Abstract
There remains an unmet clinical need to identify which patients with diffuse large B-cell lymphoma (DLBCL) would benefit from central nervous system (CNS) prophylaxis, due to the low positive predictive value (PPV; 10%-15%) of the currently available predictive models. To stratify patients at high risk of developing CNS relapse, we retrospectively analyzed 182 patients with DLBCL initially treated with rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP), or a R-CHOP-like regimen. Among them, 17 patients relapsed with CNS involvement, and the 2-year rate of CNS relapse was 7.9%. Upon carrying out multivariate analysis, ≥3 extranodal sites and elevated soluble interleukin-2 receptor (sIL-2R) levels at diagnosis were identified as independent risk factors for CNS relapse. The 2-year and 3.5-year rates of CNS relapse were 57.1% and 78.6%, respectively, in patients with both elevated sIL-2R and ≥3 extranodal sites. Furthermore, combined use of these risk factors of both elevated sIL-2R and ≥3 extranodal sites resulted in a high PPV (71.4%), negative predictive value (93.1%), and overall accuracy (92.3%) for undergoing CNS relapse. In conclusion, we propose a simple and valuable tool to predict patients with DLBCL at very high risk of CNS relapse.