Norfloxacin vs Ceftriaxone in the Prophylaxis of Infections in Patients With
Advanced Cirrhosis and Hemorrhage
JAVIER FERNÁNDEZ,* LUIS RUIZ DEL ARBOL,‡CRISTINA GÓMEZ,§ROSA DURANDEZ,?REGINA SERRADILLA,‡
CARLOS GUARNER,§RAMÓN PLANAS,?VICENTE ARROYO,* and MIGUEL NAVASA*
*IMDM and IDIBAPS, Hospital Clínic, University of Barcelona, Barcelona;‡Servicio de Digestivo, Hospital Ramón y Cajal, Madrid;§Servicio de Digestivo, Hospital
Santa Creu i Sant Pau, Barcelona; and?Servicio de Digestivo, Hospital Germans Trias i Pujol, Barcelona, Spain
See CME Quiz on page 1285.
Background & Aims: Oral norfloxacin is the standard of
therapy in the prophylaxis of bacterial infections in cirrhotic
patients with gastrointestinal hemorrhage. However, during
the last years, the epidemiology of bacterial infections in
cirrhosis has changed, with a higher incidence of infections
caused by quinolone-resistant bacteria. This randomized
controlled trial was aimed to compare oral norfloxacin vs
intravenous ceftriaxone in the prophylaxis of bacterial infec-
tion in cirrhotic patients with gastrointestinal bleeding.
Methods: One hundred eleven patients with advanced cir-
rhosis (at least 2 of the following: ascites, severe malnutri-
tion, encephalopathy, or bilirubin ?3 mg/dL) and gastroin-
testinal hemorrhage were randomly treated with oral
norfloxacin (400 mg twice daily; n ? 57) or intravenous
ceftriaxone (1 g/day; n ? 54) for 7 days. The end point of the
trial was the prevention of bacterial infections within 10 days
after inclusion. Results: Clinical data were comparable
between groups. The probability of developing proved or
possible infections, proved infections, and spontaneous bac-
teremia or spontaneous bacterial peritonitis was significantly
higher in patients receiving norfloxacin (33% vs 11%, P ?
.003; 26% vs 11%, P ? .03; and 12% vs 2%, P ? .03, respec-
tively). The type of antibiotic used (norfloxacin), transfusion
requirements at inclusion, and failure to control bleeding
were independent predictors of infection. Seven gram-nega-
tive bacilli were isolated in the norfloxacin group, and 6 were
quinolone resistant. Nonenterococcal streptococci were only
isolated in the norfloxacin group. No difference in hospital
mortality was observed between groups. Conclusions:
Intravenous ceftriaxone is more effective than oral norfloxa-
cin in the prophylaxis of bacterial infections in patients with
advanced cirrhosis and hemorrhage.
frequent. Between 25% and 65% of these patients present infec-
tions at admission or develop them during hospitalization.1–9
The incidence of infections is particularly high in patients with
advanced liver failure and/or severe hemorrhage.7,10Second, in
patients with ascites, infections may induce acute impairment
drome.11–14Finally, bacterial infection in patients with variceal
bleeding is associated with an increased rate of failure to con-
trol bleeding,15,16rebleeding,9,17,18and hospital mortality.5,16,19
acterial infection is a major problem in patients with
cirrhosis and gastrointestinal hemorrhage. First, it is very
The recent demonstration that renal failure in cirrhotic patients
with spontaneous bacterial peritonitis is associated with a
marked increase in portal pressure offers a rational explanation
for these features.13
Since the pioneer study of Rimola et al demonstrating
that oral administration of nonabsorbable antibiotics mark-
edly reduces the incidence of bacterial infections in cirrhotic
patients with gastrointestinal hemorrhage,1antibiotic pro-
phylaxis is considered a standard of care in these pa-
tients.5,20,21Selective intestinal decontamination with oral
norfloxacin, a poorly absorbable quinolone with antibacterial
activity against gram-negative bacteria but not against gram-
positive cocci or anaerobic bacteria, is the most commonly
used approach for the prophylaxis of bacterial infections in
cirrhotic patients with gastrointestinal hemorrhage.2,20Re-
cent studies, however, have presented evidence suggesting
that oral quinolone administration may not be the best
