Objective: To determine the reliability of the histopathologic diagnosis of melanocytic dysplasia among diverse dermatopathologists who had no joint training, agreed to abide by predetermined criteria, and who were provided reference photomicrographs illustrative of the criteria.
Design, setting, and participants: A stratified random sample of 112 melanocytic tumors were chosen from the files of the pathology department of a large staff-model health maintenance organization. The original diagnoses included typical and dysplastic melanocytic nevi and melanoma. A single representative slide for each case was interpreted independently by each of the 5 panel dermatopathologists and 2 melanoma specialists. They had no prior knowledge of the original diagnosis or the diagnoses of the other panel members.
Main outcome measures: Interrater reliability was measured by intraclass and Pearson correlation coefficients. Each case was graded on a 5-point scale from no dysplasia to melanoma.
Results: The intraclass correlation among the panel members was 0.67 (95% confidence interval, 0.59-0.73). The Pearson correlations of each of the 5 panel dermatopathologists with the mean of the 2 melanoma specialists ranged from 0.67 to 0.84, and the correlations of the mean of the panel with the 2 melanoma specialists were 0.79 and 0.82; the mean reading of the melanoma specialists correlated 0.89 with the mean panel reading. Apparent protocol violations occurred in 6.5% of the readings.
Conclusions: Agreement was substantial to excellent for the histopathologic diagnosis of 112 melanocytic tumors by dermatopathologists. Using predetermined criteria, melanocytic dysplasia can be reproducibly graded among diverse general dermatopathologists.