Accuracy and interobserver variability of the cytologic diagnosis of low-grade urothelial carcinoma in instrumented urinary tract cytology specimens.

Abstract

The aim of this study was to determine the accuracy of diagnosis and level of agreement of the cytologic diagnosis of low-grade urothelial carcinoma (LGUC) among experienced cytopathologists to help the development of the upcoming classification scheme ("The Paris System") of urinary cytology (UTCy). Forty ThinPrep (Cytyc, Boxborough, MA) UTCy specimens (19 with corresponding biopsy specimens diagnosed as LGUC and 21 with negative biopsy and/or cystoscopy specimens) were blindly reviewed by six cytopathologists from three institutions and diagnosed as negative for malignancy (NM), atypical urothelial cells (AUCs), and LGUC. Interobserver agreement was measured with the weighted κ statistic. Each observer's receiver operating characteristic (ROC) curves and the area under the ROC curve (AUROC) were calculated. Four to six reviewers diagnosed LGUC in five (26%) of 19 UTCy specimens, corresponding to a biopsy diagnosis of LGUC. Overall agreement was fair (κ = 0.30). After collapsing AUC with NM, the reviewers' ranges of sensitivity, specificity, positive predictive value, and negative predictive value for LGUC were 21% to 53%, 81% to 95%, 71% to 90%, and 57% to 67%, respectively. AUROC varied from 0.66 to 0.74. We found that UTCy has a low sensitivity but relatively high specificity, resulting in poor to fair accuracy for the diagnosis of LGUC; the level of agreement between reviewers was only fair to moderate.