Predicting Clinical Outcomes for Patients with Sarcomatoid Urothelial Carcinoma - Expert Commentary
The median age at diagnosis was 71 years, and 58% of patients were diagnosed after age 70. Males represented 68% of the cohort. Most patients had muscle-invasive disease. With respect to treatment, 48% had undergone radical cystectomy, 37% had received chemotherapy, and 13% had received radiation. The median follow-up time was 64 months, the median overall survival (OS) was 12 months, and the median disease-specific survival (DSS) was 17 months. Patients were divided into a training cohort (n = 507) and a validation cohort (n = 234). In a multivariate Cox regression analysis of OS in the training cohort, the following six variables were significant: age, sex, tumor size, combined SEER stage, radical cystectomy, and chemotherapy. In the equivalent analysis for DSS, the following five variables were significant: sex, tumor size, combined SEER stage, radical cystectomy, and chemotherapy. Prognostic nomograms for 3- and 5-year mortality were subsequently developed based on these regression models for OS and DSS.
In internal validation steps, C-statistic for OS and DSS nomograms in the training cohort were 0.68 (CI 95%, 0.65 – 0.72) and 0.67 (0.63 – 0.71), respectively. The corresponding C-statistics generated using an available staging system for prognostication (AJCC 8th edition) were 0.59 (0.55 – 0.62) and 0.60 (0.57 – 0.63), respectively. The likelihood ratio test (LRT) and Akaike information criterion (AIC) were used to determine the discriminative performance of the nomograms. This showed that the nomogram exhibited significantly better performance than the AJCC model. Calibration curves revealed adequate consistency between nomogram-predicted trends and actual survival at three and five years. The authors then performed external validation in the validation cohort, characterized by nomogram C-statistics of 0.66 (0.63 – 0.70) for OS and 0.67 (0.64 – 0.70) for DSS. C-statistics produced by the AJCC model were 0.59 (0.56 – 0.62) for OS and 0.61 (0.58 – 0.65) for DSS, respectively. Based on LRT calculation, nomograms were found to have significantly higher performance than the AJCC system. There was an appropriate level of consistency between nomogram-predicted survival curves and patient survival at 3- and 5-year time points. A decision curve analysis was conducted to demonstrate the two systems' clinical applicability and showed that the nomogram curves were associated with consistently higher net benefits for OS and DSS than the AJCC curves. Finally, to perform risk stratification, authors assigned each patient a set of cumulative risk points based on nomograms to yield a mean score that could be categorized into low, intermediate, high, or very high risk. In the training cohort, the calculated mean score for OS was 157 and 119 for DSS. Higher risk was associated with worse survival.
The prognostic nomograms developed for SUC patient survival exhibited significantly higher performance than existing prognostic tools such as the AJCC model. Nomograms are relatively easy to use in the clinical setting and are practical and widely applicable. These can be used as aids in risk assessment and subsequent treatment decisions. Nevertheless, due to the use of a patient database, some relevant variables, such as details regarding treatment regimens, were unavailable.
Written by: Bishoy M. Faltas, MD, Director of Bladder Cancer Research, Englander Institute for Precision Medicine, Weill Cornell Medicine
Reference:
Read the Abstract