Predicting Disease Recurrence, Early Progression, and Overall Survival Following Surgical Resection for High-risk Localized and Locally Advanced Renal Cell Carcinoma

Abstract

Background: Risk stratification for localized renal cell carcinoma (RCC) relies heavily on retrospective models, limiting their generalizability to contemporary cohorts.

Objective: To introduce a contemporary RCC prognostic model, developed using prospective, highly annotated data from a phase III adjuvant trial.

Design, setting, and participants: The model utilizes outcome data from the ECOG-ACRIN 2805 (ASSURE) RCC trial.

Outcome measurements and statistical analysis: The primary outcome for the model is disease-free survival (DFS), with overall survival (OS) and early disease progression (EDP) as secondary outcomes. Model performance was assessed using discrimination and calibration tests.

Results and limitations: A total of 1735 patients were included in the analysis, with 887 DFS events occurring over a median follow-up of 9.6 yr. Five common tumor variables (histology, size, grade, tumor necrosis, and nodal involvement) were included in each model. Tumor histology was the single most powerful predictor for each model outcome. The C-statistics at 1 yr were 78.4% and 81.9% for DFS and OS, respectively. Degradation of the DFS, DFS validation set, and OS model's discriminatory ability was seen over time, with a global c-index of 68.0% (95% confidence interval or CI [65.5, 70.4]), 68.6% [65.1%, 72.2%], and 69.4% (95% CI [66.9%, 71.9%], respectively. The EDP model had a c-index of 75.1% (95% CI [71.3, 79.0]).

Conclusions: We introduce a contemporary RCC recurrence model built and internally validated using prospective and highly annotated data from a clinical trial. Performance characteristics of the current model exceed available prognostic models with the added benefit of being histology inclusive and TNM agnostic.

Patient summary: Important decisions, including treatment protocols, clinical trial eligibility, and life planning, rest on our ability to predict cancer outcomes accurately. Here, we introduce a contemporary renal cell carcinoma prognostic model leveraging high-quality data from a clinical trial. The current model predicts three outcome measures commonly utilized in clinical practice and exceeds the predictive ability of available prognostic models.

Keywords: ASSURE trial; Disease-free survival; Prognostic model; Renal cell carcinoma.

Fig. 1 –

Kaplan-Meier curves for (A) DFS risk groups and (B) OS risk groups. CI = confidence interval DFS = disease-free survival; NA = not available; NR = not reached; OS = overall survival.

Fig. 2 –

Model discrimination measures. AUC = area under the curve; CI = confidence interval; DFS = disease-free survival; EDP = early disease progression; OS = overall survival; ROC = receiver operating characteristics.

Fig. 3 –

(A) Points associated with each risk factor, (B) estimated disease-free survival, (C) estimated overall survival based on risk groups, and (D) output from online calculator. CC = clear cell; DFS = disease-free survival; EDP = early disease progression; IVC = inferior vena cava; NA = not available; OS = overall survival; RCC = renal cell carcinoma. ^a If tumor’s papillary histology cannot be differentiated into type I, II, or mixed, the model user should use the points assigned to clear cell or papillary type II/mixed or >25% CC category. * A 55-yr-old male with a 12-cm clear cell RCC (grade 2), with evidence of renal parenchyma invasion, tumor necrosis, and node-positive disease, would be in the high-risk group for both DFS and OS, and therefore have estimated 5-yr DFS and OS probabilities of 33.1% and 49.0%, respectively. Alternatively, the web-based calculator can be used (https://studies.fccc.edu/nomograms/492).