The Universality of Parallel Increasing Quantitative Nodal Burden and Mortality Across Solid Cancers

Abstract

<h3>Purpose/Objective(s)</h3> Nodal staging systems vary substantially across solid tumors, which implies significant heterogeneity in the behavior of nodal variables across cancers. We hypothesized, in contradiction to this, that quantitative metastatic lymph node (LN) burden is a universal and dominant predictor of mortality across solid cancers, and therefore is capable of both simplifying and improving nodal staging for many cancers. <h3>Materials/Methods</h3> Retrospective cohort analysis of 1,304,498 patients in the National Cancer Database with solid cancers undergoing surgery between 2004-2015 across the 16 most common cancer sites in the United States (bladder, breast, cervix, colorectal, endometrium, esophagus, gastric, head and neck, hepatobiliary, lung, melanoma, ovary, pancreas, prostate, renal, thyroid). Restricted cubic spline modeling was used to assess the independent association of LN number and survival across a continuous nonlinear spectrum. Recursive partitioning analysis (RPA) was used to develop 16 nodal classification systems based on quantitative nodal burden. The reproducibility of these findings was assessed in 1,969,727 patients from SEER-18. <h3>Results</h3> Consistently across all disease sites and without exception, mortality risk increased continuously and steeply with each +LN to a change point of ≤5 LN, and continued to increase more modestly without plateau beyond this (<b>P</b> < 0.001 for all hazard ratios [HRs] per LN of the initial and secondary spline segments). Disease sites with the highest HR per LN of the initial spline segment were melanoma (1.92/LN), bladder (1.66/LN) and renal (1.57/LN). Conversely, sites with the lowest HR per LN of the initial spline segment were thyroid (1.13/LN), ovary (1.19/LN) and endometrium (1.21/LN). RPA-derived nodal classification systems for all 16 disease sites universally produced multiple prognostic groups (between 3-8 depending on site) with non-overlapping survival curves that captured a wide spectrum of mortality risk (all <b>P</b> < 0.001). These RPA-derived systems demonstrated improved concordance with mortality compared to current AJCC nodal staging systems in disease sites without pathologic nodal classification based on number of positive LNs including bladder, cervix, endometrium, lung, ovary, prostate, and renal. The RPA-derived systems were also validated in an independent database, producing groups with distinct all-cause mortality and cause-specific mortality (all <b>P</b> < 0.001). Defining high nodal burden as the RPA-derived group with the highest number of +LNs, this factor conveyed markedly higher mortality (HR range: 1.8-4.6) than any other nodal covariate, including extranodal extension, size, and location (HR range: 0.9-1.5) in multivariable analysis across all organ sites. <h3>Conclusion</h3> Quantitative metastatic LN burden is a fundamental driver of mortality across solid cancers. Given its reproducibility and objectiveness, +LN number should serve as a foundation for pathologic nodal staging across all solid tumors.