Predictive model of small choroidal melanoma progression after eye-saving treatment based on clinical, morphometric and immunological parameters

Abstract

Choroidal melanoma is a malignant tumor characterized by early metastasis and poor vital prognosis. Prognostic indexes for the tumor development are of importance, depending on various factors and making it possible to optimize therapeutic measures. Usage of present models for prediction of the uveal melanoma course enables optimal management of the patients with a malignant tumor upon primary admission, and to perform maximally efficient counseling. So far, however, a complex of clinical, morphometric and immunological indexes predictive for unfavorable course of initial choroidal melanoma following the eye-saving treatment remains not fully determined. Our purpose was to create a prognostic model for initial choroidal melanoma after eye-saving treatment, basing on clinical, morphometric and immunological parameters.We have performed examination and treatment of 31 patients with small choroidal melanoma (53.7 to 12.2 years old). To perform the analysis, we used clinical data (age, decreased vision, tumor localization, degree of pigmentation, presence of hemorrhages, orange pigmentation), morphometric indexes (intra- and subretinal exudate and disorganization of pigment of the retinal epithelium) and immunological parameters (serum levels of pro-inflammatory, pro-angiogenic, proliferative, metastasis-causing cytokines). Selection of variables for this model was based on assessment of significant differences between the groups with chorio-retinal scar (n = 14) and residual tumor and/or continued tumor growth (n = 17).Multivariate analysis with conditional exclusion of variables revealed prognostic significance with four markers: morphometric, i.e., disorganization of the pigment in retinal pigment epithelium – Z1 (rs = 0.455); immunological, increased blood serum concentration of hepatocyte growth factor (HGF) – Z2 (rs = 0.377); level of pro-inflammatory chemokine RANTES – Z3 (rs = 0.362), content of transforming growth factor (TGF-2â) – Z4 (rs = 0.431). A formula was calculated where P (z) is the value of the logistic function; Z, linear combination of symptoms; bo , intercept (free term), bi – regression coefficients for parameters Zi.P (z) = 1 : 1 + e – b0– b1z1– b2z2– b3z3– b4z4The logistic function increases monotonically and takes the values from 0 to 1 for any b and Z values [P∈ (0;1)]. If P (Z) is under the cutoff value, chorioretinal scar prognosis is predicted, at the higher values, a residual tumor or continued growth is expected. In ROC analysis, the area under the curve with this model was 0.891±0.11, thus providing good predictive quality.Usage of the predictive model is a possible solution for planning and correcting treatment strategy in the patients with small choroidal melanoma, in order to minimize complications and errors, and to ensure control of treatment.