Extent of lymph node resection and effect on pancreatic cancer overall survival.

Abstract

682 Background: Lymph node (LN) metastases affect overall survival (OS) in pancreatic cancer (PC). However, a LN sampling threshold does not exist. We examined the impact of nodal sampling on overall survival (OS). Methods: Patients with Stage I-III PC ≥55 years old who underwent curative resection from 2004-2016 were identified from the National Cancer Database (NCDB). After adjusting for age, gender, grade, stage, and Charlson-Deyo score, multiple binomial logistic regression analyses assessed the impact of the LN ratio (LNR) on OS. LNR was defined as the number of positive LN over the number of LN examined. Regression analyses, a Cox-Regression, and a Kaplan-Meier survival curve assessed how many LN should be sampled. Results: A total of 13,673 patients, median age 69 years (55-90), were included. Most were Caucasian (86.6%) males with Charlson-Deyo scores ≤ 1 (90.3%) and moderately to poorly differentiated PC (90.1%). Median number of LN examined was 15 (1-75) with a median of 1 positive LN (0-35). As expected, increased number of positive LNs was associated with reduced OS, p < 0.001. After data normalization, an increasing LNR was associated with a 12-fold likelihood of death [OR: 11.9, p < 0.001 (CI 6.0, 23.7)]. Subsequent regression models established evaluation of ≥ 16 LNs as the greatest predictor of OS. A regression model evaluating < or ≥ 16 lymph nodes was performed to ascertain the effects of age, gender, ethnicity, grade, stage, and LN examined on OS. The logistic regression model correctly classified 74.5% of cases with a specificity of 99.6% (p < 0.001). Examination of < 16 LN, Caucasian race, grade, stage, and higher Charlson-Deyo scores were significantly associated with decreased OS. If ≥ 16 LNs were examined, patients had a 1.5-fold likelihood of better OS, p < 0.001 (CI 1.4, 1.6). An adjusted Cox Regression showed increased HR of 1.2, p < 0.001 (CI 1.1, 1.2) and an unadjusted Kaplan Meier survival curve predicted ≥ 16 LN examined are associated with an increase in OS of 2.8 months [log-rank: 32.0, p < 0.001]. Conclusions: Patients undergoing curative intent resection for PC should have adequate nodal sampling. Stratification of patients by LNR may provide useful information of OS. Examination of ≥ 16 LNs impacts OS in patients with Stage I-III PC.