All cirrhotic patients admitted because of gastrointestinal haemorrhage were considered for entry into the trial. 45 were excluded for the following reasons: (1) they had been treated with antibiotics during the week before admission, (2) they had a history of allergy to penicillins or quinolones, or (3) they showed signs of infection on arrival.
A total of 147 were allocated into 3 groups:
Group 1 consisted of 66 patients who were Child-Pugh's class A or B with no rebleeding (i.e. low risk of infection). Patients with a Child-Pugh's class A or B who rebled were added to those admitted with a Child-Pugh's class C and randomly allocated to groups 2 or 3.
Group 2 was the control group and consisted of 41 patients.
Group 3 consisted of 40 patients.
A further 29 patients (20%) were excluded from the analysis because of having infection on admission, undergoing surgery within 24 hours after admission and because of death. For the comparison between groups 2 and 3, sample size estimation was based on an estimated incidence of infections in the control and in the antibiotic group. Group sizes of 30 were required to give a one-tailed test power of 95% and an alpha error of 5%. Informed consent was obtained for participation and no refusals were reported.