Diagnostic Accuracy of Dermoscopy for Spitzoid Tumours in Adults

Abstract

Spitzoid tumours (ST) represent a diagnostic challenge and the accuracy of dermoscopy in these cases remains unclear. This retrospective study evaluated dermoscopic images collected from 2011 to 2016. Two dermatologists, blinded to clinical/ histopathological data, identified lesions with spitzoid dermoscopic patterns, using histopathology as the gold standard. Of 4,337 dermoscopic images, 240 (5.5%) showed spitzoid dermoscopic patterns (221 patients; median age 40; 57.9% female). Histopathology revealed 23% (55/240) ST, 62% (149/240) non-spitzoid melanocytic lesions and 15% (36/240) non-melanocytic lesions. Overall, 25.4% (61/240) were malignant, mainly melanoma/melanoma metastases (50/240). Across excised tumours, 1.5% (67/4,337) were ST. Sensitivity and specificity were 82.1% (95% CI 70.8–90.4%) and 95.7% (95% CI 95–96.3%), with an AUC of 0.88 (95% CI 0.84–0.93). Positive predictive value was 22.9% (95% CI 17.8–28.8%), and negative predictive value was 99.7% (95% CI 99.5–99.9%). Starburst pattern was more frequent in ST (p=0.007). Linear vessels (OR 5.12; 95% CI 2.08–13.1) and glomerular vessels (OR 3.03; 95% CI 1.37–6.69) correlated with malignancy. Dermoscopy demonstrated high diagnostic accuracy and negative predictive value. However, low positive predictive value highlights limited specificity in adults; since 1 in 4 lesions with a spitzoid dermoscopic pattern was malignant, any spitzoid lesion in adults warrants high clinical suspicion.