Introduction

Despite therapeutic advances, a significant proportion of patients with diffuse large B-cell lymphoma (DLBCL) experience disease relapse (R). Previously, we reported a 5-year cumulative incidence of systemic relapse (Sy-R) and central nervous system relapse (CNS-R) of 21.25% and 3.76%, respectively (Klanova et al., ASH 2023). Compared to Sy-R, CNS-R is generally associated with worse outcomes. Most CNS-Rs present with isolated CNS involvement (CNS-I-R), while a smaller subset develops concurrent CNS and systemic R (CNS+Sy-R). Based on localization, CNS-R can affect parenchymal (Par), leptomeningeal, or both sites (Lep/Comb). This study compared clinical outcomes of Sy-R, CNS-I-R (further divided into CNS-I-Par and CNS-I-Lep/Comb), and CNS+Sy-R in a large real-world DLBCL cohort.

Methods Using the prospective NiHiL project (NCT03199066), we identified 1,364 patients with systemic DLBCL who experienced their first R between 2005 and 2023. Patients with CNS involvement at initial diagnosis or incomplete staging at R were excluded. Among the included patients (n=1,351), 1,186 (87%) had Sy-R, 127 (9%) had CNS-I-R (CNS-I-Par n=92; CNS-I-Lep/Comb n=34; unknown localization n=1), and 38 (3%) had CNS+Sy-R. In total, 1,153 patients received chemotherapy (CTx), including 901 treated with curative intent.

We first compared outcomes across Sy-R, CNS-I-R, and CNS+Sy-R. Subsequently, CNS-I-R cases were stratified by localization (CNS-I-Par vs CNS-I-Lep/Comb). The primary endpoint was progression-free survival (PFS); overall survival (OS) was the secondary endpoint, both measured from the date of R. Univariate and multivariate Cox regression analyses were performed.

Results When comparing Sy-R, CNS-I-R, and CNS+Sy-R, the median time from diagnosis to relapse was 12.3, 14.2, and 9.3 months, respectively (P=0.01); median age at R was 68, 68, and 64.5 years, respectively (P=0.04). ECOG performance status (PS) 2–4 was more frequent in R involving the CNS (CNS-I-R 65%, CNS+Sy-R 63%) than in Sy-R (32%; P<0.01). In contrast, elevated LDH was more common in relapses with Sy involvement (Sy-R 67%, CNS+Sy-R 61%) compared to CNS-I-R (43%; P<0.01). With a median follow-up of 7.6 years, median PFS decreased across Sy-R, CNS-I-R, and CNS+Sy-R: 8 vs 6 vs 3 months (P<0.01), and median OS was 13, 6, and 3 months, respectively (P<0.01).

We then focused on CNS-I-R and compared outcomes by localization. Patients with CNS-I-Par (73%) were older (median age 70 vs 62.5 years; P=0.048) and had longer time from diagnosis to relapse (15.75 vs 8.15 months; P<0.01) than those with CNS-I-Lep/Comb (27%). Median PFS and OS for CNS-I-Par vs CNS-I-Lep/Comb were 7 vs 4 months (HR=0.49, P<0.01) and 8 vs 4 months (HR=0.48, P<0.01), respectively.

When comparing all four relapse types (Sy-R, CNS-I-Par, CNS-I-Lep/Comb, CNS+Sy-R), Sy-R and CNS-I-Par had the most favorable outcomes with similar PFS (median 8 vs 7 months; HR=0.98, P=0.87) and modestly better OS for Sy-R (median 13 vs 8 months; HR=0.77, P=0.04). CNS-I-Lep/Comb and CNS+Sy-R had similarly poor outcomes (median PFS 4 vs 3 months, HR=0.99, P=0.95; median OS 4 vs 3 months, HR=0.85, 95% CI 0.52–1.38, P=0.51).

After adjusted for age, ECOG PS, LDH, and diagnosis-to-relapse interval, survival was similar between CNS-I-Par and Sy-R (PFS P=0.63; OS P=0.15). However, both CNS-I-Lep/Comb (PFS HR=1.82, P=0.01; OS HR=1.55, P=0.05) and CNS+Sy-R (PFS HR=1.60, P=0.01; OS HR=1.69, P=0.01) remained independently associated with worse survival compared to Sy-R. These findings were consistent in the curatively treated subgroup (n=901): median PFS 10, 8, 4, and 3 months, median OS 19, 10, 4, and 4 months, resp., for Sy-R, CNS-I-Par, CNS-I-Lep/Comb, CNS+Sy-R resp. (all P<0.01).

Conclusion While all DLBCL relapses are associated with poor prognosis, clinical outcomes vary by relapse pattern. Sy-R and CNS-I-Par showed comparable PFS, and better survival compared to other subtypes,. In contrast, CNS-I-Lep/Comb and CNS+Sy-R were associated with worst outcomes. The OS in CNS-R cases closely mirrors the PFS, suggesting that treatment remains challenging, and therapeutic failure often represents a terminal event. Novel approaches, including CNS-penetrating targeted agents and CAR-T, which show early promise in CNS-Rs, are needed for these patients.

Funding

Grant No. NU23-03-00127.

This content is only available as a PDF.
2025
Sign in via your Institution