Architectural Patterns are a Relevant Morphologic Grading System for Clear Cell Renal Cell Carcinoma Prognosis Assessment Comparisons With WHO/ISUP Grade and Integrated Staging Systems

We developed and validated an architecture-based grading for clear cell renal cell carcinoma (ccRCC) in an observational retrospective cohort study including 506 tumors (principal cohort, n=254; validation cohort, n=252). Study endpoints were disease-free survival (DFS) and cancer-specific survival (CSS). Relationships with outcome were analyzed using Harrell concordance index, time-dependent receiver operating characteristic curve, area under curve, and Cox regression model. An architecture-based grading was devised on positive likelihood ratio (LR+) for DFS at 50 months as follows: grade 1 (LR+<0.8), cystic, compact, acinar, clear cell papillary RCC-like, and/or regressive patterns; grade 2 (1.2 <= LR+< 5), large nest, alveolar, papillary, chromophobe/oncocytic cell-like, eosinophilic hyaline globule, and/or intratumoral inflammatory reaction patterns; grade 3 (5 <= LR+<10), rhabdoid, tumor giant cell, enlarged vascular space, and/or hereditary leiomyomatosis renal cell carcinoma (HLRCC)-like patterns; grade 4 (LR+>= 10), sarcomatoid, infiltrative growth patterns, and lymphatic invasion. In the principal cohort, 3-tier (grades 1-2, 3, and 4) and 4-tier architectural scores outperformed World Health Organization/International Society of Urological Pathology, and World Health Organization/International Society of Urological Pathology+necrosis gradings for DFS and CSS, and constituted an independent predictor for DFS (hazard ratio [HR]=5.91; P<6.7E-10) and CSS (HR=4.49; P=2.2E-03), retained in the localized (pT1-3N0M0) ccRCC subgroup (HR=6.10; P=1.3E-07 for DFS, and HR=20.09; P=9.4E-05 for CSS). On comparing with integrated staging systems, architectural grade with 1 morphologic datum remained an independent predictor of CSS, as did University of California Los Angeles Integrated Staging System and SSIGN, and was associated with the highest HR (HR=2.60; P=9.1E-04 in all patients; HR=4.38; P=2.0E-05 in the localized ccRCC subgroup). Architecture-based score for ccRCC outperforms all other morphologic grading systems and constitutes an independent predictor for DFS and CSS. As the predictive values of 3-tier and 4-tier architecture-based scores were similar throughout the study, we proposed to keep the simplified version as the final score, and to define 3 risk groups as follows: low risk (grades 1 to 2), intermediate risk (grade 3), and high risk (grade 4).