Histologic grading based on counting poorly differentiated clusters in preoperative biopsy predicts nodal involvement and pTNM stage in colorectal cancer patients.

Histologic grading is commonly assessed in colorectal cancer preoperative biopsies. Nevertheless, its clinical impact is limited by low interobserver reproducibility and poor concordance with grading found in the final resection specimen. In the present study, we aimed to investigate the reproducibility, accuracy, and predictive value on lymph node status or pTNM stage of a novel grading system based on the number of poorly differentiated clusters in colorectal cancer preoperative endoscopic biopsies. Grading based on counting poorly differentiated clusters was assessed in 163 colorectal cancer endoscopic biopsies and corresponding surgical specimens. With this system, 152 biopsies could be graded with good interobserver agreement (κ = 0.735). In comparison with the surgical specimens, 75% of colorectal cancers were correctly graded in the biopsy, and 81% of poorly differentiated colorectal cancers were identified at initial biopsy. High poorly differentiated clusters grade in the biopsy was significantly associated with nodal metastasis, high pTNM stage (P < .0001), or histologic features suggestive of more aggressive behavior (tumor budding, perineural invasion, vascular invasion, and infiltrating tumor border) in the surgical specimen. Furthermore, this system identified colorectal cancer with nodal involvement or high pTNM stage with a 78% positive predictive value and 71% and 69% negative predictive values, respectively. Our findings suggest that a grading system based on the quantification of poorly differentiated clusters is feasible in most colorectal cancer endoscopic biopsies. In view of its good reproducibility, accuracy, and predictive value on the anatomical extent of the disease, it may be taken into account for decision-making in colorectal cancer treatment.