Optimizing the use of Ki-67 proliferative index as a prognostic biomarker in meningiomas using digital analysis

OBJECTIVE Ki-67 immunohistochemistry is widely used as a prognostic marker in meningiomas, but visual estimations tend to be imprecise. Whether the average Ki-67 over an entire slide, a particular block, or areas of high staining (hotspots) is prognostic for recurrence-free survival (RFS) and overall survival (OS) is unknown. This study aimed to generate evidence-based recommendations for the optimal use of Ki-67 immunohistochemistry in the workup of meningiomas. METHODS All tissue blocks from a retrospective cohort of 221 patients with primary meningioma (141 WHO grade 1, 64 WHO grade 2, 16 WHO grade 3) were immunostained for Ki-67 and scanned using automated digital analysis software. QuPath was used to quantify the average Ki-67 proliferation index per slide as well as the Ki-67 hotspot in a 2.2-mm 2 area within each slide. The best block was defined subjectively by a neuropathologist as the most representative tissue block from each case. The pathology report Ki-67 was determined by visual estimation. Age, sex, WHO grade, extent of resection, tumor location, and quantitative Ki-67 labeling were tested to determine risk factors for RFS and OS. RESULTS Multivariable analyses demonstrated that WHO grade 2 (HR 2.42, p = 0.018), subtotal resection (HR 8.16, p < 0.0001), near-total resection (HR 2.24, p = 0.041), QuPath Ki-67 averaged across all blocks (HR per % increase = 1.72, p = 0.030), and pathology report Ki-67 (HR per % increase = 1.05, p = 0.0026) were associated with shorter RFS. The average Ki-67 in the best block, maximum across all slides, and maximum hotspot in the best block were not associated with RFS. Only male sex was independently associated with shorter OS (HR 8.52, p = 0.0003). The pathology report Ki-67 was, on average, 6.5 times higher than the QuPath average. CONCLUSIONS These data on Ki-67 in meningiomas indicate the following: 1) visual estimation substantially overestimates Ki-67, 2) digital quantification of average Ki-67 across all tissue blocks provides more prognostic information than small hotspot regions or an entire single block, and 3) Ki-67 is not informative for OS. The results suggest that best practices for incorporating Ki-67 into meningioma prognostication include digital quantification of average Ki-67 over multiple blocks.