Grading ovarian tumors. Evaluation of decision making by different pathologists.

Although grading of ovarian tumors is widely performed, the criteria for each grade are not well defined; as a result, pathologists tend to establish their own criteria without, however, assessing the actual predictive value of the criteria. In order to investigate this relationship, four gynecologic pathologists independently reviewed and carefully graded as benign, borderline or malignant (grade I, II or III) 40 "common" epithelial tumors of the ovary, without reference to clinical, prognostic or other findings. Intermediate grades were allowed. Subsequently, a subjective grading form was completed for each case; the form contained questions regarding the histologic and cytologic features. The sets of features with the biggest correlation with the tumor grades differed among the pathologists. This may indicate that the observers use different features in their grading processes. Moreover, the pathologist with the highest number (five) of significant microscopic features in the multivariate model had the lowest coefficient of correlation between his tumor grade and his feature set. The correlation coefficients for the other pathologists were quite similar, although the features used (no more than two or three) varied. The participants in the study felt that the methodologic approach had an educational value for them. Further investigations are required to evaluate whether the differences in the underlying decision making process also result in frank disagreement in ovarian tumor grading.