The assigned stage of a cancer is the most objective and universally recognized identifying feature that is used to guide treatment, assess response, and determine prognosis. The more reflective the assigned stage is of the underlying tumor behavior, the better it can meet these goals. Tumor characteristics and behavior change over the course of time and treatments. A perfect staging system would be reflective of tumor behavior at any time during the life of the tumor. Every new iteration of a staging system attempts to meet these goals.
Tumor-node-metastasis (TNM) esophageal cancer staging, first introduced in 1968, was developed to better provide prognosis based on clinically relevant data. The 7th edition of the American Joint Committee on Cancer (AJCC) was the first to use modern machine learning analysis to produce data-driven staging of esophageal and esophagogastric (EGJ) cancer. Building upon its predecessor, the 8th edition uses more than just anatomic and histologic considerations to generate more homogeneous groups and provide more accurate prognoses.
Historically, esophageal cancer staging was based solely on esophagectomy data, which works well for early-stage tumors but loses relevance in tumors that have been pretreated (neoadjuvant therapy). The 8th edition is designed to stage patients at different time points and to allow clinicians to determine the patient’s prognosis, which can be a moving target depending on the information available at the time. While it may seem daunting to understand this staging system, this chapter provides guidance for thinking about esophageal cancer staging in the modern era.
At the request of the AJCC, the Worldwide Esophageal Cancer Collaboration was inaugurated in 2006.1 Thirty-three institutions across six continents submitted de-identified data, generating a database of 22,654 patients.1
Staging for the 8th edition uses random forest (RF) analysis, a machine-learning technique that focuses on predictability for future patients.2 RF analysis makes no a priori assumptions about patient survival, is able to identify complex interactions among variables, and accounts for nonlinear effects. RF may be viewed as a “backward” analysis that determines the anatomic classifications (TNM) and nonanatomic cancer characteristics associated with specific survival groups.
RF analysis first isolates cancer characteristics of interest from other factors that influence survival by generating risk-adjusted survival curves for each patient. Unlike previous approaches that began by placing cancer characteristics into proposed groups, RF analysis produces distinct groups with monotonically decreasing risk-adjusted survival without regard to cancer characteristics. Then, cancer characteristics important for stage group composition are identified within these groups. Finally, homogeneity within groups guides both the amalgamation and segmentation of cancer characteristics between adjacent groups to arrive at the final stage groups.3–5
8TH EDITION TNM: CHANGES AND ADDITIONS
Nomenclature has changed, as TNM classifications are ...