Risk of central nervous system versus systemic relapse in patients with diffuse large B-cell lymphoma treated with R-CHOP: What should we focus on?

Diffuse large B-cell lymphoma (DLBCL) is the most common hematological malignancy in adults. Although clinical outcomes of patients with DLBCL have improved over the past two decades, a significant proportion of DLBCL patients remain at risk of lymphoma relapse/progression or death due to relapse.^{1, 2} Patients with high international prognostic index (IPI) or central nervous system IPI (CNS-IPI) scores face the greatest risks of systemic and CNS relapse.^3-5 The CNS-IPI model has been validated in numerous studies and is widely used in clinical practice to guide decisions regarding CNS prophylaxis.^6-8 However, the effectiveness of CNS prophylaxis has been repeatedly questioned in the recent retrospective studies focusing on both intrathecal (IT) and intravenous high-dose methotrexate (HD MTX).^9-14 There is a lack of robust, prospective clinical trials evaluating the impact of CNS prophylaxis on the incidence of CNS relapse. As such, CNS prophylaxis may still be considered for high-risk patients, in accordance with international guidelines (NCCN v5.2025). However, patients classified as high-risk by IPI or CNS-IPI are at risk of not only CNS relapses but also systemic relapses. The risk proportion of these two types of relapse in the same patient population has not been formally studied to date. To address this, we analyzed the competing risks of systemic versus CNS relapse in a real-world cohort of patients with DLBCL.

Using data from the prospective observational NiHiL project (NCT03199066), we identified consecutively diagnosed adult patients with histologically confirmed DLBCL not otherwise specified or DLBCL/high-grade B-cell lymphoma (HGBL) with BCL2/MYC or BCL6 rearrangements, HGBL NOS, or T-cell/histiocyte-rich large B-cell lymphoma. Patients were diagnosed between January 2010 and December 2021 at two academic centers in the Czech Republic and were treated with anti-CD20 monoclonal antibody plus CHOP chemotherapy as first-line therapy. Patients with biopsy-proven, treatment-naïve transformed indolent lymphoma (excluding Richter's transformation) were included. Patients with CNS involvement at diagnosis were excluded. Causes of death were uniformly categorized as CNS (±systemic) relapse-related, systemic relapse-related, first-line treatment toxicity-related, other, or unknown. Deaths were considered first-line treatment toxicity-related if they occurred during first-line therapy or shortly thereafter (up to Day +60 after the last treatment administration) and were attributed to treatment toxicity in the absence of progressive or refractory disease. Deaths occurring after second or subsequent lines of therapy were considered relapse-related (CNS ± systemic or systemic), unless there was clear documentation of a non-lymphoma-related cause unrelated to active disease or its treatment (e.g., unrelated comorbidity). In such cases, the cause of death was classified as other or unknown, depending on the available documentation. Statistical methods are described in Supporting Information.

A total of 1228 newly diagnosed DLBCL patients treated with curative-intent anti-CD20-CHOP were included. The median time in study for all patients was 5.64 years (range: 0.1–14.1). The median time in study for all living patients was 7.29 years (range: 1.4–14.1). Patient and disease characteristics are listed in Table S1. According to the CNS-IPI score (available in 1222 out of 1228 patients), 343 (28.1%), 541 (44.3%), and 338 (27.7%) patients were categorized as being at low (0–1), intermediate (2–3), and high risk (4–6) for developing CNS relapse, respectively. Overall, 300 (24.4%) patients received CNS prophylaxis with either HD-MTX, HD-AraC, IT chemotherapy, or combinations. Among CNS-IPI risk groups, 145 (42.9%) of high-risk, 109 (20.1%) of intermediate-risk, and 45 (13.1%) of low-risk patients received CNS prophylaxis. Full details are available in Table S1.

Overall, 321 (26.1%) out of 1228 patients developed lymphoma relapse; 48 of these occurred in the CNS (38 isolated CNS relapses, 10 simultaneous CNS and systemic relapses), whereas 273 lymphoma relapses occurred systemically. Within the low, intermediate, and high CNS-IPI, 7 (2.0%), 10 (1.8%), and 31 (9.2%) patients developed CNS (±systemic) relapse and 42 (12.2%), 121 (22.4%), and 108 (32.0%) patients developed systemic relapse, respectively (Table S2). In patients with any type of relapse, the median times to systemic and CNS (±systemic) relapse were 0.97 years (range: 0.1–13.1) and 0.99 years (range: 0.2–10.6), respectively. The time to systemic and CNS (±systemic) relapses differed between the CNS-IPI risk groups and is shown in Table S2. Consistent with prior retrospective studies, CNS prophylaxis had no impact on CNS relapse incidence in patients with high CNS-IPI scores (P = 0.82; Figure S1A,B).^{7, 9-14} This result must be interpreted with caution due to the retrospective nature of the study and potential selection bias, particularly in patient selection for CNS prophylaxis. Moreover, the low incidence of CNS relapse in DLBCL limits the statistical power of the analysis.

In the entire cohort, the 5-year cumulative incidence of systemic and CNS (±systemic) relapse was 21.25% and 3.76%, respectively (Figure 1A). Across all CNS-IPI risk groups, there was a higher incidence of systemic as compared to CNS (±systemic) relapse. By CNS-IPI, the 5-year cumulative incidence of systemic and CNS (±systemic) relapse was 8.34% versus 1.6% in the low-risk group, 23.12% and 1.79% in the intermediate-risk group, and 31.99% and 9.45% in the high-risk group, respectively (Figures 1B–D and S2A,B). The IPI stratification provides a very similar incidence of systemic and CNS relapses as CNS-IPI (Figure S2C,D).

