Dermatoscopic prediction of melanoma thickness using latent trait analysis and likelihood ratios

Abstract

Breslow thickness and Clark level can be used to determine surgical procedures for cutaneous malignant melanoma and patient eligibility for experimental adjuvant therapy. Efforts to predict the thickness of melanomas using dermatoscopy have focused on differences between single dermatoscopic findings. The aim of this study was to develop a method for preoperative identification of melanomas of > or = 1 mm Breslow thickness using the entire range of dermatoscopic findings. Sixty-five melanomas were assessed for the presence of 22 dermatoscopic features. Ten dermatoscopic features showed differences in thick and thin melanomas and were selected for further analysis. A latent trait analysis construct implied that a progression in dermatoscopic features was associated with advancement of melanomas. Early melanomas are characterized by a light brown colour, a pigment network and irregularity or heterogeneity. Gray-blue areas, white scar-like areas and an atypical vascular pattern gradually displace these features. Likelihood ratios were determined for these 6 dermatoscopic findings and an algorithm for calculating the probability of thick malignant melanoma was established.