Interobserver variability in the diagnosis of circumscribed sebaceous neoplasms of the skin

Abstract

Separation of sebaceous adenoma, sebaceoma and well differentiated sebaceous carcinoma is a clinically important distinction which relies on a number of subjective criteria. In routine practice we had noted significant interobserver variability in the classification of these lesions. This study sought to determine the degree of interobserver variability between general surgical pathologists and dermatopathologists in the diagnosis of well differentiated cutaneous sebaceous neoplasms. We circulated 61 examples of well circumscribed cutaneous sebaceous neoplasms to nine pathologists, including dermatopathologists and general surgical pathologists who were asked to submit a diagnosis for each case. Fleiss' kappa statistic was used for assessment of interobserver agreement. We found that only seven cases (11%) had consensus agreement across all nine pathologists. Many cases had multiple diagnoses suggested, with three or more submitted diagnoses in 26 cases (43%), while 38 cases (62%) were diagnosed as sebaceous carcinoma by at least one pathologist. There was marked variability amongst the individual pathologists in the proportion of cases diagnosed as carcinoma, ranging from 5% to 57% of cases. Fleiss' kappa statistic for all pathologists across all diagnostic categories was 0.44, amounting to only fair to moderate agreement. These data indicate that there is substantial interobserver variability in the diagnosis of well circumscribed sebaceous neoplasms. This was seen in both the separation of benign and malignant lesions, as well as in the classification of the benign entities. This interobserver variability is likely to have significant clinical implications in terms of potential for over- or under-treatment, as well as in selection of cases for mismatch repair protein evaluation.