regime for the prevention of bacterial infections in cirrhotic
patients with gastrointestinal hemorrhage. The prevalence of
quinolone-resistant bacteria in the fecal flora22,23and the
incidence of spontaneous bacterial peritonitis24and other
infections25,26caused by these organisms have increased sub-
stantially during the last years. However, a significant num-
ber of infections in cirrhotic patients with gastrointestinal
hemorrhage are caused by gram-positive bacteria related to
the invasive procedures used in these patients.24
These considerations led us to perform the current study,
which consisted of a randomized controlled trial aimed at
comparing oral norfloxacin vs intravenous ceftriaxone in the
prophylaxis of bacterial infections in cirrhotic patients with
gastrointestinal hemorrhage and severe liver failure. Intrave-
nous ceftriaxone was selected for 2 reasons. First, we have
recently shown that most quinolone-resistant bacteria isolated
in cirrhotic patients with spontaneous bacteremia, spontaneous
bacterial peritonitis, and other infections are susceptible to
third-generation cephalosporins.24Second, antibiotics admin-
istered by intravenous route are theoretically more appropriate
than those administered orally in the prophylaxis of infection
in patients with active upper gastrointestinal bleeding. The end
point of the study was to assess whether intravenous ceftriax-
one is more effective than oral norfloxacin in reducing the rate
of bacterial infections within the first 10 days after the hemor-
rhage because this is the period within which most infections
© 2006 by the American Gastroenterological Association (AGA) Institute
Materials and Methods
The study was performed in patients with cirrhosis
admitted to 4 Spanish hospitals for the treatment of an upper
gastrointestinal hemorrhage between February 2000 and April
2004. Diagnosis of cirrhosis was based on clinical, laboratory,
and ultrasonographic data or on histology. Inclusion criteria
were as follows: age 18–80 years, hematemesis and/or melena
within 24 hours prior to inclusion, and advanced cirrhosis as
defined by the presence of 2 or more of the following signs of
liver failure: severe malnutrition (as defined by the presence of
clear signs of muscle wasting), serum bilirubin ?3 mg/dL,
ascites (confirmed by paracentesis), and hepatic encephalopathy
(grade 1 or more). Diagnosis of ascites, severe malnutrition, and
encephalopathy was made clinically. Exclusion criteria were as
follows: allergy to cephalosporins or quinolones, presence of
any of the following signs of infection (fever ?37.5°C, white
blood cell count ?15,000 mm3, immature neutrophils ?500
mm3, polymorphonuclear cell count in ascitic fluid ?250/mm3,
more than 15 leukocytes/field in the fresh urine sediment, or
data compatible with pneumonia on the chest x-ray), treatment
with antibiotics within 2 weeks before the hemorrhage (exclud-
ing oral norfloxacin for prophylaxis of spontaneous bacterial
peritonitis), previously diagnosed advanced hepatocellular car-
cinoma (1 nodule greater than 5 cm, 3 nodules with 1 greater
than 3 cm, or more than 3 nodules), and human immunodefi-
ciency virus (HIV) infection.
The study was approved by the ethics committee of each
hospital participating in the study. Written informed consent
was obtained from the patients and, in those with encephalop-
athy, from their families. The protocol conformed to the Hel-
sinki Declaration and Guidelines for Good Clinical Practice in
Treatment of the Hemorrhage
Following admission, a history and physical examina-
tion were obtained, and 2 short intravenous cannulas and a
nasogastric tube were placed. A central line was also placed in
most patients. Urinary catheter was inserted only if indicated.
Laboratory measurements including standard liver and renal
function tests, blood and ascitic fluid cell count and cultures
(samples were inoculated into aerobic and anaerobic blood
culture bottles at patient’s bedside), and fresh urine sediment
and culture were then performed as well as a chest x-ray.
