Reproducibility of the diagnosis of atypical endometrial hyperplasia: a Gynecologic Oncology Group study.

BACKGROUND Most gynecologists determine therapy based on current International Society of Gynecologic Pathologists (ISGP)/World Health Organization classification of endometrial hyperplasia, the reproducibility of which has been questioned. The Gynecologic Oncology Group (GOG) initiated a protocol to assess the efficacy of hormonal therapy of atypical endometrial hyperplasia (AEH). Primary goals of the first phase (Part A) were to prospectively determine reproducibility of referring institution's pathologist's diagnosis of AEH by a panel of 3 gynecologic pathologists and to determine reproducibility of diagnoses by panel members. METHODS Three hundred six women were entered on this protocol with a referring institution's pathologist diagnosis of AEH based on biopsy or curettage. Available slides were assessed independently and interpreted by each of a panel of 3 gynecologic pathologists who used International Society of Gynecologic Pathologists (ISGP)/World Health Organization criteria. The majority diagnosis was based on diagnostic concordance by at least 2 of the 3 panelists. RESULTS The referring institution's pathologist's diagnosis of AEH was supported by the majority of the panel in only 38% of cases. Overall kappa value for the panel diagnosis of AEH was 0.28. The majority diagnosis was adenocarcinoma in 29%, cycling endometrium in 7%, and nonatypical hyperplasia in 18% of cases. Unanimous agreement for any diagnosis was reached among all 3 of the panel in 40% of cases. For the panel, paired kappa values for any diagnosis ranged 0.34–0.43, with an overall kappa value of 0.40. CONCLUSION Reproducibility of referring institution's pathologists' diagnosis of AEH by a panel of gynecologic pathologists is poor. Both underestimation and overestimation of the severity of the lesion are very common. The level of reproducibility among subspecialist panel members for diagnosis of AEH in these specimens also is poor. Better criteria and better sampling are needed to improve reproducibility of this diagnosis, particularly if it is to be used for clinical decisions. Cancer 2006. © 2006 American Cancer Society.