Leukemia & Lymphoma | 2019
Cell-of-origin and bone marrow involvement increase specificity of defining patients with diffuse large B-cell lymphoma at high risk of CNS relapse: a study of 793 patients treated with chemoimmunotherapy in Asia
Abstract
Central nervous system (CNS) relapse occurs in 2–6% of patients with diffuse large B-cell lymphoma (DLBCL) and is a dismal diagnosis with poor outcomes. The National Cancer Comprehensive Cancer Network guidelines recommend use of the prospectively validated CNS-IPI, which includes the International Prognostic Index (IPI) factors and kidney/adrenal involvement [1]. Although there is a significant improvement, CNS-IPI had a modest positive predictive value with specificity of 5.2–13%, leading to unnecessary CNS prophylaxis. We postulated that integrating cell-of-origin by the Hans algorithm and additional extranodal sites to CNS-IPI may improve stratification of patients at higher risk of developing secondary CNS involvement. While cell-of-origin by gene expression profiling (GEP) was found to be prognostic for CNS relapse [2], GEP is not used routinely in clinical practice locally. The CNS-IPI was also derived in patients without extensive (>25%) bone marrow involvement. Given these limitations, we sought to refine the prognostic index by evaluating the impact of molecular subtype by the Hans algorithm and any percentage of concordant bone marrow involvement by immunohistochemistry with CNS relapse risk. Patients with newly diagnosed DLBCL treated with RCHOP chemotherapy (n1⁄4 793) from 2001 to 2015 were included from two tertiary centers’ database after excluding patients who were immunocompromised, had CNS disease at diagnosis, or did not receive R-CHOP chemotherapy (Supplementary Figure 1). CNS relapses were diagnosed by cerebrospinal fluid cytology or flow cytometry, or brain/vitreous biopsy. No routine CNS prophylaxis was recommended for specific groups of patients. PET-CT scan was used for staging in 31.5% of patients, while the rest of the patients had conventional CT scans. The primary end-point was time to CNS disease measured from date of diagnosis. Probabilities of time to CNS disease were estimated using the Kaplan–Meier method, and compared using the log-rank test. Two-year estimates of the risk of CNS relapse were reported with 95% confidence intervals (CIs). Univariate analysis of potential risk factors including IPI factors, specific extranodal sites, and non-germinal center B-cell like (GCB) subtype were correlated with risk of relapse using the log-rank test. Multivariable analysis initially identified IPI factors that are prognostic for CNS relapse. After adjusting for prognostic IPI factors, specific extranodal sites found significant in univariable analysis and non-GCB subtype were included in the multivariable model. Factors that remained significantly associated with CNS relapse in the multivariable model were included in the prognostic model. The assumption of proportional hazards was tested using a time-dependent covariate, and all variables in the multivariable analysis satisfied the proportionality assumption. C-statistics were calculated to evaluate the discriminatory value of the prognostic index compared with the CNS-IPI. Analyses were performed using Stata 11 (StataCorp, College Station, TX). Median age of the cohort was 59 years (inter-quartile range, IQR 49–67 years), 20% were IPI high risk, and 5.2% and 14.4% had kidney/adrenal and bone marrow involvement, respectively. Fifty-eight CNS events occurred at a median of 8.1 months (range 16 days–11.9 years), with two-year CNS relapse rate of 5.1% (95% CI 3.7–7.1%). Time to death after CNS relapse was a median of 110 days (IQR 51–212 days). The median follow-up is 3.52 years (IQR 1.14–7.09 years). While individual IPI factors were prognostic for CNS relapse in univariable analysis, elevated LDH, involvement of two or more extranodal sites, and age >60 years lost its significance in multivariable analysis of the five IPI factors. Multivariable analysis including additional risk factors identified kidney/adrenal, bone marrow, breast, limited testicular involvement (stage I/II), and non-GCB subtype