RADT-86. NEUROANATOMICAL AND MOLECULAR FEATURES OF SECONDARY CNS LYMPHOMA INFLUENCE THE PATTERNS OF CNS FAILURE AFTER SALVAGE RADIOTHERAPY

Abstract INTRODUCTION Radiotherapy (RT) is a potential salvage treatment for secondary central nervous system lymphoma (SCNSL), but its optimal use is not well-defined. We analyzed response and failure patterns after salvage RT. METHODS We identified 66 patients with large B-cell SCNSL who underwent brain RT for progression after ≥1 prior CNS therapy (excluding prophylaxis). Response was assessed by IPCG or RANO criteria. Overall survival (OS) was estimated by Kaplan-Meier from RT start; CNS failure risk used competing-risk methods; relapse patterns compared by chi-square. RESULTS Median follow-up was 3.3 months. Prior to RT, 94% received methotrexate (MTX), with a median interval of 23 days from last MTX to RT. At RT, 94% were neurologically symptomatic and 42% had extracranial disease. Median RT dose was 30 Gy in 10 fractions (whole-brain 71%; partial-brain 23%; craniospinal 6%). Among 41 evaluable patients, overall response rate was 86% (complete response 32%); median time to best response was 1.7 months, 58% were bridged to subsequent therapy. Median OS was 3.5 months, differing by involved neuroanatomical compartment: leptomeningeal 1.6 months; periventricular 6.4 months; parenchymal-only 8.3 months (p < 0.0001). The 12-month CNS failure risk was 45% (30–58); failure pattern differed by compartment (p = 0.0012). In leptomeningeal cases, 8/8 failures were confined to leptomeninges. In parenchymal-only cases, 6/9 failures were parenchymal, of which 5 were local. Periventricular cases had a mixed pattern of failure. MYC-rearranged patients had a higher risk of local failure (p = 0.046). DISCUSSION In this high-risk, symptomatic SCNSL cohort, salvage RT achieved high response rates and enabled bridging to next systemic therapy. Leptomeningeal involvement was a poor prognostic factor. Local relapse was the predominant failure pattern for parenchymal-only disease and was associated with MYC-rearrangement. These findings highlight the neuroanatomical and molecular diversity of SCNSL and suggest that partial brain RT may be suitable for therapy bridging in parenchymal-only cases.