Positive Lymph Node Ratio and Number in Endometrial Cancer: A Better Predictor for Survival and Treatment Effect than FIGO Staging?

Abstract

TU_19_3508Endometrial cancer relies upon surgical staging, yet lymph node (LN) number is not incorporated as in other sites including breast, gastric, esophageal, and colorectal cancers. We hypothesized number and ratio of involved LN would better predict survival outcomes and response to therapy than FIGO stage IIIC1 vs IIIC2. We identified 15,676 patients with LN positive endometrial adenocarcinoma without distant metastases who underwent hysterectomy with known number of LN removed and positive within the National Cancer Database. A multivariable Cox proportional hazards model (MVA), using backward selection, was created based upon available demographic, pathologic, and therapeutic factors. FIGO stage, number of LN positive, number of LN examined, and positive LN ratio were each individually entered into the MVA to test the ability of each factor to predict survival. A subset analysis of those with at least 10 LN examined was performed. Outcomes of those treated with chemotherapy (CT), radiotherapy (RT), neither, or both were compared. At a median follow-up of 44.6 months, the unadjusted 5 year survival for all patients was 65.9% (95% CI = 65.1%-66.7%). Factors predictive of survival on MVA included later year of diagnosis, younger age, lower comorbidity score, race, no prior malignancy, insurance status, residential area education, academic facility, T stage, lower grade, more than subtotal hysterectomy, radiotherapy, and chemotherapy (all p<0.05). Five year survival for IIIC1 vs IIIC2 patients was 69.6% (67.6-71.6%) vs. 64.3% (61.3-67.3%) (p=8.8E-11). This correlated with a hazard ratio of death (HR) of 1.16 (1.02-1.32) on MVA (p=0.022). When grouped by number of positive LN, 5 year survival was 70.0% (68.6-71.4%) for 1 LN, 67.0% (65.4-68.6%) for 2-3 LN, 60.9% (58.5-63.3%) for 4-6 LN, and 46.4% (43.0-49.8%) for 7+ LN (p=1.3E-53). On MVA this corresponded to HRs in reference to 1 positive LN of 1.10, 1.30, and 1.61 respectively (p=4.1E-16). When stratified by LN ratio, 5 year survival was 75.1% (73.7-76.5%) for ≤10%, 68.2% (66.6-69.8%) for 10.1-25%, 53.9% (51.9-55.9%) for 25.1-99%, and 38.7% (34.7-42.7%) for 100% (p=5.2E-166). On MVA this corresponded to HRs in reference to LN ratio of ≤10% of 1.20, 1.64, and 2.21 respectively (p=5.0E-43). Number of LN examined was significant within the MVA on its own, but lost statistical significance when LN ratio was also included in the model (p=0.143). In patients with an adequate dissection (≥ 10 nodes examined), all subsets of LN number demonstrated an improved survival in those who received both CT and RT vs CT alone. In node positive endometrial cancer, nodal ratio and number of involved nodes (especially after adequate dissection) offer improved prognostic precision compared with FIGO staging. A subset that did not benefit from the combination of CT and RT was not identified.