Integration of COO into the clinical CNS International Prognostic Index could improve CNS relapse prediction in DLBCL

Magdalena Klanova, Laurie H. Sehn, Isabelle Bence-Bruckler, Federica Cavallo, Jie Jin, Maurizio Martelli, Douglas Stewart, Umberto Vitolo, Francesco Zaja, Qingyuan Zhang, Federico Mattiello, Gila Sellam, Elizabeth A. Punnoose, Edith Szafer-Glusman, Christopher R. Bolen, Mikkel Z. Oestergaard, Guenter R. Fingerle-Rowson, Tina Nielsen and Marek Trneny

Key points

  • High CNS-IPI score and ABC/unclassified COO subtypes were independent risk factors for CNS relapse in DLBCL.

  • Combining CNS-IPI score and COO improved identification of DLBCL patients with different CNS relapse risks.


Central nervous system (CNS) relapse carries a poor prognosis in diffuse large B-cell lymphoma (DLBCL). Integrating biomarkers into the CNS International Prognostic Index (CNS-IPI) risk model may improve identification of patients at high risk of developing secondary CNS disease. CNS relapse was analyzed in 1,418 DLBCL patients treated with obinutuzumab or rituximab plus CHOP (cyclophosphamide, doxorubicin, vincristine, prednisone) chemotherapy in the phase III GOYA study. Cell-of-origin (COO) was assessed using gene expression profiling. BCL2 and MYC protein expression were analyzed by immunohistochemistry. The impact of CNS-IPI, COO, and BCL2/MYC dual-expression status on CNS relapse was assessed using a multivariate Cox regression model (data available in n = 1,418, n = 933, and n = 688, respectively). High CNS-IPI score (hazard ratio [HR], 4.0; 95% confidence interval [CI], 1.3–12.3; P = .02) and activated B-cell–like (ABC) (HR, 5.2; 95% CI, 2.1–12.9; P = .0004) or unclassified COO subtypes (HR, 4.2; 95% CI, 1.5–11.7; P = .006) were independently associated with CNS relapse. BCL2/MYC dual-expression status did not impact CNS relapse risk. Three risk subgroups were identified according to the presence of high CNS-IPI score and/or ABC/unclassified COO (CNS-IPI-C model): low risk (no risk factors, n = 450 [48.2%]); intermediate risk (one factor, n = 408 [43.7%]); and high risk (both factors, n = 75 [8.0%]). Two-year CNS relapse rates were 0.5%, 4.4%, and 15.2% in respective risk subgroups. Combining high CNS-IPI and ABC/unclassified COO improved CNS relapse prediction and identified a patient subgroup at high risk of developing CNS relapse. The study is registered to as NCT01287741.

  • Submitted July 19, 2018.
  • Accepted December 6, 2018.