OBJECTIVE:To compare the effectiveness of a new method of closed-loop (automatic) control of BP with usual manual control. DESIGN:Within-subject crossover trial. SETTING:General and open heart ICUs in a university teaching hospital. PATIENTS:Convenience sample of 74 seriously ill patients aged 17 to 88 yr. Twenty had hypertension: postoperative (n = 7), neurologic damage (n = 6), miscellaneous (n = 7), and were prescribed nitroprusside (n = 11) or nitroglycerin (n = 9). Fifty-four had hypotension: septic (n = 33), other (n = 21), and were prescribed norepinephrine (n = 30), dopamine (n = 11), epinephrine (n = 8), or dobutamine (n = 5). An additional 21 trials were omitted from analysis because of technical (n = 7), clinical (n = 1), or drug-related problems (n = 13). INTERVENTION:Closed-loop and manual drug administrations 1-hr were studied during periods in each patient. The target and observed mean arterial pressure (MAP) and drug infusion rate were recorded electronically every 30 sec. MAIN OUTCOME MEASUREMENTS:Time taken to achieve initial control (min); fidelity of control (the integral of size and duration of error from target MAP +/- 10% in mm Hg.hr/hr); and average drug dose administered (microgram/min as % maximum possible per drug). RESULTS:Compared with manual control, closed-loop achieved faster initial control (log-rank chi 1 = 5.04, p less than .05) and greater fidelity (mean 1.37 vs. 2.36; F = 7.15, p less than .01). There was no difference in average drug dose administered. There was uniformity in the efficacy advantage of closed-loop drug administration across drugs and patient classifications. CONCLUSION:The new closed-loop system is more effective than the usual manual control in managing acute BP disturbances in the seriously ill patient.