Abstract
PURPOSE: To predict monosomy 3 (M3) in uveal melanoma with clinical and histopathological features, including nuclear BRCA1-associated protein 1 immunohistochemistry (nBAP1-IHC), and to evaluate its prognostication performance in the Liverpool Uveal Melanoma Prognosticator Online (LUMPO3) algorithm. METHODS: Patients with uveal melanoma excluding iris melanoma were included. Clinical, genetic, and histopathological features, including nBAP1-IHC, were obtained and used to construct an M3 prediction model. Two simplified models mimicking the scenario where only histology or cytology specimens are available were obtained by simplification of the full model; the M3 status predicted by these models was introduced into LUMPO3 to evaluate the performance against the standard LUMPO3 with genetic testing-based M3 status. RESULTS: Data from 2521 patients were used to construct the M3-prediction model. nBAP1-IHC was the most significant predictor (χ2 = 128, P < 0.0001). The simplified M3-prediction models maintained the area under the receiver operating curve at 88%, with Youden sensitivity at 79% and specificity at 82%. Predicted-M3-based LUMPO3 showed that 83% (cytology-based) and 87% (histology-based) of the prognostications fell within ±10% on the metastatic mortality scale of the genetic-based LUMPO3. The designations of metastatic surveillance group, using 3%, 5%, or 10% metastatic mortality prognostication as thresholds, were identical in 96.9%, 95.8%, and 92.5%, respectively (cytology-based), and 97.1%, 95.9%, and 93.8%, respectively (histology-based), using either the genetic-M3-based LUMPO3 or the predicted-M3-based LUMPO3. CONCLUSIONS: M3 can be accurately predicted using clinical and histopathological features, of which nBAP1-IHC is the most significant predictor. The predicted-M3-based LUMPO3 and the genetic-M3-based LUMPO3 are highly consistent in terms of survival prognostication and surveillance planning.