Nucleolar diameter and microvascular factors as independent predictors of mortality from malignant melanoma of the choroid and ciliary body.

PURPOSE To determine whether nucleolar diameter and microvascular factors are independent predictors of mortality in malignant uveal melanoma of the choroid and ciliary body. METHOD A population-based retrospective cohort study was conducted of melanoma-specific and all-cause mortality in 167 consecutive patients who had an eye enucleated because of choroidal and ciliary body melanoma from 1972 through 1981. The largest nucleoli were measured from digital photographs of silver-stained tumors along a central 5-mm-wide linear field parallel to the base of the tumor. The mean of the 10 largest nucleoli (MLN) was calculated. Microvascular loops and networks and microvascular density (MVD) were assessed. Kaplan-Meier and Cox regression analyses were performed. Associations between MLN and other variables were determined. RESULTS The MLN could be determined in 126 (75%) melanomas. It ranged from 2.60 to 6.18 microm (median, 4.05). The association of large MLN with the presence of epithelioid cells (P = 0.017) and high MVD (P = 0.0053) was statistically significant. MLN was not significantly associated with tumor diameter and microvascular loops and networks. The 10-year melanoma-specific survival decreased with MLN (0.74, 0.60, and 0.42, arranged in tertiles; P = 0.0060), presence of loops and networks (P = 0.0001), and increasing MVD (P = 0.0001). By Cox regression, MLN was an independent predictor of survival, when adjusting in turn for presence of epithelioid cells, loops and networks, and MVD. In multivariate models with MVD, the independent prognostic information carried by MLN decreased, but the model as a whole was a better predictor of survival. The magnitude of this effect depended on whether MLN was modeled as a continuous or categorical variable. CONCLUSIONS In this population-based data set, MLN and microvascular loops and networks were unrelated, independent predictors of survival. MLN and MVD were found to be partially interrelated. Multivariate models that included MVD in addition to MLN fitted better with observed melanoma-specific survival than models that excluded MVD.