Reflectance Confocal Microscopy Integrated With Dermoscopy in a Real-World Tertiary Referral Cohort

BACKGROUNDAccurate non-invasive assessment of cutaneous lesions requires integration of diagnostic information with real-world management outcomes. Reflectance confocal microscopy (RCM) is increasingly used as a second-level tool after dermoscopy, but evidence linking diagnostic agreement, histological verification, and management across melanocytic and non-melanocytic lesions remains limited.

METHODSThis retrospective observational study included 886 patients with 1057 cutaneous lesions evaluated by combined dermoscopy and RCM at a tertiary referral center between October 2023 and September 2024. Lesions were classified as benign melanocytic, malignant melanocytic, benign non-melanocytic, or malignant non-melanocytic. RCM-dermoscopy agreement was assessed using observed agreement and Cohen's kappa; chi-square testing evaluated diagnostic association. Histology-based performance analyses were restricted to histologically confirmed lesions.

RESULTSAmong 931 lesions with both classifications available, observed agreement was 37.8%, with fair agreement by Cohen's kappa (κ = 0.211; 95% confidence interval [CI], 0.172-0.251). Diagnostic distributions were significantly associated (χ2 = 571.3, df = 9, p < 0.001), with substantial discordance, particularly within melanocytic categories. Histopathological confirmation was available for 225 lesions (21.3%), indicating verification bias. In the dedicated melanocytic excision subset, 162 excised lesions yielded 41 melanomas, resulting in a numbers needed to excise (NNE) value of 3.95. Malignant non-melanocytic detection yield was 39.0% among histologically confirmed excised non-melanocytic lesions.

CONCLUSIONSIn this real-world tertiary referral cohort, RCM integrated with dermoscopy supported lesion stratification and management across diagnostic categories. The significant association but fair agreement between modalities highlights their complementary role. Diagnostic performance estimates should be interpreted within the histology-confirmed subset and in light of verification bias.

Abstract from PubMed