RESUMEN
BACKGROUND: To evaluate reproducibility of a reporting system for endometrial cytology. METHODS: Cytologic slides from 49 patients, prepared via liquid based cytology, were blindly examined by five cytopathologists of various experience levels, applying a recently introduced reporting system as previously reported. The agreement among cytopathologists was evaluated via Kappa (κ) statistics and the Kendall's Coefficient of Variation (W); cytologic results were compared with the relevant histologic report. RESULTS: Substantial agreement among all five raters was found in the benign, ACE-L and malignant categories, fair agreement in inadequate and ACE-H categories, whereas only slight agreement in ACE-U. For the three more experienced cytopathologists, an almost perfect agreement was found in inadequate, benign, and ACE-L categories, substantial agreement in ACE-H and malignant categories and fair agreement in ACE-U category. Overall agreement for all five cytopathologists and for all categories was moderate, whereas it was very high for the three senior raters. Using the Kendall's test, both five cytopathologists (W = 0.81) and the three senior ones (W = 0.93) had very high agreement. Sensitivity: 83.33-92.59%, specificity: 83.33-94.74%, ROC area: 71.72-90.3%. CONCLUSION: Application of appropriate statistical tests shows that integration of a new reporting cytologic system is effective with an overall accuracy around 90%. Both statistical tests applied disclosed lower agreement rates among both all five raters and the three most experienced ones in the intermediate categories constituting the gray zone, thus delineating the need for better training of cytopathologists to correctly identify diagnostic criteria for classification of a given case into these categories.