Objective The aim of this study was to investigate the frozen section (FS) accuracy in tailoring the surgical staging of patients affected by endometrial cancer, using 2 different risk classifications.

Methods/Materials A retrospective analysis of 331 women affected by type I endometrial cancer and submitted to FS assessment at the time of surgery. Pathologic features were examined on the frozen and permanent sections according to both the GOG33 and the Mayo Clinic algorithms. We compared the 2 models through the determination of Landis and Koch kappa statistics, concordance rate, sensitivity, specificity, positive predictive value, and negative predictive value for each risk algorithm, to assess whether there are differences in FS accuracy depending on the model used.

Results The observed agreement between the frozen and permanent sections was respectively good (k = 0.790) for the GOG33 and optimal (k = 0.810) for the Mayo classification. Applying the GOG33 algorithm, 20 patients (6.7%) were moved to an upper risk status, and 20 (6.7%) were moved to a lower risk status on the permanent section; the concordance rate was 86.5%. With the Mayo Clinic algorithm, discordant cases between frozen and permanent sections were 19 (7.6%), and the risk of lymphatic spread was underestimated only in 1 case (0.4%); the concordance rate was 92.4%. The sensitivity, specificity, positive predictive value, and negative predictive value for the GOG33 were 92%, 94%, 92%, and 93%, whereas with the Mayo algorithm, these were 98%, 91%, 77%, and 99%, respectively.

Conclusions According to higher correlation rate and observed agreement (92.4% vs 86.5% and k = 0.810 vs 0.790, respectively), the Mayo Clinic algorithm minimizes the number of patients undertreated at the time of surgery than the GOG33 classification and can be adopted as an FS algorithm to tailor the surgical treatment of early-stage endometrial cancer even in different centers.