Quality Assessment of Tumor Size Data Collection for Pancreatic and Breast Cancer in SEER

In 2017, the Surveillance, Epidemiology, and End Results (SEER) program piloted a reactive quality audit plan (r-QAP) to analyze Collaborative Stage (CS) tumor size in breast and pancreatic cancer. Preevaluation objectives were to establish procedures and analytic scope for SEER quality audits, cutoffs for data completeness/accuracy, and key decision checkpoints.Tumor size data between 2004-2014 were selected from SEER registries for pancreatic and breast cancers, and initially assessed by site and registry for completeness. Further exploration was undertaken via cross tabulation in SEER with the American Joint Committee on Cancer (AJCC) 6th edition derived T data item to evaluate discrepancies between these closely related variables.For both cancer sites, completeness improved between 2004 and 2014, with the proportion of known tumor size values increasing from 60.6% to 79.2% in pancreatic cancer and from 94.0% to 95.9% in breast cancer. Tumor size plausibility categories were established wherein any tumor over 100 mm for pancreatic cancer or over 200 mm for breast cancer were considered highly unlikely. Only 2% of pancreas tumors and 0.1% of breast tumors were implausibly large per site-specific cutoffs. Less than 2% of all tumor size values were potentially discrepant in cross-tabulation with AJCC 6th edition derived T for each site.Most tumor size values appear to fall within acceptable ranges based on r-QAP activities, and implausibly large tumor size values are rare. Different natural histories and clinical presentation for pancreatic and breast cancer illustrate the need for site-specific cutoffs. Our results indicate that there are no major quality issues in the SEER research database for the CS tumor size data item in either pancreatic or breast cancer.