Objectives. A phase I pretargeted radioimmunotherapy trial (EudractCT 200800603096) was designed in patients with metastatic lung cancer expressing carcinoembryonic antigen (CEA) to optimize bispecific antibody and labelled peptide doses, as well as the delay between their injections.Methods. Three cohorts of 3 patients received the anti-CEA x anti-histamine-succinyl-glycine (HSG) humanized trivalent bispecific antibody (TF2) and the IMP288 bivalent HSG-peptide. Patients underwent a pre-therapeutic imaging session S1 (44 or 88 nmol/m2 of TF2 followed by 4.4 nmol/m2, 185 MBq, of 111In-labelled IMP288), and, 1-2 weeks later, a therapy session S2 (240 or 480 nmol/m2 of TF2 followed by 24 nmol/m2, 1.1 GBq/m2, 177Lu-labeled IMP288). The pretargeting delay was 24 or 48 hours. The dose schedule was defined based on pre-clinical TF2 pharmacokinetic studies, on our previous clinical data using the previous anti-CEA pretargeting system and on clinical results observed in the first patients injected using the same system in the Netherlands.Results. TF2 pharmacokinetics (PK) was represented by a two-compartment model in which the central compartment volume was linearly dependent on the patient's surface area. PK were remarkably similar, with a clearance of 0.33 +/- 0.03 L/h per m2. 111In- and 177Lu-IMP288 PK were also well represented by a two-compartment model. IMP288 PK were faster (clearance 1.4 to 3.3 l/h). The central compartment volume was proportional to body surface area and IMP288clearance depended on the molar ratio of injected IMP288 to circulating TF2 at the time of IMP288 injection. Modelling of image quantification confirmed the dependence of IMP288 kinetics on circulating TF2, but tumour activity PK were variable. Organ absorbed doses were not significantly different in the 3 cohorts, but the tumour dose was significantly higher with the higher molar doses of TF2 (p < 0.002). S1 imaging predicted absorbed doses calculated in S2. Conclusion. The best dosing parameters corresponded to the shorter pretargeting delay and to the highest TF2 molar doses. S1 imaging session accurately predicted PK as well as absorbed doses of S2, thus potentially allowing for patient selection and dose optimization.