Coded tumor size may be unreliable for small metastatic renal cancers in the Surveillance, Epidemiology, and End Results dataset

Abstract

Objectives: To report a weakness in the April 2006 release of the Surveillance, Epidemiology, and End Results (SEER) dataset, in which the primary tumor size of small (< 1.8 cm) metastatic renal cancers was often incorrectly coded into the dataset from the measurement as listed in the patient's chart.

Methods: In the SEER dataset, 167 patients with tumor size < or = 2.5 cm had metastatic disease at presentation in 1998-2003. Each patient's chart was individually re-examined by SEER registries to determine the correct primary tumor size. This confirmed data were compared with the coded tumor size in the SEER dataset.

Results: Of the 167 re-examined cases, 2 had incorrect histology and 6 could not be verified. Of the remaining 159 cases, 87 (55%) were correctly coded for primary tumor size while 72 (45%) were incorrect. The error rate decreased with increasing size; for tumors < or = 1 cm, > 1-2 cm, and > 2-2.5 cm, error rates were 88%, 53%, and 6.8%, respectively (P < .001). A breakpoint in error rate occurred between tumor sizes < 1.8 cm (78%) and > or = 1.8 cm (10%) (P < .001). Most errors (72%) were miscoded by a factor of 10. Analysis of the latest April 2009 release suggests that most corrections have been incorporated into the public access dataset.

Conclusions: Coded primary tumor sizes in the April 2006 release SEER dataset for metastatic renal tumors < 1.8 cm from 1998 to 2003 were often inaccurate. Verification of tumor size in this subset was essential to insure data accuracy and quality of research. Researchers should recognize potential limitations of population-based cancer registries.