A prospective naturalistic multicentre study for deep sedation was conducted in intensive care with continuous electrocardiogram (ECG) monitoring. Clinical purpose was enough sedation, which made uncooperative and disrupted patients receive brain computed tomography (CT), magnetic resonance imaging (MRI), or fluid therapy, with minimum drug doses. A first infusion was either haloperidol (HAL group) or flunitrazepam (FNP group). If enough sedation was not achieved, a second infusion, which was the opposite drug to the first infusion, was given. The proportion requiring a second infusion was higher in the HAL group than in the FNP group (82% vs. 36%, P<0.0001). The mean reduction of the Excited Component for Positive and Negative syndrome scale at 15 min was greater for the FNP first group (FNP+HAL group) than the HAL first group (HAL+FNP group) (68% [S.D. 17] vs. 54% [S.D. 31], P=0.02). The mean dose of flunitrazepam in the HAL+FNP group was significantly lower than that in the FNP+HAL-group (1.3 mg vs. 3.5 mg, P=0.0003). Thus, in terms of monotherapy and speed of action, flunitrazepam has advantages over haloperidol as a first infusion for deep sedation. Regarding drug dosages, haloperidol has an advantage over flunitrazepam as a first infusion in safety.