Predicting Renal Cell Carcinoma Progression after Surgery
Sources of Funding: None
Introduction
Multiple algorithms exist for the prediction of progression after surgical treatment of localized renal cell carcinoma (RCC); however, most are limited to clear cell (ccRCC) only, and have not been updated with contemporary pathologic assessment. We therefore sought to develop predictive models for progression in ccRCC, papillary RCC (papRCC), and chromophobe RCC (chrRCC).
Methods
Binephric patients treated with radical or partial nephrectomy for sporadic, unilateral M0 ccRCC, papRCC, or chrRCC between 1980 and 2010 were identified. All patients had their pathology slides re-reviewed by one pathologist, blinded to patient outcome. Associations with time to progression (defined as local recurrence, distant metastasis, or death from RCC) were evaluated with multivariable Cox proportional hazards regression with stepwise selection using a 500-sample bootstrap resampling approach.
Results
In total, 3,549 patients were identified: 2,726 (76.8%) with ccRCC, 601 (16.9%) with papRCC, and 222 (6.3%) with chrRCC. For patients with ccRCC, median follow-up was 9.9 years during which time 862 progressed. Features independently associated with ccRCC progression were constitutional symptoms, grade, coagulative necrosis, sarcomatoid differentiation, tumor size, fat invasion, tumor thrombus level, extension beyond Gerota's fascia, and pN classification. The c-index of this model was 0.83. For papRCC patients, median follow-up was 10.3 years during which time 66 had progressed. Features associated with papRCC progression were grade, fat invasion, and tumor thrombus level, resulting in a c-index of 0.77. For chrRCC patients, median follow-up was 9.1 years during which time 35 had progressed. Features associated with progression included sarcomatoid differentiation, fat invasion, and pN classification, resulting in a c-index of 0.77. Predicted 10-year progression-free survivals for patients without any risk factors were 96%, 96%, and 91% for ccRCC, papRCC, and chrRCC, respectively.
Conclusions
Using routine clinical and pathologic data, we generated 3 histology-specific predictive models for progression after surgical management of RCC. These models have excellent discrimination and may prove important in patient counseling and follow-up planning after surgical intervention.
Funding
None
Christine Lohse
John Cheville
Harras Zaid
Stephen Boorjian
Igor Frank
R. Houston Thompson
Bradley Leibovich