Background <p>EPOCH demonstrated superior progression free survival using yttrium-90 [⁹⁰Y] glass microspheres radioembolization plus chemotherapy versus chemotherapy in colorectal cancer patients with liver metastases who had progressed following first-line oxaliplatin or irinotecan-based chemotherapy. Exploratory post-hoc analyses and theoretical modeling assessed the potential impact of dosimetry on safety outcomes.</p> Patients and methods <p>Liver-related treatment-emergent adverse event (TEAE) frequency and/or severity, and bilirubin increase were explored in participating patients. For hypothesis generation, a theoretical model was constructed to understand EPOCH dosimetry in the absence of multicompartment dosimetry data.</p> Results <p>Liver-related TEAEs were reported in 93/186 (50%) patients, TEAEs ≥ grade 3 in 44/186 (24%) patients. A total of 26/186 patients (14%) experienced a bilirubin increase TEAE of any grade, in five of which assessed as related to glass [⁹⁰Y]Y-radioembolization with a median fractional tumour involvement of 2.3% (range 1.4–4.1%). Under the theoretical assumption of a T/N ratio of 2–10, most patients in EPOCH (62% had a fractional tumour involvement &lt; 10%) received a normal liver absorbed dose of at least 75 Gy. In this model, at &lt; 4.1% fractional tumour involvement, the normal liver absorbed dose approximates 120 Gy at any T/N ratio.</p> Conclusion <p>When treatment is same-day whole liver at 120 Gy average absorbed dose, the normal liver absorbed dose is dependent on fractional tumour involvement.</p> Trial Registration <p>ClinicalTrials.gov NCT01483027.</p>