Classifying dermoscopic patterns of naevi in a case-control study of melanoma

Changes in dermoscopic patterns of naevi may be associated with melanoma; however, there is no consensus on which dermoscopic classification system is optimal. To determine whether different classification systems give comparable results and can be combined for analysis, we applied two systems to a case-control study of melanoma with 1037 participants: 573 classified using a "1/3 major feature" system, 464 classified based on rules of appearance, and 263 classified with both criteria. There was strong correlation for non-specific (Spearman R = 0.96) and reticular (Spearman R = 0.82) naevi, with a slight bias for globular naevi with the rules of appearance system. Inter-observer reliability was high for the rules of appearance system, particularly for reticular naevi (Pearson >0.97). We show that different classification systems for naevi can be combined for data analysis, and describe a method for determining what adjustments may need to be applied to combine data sets.