Emergency endoscopy and endoscopic treatment (if indicated,
sclerotherapy or banding) were done within the first 24 hours
after onset of the hemorrhage in all cases. Patients with severe
hepatic encephalopathy (grade 3 or 4) were intubated prior to
endoscopy. Patients with bleeding from esophageal or gastric
varices or from portal hypertensive gastropathy received soma-
tostatin or terlipressin. In cases with uncontrolled variceal
bleeding, ballon tamponade was applied after prophylactic oro-
tracheal intubation. Endoscopic treatment was repeated during
the hospital admission until eradication of varices if indicated.
A transjugular intrahepatic portosystemic shunt (TIPS) or sur-
gical portacaval shunt were performed in patients in whom
other treatments failed to control the bleeding. Patients with
peptic ulcer or esophagitis received proton-pump inhibitors.
Blood transfusions were given to maintain hematocrit levels
between 25% and 30%.
Definitions regarding the course of the hemorrhage (failure
to control the bleeding and early rebleeding) were based on
those proposed by the Baveno III Consensus Workshop.27Fail-
ure to control the bleeding within the first 6 hours after inclu-
sion was considered when transfusion requirements were equal
to or greater than 4 units of blood together with an inability to
achieve an increase in systolic arterial blood pressure by 20 mm
Hg or to 70 mm Hg or higher. Failure to control the bleeding
within the period from 6 to 24 hours after inclusion was
considered if there was a new hematemesis together with a
decrease in systolic arterial blood pressure equal to or greater
than 20 mm Hg and/or transfusions requirements equal to or
greater than 2 units to increase hemoglobin levels to 9 g/dL.
Whenever balloon tamponade had to be used within these 2
periods, failure to control the bleeding was also considered.
Early rebleeding and study rebleeding were defined as new
hematemesis or melena from 24 hours to 5 days after inclusion
and within the study period (10 days), respectively, with trans-
fusion requirements equal to or greater than 2 units of blood in
any of the 24-hour periods and at least 1 of the following:
systolic arterial blood pressure lower than 100 mm Hg, decrease
in arterial pressure after postural change greater than 20 mm
Hg, or heart rate greater than 100 beats/min.
Randomization and Infection Diagnosis
Patients who fulfilled the inclusion criteria were ran-
domly allocated into 2 groups. Patients in the first group
received oral norfloxacin 400 mg every 12 hours during 7 days.
Patients in the second group received intravenous ceftriaxone
1 g per day during 7 days. Antibiotics were initiated following
the emergency endoscopy and always within the first 12 hours
after admission into the hospital. Randomization was done
using consecutively numbered computer-generated envelopes
containing treatment assignment. Randomization was indepen-
dent at each hospital.
Patients were followed up closely, with special emphasis on
the detection of bacterial infections. Physical examination and
blood cell count were performed daily, urinary sediment every
48 hours, and chest x-ray every 3 days for 10 days. Blood, ascitic
fluid, and urine cultures were taken whenever a patient devel-
oped signs of infection.
Diagnosis of proved infection was performed as follows: (1)
spontaneous bacteremia: positive blood cultures in the absence
of any potential source of infection; (2) spontaneous bacterial
peritonitis: ascitic fluid polymorphonuclear count equal to or
greater than 250/mm3(Rimola et al20); (3) urinary tract infec-
tion: urinary leukocyte count greater than 15 cells per high-
power field and positive urine culture28; (4) other infections
were diagnosed according to clinical, radiologic, and bacterio-
logic data. Possible infection was diagnosed in patients with
fever (?37.5°C during more than 6 hours), leukocytosis
(?15,000 mm3) or increased concentration of immature neu-
trophils in blood (?500/mm3), negative cultures, and no other
signs of infection. Analysis was performed classifying the pa-
tients into 3 groups: (1) patients not developing infections; (2)
patients developing proved infections; and (3) patients devel-
oping proved or possible infections.
The sample size was calculated on the basis of an
expected incidence of bacterial infections (proved plus possible)
1050 FERNÁNDEZ ET AL GASTROENTEROLOGY Vol. 131, No. 4
of 30% in the norfloxacin group and of 10% in the ceftriaxone
group. Sixty-one patients had to be included in each group to
obtain a P value ? .05 with an ? error of 5% and a ? error of
Continuous variables were compared by the Student t test.
