A three-step avidin-biotin approach has been applied as a pretargeting system in radioimmunotherapy (RIT) as an alternative to conventional RIT with directly labelled monoclonal antibodies (MoAbs). Although dosimetric and toxicity studies following conventional RIT have been reported, these aspects have not previously been evaluated in a three-step RIT protocol. This report presents the results of pharmacokinetic and dosimetric studies performed in 24 patients with different tumours. Special consideration was given to the dose delivered to the red marrow and to the haematological toxicity. The possible additive dose to red marrow due to the release of unbound yttrium-90 was investigated. The protocol consisted in the injection of biotinylated MoAbs (first step) followed 1 day later by the combined administration of avidin and streptavidin (second step). After 24 h, biotin radiolabelled with 1.85-2.97 GBq/m2 of 90Y was injected (third step). Two different chelating agents, DTPA and DOTA, coupled to biotin, were used in these studies. Indium-111 biotin was used as a tracer of 90Y to follow the biodistribution during therapy. Serial blood samples and complete urine collection were obtained over 3 days. Whole-body and single-photon emission tomography images were acquired at 1, 16, 24 and 40 h after injection. The sequence of images was used to extrapolate 90Y-biotin time-activity curves. Numerical fitting and compartmental modelling were used to calculate the residence time values (tau) for critical organs and tumour, and results were compared; the absorbed doses were estimated using the MIRDOSE3.1 software. The residence times obtained by the numerical and compartmental models showed no relevant differences (<10%); the compartmental model seemed to be more appropriate, giving a more accurate representation of the exchange between organs. The mean value for the tau in blood was 2.0+/-1.1 h; the mean urinary excretion in the first 24 h was 82.5%+/-10.8%. Without considering any contribution of free 90Y, kidneys, liver, bladder and red marrow mean absorbed doses were 1.62+/-1.14, 0.27+/-0.23, 3.61+/-0.70 and 0. 11+/-0.05 mGy/MBq, respectively; the effective dose was 0.32+/-0.06 mSv/MBq, while the dose to the tumour ranged from 0.62 to 15.05 mGy/MBq. The amount of free 90Y released after the injection proved to be negligible in the case of 90Y-DOTA-biotin, but noteworthy in the case of 90Y-DTPA-biotin (mean value: 5.6%+/-2.5% of injected dose), giving an additive dose to red marrow of 0.18+/-0.08 mGy per MBq of injected 90Y-DTPA-biotin. Small fractions of free 90Y originating from incomplete radiolabelling can contribute significantly to the red marrow dose (3.26 mGy per MBq of free 90Y) and may explain some of the high levels of haematological toxicity observed. These results indicate that pretargeted three-step RIT allows the administraton of high 90Y activities capable of delivering a high dose to the tumour and sparing red marrow and other normal organs. Although 90Y-biotin clears rapidly from circulation, the use of DOTA-biotin conjugate for a stable chelation of 90Y is strongly recommended, considering that small amounts of free 90Y contribute significantly in increasing the red marrow dose.