Presentation Authors: Michael Siev*, Audrey Renson, Stella Kang, William Huang, New York, NY, Marc Bjurlin, Chapel Hill, NC
Introduction: Renal cell carcinoma (RCC) has several histologic subtypes, which have shown potential for differential clinical outcomes. Our study objective was to evaluate overall survival (OS) of T1a kidney cancers stratified by histologic subtype and treatment including partial nephrectomy (PN), percutaneous ablation (PA), and radical nephrectomy (RN).
Methods: We queried the National Cancer Data Base (2004-2015) for patients with T1a RCC who were treated surgically. OS was estimated by Kaplan-Meier curves based on histologic subtype and management. Cox proportional regression models were used to determine whether histologic subtypes and management procedure predicted OS. Our adjusted model included age, sex, race/ethnicity, insurance status, median income, proportion without high school diploma, urbanicity, Charlson-Deyo index, tumor grade, and facility volume.
Results: Of the 52,245 T1a masses that met inclusion criteria, 51.1% were clear cell, followed by papillary (34.8%), chromophobe (11.8%), cystic (1.8%), sarcomatoid (0.4%) and collecting duct (0.2%). PN was performed in 51%, RN in 31%, and 18% had PA. Kaplan Meier curves demonstrated differences in survival by histology among clear cell, papillary, chromophobe, and cystic subtypes (all p < 0.001), but not for sarcomatoid (p=0.110) or collecting duct (p=0.392) (Figure 1a). Adjusted Cox regression showed worse OS for PA than PN among patients with clear cell (HR 1.58, 95%CI [1.44-1.73], papillary RCC (1.53 [1.34-1.75]), and chromophobe RCC (2.19 [1.64-2.92]). OS was worse for RN than PN for clear cell (HR 1.38 [1.28-1.50]) papillary (1.34 [1.15-1.56]) and chromophobe RCC (1.92 [1.43-2.58]). Predictive models using cox proportional hazards incorporating histology and surgical procedure alone were limited (c-index 0.63) while adding demographics demonstrated fair predictive power for OS (c-index 0.73) (Figure 1b).
Conclusions: In patients with T1a RCC, OS differs by histologic subtype. PN appears to be superior to both PA and RN. Incorporating histologic subtype and treatment modality into a risk stratification model to predict OS appears to have limited utility compared with variables representing competing risks. Given the preponderance of T1a RCC, more systematic methods of weighing competing risks are needed for deciding upon optimal management.