Changes in cancer registry coding for lymphoma subtypes: reliability over time and relevance for surveillance and
Christina A Clarke1, Dawn M Undurraga, Patricia J Harasty
1Northern California Cancer Center, Fremont, CA 94538, USA. tina@nccc.org
Abstract:
Because lymphoma comprises numerous histologic subtypes, understanding the reasons for ongoing increases in its incidence requires surveillance and etiologic study of these subtypes. However, this research has been hindered by many coexisting classification schemes. The Revised European American classification of Lymphoid Neoplasms (REAL)/WHO system developed in 1994 and now used in clinical settings was not incorporated into the International Classification of Diseases-Oncology (ICD-O), used by cancer registries, until the release of the third edition (ICD-O-3) in 2001. Studies including patients diagnosed before 2001 may have codes from earlier ICD-O versions that must be converted to ICD-O-3 and have higher proportions of unclassified (e.g., lymphoma and not otherwise specified) cases. To better understand (a) the agreement of computer-converted ICD-O-3 codes to ICD-O-3 codes generated directly from diagnostic pathology reports and (b) the reproducibility of unclassified status, we reviewed a population-based series of diagnostic pathology reports for lymphoma patients diagnosed before (1988-1994; n = 1,493) and after (1998-2000; n = 1,527) the REAL/WHO scheme was introduced. Overall, computer- and coder-assigned ICD-O-3 codes agreed for 77% of patients in both groups and improved slightly (82%) when codes were grouped. The most common lymphoma subtypes, diffuse large B cell and follicular, had relatively good reliability (84-89%) throughout the study period. T-cell and natural killer cell lymphomas had worse agreement than B-cell lymphomas, even when grouped. Many (42-43%) lymphomas reported as unclassifiable could be assigned a subtype upon pathology report review. These findings suggest that the study of lymphoma subtypes could be improved by (a) use of more standardized terminology in pathology reports, (b) grouping individual ICD-O-3 codes to reduce misclassification bias, and (c) routine secondary editing of unclassified lymphomas by central cancer registries.