Updated Review: Using the National Cancer Database for Outcomes-Based Research.

Abstract

The National Cancer Database (NCDB) captures 73.7% of newly diagnosed cancers in the USs and underpins thousands of outcomes studies informing oncologic practice. To remain relevant amid rapid therapeutic and policy changes, the NCDB has undergone substantial structural and variable-level revisions. We characterized major updates during the past decade and their implications for research. We performed a narrative review of annual NCDB data dictionary revisions, American College of Surgeons bulletins, and internal program updates in collaboration with NCDB leadership. Structural modifications, variable additions, and policy changes affecting data capture, follow-up, staging, and accessibility were systematically summarized. The NCDB now includes data from 1,413 Commission on Cancer-accredited hospitals and more than 55 million records. Since 2020, the Rapid Cancer Reporting System enables near-real-time monthly submissions. Embargo periods were reduced from 5 to 3 years for survival data and 2 years for other variables, increasing analytic timeliness. Follow-up was limited to 15 years beginning January 1, 2022. Variable refinements include continuous tumor size in millimeters (since 2016), separation of tumor grade into clinical, pathologic, and posttherapy fields (since 2018), phased radiation treatment reporting (since 2018), and American Joint Committee on Cancer eighth edition staging implementation in January 2018 with nineth edition rollout ongoing. New data elements include Medicaid expansion status (2020), COVID-19 variables (2020 to 2021; 12.4% reduction in cases in 2020), smoking status (2023), and planned programmed death-ligand 1 reporting for non-small cell lung cancer beginning in 2025. The NCDB has evolved toward more granular, contemporary, and policy-relevant data capture while maintaining broad national coverage. Investigators must account for staging transitions, variable maturation, follow-up limits, and registry-specific biases to ensure valid interpretation of NCDB-based research.