Discontinuous variables were compared by the ?2test with the
Yates correction when indicated. Probability curves were ob-
tained by the Kaplan–Meier method and compared by the
log-rank test. Univariate analysis using the Student t test and
the Kaplan–Meier curves was performed to identify possible
predictors of infection. Variables showing significant differ-
ences were subsequently introduced in a Cox proportional haz-
ards regression model to identify independent predictors of
infection. Results are given as relative hazard plus 95% confi-
dent interval (CI). The median values of the independent pre-
dictors of infection were considered as cut-off levels to identify
groups of patients with different risks of infection. Calculations
were performed with the SPSS Statistical Package (SPSS Inc.
Version 11.0, 2000, Chicago, IL). Differences were considered
significant at the level .05. Results are expressed as mean ? SD.
A total of 1369 cirrhotic patients with gastrointestinal
hemorrhage were screened. Of these, 1245 patients were not
included because of absence of 2 signs of advanced liver failure
(884 patients), presence of infection at admission (253 pa-
tients), presence of advanced hepatocellular carcinoma or other
neoplasia (106 patients), antibiotic treatment other than oral
norfloxacin at admission (54 patients), age over 80 years (21
patients), HIV infection (19 patients), and other causes includ-
ing refusal to participate in the trial (32 patients). Of the 124
patients randomized, 63 in the norfloxacin group and 61 in the
ceftriaxone group, 7 patients (4 in the norfloxacin group and 3
in the cefriaxone group) were excluded because of violation of
the protocol (presence of only 1 sign of advanced liver failure)
and 6 (2 in the norfloxacin group and 4 in the ceftriaxone
group) because of occult infection (positive blood cultures
obtained prior to randomization). Thus, 111 patients (57 in the
norfloxacin group and 54 in the ceftriaxone group) were con-
sidered in the final analysis of the results (Figure 1).
Clinical Characteristics of Patients
The median age of the patients was 58 ? 12 years, 77%
were male, and the most frequent etiology was alcoholism
(68%). Patients had advanced liver insufficiency with high se-
rum bilirubin (4.4 ? 4.0 mg/dL), low serum albumin (26 ? 5
g/L), high INR (1.56 ? 0.39), and high Child–Pugh score (9.8 ?
1.5 points) and Model for End Stage Liver Disease (MELD)
score (17.1 ? 4.8 points). Forty-seven percent of the patients
were grade B and 53% grade C of the Child–Pugh classification.
Seventy-nine percent of patients had ascites, 72% signs of severe
malnutrition, 58% a serum bilirubin ?3 mg/dL, and 37% he-
patic encephalopathy. Forty-one percent of the patients had
more than 2 signs of liver failure, and 12% had renal failure at
inclusion. Hepatocellular carcinoma was present in 18% of
patients and diabetes mellitus in 20%. Only 9% of the patients
were receiving oral norfloxacin for prophylaxis of spontaneous
bacterial peritonitis at inclusion.
Table 1 shows that with the exception of the frequency of
alcoholic cirrhosis, higher in the norfloxacin group, and the
incidence of renal failure at inclusion, higher in the ceftriaxone
group, there were no significant differences between groups in
clinical data and laboratory measurements.
Figure 1. Flow diagram of patient allocation. More detailed infor-
mation is shown in the text.
Table 1. Baseline Clinical and Analytical Characteristics
(n ? 54)
(n ? 57)
Alcoholic cirrhosis (%)
Active alcoholism (%)b
Serum bilirubin (mg/dL)
Serum albumin (g/L)
Prothrombin time international
Child-Pugh score (points)
Child-Pugh score (% B/C)
MELD score (points)
Serum bilirubin ? 3 mg/dL (%)
Hepatic encephalopathy (%)
Severe malnutrition (%)
Signs of liver failure (2/3/4) (%)d
Serum creatinine (mg/dL)
Serum sodium (mEq/L)
Ascitic fluid protein (g/L)
Renal failure (%)c
Hepatocellular carcinoma (%)
Diabetes mellitus (%)
Norfloxacin prophylaxis (%)
58 ? 12
3.8 ? 3.5
26 ? 5
1.56 ? 0.41
57 ? 12
4.9 ? 4.4
26 ? 5
1.55 ? 0.37
9.7 ? 1.6
17.1 ? 4.9
1.2 ? 0.5
134 ? 5
11 ? 7
9.8 ? 1.5
17.1 ? 4.7
1.0 ? 0.4
133 ? 16
11 ? 6
NOTE. Values represent mean ? standard deviation.
