Statistical analysis of cytologic features useful in separation of metastatic urothelial carcinoma from other metastatic epithelial malignancies.

Journal Article (Journal Article)

The identification of a site of origin and direction of differentiation for metastatic neoplasms is clinically important, but is often difficult purely by cytologic analysis of aspirated material. Cytologic separation of metastatic urothelial carcinoma (UC) from other moderate or poorly differentiated epithelial malignancies is difficult, with few cytologic criteria identified in the literature as valuable for this distinction. Several investigators have suggested that "cercariform cells" (CCs) are highly correlated with the presence of metastatic UC. We statistically analyzed the utility of 37 cytomorphologic features for the recognition of urothelial differentiation in a series of 26 metastatic UCs, 10 metastatic squamous cell carcinomas (SCCs), and 15 metastatic adenocarcinomas (ADCs). All specimens had been obtained from metastatic deposits in the lung, liver, lymph nodes, or soft tissues. Stepwise discriminate function analysis with all three diagnoses showed that the strongest discrimination could be made using the findings of waxy metaplastic cytoplasm, with significant increments in prediction added by analysis for (CCs) followed by spindle cells, multiple nucleoli, and columnar-shaped cells. The combination of these five variables accurately predicted 90% of all diagnoses, including 26 accurate diagnoses of UC, 9 accurate diagnoses of SCC, and 11 accurate diagnoses of ADC. CCs were present in highest numbers in UCs and present least frequently in cases of poorly differentiated ADC. While CCs were useful in the identification of UC, CCs occurred in a significant number of SCCs, limiting their diagnostic value as a single variable.

Full Text

Duke Authors

Cited Authors

  • Layfield, LJ; Jones, C; Hirschowitz, S

Published Date

  • December 2003

Published In

Volume / Issue

  • 29 / 6

Start / End Page

  • 334 - 338

PubMed ID

  • 14648790

International Standard Serial Number (ISSN)

  • 8755-1039

Digital Object Identifier (DOI)

  • 10.1002/dc.10380


  • eng

Conference Location

  • United States