Does subdivision of the "atypical" urine cytology increase predictive accuracy for urothelial carcinoma?

Urine cytology is routinely used for early diagnosis and monitoring of patients with hematuria or a history of urothelial carcinoma, but its clinical utility is greatly diminished by a high frequency of "atypical" specimens, reportedly around 20% in the literature. We compared our results with double-stained urine cytology specimens (papanicolaou and acid hematoxylin stains) with published results with only a single or double papanicolaou stain. The acid hematoxylin stain enhanced nuclear chromatin staining, eliminated significant background debris, and improved visibility of diagnostic cells in the presence of obscuring blood. Medical records of all urine cytologies received between 2005 and 2012 in our laboratories were reviewed. The study group consisted of all cases with bladder biopsy follow-up within one year of cytology. Of 43,131 urine cytologies diagnosed in our laboratories, biopsy follow-up results were available within one year in 10,473 cases, including 852 for symptoms and 1,461 for follow-up of bladder cancer. An additional 6,427 cases had cystoscopy results in which no biopsy was obtained. Cases were classified as negative (81.6%), atypical, favor reactive (2.9%), atypical, favor neoplastic (7.3%), suspicious (5.7%), and malignant (2.5%), with subsequent frequencies for urothelial cancer on biopsy of 13.3%, 31.1%, 37.6%, 53.6%, and 74.3%, respectively. No significant difference was found if atypical was subdivided into two categories: favor reactive and favor neoplastic. Subdivision of the atypical category did not improve diagnostic accuracy. Addition of the acid hematoxylin stain decreased the incidence of atypical urine cytologies from about 20% to 10.2%.