Despite improvements in surgical treatment and intensive care, mortali
ty from severe acute pancreatitis remains high. We have carried out a
randomised study of 60 consecutive patients with alcohol-induced necro
tising pancreatitis to find out whether early antibiotic treatment can
improve outcome. 30 patients were assigned cefuroxime (4.5 g/day intr
avenously) from admission. In the second group, no antibiotic treatmen
t was given until clinical or microbiologically verified infection or
after a secondary rise in C-reactive protein. The inclusion criteria w
ere C-reactive protein concentration above 120 mg/L within 48 h of adm
ission and low enhancement (<30 Hounsfield units) on contrast-enhanced
computed tomography. There were more infectious complications in the
non-antibiotic than in the antibiotic group (mean per patient 1.8 vs 1
.0, p=0.01). The most common cause of sepsis was Staphylococcus epider
midis; positive cultures were obtained from pancreatic necrosis or the
central venous line in 14 of 18 patients with suspected but blood-cul
ture-negative sepsis, Mortality was higher in the non-antibiotic group
(seven vs one in the antibiotic group; p=0.03). Four of the eight pat
ients who died had cultures from pancreatic necrosis positive for Stap
h epidermidis. We conclude that cefuroxime given early in necrotising
pancreatitis is beneficial and may reduce mortality, probably by decre
asing the frequency of sepsis.