Internal Staging Discordance in National Cancer Databases for Non-Small Cell Lung Cancer

Abstract

BackgroundThe Surveillance, Epidemiology and End Results (SEER) and the National Cancer Database (NCDB) are databases for cancer analysis that may be subject to error in data reporting. This study examined the rates and impact of discordant data for non-small cell lung cancer. MethodsNCDB and SEER were queried for non-small cell lung cancer pathologic tumor, node, metastasis data (NCDB) or “derived” data (SEER). Discordancy between descriptors with stage and impact of outlier data were analyzed. ResultsIncomplete staging was noted in 71.5% of the NCDB and 10.3% of SEER patients. A total of 174 829 patients from the NCDB and 117 114 from SEER were analyzed. The NCDB had 97 cases with ≥20 positive lymph nodes recorded vs 27 in SEER (P < .001). Mean and median sampled lymph nodes were skewed with inclusion of these data points (P < .001). The NCDB misclassified 0.99% tumors >5 cm as stage I vs 0.04% in SEER (P < .001). The NCDB misstaged positive lymph nodes as pathologic N0 (0.59%) or stage 0 or stage I (0.65%). The NCDB misclassified pathologic N1 as lower than stage II (0.91%) or N2 as lower than stage III (0.36%). The NCDB misclassified stage I with documentation of pathologic N1 or N3 disease (0.24%) or stage II with evidence of N2 or N3 disease (0.50%). The NCDB misclassified pathologic M1 as pathologic stage <IV in 0.9% of cases and misclassified 19.8% of stage IV as pathologic M0. SEER collaborative staging had no discordance (P < .001). ConclusionsThe NCDB and SEER are 2 powerful cancer databases. However, the cumulative discordancy rate was 4.9% for the NCDB and 0.008% for SEER, with more misstaging and outliers in the NCDB.