High-Grade B-Cell Lymphoma, Not Otherwise Specified (HGBL, NOS): Central Nervous System (CNS) Involvement, Prophylaxis, and Recurrence Risk in a Multi-Institutional Series

Background: HGBL, NOS encompasses aggressive B-cell lymphomas with Burkitt lymphoma-like (BLL) or blastoid cytomorphology which are otherwise unclassifiable. We have previously described characteristics and outcomes of HGBL, NOS treated in 20 academic centers across the United States (Zayac et al, ASH 2021). Here, we report the CNS-related outcomes from this multi-institutional dataset.

Methods: We collected retrospective data on adult patients (pts) with HGBL, NOS diagnosed by academic hematopathologists in 2017-2021, excluding cases with unknown MYC rearrangement (MYC-R) status, double/triple-hit lymphomas, any diffuse large B-cell lymphoma (DLBCL, NOS) or other specified entities. Ten participating institutions conducted formal local pathology review; immunohistochemistry and cytogenetics were performed locally. Outcomes included progression-free (PFS) and overall survival (OS), and cumulative incidence function (CIF) of CNS recurrence (CNSrec accounting for competing risk of systemic relapse), estimated with 95% confidence intervals (CI).

Results: Among 160 pts (median age 64 years, 68% male, 82% germinal center B-cell [GCB] tumors), the CNS International Prognostic Index (IPI) was low in 21%, intermediate (int) in 44%, high in 30%, and missing in 5%. Baseline CNS involvement (CNSinv) was present in 11 pts (7%), including leptomeningeal in 6, parenchymal in 4, and both in 1 case. Only 2 factors were significantly associated with CNSinv: testis or female pelvis involvement (odds ratio [OR]=6.40; 95%CI: 1.09-37.7) and MYC-R (OR=3.51; 95%CI: 1.01-12.2). Pts with CNSinv were treated with R-CODOX-M/IVAC (27%), R-EPOCH (18%), R-hyperCVAD (18%), or R-CHOP + high-dose methotrexate (HDMTX, 18%). CNSinv was associated with numerically lower complete response rate (55% with vs 70% without, P=.32), PFS (3-year PFS: 53% and 55%, respectively; log-rank P=.45), and OS (2-year OS, 51% and 69%, respectively, P=.54).

Data on CNS-directed prophylaxis (ppx) were available for 139 pts without baseline CNSinv. CNS ppx was administered to 48%, including intrathecal (IT)-only in 29% (median 4 doses, range 1-10), intravenous HDMTX in 7% (median 2 doses, range 1-5), and both IT and HDMTX in 12%.

Among 59 recurrences after initial therapy, 16 involved the CNS: 7 leptomeningeal-only, 4 parenchymal-only, 4 involving both compartments, and 1 unspecified. Median time to CNSrec was 4.4 months (95% CI: 2.6-9.0), not significantly different from systemic recurrences (median 5.3 months, P=0.58). Median OS after CNSrec was 5.3 months (95% CI: 2.9-13.6), also not significantly different from systemic recurrences (median 7.7 months, P=.18).

For all 160 pts, the CIF of CNSrec at 3 years was 11.7% (95% CI: 6.9-17.9), but pts with baseline CNSinv were at highest risk (CIF=57.6% vs. 8.0% for others; P<0.001 on Gray’s test) and were excluded from further analysis of risk factors. Among other CNSrec events, 4/10 (40%) were in non-GCB HGBL, 5 of 6 tested (83%) had dual MYC/BCL2 expression (DEL), and 4 of 4 tested carried a TP53 alteration. CNSrec was associated with high-int/high IPI (CIF=13.4% vs. 1.7% for low/int; P=.020) but not CNS-IPI (7.5% high vs. 6.8% low/int; P=.85). Other significant risk factors for CNSrec included: involvement of bone marrow (CIF=20.5% vs. 4.5% for none, P=.002) or blood (47.2% vs. 5.3% for none, P<.001), non-GCB (20.5% vs. 5.9% for GCB, P=.021, Fig. A), DEL (10.8% vs. 1.5% for others, P=.019), and CD5+ HGBL (27.5% vs. 6.0% for CD5-, P=.013).

The risk of subsequent CNSrec did not differ according to first-line therapy, including comparing R-CHOP (CIF=10.2%) vs. more intensive regimens (7.0%, P=0.77), or R-CHOP vs. R-EPOCH specifically (CIF=6.7%, P=.74). We also observed no difference in CNSrec after receipt of any form of CNS ppx (P=.59; Fig. B), or specifically HDMTX (P=.18) or IT methotrexate MTX (P=.94).

Conclusions: In HGBL, NOS baseline CNSinv is uncommon, but highly prognostic for future CNSrec. Both CNS and systemic recurrences portend poor prognosis. CNS-IPI may not predict higher risk of CNSrec in HGBL, NOS. Although our study is limited by the sample size in this rare and heterogeneous disease, the CNSrec was higher in pts with blood/marrow involvement or the non-GCB or MYC/BCL2 DEL HGBL. Because the >10% risk of CNSrec is not mitigated by standard CNSppx modalities, HGBL, NOS should be included in trials of novel CNSppx approaches along with high-risk DLBCL or double-hit lymphoma.