Analysis of Lymph Node Sampling Minimums in Early Stage Non-Small-Cell Lung Cancer.

Analyze "number of nodes" as an integer-valued variable to identify possible minimum lymph node (LN) number to sample during lung cancer resection. The National Cancer Database (NCDB) queried 2004-14 for surgically treated clinical stage I/II non-small-cell lung cancer (NSCLC). Overall survival (OS) by number of LN sampled was examined for the complete dataset, by adenocarcinoma, and by degree of resection using number of sampled LN both as integer-valued (0-30 nodes) variable and collapsed into classes. A total of 102,225 patients were analyzed. Median sampled LNs were 7. Median overall survival was 59 months if no LNs were sampled (95% confidence interval [CI]: 57.0-62.4), 74.7 months for 1 sampled LN (95% CI: 69.6-78.1), 80.2 (95% CI: 74.2-85.6) for 2 sampled LN, up to 81.5 mos. for 29 sampled LN. A Cox regression model using "0 LN" as baseline level, showed association with increased overall survival starting at 1 LN (hazard ratio [HR] 0.81, 95% CI 0.76-0.87; P <0.001). A "moving baseline" Cox regression model, showed no additional benefit when sampling additional nodes. We noticed a decreasing, linear association between OS and a number of 0-5 sampled LNs, most pronounced between 0 and 1 LN sampled, using a martingale residual plot from a null Cox model; no association was observed for more sampled LNs. For patients undergoing lobectomy, difference in OS was noted between 0 and 1LN sampled but not between 2 and 30 LN. These differences were not statistically significant until the number of 4 removed LN (respectively 3 for wedge-resections). For segmentectomies, median survival was not statistically associated with number of LN sampled. Based on NCDB data, LN sampling for lung cancer resections is recommended. Lobectomy survival is positively associated with 4 LN sampled, but ideal sampling may lie at 5LN in most cases. NCDB data does not seem to justify the quality metric of minimum 10 LNs.