Interobserver concordance in discriminating clinical atypia of melanocytic nevi, and correlations with histologic atypia

Background: The clinical features attributed to atypical (formerly ''dysplastic'') nevi and to the atypical multiple mole melanoma syndrome have been used in clinical practice, as well as experimentally, to assign melanoma risk. Little information is available, however, on the interobserver reliability in assessing those features. Objective: Our purposes were to quantify interobserver and intraobserver concordances in recognizing certain atypical characteristics of nevi and to correlate the clinical assessments with the histologic characteristics. Methods: Three observers evaluated clinical photographs of 100 pigmented lesions (predominantly melanocytic nevi, with some lentigines and seborrheic keratoses) from 95 subjects, of whom 85 were family members of four multiple melanoma kindreds and 10 were spouses. Each lesion was rated for border irregularity, color variegation, surface contour irregularity, pigment diffusion, and macularity versus papularity. Predictions were made as to the histologic diagnoses and presence of melanocytic atypia for those lesions judged to be nevi. Results: The pair wise concordances before agreement on specific criteria were quantified by kappa statistics, which indicated slight to fair agreement in judging the atypical clinical characteristics; concordances increased to moderate levels after consensus development of criteria for color variegation and assessment of macularity, but agreement on the other features remained limited. Whereas macularity and color variegation did correlate somewhat with higher grades of histologic atypia, correlations were generally low between the clinical and histologic diagnoses. Conclusion: There is limited interobserver reliability in the clinical assessment of nevus atypia, although correlations do exist between some atypical characteristics and grades of histologic atypia. Because of the low concordances, the clinical discrimination of the melanoma-associated atypical nevus phenotype should rely more on quantitative aspects of the trait, such as total numbers or maximal sizes of nevi, rather than on the subjective determinations of atypia.