OBJECTIVES--(a) To test the hypothesis that a fixed 3 day course of the combination of a thiazide and loop diuretic is as effective as more prolonged treatment in the management of severe resistant cardiac failure. (b) To compare two thiazide diuretics (bendrofluazide and metolazone) in combination with loop diuretics in the treatment of severe resistant cardiac failure. DESIGN--Randomised study with a 2 x 2 factorial design. SETTING--Provincial teaching hospital. PATIENTS--33 consecutive patients (40 episodes) admitted with severe congestive cardiac failure (New York Heart Association class III or IV) unresponsive to intravenous loop diuretics for 48 hours. MAIN OUTCOME MEASURES--Change in daily weight and serum electrolytes and clinical improvement in heart failure. RESULTS--Diuresis was established during 37 of 40 episodes; of the rest two patients died in hospital. On 36 occasions improvement was sufficient to allow discharge from hospital. Median (range) maximal weight loss was -5.05 (-11.3 to 1.6) kg after the addition of bendrofluazide and -5.6 (-12.2 to 4.8) kg after the addition of metolazone (NS). Area under the body weight loss against time curves showed no significant difference between the two thiazide diuretics. Median (range) maximal weight loss after three days of treatment was -5.4 (-12.2 to 4.8) kg and -5.5 (-10.3 to 1) kg after a more prolonged course of median (range) 5.6 (1 to 13) days (NS). Area under the body weight loss time curves showed no significant difference between the two durations of treatment. Bendrofluazide was associated with fewer electrolyte disturbances. CONCLUSIONS--Bendrofluazide and metolazone were equally effective in establishing a diuresis in patients with severe congestive cardiac failure resistant to loop diuretics. A fixed three day course of the combination was as effective as a longer course.