Session Item

Despite high response rates to chemotherapy, many patients with primary central nervous system lymphoma (PCNSL) relapse. Detailed patterns of failure analyses are lacking particularly following contemporary consolidation strategies. We analyzed patient-, disease-, and treatment-related factors associated with relapse patterns in a large cohort.

Patients diagnosed with PCNSL (diffuse large B-cell subtype) from 1983-2020 were identified from an institutional database. Initial T1 post-contrast-enhancing disease was characterized by site and largest dimension. Patients were stratified by response (complete or unconfirmed [uCR], partial [PR], stable, progression [POD]) to induction and consolidation (autologous stem cell transplant [ASCT], whole-brain radiotherapy [≤24Gy reduced-dose, RD-WBRT vs >24Gy standard-dose, SD-WBRT], cytarabine, other). Refractory was defined as POD on induction or relapse within 3 months of induction completion. Site of initial relapse was characterized as local (involving/adjacent to baseline site) vs distant intraparenchymal, leptomeningeal (CSF and/or MRI), or other. Progression-free survival (PFS) was assessed from diagnosis using Kaplan Meier and proportional hazards. Cumulative incidence with death as a competing risk was used to identify factors associated with local relapse.

Of 645 patients evaluated, 559 were eligible. Median follow-up was 7.4 years and median overall survival was 4.8 years. Median age was 63 years (IQR 54-72); most (n=321, 57%) patients were MSKCC recursive partitioning analysis (RPA) class 2 (age≥50, KPS≥70; Table). Most presented with supratentorial (n=420, 81%), multifocal (274, 53%), bilateral (224, 43%), and deep structure involvement (n=314, 56%). Nearly all received methotrexate-based induction (n=532, 95%). Among the 351 patients (91% of 385) who achieved CR to consolidation, PFS at 1- and 5-years was 93% (95%CI 90-95) and 66% (95%CI 60-72%), respectively; relapse patterns were mostly distant from the initial site, with a 1-year cumulative incidence from consolidation of non-local vs local POD of 15% vs 1.8%, respectively (Figure). Of the 97 refractory patients, the 1-year cumulative incidence from diagnosis of local vs non-local POD was 57% vs 42%, respectively with median OS of only 1.0 year from diagnosis (95%CI 0.8-2.2). Tumor (HR 1.21, 95%CI 1.00-1.46, p=0.06) and deep structure involvement (HR 1.77, 95%CI 1.03-3.05, p=0.04) were associated with local POD in refractory patients.

We report the first comprehensive patterns of relapse in a large PCNSL cohort. For a disease historically considered inherently multifocal, we confirmed that relapses after CR to consolidation are often distant and unpredictable; however, we found a particularly high cumulative incidence of local relapse among refractory and early-relapsing patients. These findings suggest a potential role for local therapy (i.e., involved site RT) in this high-risk subset.