Prognostic significance and interobserver variability of histologic grading systems for endometrial carcinoma

BACKGROUND. The most widely used histologic grading system for endometrial carcinoma is the three-tiered International Federation of Gynecology and Obstetrics (FIGO) system. Although FIGO grading has significant predictive value, the reproducibility of Grade 2 is limited. Recently, a binary grading system was proposed based on the amount of solid growth, the pattern of myometrial invasion, and the presence of tumor cell necrosis. The authors analyzed and compared the prognostic significance and the interobserver variability of both grading systems and of the three criteria for the binary grading system. METHODS. Eight hundred patients with Stage I-III endometrioid endometrial carcinoma were reviewed and graded independently by two pathologists according to the three-tiered FIGO grading system and the novel binary grading system. RESULTS. The interobserver agreement for both-systems was moderate, with 70% and 73% agreement rates for the FIGO (kappa = 0.41) and binary (K = 0.39) grading systems, respectively. when converting the FIGO grading system into an artificial, 2-tiered grading system (Grade 3 vs. Grades 1-2), the agreement was much better (agreement rate, 85%; kappa = 0.58). Of the 3 criteria for the binary grading system, amount of solid growth (less than or equal to 50% vs. > 50%) had the greatest reproducibility (agreement rate, 80%; kappa = 0.50). Both the 2-tiered FIGO grading system and the binary grading system were significant predictors of local recurrence, distant recurrence, and disease-specific survival (hazard ratios [HRs]: 1.7, 2.5, and 2.6, respectively, for FIGO and 2.1, 4.1, and 3.8, respectively, for the binary grading system). The amount of solid growth also was a strong prognostic factor for these three endpoints (HRs: 2.4, 3.9, and 3.8, respectively). CONCLUSIONS. Both the binary grading system and the FIGO grading system had strong prognostic significance. Their reproducibility, however, was limited. A simple architectural binary grading system that divided tumors into low-grade lesions and high-grade lesions based on the proportion of solid growth (less than or equal to 5 50% or > 50%) had superior, prognostic power and greater reproducibility. (C) 2004 American Cancer Society.