aP ? .05 norfloxacin vs ceftriaxone group.
bArbitrarily defined as a daily alcohol intake over 20 g in patients with
cSerum creatinine ?1.5 mg/dL.
dTwo, 3, or 4 signs of liver failure (severe malnutrition, serum bilirubin
?3 mg/dL, ascites, and hepatic encephalopathy).
October 2006 CEFTRIAXONE IN BLEEDING CIRRHOTIC PATIENTS 1051
Characteristics of the Hemorrhage
In the whole series of patients, the time elapsed
between the initiation of bleeding and the emergency endos-
copy was 4.6 ? 5.1 hours. The site of bleeding was esopha-
geal varices in 64% of patients, gastric varices in 5%, peptic
ulcer in 10%, portal hypertensive gastropathy in 8%, and
other in 13%. There was active bleeding at the time of
endoscopy in 27% of cases. Vasoactive drugs (somatostatin or
terlipressin) were the treatment most commonly used. It was
applied to patients bleeding from esophageal varices, gastric
varices, or portal hypertensive gastropathy (77% of the cases).
Emergency sclerotherapy or banding was used in 61% of
cases. Only 6% of patients received balloon tamponade. The
median amount of blood transfused was 1.6 ? 1.8 units.
Hemorrhagic shock, as defined by a systolic blood pressure
?90 mm Hg and a heart rate ?100 beats/min, was present at
admission in only 7% of the patients. Failure to control
bleeding, early rebleeding, and study rebleeding occurred in
13%, 9%, and 11% of the patients, respectively. Seven percent
of patients required TIPS and 5% a surgical shunt. The time
elapsed between the initiation of bleeding and the first dose
of antibiotics was 6.8 ? 2.9 hours.
Table 2 shows that there were no significant differences
between the study groups in the characteristics of the hemor-
rhage at inclusion, etiology of the hemorrhage, time elapsed
between the initiation of the bleeding and the emergency en-
doscopy, or prophylactic antibiotic administration and course
of the hemorrhage (failure to control the bleeding, early and
study rebleeding rates, and need for TIPS or surgical shunt).
Efficacy of Oral Norfloxacin and IV
Ceftriaxone in the Prevention of Bacterial
Nineteen patients (33%) in the norfloxacin group and 6
(11%) in the ceftriaxone group developed proved or possible
infections within the first 10 days after the initiation of the
hemorrhage (P ? .01) (Table 3). The corresponding figures for
proved infections were 15 cases (26%) and 6 cases (11%), respec-
tively (P ? .07). Seven patients (12%) in the norfloxacin group
and only 1 (2%) in the ceftriaxone group developed spontane-
ous bacteremia or spontaneous bacterial peritonitis (P ? .06).
Urinary tract infection developed in 8 patients in the norfloxa-
cin group and in 3 in the ceftriaxone group. Finally, 1 patient in
the norfloxacin group and 2 in the ceftriaxone group developed
pneumonia. The isolated organisms were gram-negative bacilli
(mainly Escherichia coli) in 8 patients and gram-positive cocci in
9 patients. Gram-negative bacilli were cultured in 7 patients
receiving norfloxacin and in only 1 receiving ceftriaxone (P ?
.03). Six out of the 7 gram-negative bacilli isolated in the
norfloxacin group were resistant to quinolones, whereas that
isolated in the ceftriaxone group was susceptible to quinolones.
