Objective: To analyse and compare the infusion of frusemide by different modes of administration. Design: Prospective clinical survey of patients suffering from acute respiratory failure due to cardiogenic pulmonary oedema. Setting: A seven-bed general intensive care unit. Patients: Group A comprised 10 patients treated with intravenous frusemide several times daily at a dose of 1 mg/kg. Group B comprised 10 patients treated with continuous infusion of frusemide for 24 hours at an initial dose of 1 mg/kg, and 0.1 mg/kg/hr thereafter, with an option of increasing the dose every 2 hours. Group C comprised 10 patients treated with albumin and frusemide combined in continuous infusion over 24 hours at a dose of 250 mg frusemide diluted in 12.5 g albumin at a rate of 0.1 mg frusemide/kg/hr. Measurements: i) Serum electrolytes, creatinine, blood urea nitrogen (BUN) and protein; ii) urinary electrolytes; iii) hourly urinary output during the first 24 hours after admission. Results: There were no differences in the total dose of frusemide between the three groups. Urinary output of Group B was significantly greater than Groups A or C. There was less variability in hourly urinary output compared with the other groups. There was no significant difference in urinary electrolyte excretion among the three groups. Conclusion: The continuous infusion of frusemide in patients with cardiogenic pulmonary oedema results in less variability of hourly urinary output, and is much easier to administer compared with the bolus dosing method. Steady urinary output is particularly advantageous in haemodynamically labile patients, and may thus be the preferred method of frusemide administration.