Population-based genomic detection of childhood cancer predisposition using newborn dried blood spots

Population-based genomic newborn screening identifying newborns at risk for early-onset cancers has not been evaluated. Here, within a Michigan birth cohort (1987-2020), we identify all children developing a solid or central nervous system malignancy by age 8 years (n = 1948). We perform targeted sequencing of 11 cancer predisposition genes using archived newborn dried blood spot DNA. We find pathogenic or likely-pathogenic germline variants (P/LP) in 6.8% of cases (n = 132): RB1 (n = 69), TP53 (n = 24), SMARCB1 (n = 8), WT1 (n = 7), RET (n = 6), SUFU (n = 6), PTCH1 (n = 4), DICER1 (n = 4), APC (n = 3) and PHOX2B (n = 1). We show approximately 1/27,000 newborns develop an early-onset malignancy with an associated pathogenic or likely-pathogenic variant. Germline variant prevalence is 100% in medullary thyroid carcinoma, 40% in retinoblastoma, and 11-30% across five additional diagnoses, with strong gene-tumor specificity (p < 0.001). P/LP variants are rare in comparison datasets from healthy newborns and gnomAD. Our data support newborn screening for selected cancer-risk genes.