Histologic regression in malignant melanoma: an interobserver concordance study.

Abstract

Histologic evidence for regression as a prognostic indicator in melanoma has shown conflicting results. To assess if melanoma regression is a consistently identifiable histologic feature, an interobserver concordance study was undertaken. Fifty histologic slides of melanoma with Breslow thickness of 1 mm or less were non-randomly selected from the Massachusetts General Hospital Melanoma Registry. The selection was in favor of those with regression present in 44 of 50 slides (88%). Two dermatopathologists working independently evaluated a 2 mm wide bracketed area in each tissue section and then the remainder of the non-bracketed tissue section for regression. Before and after each slide review, the evaluators were required to read criteria for early, intermediate, and late regression explicitly outlined. The overall concordance rate for the presence and absence of regression was 96% (48/50) in the bracketed area and 90% for the outside area. When the group with regression was subdivided into three stages, interobserver agreement fell to 86% for the bracketed, and 66% for the outside area. In at least 30% of cases where regression was detected in the bracketed area, both reviewers observed two or more stages of regression in the remainder of tissue section. Therefore, subdividing the process is impractical and unrealistic. Consistent histologic identification of regression in melanoma requires a simple and systematic approach which should be applied to future studies that include melanoma regression as a prognostic factor.