Abstract
Central nervous system (CNS) relapse in diffuse large B-cell lymphoma (DLBCL) is a rare but devastating event, with dismal prognosis. It is observed particularly in patients with high CNS-International Prognostic Index (CNS-IPI) scores, where 2-year CNS relapse rates can exceed 10%. Despite intensive salvage therapies, median overall survival (OS) often remains below 6 months. CNS prophylaxis, notably with high-dose methotrexate (HD-MTX) or intrathecal methotrexate (IT-MTX), is widely practiced, though its true efficacy on CNS relapse remains debated. This study aimed to evaluate whether CNS prophylaxis is associated with reduced CNS relapse and progression-free survival (PFS) in high-risk DLBCL patients in a real-world setting.
A retrospective analysis was performed including patients diagnosed with aggressive B-cell lymphoma and high CNS relapse risk (CNS-IPI ≥ 4), treated with anti-CD20-based chemoimmunotherapy between 2015 and March 2025 at two Brazilian centers. Patients with Burkitt lymphoma or transformed indolent lymphomas were excluded. The primary endpoint was CNS relapse, with PFS, OS and cumulative incidence of relapse (CIR) as secondary endpoints. Fisher's exact test was used for categorical comparisons given low expected frequencies. Logistic regression was used to evaluate the association between CNS prophylaxis and CNS relapse, with odds ratios (ORs) and 95% confidence intervals (CIs) reported. PFS and OS were analyzed using Kaplan-Meier and Cox proportional hazards models, while CIR was estimated using cumulative incidence functions. HD-MTX was modeled as a time-dependent covariate using Cox regression to eliminate immortal time bias. Simon-Makuch survival curves accounted for treatment switching over time. Analysis used timereg package in R, with p<0.05 considered significant.
Our cohort comprised 98 patients (median age: 67 years; 52% female; median CNS-IPI: 4, range 4–6). Cell-of-origin (COO) was ABC in 50% and GCB in 48%. High-risk extranodal involvement (adrenal, kidney, breast, testicles) was present in 28.5%. Among the 38% who received CNS prophylaxis, 47% received HD-MTX, 42% IT-MTX, and 10% both. HD-MTX was primarily administered at 3 g/m² for 2 cycles, and IT-MTX typically involved 4 administrations. Median follow-up was 26 months. Ten patients (10.2%) developed CNS relapse, leading to death in 80% of these cases. The CIR was approximately 14% for CNS involvement and 45% for disease progression. No significant association was found between HD-MTX prophylaxis (with or without IT) and CNS relapse (OR = 0.85; 95% CI: 0.08–4.77; p = 1.00). Individually, neither HD-MTX (OR = 0.90; 95% CI: 0.12–4.29; p = 0.902) nor IT-MTX (OR = 1.29; 95% CI: 0.18–6.31; p = 0.773) significantly reduced CNS relapse. Similarly, no significant OS difference was observed between prophylaxis groups (p = 0.2). Among the 21 patients who received HD-MTX, 6 (28%) discontinued treatment due to toxicity. Conversely, patients who received HD-MTX had a significantly lower risk of disease progression compared to those without prophylaxis (OR = 0.26; 95% CI: 0.07–0.86; p = 0.035). This association remained significant even after adjusting for CNS-IPI, extranodal involvement, and COO. The cumulative incidence of disease progression was 52% for patients without HD-MTX prophylaxis and 22% for those who received it (p = 0,008). To mitigate immortal time bias, we modeled HD-MTX as a time-dependent covariate and HD-MTX remained significantly associated with a lower risk of disease progression (HR = 0.43; 95% CI: 0.19–0.99; p = 0.048).
Consistent with recent findings, our data demonstrate that CNS relapses occurred despite prophylaxis in patients with high risk DLBCL. However, our multivariable analysis uniquely demonstrates a significant association between HD-MTX and reduced risk of systemic disease progression, even after mitigating the risk of immortal time bias. Potential selection bias inherent in retrospective studies, due to unmeasured factors, is a limitation.
