Keratin 17 immunocytochemistry in urine cytology for urothelial carcinoma detection: Diagnostic performance and interobserver agreement in a prospective real‐world cohort

BACKGROUND: Urine cytology is recommended for noninvasive detection of urothelial carcinoma (UC) but shows limited sensitivity, particularly in low-grade disease. Keratin 17 (K17) has emerged as a promising immunocytochemical marker. This study evaluated the diagnostic performance and interobserver agreement of K17 in a routine setting. METHODS: This prospective cross-sectional study included 184 urine samples from patients undergoing diagnostic evaluation for suspected UC. K17 immunocytochemistry was performed with a fully automated platform and independently evaluated by two blinded raters. Cytology was classified according to The Paris System for Reporting Urinary Cytology. UC status was determined by histology when available or by follow-up and multidisciplinary assessment. Diagnostic accuracy was assessed by receiver operating characteristic (ROC) analysis; interobserver reliability was assessed with the Cohen κ and intraclass correlation coefficient (ICC). Multivariable logistic regression evaluated combined diagnostic approaches. RESULTS: K17 expression was significantly higher in UC than in non-UC samples (p < .001). Across raters, area under the ROC curve (AUC) values ranged from 0.68 to 0.71. Sensitivity was 52.9%-65.9%, with a specificity of 73.7%-82.8%. Sensitivity ranged from 45.5% to 54.5% for low-grade tumors and from 57.7% to 73.1% for high-grade tumors. In stage-based analyses, sensitivity was 49.2%-61.5% for non-muscle-invasive UC and 65%-80% for muscle-invasive UC. Combining K17 with cytology numerically improved diagnostic discrimination (AUC value, 0.76; 95% CI, 0.69-0.83). Interobserver agreement was moderate to substantial (κ = 0.684; ICC, 0.653). Limitations included single-center design, partial histological verification (81%), and post hoc optimized cutoffs. CONCLUSIONS: K17 immunocytochemistry is a reproducible adjunct to urine cytology, with moderate diagnostic performance. It may warrant further evaluation as an adjunctive marker, particularly in low-grade and cytologically equivocal cases.