Abstract
Background: DLBCL is a heterogeneous entity with patients exhibiting a wide range of clinical outcomes. A reliable prediction tool is vital for optimizing patient care. We have previously demonstrated that a revised International Prognostic Index (R-IPI) is a better predictor of outcome than the standard IPI following treatment with rituximab and CHOP (CHOP-R). (Sehn, ASH 2005). Further exploration of clinical prognostic factors in an expanded cohort of patients yields a simplified prediction model based on stage and LDH alone.
Methods: We performed a retrospective analysis to assess the predictive value of recognized clinical prognostic factors in an unselected population of patients with DLBCL treated with CHOP-R. Patients were identified using the Lymphoid Cancer Database of the British Columbia (BC) Cancer Agency and included all patients >16 y of age who were newly diagnosed with DLBCL prior to Jan 15, 2006, treated in BC with a CHOP-R protocol. Patients were excluded if they were HIV positive, had evidence of an active second malignancy or an underlying indolent NHL.
Results: 523 patients were identified with the following characteristics: median age 61 y (16–93); male, 60%; stage III/IV, 55%; elevated LDH, 53%; performance status (PS) ≥2, 41%; ≥2 extranodal sites, 29%. Patients with advanced stage lymphoma (stage III/IV, B symptoms or bulky disease) or patients who were not candidates for radiation therapy were treated with 6–8 cycles of CHOP-R (91%) while those with limited stage were treated with CHOP-Rx3 and involved field radiation therapy (9%). Median follow-up for living patients is 24 months. On univariate analysis, stage, PS and LDH remain predictive for both PFS and OS. Number of extranodal sites is a borderline predictor of PFS (p=0.05) and no longer predictive of OS (p=0.276). Although age no longer predicts for PFS (p=0.17), it remains predictive of OS (p=0.001) since elderly patients are less likely to be treated successfully with secondary therapy and more likely to die from unrelated causes. On multivariate analysis, only stage (p<0.001) and LDH (p<0.001) remain significant predictors of PFS with similar predictive capacity (relative risk 2.7 and 3.0 respectively). A prediction model based on these two factors alone can predict outcome (see Table) as effectively as the full 5-factor IPI.
Conclusions: A prediction model based on stage of disease and serum LDH level at presentation provides an accurate estimate of outcome following primary therapy with CHOP-R for patients with DLBCL.
Disclosures: Unconditional research grant from Roche, Canada.
