To understand which factors could affect the assessment of anti-vascular treatment by DCE-MRI, we investigated possible causes that could have hampered the selection of an optimal biological dose in humans of the vascular targeted agent NGR-hTNF by DCE-MRI: (1) insufficient reproducibility of DCE-MRI; (2) less specific targeting of NGR-hTNF; (3) interference of vessel characteristics with NGR-hTNF efficacy; (4) interfering pharmacodynamic effects.In a phase I study NGR-hTNF, DCE-MRI was performed at baseline and 2h after NGR-hTNF administration in 31 patients with advanced solid cancer. Reproducibility measurements were performed in 5 other non-treated patients with metastatic disease. Mean k(ep), K(trans) values and their histogram distribution were determined in metastases and healthy liver tissue. The correlation between tumour size and DCE-MRI parameters was determined. Kinetics of soluble TNF receptors and the development of anti-TNF antibodies were assessed.Reproducibility of the DCE-MRI technique was adequate. Mean DCE-MRI parameters did not significantly change after NGR-hTNF administration, but histogram analyses showed significant changes in metastases and healthy liver tissue in some patients. The anti-vascular effects of NGR-hTNF were larger in smaller tumours, which have less mature neovasculature. Soluble TNF receptors were released.The difficulty to find an optimal biological dose of NGR-TNF by DCE-MRI is likely caused by a combination of factors: (i) different profiles of early anti-vascular effects in tumours and healthy liver tissue, (ii) dependence of the magnitude of the anti-vascular effect of NGR-hTNF on tumour size and (iii) shedding kinetics of soluble TNFa receptors.