Consistency of primary brain tumor diagnoses and codes in cancer surveillance systems

Abstract

High-quality cancer registry data are essential for assessing trends in incidence rates. This study evaluated the consistency of brain tumor surveillance data using a random sample of cases from the Connecticut Tumor Registry. Three neuropathologists independently and blindly reviewed tumor slides from 204 cases and a nosologist blindly reviewed and assigned International Classification of Diseases for Oncology (ICD-O) codes to 326 cases. For the pathology review, absolute concordance was as high as 81% for all primary brain tumors. Absolute concordance rates were high for nerve sheath (89%), meningioma (95%), and pituitary (95%) tumors. Rates were much lower for malignant tumors. ICD-O coding of malignant brain tumors is of relatively high quality with the exception of mixed gliomas and unspecified tumors. A high level of consistency for nonmalignant brain tumor diagnoses suggests that rates for these tumors, when actively reported to a surveillance system, can be of high quality.