The purpose of this study was to determine if a computer-directed weaning system could wean patients with complex medical problems from mechanical ventilation and appropriately respond to signs of respiratory failure more effectively than traditional physician-directed weaning methods. In a prospective, randomized, controlled study, we tested the system in 15 patients who had required prolonged mechanical ventilation and met predetermined weaning tests. Patients were randomly assigned to one of two groups: (1) automatic, computer-directed weaning or (2) physician-controlled weaning. The computer-directed system decreased or increased the synchronized intermittent mandatory ventilation rate and pressure support (PS) based on predetermined limits of patient respiratory rate (RR) and tidal volume (TV). Pulse oximeter oxygen saturation was monitored, and alarms were triggered by an oxygen saturation of <90 percent. In the physician control group, weaning progressed with SIMV rate and PS reduction, as judged appropriate by the physician. Nine patients were assigned to the computer group; six patients were assigned to the control group. The average patient age was 64.8±14.9 years for the computer group, 65.2±22.7 years for the control group. The average time on mechanical ventilation prior to weaning was 13.4±7.8 days for the computer group and 14.5±11.1 days for the control group. Seven of the nine computer group patients weaned within 48 h of the study, with an average time to wean of 18.7±5.9 h. All seven were breathing spontaneously 48 h after weaning. Two of the six control group patients weaned within 48 h of the study, with an average time of 25.6±5.6 h. Both patients who weaned were placed back on mechanical ventilation within 30 h. The number of arterial blood gas samples drawn during the study was 1.4±0.7 for the computer group, 7.2±4.3 for the control group. The number of minutes per hour outside acceptable limits of RR>30, RR<8; or TV<5 ml/kg was 3.2±2.8 min for the computer group and 6.6±4.1 min for the control group. The study suggests that use of the computer-directed weaning system results in fewer arterial blood gas samples, shorter weaning times, and less time spent outside acceptable RR and TV parameters.