Atypical Category of the Johns Hopkins Template Has Higher Risk of Malignancy than the Paris System but the Paris System Is More Applicable for Suspicious Category

Abstract

Introduction: To compare performance of indivisual categories between the Johns Hopkins template and the Paris system for reporting urinary cytology. Methods: Medical records of patients with bladder biopsy and relevant cytology slides were obtained from archived material. Slides were reclassified according to Johns Hopkins Template and the Paris. Results were compared to histologic diagnoses. Results: BD SurePath preparations from 205 cases with biopsy follow up (118 benign, 5 dysplasia, 23 low-and 59 malignant urothelial carcinoma (UC)) were reviewed. There were 2 inadequate specimens in each system. According to the Johns Hopkins Template, there were 96 (46.8%) no urothelial atypia or malignancy, 37 (18%) atypical urothelial cells of uncertain significance (AUC-US), 21 (10.2%) atypical urothelial cells, cannot exclude HGUC (AUC-H), 38 (18.5%) high grade urothelial carcinoma (HGUC) , and 11 (5.4%) low grade urothelial carcinoma (LGUC). The Paris System categorized 111 (54.1%) negative for high grade urothelial carcinoma, 29 (14.1%) atypical urothelial cells (AUC), 25 (12.2%) suspicious for HGUC (SHGUC), 36 (17.6%) HGUC and 2 (1%) LGUC. The Johns Hopkins Template had a sensitivity of 95.6%, specificity of 73.6%, Positive Predictive Value of 61.5%, Negative Predictive Value of 96.3, with an overall diagnostic accuracy of 79.8%. The Paris System had a sensitivity of 93.6%, specificity of 77.9%, Positive Predictive Value of 65.6%, Negative Predictive Value of 96.5, with an overall diagnostic accuracy of 82.8%. The Risk of Malignancy (ROM) for atypical category (AUC-US/AUC) in the Johns Hopkins Template was 43.2%, while it has been 24.1% for the Paris System. The ROM for suspicious category was 47.6% and 68.0% respectively. There was a statistically significant differences between negative and atypical, suspicious, and HGUC categories in each system (p<0.0001). Conclusions: Discrete negative or benign urine cytology had the same sensitivity and specificity between two systems. Although atypical category associated with a higher ROM with the Hopkins template, the ROM for the suspicious category yielded better result with the Paris system.

Examining Longitudinal Markers of Bladder Cancer Recurrence Through a Semi-Autonomous Machine Learning System for Quantifying Specimen Atypia from Urine Cytology

“…After a positive diagnosis of UBC, urine cytology remains an essential longitudinal monitoring tool for patients. However, urine cytology suffers from susceptibility to issues such as specimen quality, inter/intra-observer variability, and 'hedging' towards atypical diagnosis, making it a semiqualitative assessment and vulnerable to individual biases [24][25][26][27][28] . Such factors restrict the predictive value of urine cytology therefore increasing reliance on invasive cystoscopy.…”

Examining longitudinal markers of bladder cancer recurrence through a semiautonomous machine learning system for quantifying specimen atypia from urine cytology

“…However, UC suffers from susceptibility to issues such as specimen quality, inter-/ intraobserver variability, and "hedging" toward atypical diagnosis, making it a semiqualitative assessment and vulnerable to individual biases. [24][25][26][27][28] Such factors restrict the predictive value of UC, therefore increasing reliance on invasive cystoscopy. Cytology specimens have historically been tedious to screen, in part because of the sheer volume of specimens to examine, resulting from regular periodic follow up and the highly variable specimen cellularity.…”

“…After a positive diagnosis of UBC, UC remains an essential longitudinal monitoring tool for patients. However, UC suffers from susceptibility to issues such as specimen quality, inter‐/intraobserver variability, and “hedging” toward atypical diagnosis, making it a semiqualitative assessment and vulnerable to individual biases 24–28 . Such factors restrict the predictive value of UC, therefore increasing reliance on invasive cystoscopy.…”