Non-enterococcal streptococci were only isolated in the nor-
floxacin group. Figures 2 and 3 show the probability curves of
infection (proved and possible), proved infection, and sponta-
neous bacterial peritonitis or spontaneous bacteremia develop-
ment in the 2 groups. Probabilities were significantly higher in
the norfloxacin group. Median time of infection was 3 days
(range, 1–9 days) in the norfloxacin group and 6 days (range,
2–7 days) in the ceftriaxone group (P ? ns). No adverse effects
related to norfloxacin or ceftriaxone administration were ob-
served during the study period.
Predictive Factors of Bacterial Infection
Proved and possible infections. Mean arterial
blood pressure (71 ? 19 mm Hg in patients with infection vs 81
? 15 mm Hg in patients without infection, P ? .007), blood
transfusion (2.4 ? 2.3 vs 1.4 ? 1.5 units, P ? .01) and synthetic
plasma expander and/or plasma requirements at inclusion (2.0
? 2.9 vs 0.7 ? 1.3 units, P ? 05), failure to control bleeding
(28% vs 9%, P ? .001), and norfloxacin prophylaxis (76% vs 42%,
P ? .003) were found to be predictors of proved or possible
bacterial infection development in the univariate analysis. Of
these, the Cox regression analysis identified norfloxacin prophy-
laxis (hazard ratio [HR]: 3.71, 95% CI: 1.47–9.34; P ? .005),
blood transfusion requirements at inclusion (HR: 1.36, 95% CI:
1.14–1.61; P ? .0001), and failure to control bleeding (HR: 3.18;
95% CI: 1.32–7.68; P ? .01) as independent predictors for
proved or possible infection.
Proved infections. Mean arterial blood pressure (66
? 16 mm Hg in patients with infections vs 81 ? 15 mm Hg in
Table 2. Characteristics of Hemorrhage at Inclusion and
Course of the Bleeding
(n ? 54)
(n ? 57)
Characteristics of hemorrhage
Mean arterial pressure (mm Hg)
Heart rate (beats/min)
Hypovolemic shock (%)a
Blood units transfused at
Time to endoscopy (h)b
Time to antibiotic
Source of bleeding (%)
Mallory Weiss tear
Active bleeding at endoscopy (%)
Vasoactive therapy (%)
Sclerotherapy or banding (%)
Patients submitted to (%)
Urinary catheter insertion
Central line insertion
Course of bleeding
Failure to control bleeding (%)
Early rebleeding (1–5 days) (%)
Study rebleeding (10 days) (%)
Balloon tamponade (%)
TIPS insertion (%)
Surgical shunt (%)
76 ? 14
89 ? 20
28 ? 6
81 ? 18
93 ? 18
28 ? 7
1.6 ? 1.6
4.8 ? 5.1
1.6 ? 2.0
4.5 ? 5.3
7.1 ? 2.9 6.5 ? 2.9
NOTE. Values represent mean ? standard deviation.
aSystolic pressure ? 90 mm Hg and heart rate ? 100 b/min.
bTime from bleeding to gastroscopy.
cTime from bleeding to antibiotic prophylaxis. No differences were
observed between groups.
1052 FERNÁNDEZ ET ALGASTROENTEROLOGY Vol. 131, No. 4
patients without infections, P ? .0001), hematocrit (25% ? 6%
vs 29% ? 7%, P ? .03), blood transfusion (2.7 ? 2.4 vs 1.4 ? 1.5
units, P ? .003) and synthetic plasma expander and/or plasma
requirements at inclusion (2.4 ? 3.1 vs 0.7 ? 1.3 units, P ? .02),
failure to control the bleeding (29% vs 10%, P ? .003), and
norfloxacin prophylaxis (71% vs 47%, P ? .03) were found to be
predictors of bacterial infection development in the univariate
analysis. Of these, the Cox regression analysis identified nor-
floxacin prophylaxis (HR: 3.21, 95% CI: 1.24–8.32; P ? .02),
blood transfusion requirements (HR: 1.22; 95% CI: 1.01–1.47; P
? .04), and mean arterial pressure at inclusion (HR: 0.96; 95%
CI: 0.93–0.99; P ? .005) as independent predictors for proved
Efficacy of Oral Norfloxacin and IV
Ceftriaxone in the Prevention of Bacterial
Infections in Patients With High Risk of
Thirty-five patients (20 in the norfloxacin group and
15 in the ceftriaxone group) had at least 1 independent
predictor of proved or possible infection development and
were considered to have a high risk of infection. Twenty-nine
patients (16 in the norfloxacin group and 13 in the ceftriax-
one group) had transfusion requirement ?2 units and 15 (9
and 6, respectively) had failure to control the bleeding within
the first 24 hours.
