Craniospinal Irradiation for CNS Leukemia: Rates of Response and Durability of CNS Control.

Abstract

Craniospinal irradiation (CSI) is used in the management of leukemia patients with central-nervous-system (CNS) involvement, though the data on response and local control are limited. Given the radioresponsiveness of leukemia, we hypothesized that response to CSI would be high, but CNS control would be influenced by control of systemic disease. This retrospective, single-institution analysis included consecutive pediatric and adult patients between 2009-2021 with leukemia that underwent CSI for CNS involvement, defined as presence of blasts (i.e., >0%) on CSF flow cytometry. Endpoints included CNS response rate (RR), CNS local recurrence (LR), progression-free survival (PFS), and overall survival (OS), which were estimated from start of CSI. The probability of CNS LR was summarized using a cumulative incidence estimate, where death without LR was considered a competing risk. The probabilities of OS and PFS were obtained using Kaplan-Meier estimates. Among the 39 eligible patients (43% AML, 49% ALL, 8% blast-phase CML), most were male (59%). All had CSF confirmation of disease. Median age at CSI was 31 years (range 7-67). CSI (protons 54%, photons 46%) was utilized early within the CNS disease course (median 0 CNS relapses prior to CSI). Twenty-five patients (64%) received CSI immediately prior to a stem-cell transplant (SCT), of which 21 (84%) had TBI conditioning to a median dose of 12 Gy (range 2-13.2). Patients treated with CSI alone received a higher CSI dose (median 18 Gy; range 10.8-24) than those treated with SCT consolidation (median 12 Gy; range 10.8-24). Fifteen patients had CSF-positive disease immediately prior to CSI; all 14 of those assessed for response (RR 100%) had confirmed clearance of blasts at a median of 23 days (range 7-197) from CSI start. With a median follow-up of 48 months (range 0.4-123) for survivors, 2-year PFS and OS estimates were 32% and 43%, respectively. Only 5 CNS relapses were noted (2-year CNS LR of 14%). All CNS relapses either occurred after (n = 4) or concurrently (n = 1) with a systemic relapse. In Cox regression univariate models, age, sex, time to CNS disease, positive CSF immediately prior to CSI, and SCT did not show demonstrable evidence of association with CNS LR. However, systemic relapse after CSI (HR 5.9, 95% CI 2.5-13.8, P<0.0001) and systemic disease at the time of CSI (HR 3.9, 95% CI 1.6-9.5, P = 0.003) were associated with higher risk of CNS LR. No grade-3+ acute toxicity was seen during CSI. CSI is a well-tolerated and effective treatment option for patients with CNS leukemia. Though CNS local recurrence was modest, there was a high risk of systemic relapse and/or death. Control of systemic disease, both before and after CSI, may be important for CNS local control, and raises consideration that CNS recurrence may reflect reseeding from the systemic space.