In neuroblastoma, population screening during infancy failed to lower mortality because it primarily detected biologically indolent tumors rather than aggressive, life-threatening disease. The failure may reflect limited screening sensitivity, or disease onset after infancy among aggressive tumors. Here, we used an epigenetic mitotic clock based on fluctuating CpG DNA methylation to estimate patient-specific tumor mitotic ages and calendar ages in a cohort of unscreened children diagnosed with neuroblastoma. Aggressive cancers (stage 4) primarily started growing after the first year of life, making them undetectable by screening during infancy. In contrast, biologically more indolent tumors (stages 1, 2, 3 and 4S) often started growing in utero or during the first year of life, and affected children had better survival outcomes. Due to a short preclinical detectable phase of aggressive neuroblastomas, reducing mortality through screening is impractical as it would require frequent screening among older children. Patient-specific tumor-age estimation may help refine screening windows and improve early-detection strategies in other cancers where screening has so far failed to yield substantial mortality reductions.