PB2103: CAN WE ACCURATELY PREDICT SURVIVAL IN DIFFUSE LARGE B CELL LYMPHOMA IN THE RITUXIMAB ERA? RETROSPECTIVE COMPARISON OF IPI, AAIPI, R-IPI AND NCCN-IPI

Abstract

Background: In diffuse large B cell lymphoma (DLBCL), the value of the International Prognostic Index (IPI) and age-adjusted IPI (aaIPI) changed with rituximab, cyclophosphamide, doxorubicin, vincristine, prednisone (R-CHOP) as standard first line. The revised (R-IPI) and enhanced (NCCN-IPI) indices have emerged to improve prognostic ability. An analysis from Surrogate Endpoints for Aggressive Lymphoma (SEAL) database suggests NCCN-IPI better discriminates prognosis. Aims: In light of data scarcity and future changes in second line therapy for relapsed/refactory patients, we aimed to analyse the suitability of 4 survival prognostic indices in a DLBCL cohort. Methods: Single centre analysis of 169 patients diagnosed with de novo DLBCL from 2008 to 2016 and treated with R-CHOP. Primary central nervous system DLBCL or upfront stem cell transplant were exclusion criteria. IPI, aaIPI, R-IPI and NCCN-IPI scores at diagnosis were calculated. Analysis for 5-year overall survival (5y-OS) and progression free survival (5y-PFS) were performed using Kaplan-Meier curves and compared using the log-rank test. For each prognostic index, C-index test from receiver operating characteristic curves was calculated and area under the curve comparisons between prognostic indices were made using DeLong’s test. Results: Median age was 61 years (interval 18-82), 55% of patients had >60 years and 54% were male. At diagnosis, 23% had an ECOG performance status ≥2, 68% had Ann-Arbor stage III-IV, 63% had increased lactic dehydrogenase levels and 77% had extranodal disease (24% >1 site). Median follow-up was 71 months and 5y-OS and 5y-PFS were 74% and 67%, respectively. For the whole cohort, C-index for 5y-OS was 0.701, 0.629 and 0.751 for IPI, R-IPI and NCCN-IPI, respectively, and R-IPI was outperformed by IPI and NCCN-IPI (p=0.007 and p<0.001, respectively). For 5y-PFS C-index was 0.673, 0.695 and 0.689 for IPI, R-IPI and NCCN-IPI, with no statistically significant differences between indices. NCCN-IPI managed to discriminate the intermediate low (IL) and intermediate high (IH) risk groups for both 5y-OS (IL 89% versus (vs) IH 64%, p=0.001) and 5y-PFS (IL 77% vs IH 60%, p=0.029), contrary to IPI (5y-OS IL 69% vs IH 74%, p=0.552; 5y-PFS IL 63% vs HI 64%, p=0.925). The cohort was divided into age groups so that R-IPI and NCCN-IPI could be compared with aaIPI. For patients with ≤60 years, C-index for 5y-OS was 0.722, 0.585 and 0.724 for aaIPI, R-IPI and NCCN-IPI, respectively, and R-IPI was outperformed by NCCN-IPI but not by aaIPI (p=0.023 and p=0.077, respectively). For 5y-PFS C-index was 0.705, 0.622 and 0.625 for aaIPI, R-IPI and NCCN-IPI, with no statistically significant differences between indices. IL and IH risk groups could not be differentiated neither by aaIPI (5y-OS IL 85% vs IH 82%, p=0.725; 5y-PFS 76% vs 69%, p=0.556) nor NCCN-IPI (5y-OS IL 85% vs IH 70%, p=0.162; 5y-PFS IL 70% vs IH 70%, p=0.905), although in the last index there were no high-risk patients. Summary/Conclusion: NCCN-IPI seems the most robust prognostic tool for OS with frontline R-CHOP. Our results are in accordance with the SEAL analysis, although our sample was significantly smaller and SEAL included other immunochemotherapy regimens. However, improvements are needed for PFS prognostication, especially in younger patients within intermediate risk groups, and all indices performed similarly. Accurate prediction of PFS will ultimately allow for early benchmarking of patients for advanced or more aggressive therapies, and future research will be needed from bigger cohorts.