Gene scoring model to predict recurrence in low- and intermediate-risk uterine endometrial cancer: Establishment of uterine print.
5590 Background: Endometrial cancers (ECs) classified as low-, intermediate-, and high-risk, based on clinical and pathological features (CPF: Lurain, 2007) associated with 5%, 15%, and 25% risk of recurrence, respectively. The need for adjuvant chemotherapy in intermediate-risk patients is controversial. We examined whether gene expression profiling can more accurately predict the prognosis of ECs, excluding the CPF-based high-risk group. Methods: Tumor specimens were obtained from 136 ECs including 14 recurrences, excluding high-risk cases. Gene expression profiles were achieved using a custom array consisting of 85 genes associated with EC recurrence and 20 internal controls that were previously screened. We established the gene scoring model (GSM) for recurrence by the logistic regression model in randomly selected 68 ECs including 7 recurrences, and evaluated the accuracy of GSM in other 68 ECs including 7 recurrences. This process was repeated 100 times. We calculated the mean accuracy of GSM and compared it with the accuracy of CPF. We also compared GSM and CPF with respect to progression-free survival (PFS) by use of the log-rank test. Results: Median age of all cases was 58 (29-86) years, and stage, histologic grade, and risk classification based on CPF were as follows: (I, 107; II, 15; III, 14), (G1, 69; G2, 57; G3, 10), and (low, 67; intermediate, 69). The median follow-up period was 1830 (1626-3444) days. The GSM was established based on the expression of 4 genes (PRCC, SPC25, PXDN, and LBXCOR1) and 10 internal controls. The area under the receiver operating characteristic curve of GSM to predict recurrence was 0.87 in 68 test cases. Based on the CPF, 68 cases were classified as 30 low-risk and 38 intermediate-risk, and the sensitivity and specificity of CPF was 86% and 48% each in the 68 test cases. When sensitivity of GSM was fixed at 86%, specificity of 67% was achieved, and 68 cases were classified as 42 risk (-) and 26 risk (+). PFS was significantly related with GSM (p = 0.006); however, it was not related with CPF (p = 0.09). Conclusions: GSM can predict the prognosis of ECs (low- and intermediate-risk) more precisely than CPF.