Implications of missing data on reported breast cancer mortality

Background National cancer registries are valuable tools to analyze patterns of care and clinical outcomes; yet, missing data may impact the accuracy and generalizability of these data. We sought to evaluate the association between missing data and overall survival (OS). Methods Using the NCDB (National Cancer Database) and SEER (Surveillance, Epidemiology, End Results Program), we assessed data missingness among patients diagnosed with invasive breast cancer from 2010 to 2014. Key variables included demographic (age, race, ethnicity, insurance, education, income), tumor (grade, ER, PR, HER2, TNM stages), and treatment (surgery in both databases; chemotherapy and radiation in NCDB). OS was compared between those with and without missing data using Cox proportional hazards models. Results Overall, 775,996 patients in the NCDB and 263,016 in SEER were identified; missing at least 1 key variable occurred for 29% and 13%, respectively. Of those, the overwhelming majority (NCDB 80%; SEER 88%) were missing tumor variables. When compared to patients with complete data, missingness was associated with a greater risk of death: NCDB HR 1.23 (99% CI 1.21-1.25) and SEER HR 2.11 (99% CI 2.05-2.18). Patients with complete tumor data had higher unadjusted OS estimates than that of the entire sample: NCDB 82.7% vs 81.8% and SEER 83.5% vs 81.7% for 5-year OS. Conclusions Missingness of select variables is not uncommon within large national cancer registries and is associated with a worse OS. Exclusion of patients with missing variables may introduce unintended bias into analyses and result in findings that underestimate breast cancer mortality.

Automated summary

This study evaluated the association between missing data and overall survival in national cancer registries and found that missingness of select variables is not uncommon and is associated with worse overall survival.