Ten out of 20 high-risk patients in the norfloxacin group (50%)
developed proved or possible bacterial infections within the first
10 days following inclusion. In contrast, this only occurred in 2
out of 15 (13%) patients in the ceftriaxone group (P ? .02).
Probability curves of proved plus possible and proved bacterial
infections in these 2 groups are shown in Figure 4. Significant
differences were observed between groups.
Figure 2. Probability of remaining free of proved and possible in-
fections (A) and proved infections (B) in patients receiving ceftriaxone
(continuous line) and norfloxacin (dotted line). There were significant
differences between groups.
Figure 3. Probability of remaining free of spontaneous bacterial
peritonitis or bacteremia in patients receiving ceftriaxone (continuous
line) and norfloxacin (dotted line). There were significant differences
Table 3. Proved and Possible Bacterial Infections and
Organisms Isolated in the Study
(n ? 54)
(n ? 57)
Infections, n (%)
Patients with proved or
Patients with proved
Type of infection:
Non fermentative gram-
6 (11) 19 (33).01
6 (11) 15 (26).07
October 2006CEFTRIAXONE IN BLEEDING CIRRHOTIC PATIENTS 1053
There was no significant difference in mortality within
the 10 days after inclusion between patients treated with nor-
floxacin and those treated with ceftriaxone (5 and 6 patients,
respectively, died). Mortality during hospitalization was also
similar in the 2 groups (6 and 8 patients, respectively, died
during hospital admission). Causes of death were hepatic fail-
ure in 7 patients (4 in the ceftriaxone group and 3 in the
norfloxacin group), uncontrolled bleeding in 4 patients (3 and
1 patients, respectively), septic shock in 2 patients (1 in each
group), and hepatorenal syndrome in 1 patient from the nor-
The results of the current study confirm that patients
with cirrhosis, gastrointestinal hemorrhage, and advanced liver
failure are at great risk of developing bacterial infections. De-
spite the prophylactic administration of antibiotics, 25 of the
111 patients (23%) included in the study developed bacterial
infections within 10 days after inclusion. In 8 patients, the
infection was severe (spontaneous bacteremia or spontaneous
An important finding of the current study is that the efficacy
of oral norfloxacin in the prophylaxis of bacterial infection in
patients with gastrointestinal hemorrhage and severe liver fail-
ure is relatively poor. Nineteen of the 57 patients treated with
norfloxacin in this study (33%) developed bacterial infections
(either proved or possible), 15 (26%) proved infections and 7
(12%) spontaneous bacteremia or spontaneous bacterial perito-
nitis. Most organisms isolated in these patients were gram-
negative bacilli or non-enterococcal streptococci of probably
cutaneous or respiratory origin resistant to quinolones.
This low efficacy of norfloxacin is consistent with the
changes in the epidemiology of bacterial infections in cirrhosis
detected during the last few years. Traditionally, bacterial in-
fections in cirrhosis were mainly caused by gram-negative bacilli
susceptible to quinolones.29–36Only in patients receiving pro-
phylaxis with norfloxacin was the incidence of infections caused
by gram-positive bacteria higher than that caused by gram-
negative bacilli.37In contrast, at present, bacterial infections in
cirrhosis are caused by both gram-negative bacilli and gram-
positive cocci in a similar proportion. In patients receiving
long-term norfloxacin prophylaxis, the most commonly iso-
lated bacteria are gram-negative organisms resistant to quino-
lones and not gram-positive cocci. Finally, in patients not re-
ceiving norfloxacin, the incidence of infections caused by
quinolone-resistant gram-negative bacilli is also relatively
high.24The frequent use of invasive procedures in patients with
cirrhosis, which predispose to infections by gram-positive cocci,
and the high prevalence of quinolone-resistant gram-negative
bacilli in the fecal flora in the general population and in cir-
rhotic patients owing to the widespread use of these antibiotics
are the most likely mechanisms of this change in epidemiology.
