Gastrointestinal Stromal Tumors Risk Stratification Utilizing Phospho-Histone H3 Evaluated by Manual Counting and Computer-Assisted Image Analysis.

Abstract

Background. Risk of progressive disease of gastrointestinal stromal tumors (GISTs) relies on mitotic index, size, and location of the tumor. However, manual mitotic counting on hematoxylin and eosin-stained slides (MMC-HE) is inefficient with low reproducibility. Manual count of phospho-histone H3 (MC-PHH3)-positive cells on immunohistochemical stained slides has been shown to have comparable reliability with MMC-HE. This study aims to confirm the reliability of MC-PHH3 in GISTs compared with MMC-HE and then to further compare MC-PHH3 with computer-assisted image analysis of PHH3-positive cells (Comp-PHH3). Methods. The study included 119 patients with GISTs. PHH3 stains were performed. MC-PHH3 was assessed as counts/5 mm2 high-power fields. Whole slide images were captured and the tumor area with greatest mitotic activity was manually identified. The PHH3-positive cells were automatically counted in 0.5 mm2 using Ventana Virtuoso software. Results. MMC-HE ranged from 0 to 157/5 mm2. MC-PHH3 ranged from 0 to 35.6/5 mm2. Comp-PHH3 ranged from 0 to 66/0.5 mm2. Interclass correlation coefficient (ICC) indicates good agreement between the 3 pathologists for MC-PHH3 (ICC = 0.74, P = .42). There is a strong correlation between MMC-HE and MC-PHH3. The Spearman correlation coefficient was 0.63 (P < .0001). Lin's concordance further indicated a moderate diagnostic agreement between MC-PHH3 and Comp-PHH3. Conclusion. MC-PHH3 is proposed as a superior alternative to MMC-HE with potential application in GIST reporting and prognostication. Furthermore, Comp-PHH3 may be a valid alternative to MC-PHH3.