Multi-institution analysis of tumor mutational burden and outcomes in pediatric central nervous system tumor patients.

ABSTRACT

BACKGROUND: Pediatric central nervous system (CNS) tumors are the leading cause of pediatric cancer mortality. Research addressing genomic biomarkers and clinical outcomes is needed to inform therapeutic decision-making. METHODS: We conducted a retrospective analysis of pediatric patients (age <21) diagnosed with a primary CNS tumor at four upstate New York hospitals from 2008 to 2021. Clinical and histopathologic data were identified from each patient, including genomic analysis of somatic mutations and tumor mutational burden (TMB) where available. These variables were each compared with overall survival using Cox regression analyses. Multivariable analysis was conducted to identify patient characteristics that may independently predict survival. RESULTS: We identified 119 patients. Common tumor types included low-grade glioma (N = 51), high-grade glioma (N = 29), and medulloblastoma (N = 11). Common driver mutations included TP53 inactivation (N = 16), BRAF-KIAA1549 fusion (N = 16), FGFR1 amplification (N = 12), BRAF V600E mutation (N = 12), NF1 loss (N = 12), and H3F3A K28M mutation (N = 6). Median TMB was one mutation/megabase (mut/Mb, range = 0-132). Overall survival was 79.9%. Variables associated with poorer survival on univariable analysis were higher TMB (p = .002, HR 4.97), high-grade tumors (p = .009, HR 84.3), and high-grade glioma histology (p = .021, HR 3.14). Multivariable analyses further identified TMB (p = .011, HR 4.46) and high-grade histology (p = .015, HR 5.28) as independently predictive of worse survival. Tumor progression was more common in high-TMB (N = 15, 44%) than in low-TMB tumors (N = 19, 35%). CONCLUSIONS: High TMB is correlated with higher rates of progression and death as compared to low-TMB tumors. These findings may help identify patients who may benefit from alternative treatments, such as immunotherapies.