Intrathecal chemotherapy and central nervous system relapse in acute lymphoblastic leukemia.

e23063 Background: Central nervous system (CNS) involvement is a rare but devastating complication of acute lymphoblastic leukemia (ALL). While intrathecal (IT) chemotherapy is routinely used for CNS prophylaxis, substantial variation exists across protocols regarding dosing and timing, and the relationship between IT chemotherapy administration patterns and CNS relapse risk remains poorly described. Methods: To address this gap, we conducted a single-institution retrospective study of factors associated with CNS relapse in patients with ALL. Patients were included if they received IT chemotherapy between 2014 and 2024 at the University of Chicago and all first-line (1L) IT doses were known, regardless of age, cell lineage (B-ALL vs T-ALL/T-LBL) or Philadelphia chromosome status (Ph-negative, Ph-like, or Ph-positive). Total 1L IT doses and CNS relapse status were compared by sociodemographic, biologic, and clinical factors using Wilcoxon rank-sum, ANOVA, and Fisher’s exact tests. Multivariable logistic regression model assessed which factors were associated with CNS relapse, and Youden index predicted an IT dose cutoff. Results: Among 321 patients, median age was 24 years (IQR 8–54); 45% were female, 21% Black, 25% Hispanic, and 31% uninsured or Medicaid-insured. Clinically, 84% had B-ALL (69% Ph-negative, 9% Ph-like, and 22% Ph-positive), while 19% had cytogenetic risk factors for CNS relapse (abnormal 17p, MLL rearrangement, or translocation 9;22) and 4% had CNS disease at diagnosis. Patients received a median of 14 1L IT doses (IQR 7-21) with fewer doses for the 16% who had a stem cell transplant (median 7 vs 16, p<0.001). Patients with T-cell lineage received more IT doses than B-cell (median 19 vs 13, p=0.004), even when accounting for transplant. B-cell had fewer IT doses for Ph-like (8) and Ph-positive (9) than Ph-negative (17, p<0.001). CNS relapse occurred in 18 patients (6%) but did not significantly differ by cell lineage, Ph status, cytogenetic features, or CNS disease status at diagnosis. CNS relapse was higher among males (10% vs 0%, p< 0.001) and Hispanic patients (12% vs 3%, p=0.02). CNS relapse was associated with receipt of fewer 1L IT doses (median 7.5 vs 14, p=0.049). In our multivariable model, only number of 1L IT doses was associated with CNS relapse (aOR 0.90, p=0.02). A cutoff of 15 IT doses had a sensitivity of 75% and specificity of 64%. Conclusions: Only 6% of patients with ALL developed CNS relapse, with higher risk among males and Hispanic patients. Patients with T-ALL did not have higher CNS relapse rate, possibly due to receiving more 1L IT doses. CNS relapse was associated with fewer 1L IT doses, with a predicted cutoff of 15 doses having limited discrimination. Future work is needed to evaluate protocol adherence to IT chemo doses to determine the optimal number to prevent CNS relapse while avoiding unnecessary neurocognitive toxicity.