A Phase 1/2 Study of the Angiogenesis Inhibitor Sorafenib in Cervical Cancer Patients Treated With Radiation Therapy

Abstract

randomized trials have shown no survival benefit. However, identification of lymph node (LN)-positive women is important for directing adjuvant therapy. We developed a nomogram based on pathologic characteristics from surgical hysterectomy specimens to predict for LN metastasis. Materials/Methods: We retrospectively examined pathological data from all hysterectomies with LN sampling performed for endometrioid endometrial adenocarcinoma at our institution. A multivariate logistic regression analysis of selected features was performed, and a nomogram to predict LN metastasis was constructed. Results: From 1996 to 2013, 301 patients underwent hysterectomy with LN sampling for endometrial cancer. Surgical staging revealed the following staging distribution: FIGO IA (n Z 163), FIGO IB (n Z 57), FIGO II (n Z 22), FIGO IIIA-B (n Z 17), FIGO IIIC (n Z 33), and FIGO IV (n Z 9). One hundred five patients had grade 1, 123 patients had grade 2, and 73 patients had grade 3 tumors. No patient had prior treatment for endometrial cancer. While pelvic nodes were sampled in all cases, paraortic nodes were sampled in 135 cases. Median number of LNs removed was 13 (range, 1-72). On univariate analysis, tumor size 4 cm, grade, lymphovascular space involvement (LVSI), cervical stromal involvement, adnexal involvement, serosal involvement, positive pelvic washings, parametrial involvement, > one-half myometrial invasion, and number of LNs removed all significantly predicted for LN involvement. Age and ER/PR status did not. In a multivariate model, LVSI (p Z 0.0003, OR Z 6.5), > one-half myometrial invasion (p Z 0.04, OR Z 2.7), and cervical stromal involvement (p Z 0.003, OR Z 1.3) remained significant predictors of LN involvement, while tumor size 4 cm was borderline significant (p Z 0.07, OR Z 2.17). In particular, grade was no longer significant (p Z 0.56) when considered together with other factors. We constructed a prognostic nomogram (AUC 0.87), including LVSI, myometrial invasion, cervical stromal involvement, and tumor size. Of these factors, LVSI was most predictive of LN involvement, with a standardized coefficient of 49% in our model. Of 100 patients who had LVSI in this study, 30% had LN metastases. Conclusions: Studies have reported factors such as tumor size, grade, and depth of invasion as predictive of LN metastases, which we confirm. However, when looking at these factors together, LVSI was most predictive of LN involvement, followed by myometrial invasion and cervical stromal involvement, regardless of tumor grade. LVSI should be one of the main prognostic factors used to help direct adjuvant therapy in patients without LN surgery. Author Disclosure: E. Pollom: None. C. Conklin: None. R. Von Eyben: None. A. Folkins: None. E. Kidd: None.