Interobserver reproducibility in the diagnosis of invasive micropapillary carcinoma of the urinary tract among urologic pathologists.

Invasive micropapillary carcinoma (IMPC) of the urinary tract is a well-described variant of the urothelial carcinoma with aggressive clinical behavior. Recent studies have proposed that patients with IMPC on transurethral resection should be treated with radical cystectomy regardless of the pathologic stage. Despite the potentially important therapeutic implications of this diagnosis, interobserver variation in the diagnosis of IMPC has not been studied. Sixty digital images, each from hematoxylin and eosin-stained slides, representing 30 invasive urothelial carcinomas (2 images per case), were distributed to 14 genitourinary subspecialists and each pathologist was requested to classify cases as IMPC or not. These cases included "classic" IMPC (n=10) and urothelial carcinoma with retraction and variably sized nests that might potentially be regarded as IMPC (n=20). The following 13 morphologic features were recorded as positive/negative for all cases independent of the reviewers' diagnoses: columnar cells, elongate nests or processes, extensive stromal retraction, lumen formation with internal epithelial tufting, epithelial ring forms, intracytoplasmic vacuolization, multiple nests within the same lacunar space, back-to-back lacunar spaces, epithelial nest anastomosis/confluence, marked nuclear pleomorphism, peripherally oriented nuclei, randomly distributed nuclei, and tumor nest size. In addition, a mean tumor nest size was calculated for each image based on the number of nuclei spanning the width of the nests. Interobserver reproducibility was assessed and the morphologic features were correlated with the classic IMPC and nonclassic/potential IMPC groups. In addition, the relationships between morphologic features, pathologists' interpretations, and case type (classic IMPC vs. nonclassic/potential IMPC) were evaluated using unsupervised hierarchical clustering analysis. Interobserver reproducibility for a diagnosis of IMPC in the 30 study cases was moderate (kappa: 0.54). Although classification as IMPC among the 10 "classic" IMPC cases was relatively uniform (93% agreement), the classification in the subset of 20 invasive urothelial carcinomas with extensive retraction and varying sized tumor nests was more variable. Multiple nests within the same lacunar space had the highest association with a diagnosis of classic IMPC. These findings suggest that more study of IMPC is needed to identify the individual pathologic features that might potentially correlate with an aggressive outcome and response to intravesical therapy.