were developed for prediction of short(≤6months)and long-term survival(≥12months),respectively.Nomograms were developed based on logistic regression analysis.The area under the ROC curve(AUC)was obtained using Receiver Operating Characteristics(ROC)analysis.The Hosmer-Lemeshow test was used to determine the goodness of fit of the nomograms.The decision curve analysis was applied to evaluate the clinical benefit.In the third part,a total of 63 patients were treated with a second course of Cyberknife radiosurgery for intracranial progression,including recurrences and new metastases after initial stereotactic radiosurgery.A Cox proportional hazards model was used to analyze predictive factors for survival.The fourth part retrospectively analyzed the clinical data of patients with intracranial progression following initial Cyberknife radiosurgery from September 2006 to December 2015 in Tianjin medical university cancer institute and hospital.The factors influencing the overall survival after intracranial progression were analyzed.Kaplan-Meier and the Cox proportional hazard model was used for univariate and multivariate analysis.Differences in treatment modalities between groups were compared using a chi-square test.Results The results of the first part revealed that in these patients,11.4% survived no more than 3 months and 40.4% survived more than 12 months.There was no significant difference of discrimination between RPA and GPA in predicting the early death.The area under the curve(AUC)was 0.638 and 0.580,respectively(P> 0.05).The multi-factor model exhibited a higher AUC than the two systems(P<0.01).The decision curve analysis indicated that the multi-factor model showed better clinical usefulness than GPA and RPA.GPA exhibited superior discriminative ability for predicting long-term survival than RPA.The AUC was 0.638 and 0.580,respectively(P< 0.01).The decision curve analysis also suggested the favorable clinical usefulness of GPA and the multi-factor model.In the second part,the median overall survival of 403 patients were 14 months.In the training cohort,the following characteristics of patients including KPS,extracranial metastases,primary status,EGFR status and primary resection were used to develop two nomograms.The ROC analysis of the training cohort showed that the AUC of the nomograms for short(≤6months)and long-term(≥12months)survival were 0.69 and 0.67,respectively,which were significantly better than scoring systems of RPA,GPA,GGS and Rades(AUC 0.52-0.61 and 0.54-0.61,respectively),P<0.05.Similar outcomes were observed in the validation cohort.With AUC of 0.75 and 0.73,nomograms predicted the short and long-term survival significantly better than RPA,GPA,GGS,BSBM and Rades(AUC 0.56-0.62 and 0.57-0.66,respectively),P< 0.05.In the training cohort,the calibration plots of the nomograms for predicting both short and long-term survival were ideal.Hosmer-Lemeshow test also showed good fit.The decision curve analysis indicated that the nomograms showed better clinical usefulness than scoring systems in predicting short and long-term survival.Similar results were observed in the validation cohort.The third part of the study revealed that the median overall survival following the second course of Cyberknife radiosurgery were 18 months.On multivariate analysis,total plan target volume(t PTV)and minimum dose were associated with overall survival significantly.Results of the fourth part: The interval to intracranial progression,salvage treatment and the number of extracranial organs/regional metastatic are independent prognostic factors for overall survival after intracranial progression.Conclusions The predictive value of GPA and RPA for early death in patients with intracranial metastases underwent Cyberknife radiosurgery,while GPA outperformed RPA in predicting long-term survival probability.Further exploration of models with multiple factors including various clinical and therapy indicators may improve the predictive power.This study developed and validated two clinical nomograms for predicting short(≤6 months)and long-term(≥12 months)survival in patients with brain metastases of lung cancer.These models not only showed better predictive power than the scoring system,but also could assessed the individualized survival probability of patients.A second course of Cyberknife radiosurgery appears to be an effective salvage option for brain progression following initial SRS.The t PTV shows prediction for overall survival.Tumor volume of initial SRS may influence selection of the potential population that may benefit from salvage radiosurgery.For patients with brain metastases underwent Cyberknife radiosurgery,the longer interval to intracranial progression,aggressive treatment and fewer extracranial metastases were significantly associated with higher overall survival rates after intracranial progression. |
PDF Full Text Request