A detailed analysis of the main causes of death was performed in the entire DLBCL cohort and within the specific CNS-IPI risk groups. At the time of analysis, 429 (34.9%) out of 1228 patients had deceased. In the entire cohort, the most common cause of death was systemic lymphoma relapse (157 cases, 36.6%), followed by other causes of death (131, 30.5%), first-line treatment toxicity (37, 8.6%), and CNS (±systemic) relapse (33, 7.7%). The cause of death was unknown in 71 (16.6%) out of 429 patients. At the time of analysis, 57 (16.6%) out of 343 patients with low, 194 (35.9%) out of 541 patients with intermediate, and 176 (52.1%) out of 338 patients with high CNS-IPI had deceased. The causes of death within the CNS-IPI risk groups are shown in Table S3. CNS (±systemic) relapse was associated with significantly shorter post-relapse overall survival (OS) as compared to systemic relapse (P = 0.0003, Figure S3). Median OS after CNS (±systemic) and systemic relapse was 0.31 versus 1.04 years, respectively. In the entire cohort, the 5-year cumulative incidence of lymphoma-associated death was 16.8%, with 11.3%, 3.0%, and 2.5% of death due to systemic relapse, first-line treatment toxicity, and CNS (±systemic) relapse, respectively (Figure S4A,B). The 5-year cumulative incidences of death unrelated to lymphoma and unknown causes were 6.2% and 3.5%, respectively (Figure S4A,B). In the high CNS-IPI risk group, the 5-year cumulative incidence of lymphoma-associated death was 32.3%, with 20.2%, 5.3%, and 6.9% of death due to systemic relapse, first-line treatment-related toxicity, and CNS (±systemic) relapse, respectively. The 5-year cumulative incidence of death unrelated to lymphoma and unknown causes was 7.3% and 4.5%, respectively (Figure S4G,H). In patients with high CNS-IPI, systemic relapse was the leading cause of death, occurring more frequently than CNS relapse, with the majority of systemic relapse-related deaths taking place within the first 2.5 years following lymphoma diagnosis (Figure 2D).

These data highlight the key risk events in patients with DLBCL and may guide future treatment strategies. Our results show that patients with high CNS-IPI scores are at increased risk of both systemic and CNS relapse, and are often treated with CNS prophylaxis. We lack prospective studies demonstrating the benefit of CNS prophylaxis in reducing the risk of CNS relapse. Retrospective studies, including our findings, despite their limitations, have failed to show a clinically significant benefit of CNS prophylaxis.^9-14 Wilson et al. demonstrated that HD-MTX, administered either intercalated with R-CHOP (intercalated MTX [i-MTX]) or at the end of treatment (end of treatment MTX [EOT-MTX]), is associated with increased toxicity. Notably, i-MTX led to a higher risk of ≥ 7-day delays in R-CHOP cycles, which may compromise systemic disease control.¹⁵ Results of this study demonstrate that patients with high CNS-IPI have a significantly higher risk of systemic relapse compared to CNS relapse, with systemic relapse being the leading cause of death in this population (Figures 1 and 2). These findings strongly support a treatment strategy focused on controlling systemic disease through the timely delivery of first-line therapy. Nonetheless, CNS relapse—although less common than systemic relapse—remains a serious therapeutic challenge, primarily due to the lack of effective treatment options, which contributes to its significantly inferior post-relapse survival. Prior studies have shown no difference in CNS relapse incidence between IT MTX and systemic HD-MTX, or between EOT-HD-MTX and i-HD-MTX.^{15, 16} Accordingly, EOT-HD-MTX or IT MTX may be considered safer options in patients still deemed to benefit from CNS prophylaxis. Optimally, novel first-line regimens capable of simultaneously reducing both systemic and CNS relapse risks are needed to improve outcomes, particularly in patients with high CNS-IPI scores.

Magdalena Klanova: Conceptualization; writing—original draft; investigation; funding acquisition; data curation; supervision; project administration. Samuel Hricko: Data curation; investigation; writing—review and editing. Michal Masar: Data curation; investigation; writing—review and editing. Prokop Vodicka: Data curation; investigation; writing—review and editing. David Salek: Writing—review and editing; resources. Natasa Kopalova: Data curation. Petra Blahovcova: Formal analysis; writing—review and editing; visualization. Daniela Kuruczova: Formal analysis; writing—review and editing; visualization. Katerina Benesova: Resources; writing—review and editing. Jozef Michalka: Resources; writing—review and editing. Jan Koren: Writing—review and editing; resources. Pavel Klener: Writing—review and editing; resources. Andrea Janikova: Conceptualization; investigation; funding acquisition; resources; writing—review and editing. Marek Trneny: Conceptualization; funding acquisition; resources; writing—original draft; investigation; supervision.

The authors declare no conflicts of interest.

This study was approved by the Ethics Committee of the General University Hospital in Prague (approval no. 40/22 Grant AZV VES) and the Ethics Committee of the University Hospital Brno (approval no. 44-220622/EK).

This study was supported by AZV NU23-03-00127 and NU21-03-00411 grants, by Charles University Hematology-Oncology Cooperatio Program, and research project BBMRI_CZ LM2023033.