Background: The prognosis of brain metastases (BM) in colorectal cancer (CRC) is extremely poor, but the incidence is increasing. The performance of existing prognostic classifications such as recursive partitioning analysis (RPA) and graded prognostic assessment (GPA) has never been evaluated in this specific setting. Moreover, the development of nomograms for estimating survival in such patients could be extremely helpful for treating physicians. Patients and methods: Between 2000 and 2013, data from 227 patients with BM from CRC were collected at 8 Italian institutions. Overall survival (OS) was estimated with the Kaplan-Meier method and statistical comparison between curves was performed using the log-rank test. The discriminative ability for OS of RPA and GPA was assessed by the Harrell C-index from univariable Cox models. Putative prognostic factors for OS were also studied by multivariable Cox analysis, using the Harrell C index to evaluate the model discriminative ability. After a backward variable selection, a nomogram was developed to predict median survival time from individual patient- and tumor-related characteristics. The nomogram was externally validated on an independent series. Results: After a median follow-up of 59 months, fifty percent of patients were still at risk at 5 months. The C index was 0.594 and 0.607 for the RPA and GPA classifications, respectively. The C-index associated with the final multivariable Cox model used for developing the nomogram was 0.643; the favorable prognostic factors for survival were lower age (p = 0.061), better Karnofsky performance status (p < 0.001), supratentorial site of BM (p < 0.001), and lower number of BM (p = 0.035). The C index evaluated on the validation series was 0.733, even better than in the development series; also, the calibration of nomogram predictions was good. Conclusion: The C-index associated to the nomogram model was slightly higher than that obtained for the RPA and GPA classifications. Most importantly, the very satisfactory results of nomogram validation on the external series, make us confident that our instrument may assist in prognostic assessment, treatment decision making, and enrollment into clinical trials. (C) 2015 Elsevier Ireland Ltd. All rights reserved.
A new nomogram for estimating survival in patients with brain metastases secondary to colorectal cancer
Rimassa L;
2015-01-01
Abstract
Background: The prognosis of brain metastases (BM) in colorectal cancer (CRC) is extremely poor, but the incidence is increasing. The performance of existing prognostic classifications such as recursive partitioning analysis (RPA) and graded prognostic assessment (GPA) has never been evaluated in this specific setting. Moreover, the development of nomograms for estimating survival in such patients could be extremely helpful for treating physicians. Patients and methods: Between 2000 and 2013, data from 227 patients with BM from CRC were collected at 8 Italian institutions. Overall survival (OS) was estimated with the Kaplan-Meier method and statistical comparison between curves was performed using the log-rank test. The discriminative ability for OS of RPA and GPA was assessed by the Harrell C-index from univariable Cox models. Putative prognostic factors for OS were also studied by multivariable Cox analysis, using the Harrell C index to evaluate the model discriminative ability. After a backward variable selection, a nomogram was developed to predict median survival time from individual patient- and tumor-related characteristics. The nomogram was externally validated on an independent series. Results: After a median follow-up of 59 months, fifty percent of patients were still at risk at 5 months. The C index was 0.594 and 0.607 for the RPA and GPA classifications, respectively. The C-index associated with the final multivariable Cox model used for developing the nomogram was 0.643; the favorable prognostic factors for survival were lower age (p = 0.061), better Karnofsky performance status (p < 0.001), supratentorial site of BM (p < 0.001), and lower number of BM (p = 0.035). The C index evaluated on the validation series was 0.733, even better than in the development series; also, the calibration of nomogram predictions was good. Conclusion: The C-index associated to the nomogram model was slightly higher than that obtained for the RPA and GPA classifications. Most importantly, the very satisfactory results of nomogram validation on the external series, make us confident that our instrument may assist in prognostic assessment, treatment decision making, and enrollment into clinical trials. (C) 2015 Elsevier Ireland Ltd. All rights reserved.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
