Interim analysis of the DOSISPHERE-01 study demonstrated a strong improvement in response and overall survival (OS) on using 90Y-loaded glass microspheres with personalized dosimetry compared with standard dosimetry in patients with nonoperable locally advanced hepatocellular carcinoma. This report sought to provide a long-term analysis of OS. Methods: In this phase II study (ClinicalTrials.gov identifier NCT02582034), treatment was randomly assigned (1:1) with the goal to deliver either at least 205 Gy (if possible >250-300 Gy) to the index lesion in the personalized dosimetry approach (PDA) or 120 ± 20 Gy to the treated volume in the standard dosimetry approach (SDA). The 3-mo response of the index lesion was the primary endpoint, with OS being one of the secondary endpoints. This report is a post hoc long-term analysis of OS. Results: Overall, 60 hepatocellular carcinoma patients with at least 1 lesion larger than 7 cm and more than 30% of hepatic reserve were randomized (intent-to-treat population: PDA, n = 31; SDA, n = 29), with 56 actually treated (modified intent-to-treat population: n = 28 in each arm). The median follow-up for long-term analysis was 65.8 mo (range, 2.1-73.1 mo). Median OS was 24.8 mo and 10.7 mo (hazard ratio [HR], 0.51; 95% CI, 0.29-0.9; P = 0.02) for PDA and SDA, respectively, in the modified intent-to-treat population. Median OS was 22.9 mo for patients with a tumor dose of at least 205 Gy, versus 10.3 mo for those with a tumor dose of less than 205 Gy (HR, 0.42; 95% CI, 0.22-0.81; P = 0.0095), and was 22.9 mo for patients with a perfused liver dose of 150 Gy or higher, versus 10.3 mo for those with a perfused liver dose of less than 150 Gy (HR, 0.42; 95% CI, 0.23-0.75; P = 0.0033). Lastly, median OS was not reached in patients who were secondarily resected (n = 11, 10 in the PDA group and 1 in the SDA group), versus 10.8 mo in those without secondary resection (n = 45) (HR, 0.17; 95% CI, 0.065-0.43; P = 0.0002). Only resected patients displayed favorable long-term OS rates, meaning an OS of more than 50% at 5 y. Conclusion: After longer follow-up, personalized dosimetry sustained a meaningful improvement in OS, which was dramatically improved for patients who were accurately downstaged toward resection, including most portal vein thrombosis patients.