Assessing disparities in treatment outcomes for patients with diffuse large B-cell lymphoma Free

A67

Background: Diffuse large B-cell lymphoma (DLCL) is the most common lymphoma in the United States, with a median survival of <1 year in untreated patients. Recent randomized clinical trials demonstrated that adding rituximab to CHOP chemotherapy (RCHOP) significantly improves complete response rates and overall survival (OS; 5-year OS 45% CHOP vs. 58% RCHOP; Feugier, J Clin Oncol 2005). To assess whether disparities exist in DLCL treatment and evaluate whether RCHOP use has reduced disparities, we embarked on a retrospective study linking cancer outcomes and pharmacy data. As an initial step, we assessed the quality of data sources for examining disparities in lymphoma care. Methods: Utilizing previously published methods (Graiser, Cancer Informatics 2007), we identified pts with DLCL from: cancer registry data using histology codes (ICD-O 9680 & 9684), free-text searches of electronic medical records (EMR) and pathology notes, laboratory, and pharmacy data sources. We reviewed medical records to confirm each diagnosis and calculated sensitivity/specificity for each data source. The final population was examined for available data on race, other demographics, components of the International Prognostic Index for DLCL (IPI: age, performance status, serum lactate dehydrogenase, number of extranodal sites involved, stage), insurance status, employment status, and survival data. Results: Together the sources identified 885 potential cases of DLCL, with 521 cases confirmed by pathology/EMR review. Data sources ranged widely in their sensitivity (38%-78%) and specificity (50%-78%) for identifying cases of DLCL. Free-text searches of all EMR had the greatest sensitivity (specificity 55%) and searches of pathology notes had the greatest specificity (sensitivity 55%). Cancer registry ICD-O codes simultaneously maximized sensitivity (65%) and specificity (60%). 4 pts were removed from the final population due to substantial missing data. Among the remaining 517 pts with DLCL, 91% had data on race and 99.6% on marital status, but <40% of cases had insurance status or employer data. 81% of cases had treatment data. Among these: 294 pts were White, 76 Black, 6 Asian, 5 Hispanic, and 34 of Other/Unknown race. All pts had readily available data on date of last contact, survival status, and age at diagnosis, but other IPI data were missing for 31%-61% of cases. Between 1990 and 2006, Black and White pts received similar treatment regimens (CHOP 54% vs. 53%; RCHOP 20% vs. 21%). Conclusions: Linking pharmacy and cancer outcomes data provides a useful resource for research on treatment disparities. Additional data from chart reviews or prospective studies are needed to examine psychosocial and prognostic factors. However, this likely represents the largest cohort of Black pts with DLCL ever evaluated, and will provide meaningful data on disparities in treatment and treatment outcomes for this curable cancer.