Prognostic value of morphometric features and the callender classification in uveal melanomas.

Abstract

Cytomorphometric analysis of uveal melanoma has focused on nucleolar characteristics, although nuclear measurements are easier and more reproducible. In this study, we examine a restricted set of nuclear and nucleolar cytomorphometric variables in the search for the most quantifiable and reproducible prognostic parameter among these and discover a means of assessing the malignancy of tumors within the different Callender subgroups. Uveal melanomas from 94 patients with a minimum follow-up of 5 years underwent nuclear and nucleolar morphometric analysis using a digitized interactive video overlay system. Statistical analysis included univariate and multivariate survival analysis on morphometric variables and classic prognostic parameters (such as Callender cell type, size, etc.). Univariate analysis of cytomorphometric parameters showed the standard deviation of nuclear area (NASD) to be the most significant variable (P < 0.0001: Mantel-Cox, 15.2). This was followed by the difference between nuclear and nucleolar areas (P < 0.002; Mantel-Cox, 9.5), the standard deviation of nucleolar area (SDNA) (P < 0.01; Mantel-Cox, 6.4), and the standard deviation of the shortest nuclear axis (P < 0.02; Mantel-Cox, 6.4). There was a trend for significance with the mean of the ten largest nucleoli (P < 0.07). Multivariate analysis of clinical and cytomorphometric variables showed improvement of prognostic prediction with a combination of the NASD, largest tumor dimension, and glaucoma (P < 0.0001; Mantel-Cox, 26.3). Callender cell type was strongly correlated with the NASD (P < 0.0002), the mean axis ratio of the nuclear area (P < 0.0007), and the mean of the ten largest nucleoli (P < 0.0007). Cytomorphometric analysis of uveal melanoma should include nuclear area and may identify tumors of differing malignancy within the same Callender class.