Abstract
Purpose:
Vascular endothelial growth factor (VEGF)-A expression in uveal melanoma (UM) has previously been described; however, there is a lack of consensus about whether or not expression correlates with metastasis. In the present study, we developed a custom algorithm for objectively quantifying VEGF-A immunostaining in UM and determined if VEGF-A expression correlates with metastasis.
Methods:
The inclusion criteria for patient samples were: choroidal UM diagnosis, minimum 5-year follow-up, data on the presence of metastases, and sufficient tumor size to obtain 3 non-overlapping images occupying 90% of the image at 200× magnification. Tumors were stained with a monoclonal VEGF-A antibody and all images were obtained using the identical camera and settings. A custom algorithm was created using ImageJ software to identify and quantify the number of positively stained UM cells. The hue, saturation, and brightness (HSB) were adjusted in order to isolate positive cells. HSB values were determined by trial and error; however, once set, they remained consistent for all images. The percentage of the image that was immunostained and the number of positively stained cells were determined using the algorithm. An ocular pathologist evaluated the slides to confirm that the majority of cells that stained positive and were isolated by the algorithm were UM cells. A binary logistic regression was performed with the presence of metastasis as the dependent variable and average cell count and average positive staining fraction as the independent variables.
Results:
Of the 29 cases that met the inclusion criteria, eight UM patients had metastases (28%). The quantification algorithm was capable of identifying the positively stained cells and rendered cell counts and area stained even in samples with strong melanin presence. Using positive immunostaining values generated by the algorithm, we determined that VEGF-A presence in the primary tumor was significantly associated with metastasis; the predictive accuracy of the model was 79.3% using positive VEGF-A staining average cell count and average positive area (P<0.05).
Conclusions:
The presence of VEGF-A in primary uveal melanoma can be used to retrospectively predict metastasis. The custom algorithm used in this study facilitates meta-analyses and eliminates potential bias when grading immunohistochemical staining. This information may help clinicians to select patients for anti-VEGF therapy.
Keywords: 589 melanoma •
554 immunohistochemistry •
744 tumors