Colorectal carcinoma grading by quantifying poorly differentiated cell clusters is more reproducible and provides more robust prognostic information than conventional grading.

The most widely used system to define the histological grade of colorectal carcinoma (CRC) is based on the degree of gland formation. This system suffers from significant interobserver variability which may limit its prognostic value and consequently better standardized criteria for the assessment of histological grading of CRC are needed. The present study aims to evaluate and to compare, in a cohort of postsurgical pTNM stage I CRC, conventional histological grading, and a novel grading system based on the number of poorly differentiated clusters of neoplastic cells, in terms of interobserver reproducibility, prognostic significance on progression-free survival, and association with other clinicopathological characteristics. Grading with both systems was performed by two pathologists independently and blinded to the clinicopathological data. Interobserver agreement was higher when grade was assessed by counting poorly differentiated clusters than by assessing the relative proportion of the glandular component. Contrary to conventional grading, the novel system provided significant prognostic information in terms of progression-free survival and was significantly associated with budding, invasive growth, lymphatic invasion, and occult nodal metastases of CRC. In conclusion, our findings suggest that a tumor grading system based on the number of poorly differentiated clusters is more reproducible and provides better prognostic stratification of pTNM stage I CRC patients than conventional grading.