Background: Uveal melanoma is a rare disease characterized by liver metastasis and a very poor prognosis. In patients with metastatic UM, a single treatment with isolated hepatic perfusion (IHP) with high dose melphalan has shown response rates of 40%, while immune checkpoint blockade with ipilimumab (3 mg/kg) in combination with nivolumab (1 mg/kg) every third week (IPI3/NIVO1 q3w) for four cycles followed by nivolumab monotherapy has shown response rates of 11-18%. The impact of these treatments on overall survival, especially if combined, is unclear. This phase I trial investigates the safety and tolerability of the combination of IHP and IPI3/NIVO1. Methods: Eligible in this multicenter open, randomized, controlled, phase I trial, were patients with liver dominant metastatic uveal melanoma who had not received previous systemic treatment. Patients were randomized to receive either (Arm A) IHP followed by combination immunotherapy (four cycles IPI3/NIVO1 q3w) or (Arm B) one neoadjuvant cycle of IPI3/NIVO1 prior to IHP followed by three cycles IPI3/NIVO1 q3w. Thereafter, both Arm A and B received monotherapy with nivolumab (480 mg q4w) for up to 1 year. IHP was performed using melphalan 1mg/kg perfused through the liver for 60 minutes under hyperthermia (40°C). The primary endpoint was incidence and severity of adverse events while, secondary endpoints included response according to RECIST v1.1 criteria reported according by the local investigator. Results: A total of 18 patients were included and randomized, nine to Arm A and nine to Arm B. Three patients did not undergo IHP as planned, one due to perioperative complications (Arm A), one due to extensive metastatic liver infiltration ( > 50% liver volume) (Arm B) and one due to encephalitis related to neoadjuvant IPI3/NIVO1 (Arm B). A total of 20 serious adverse events (SAEs) were reported in eleven patients. There were ten SAEs in each arm, with no treatment related deaths. In total, 11 of 18 patients (six in Arm A and five in arm B) did not complete the planned four cycles of IPI3/NIVO1, with a mean of 2.4 cycles in Arm A and 3.0 cycles in Arm B. Response was evaluable in 17 patients, with the best clinical responses reported as three complete responses (18%), four partial responses (24%), seven stable disease (41%) and three progressive disease (18%). The overall response rate was 63% in Arm A (5/8) and 22% in Arm B (2/9). Conclusions: For previously untreated patients with liver dominant metastatic uveal melanoma, treatment with IHP in combination with IPI3/NIVO1 had a high, yet manageable toxicity profile. The efficacy of this combination treatment is encouraging, however, one cycle of neoadjuvant IPI3/NIVO1 given before IHP did not seem beneficial, neither in regards to safety nor efficacy. Clinical trial information: NCT04463368.