Structural asymmetry as a dermatoscopic indicator of malignant melanoma--a latent class analysis of sensitivity and classification errors

Abstract

Asymmetry of pigmented skin lesions is an important indicator of possible malignant melanoma and contributes substantially to the diagnosis of melanoma in the dermatoscopic ABCD rule for melanocytic lesions and other algorithms. However, it may be observer dependent. Dermatoscopic asymmetry cannot be assessed objectively and no golden standard of asymmetry diagnosis exists. The aim of this study was to assess the sensitivity of axis (a-) symmetry using latent class analysis. We analysed ratings from four experts in dermatoscopy of 232 pigmented lesions by latent class analysis (LCA). Possible ratings were 'no asymmetry', 'asymmetry in one axis' and 'asymmetry in two axes'. A subset of melanocytic lesions (blue naevi excluded) was analysed. Based on LCA, the asymmetry of the singular lesion was determined. The sensitivity of 'no asymmetry' was 40-77%, 40-70% for one-axis asymmetry, and 77-92% for two-axes asymmetry. Overestimation of asymmetry was more common than underestimation. Melanomas were significantly more asymmetric than pigmented naevi, atypical naevi and papillomas, but not basal cell cancers. Analysis of the melanocytic subset gave similar results. The median asymmetry of malignant melanomas (1.67, interquartile range 1.81-1.99) was higher than for melanocytic naevi. In conclusion, asymmetry and symmetry are important criteria for diagnosing or excluding malignant melanoma using the dermatoscopic ABCD rule, risk stratification and other diagnostic rules. Using LCA, we minimized observer dependence in the assessment of axis (a-) symmetry. LCA, besides conceptualizing the diagnostic process, enables the assignment of lesions to their true diagnostic class.