Antidepressant effects of mirtazapine and imipramine were compared in a randomized, double blind, fixed blood-levels study with in-patients in a single centre. Patients with a DSM-III-R diagnosis of major depression and a Hamilton (17-item) score of > or = 18 were selected. After a drug-free and a placebo-washout period of 7 days in total, 107 patients still fulfilling the HRSD criterion of > or = 18, started on active treatment. The dose was adjusted to a predefined fixed blood level to avoid suboptimal dosing of imipramine. Concomitant psychotropic medication was administered only in a few cases because of intolerable anxiety or intolerable psychotic symptoms. Eight patients dropped out and two were excluded from analyses because of non-compliance; 97 completed the study. According to the main response criterion (50% or more reduction on the HRSD score) 11/51 (21.6%) patients responded on mirtazapine and 23/46 (50%) on imipramine after 4 weeks' treatment on the predefined blood level. Such a dramatic difference in efficacy between antidepressants has not often been reported before. The selection of (severely ill) in-patients, including those with suicidal or psychotic features, may have significance in this respect. Optimization of treatment with the reference drug imipramine through blood level control, exclusion of non-compliance for both drugs, exclusion of most concomitant medication and a low drop-out rate may also have contributed. It is concluded that imipramine is superior to mirtazapine in the patient population studied.