Other features that could explain the low efficacy of oral
norfloxacin observed in the study may be related to some
specific characteristics of this treatment. Selective intestinal
decontamination with oral norfloxacin is probably not achieved
until several days after the initiation of treatment; thus, there is
an initial period in which patients may be less protected against
infections. Moreover, the oral route is probably not adequate to
treat patients with active gastrointestinal hemorrhage, vomit-
ing, submitted to periodic aspiration through a nasogastric
tube, and with an extremely rapid intestinal transit. The obser-
vation that bacterial infections in our patients treated by nor-
floxacin tended to occur earlier than those in patients treated
with IV ceftriaxone supports this contention.
The most important result of our trial is that IV ceftriaxone
was significantly more effective than oral norfloxacin in the
prophylaxis of bacterial infections in cirrhotic patients with
gastrointestinal hemorrhage and severe liver failure. Only 6 of
57 patients (11%) treated with IV ceftriaxone developed infec-
tions, only 1 developed spontaneous bacterial peritonitis, and
none developed spontaneous bacteremia. The higher efficacy of
ceftriaxone over norfloxacin was due to the fact that both
non-enterococcal streptococci and quinolone-resistant gram-
negative bacteria, the most common organisms causing infec-
tion in patients treated with norfloxacin, are highly susceptible
to third-generation cefalosporins. The incidence of infections
caused by enterococci, a bacteria resistant to both quinolones
and third-generation cephalosporins, was comparable in the 2
Figure 4. Probability of remaining free of proved and possible in-
fection (A) and proved infections (B) in patients at high risk for infection
ing) who received ceftriaxone (continuous line) or norfloxacin prophy-
laxis (dotted line). There were significant differences between groups.
1054FERNÁNDEZ ET AL GASTROENTEROLOGY Vol. 131, No. 4
In the current series of patients with severe liver failure, only
the type of the prophylactic antibiotic used and 2 parameters
related to the severity of the hemorrhage (transfusion require-
ments at inclusion and failure to control bleeding) were found
to be independent predictors of infection development. Using
these 2 predictive factors, a subset of patients with high risk of
bacterial infections was identified. Intravenous ceftriaxone was
also much more effective than oral norfloxacin in the prophy-
laxis of bacterial infections in these patients, further indicating
that it is an excellent antibiotic for the prevention of bacterial
infections in cirrhotic patients with advanced liver failure and
Allergy to ?-lactamic antibiotics is relatively common in the
general population. In these cases, two possible alternatives to
third-generation cephalosporins exist. The first is the intrave-
nous administration of quinolones or trimethoprim sulfameth-
oxazole, which are effective in the prophylaxis of spontaneous
bacterial peritonitis by oral route. However, they do not prevent
infections caused by quinolone-resistant bacteria, which are
also frequently resistant to trimethoprim sulfamethoxazole.24
The second alternative is aztreonam, an antibiotic active against
gram-negative bacilli, plus a glycopeptide (vancomycin or teico-
planin), active against gram-positive cocci. The nephrotoxic
potential of glycopeptides may be a problem of this combina-
tion. Further studies assessing possible alternatives to ?-lac-
tamic antibiotics in the prophylaxis of bacterial infections in
cirrhotic patients with gastrointestinal bleeding are clearly
We could not confirm recent studies suggesting that bacte-
rial infections in patients with cirrhosis are associated with a
higher rate of failure to control bleeding,15,16higher rate of
rebleeding,9,17,18and higher hospital mortality.5,16,19Differ-
ences in the designs of the studies probably account for this
discrepancy. Whereas, in these studies, both patients admitted
to hospital with infections and patients developing infections
during hospitalization were considered, in our trial, only pa-
tients without infection at inclusion were included into the
study. This is a fundamental difference because, in our series, all
patients developed infections when the endoscopic and/or the
pharmacologic treatment of the hemorrhage had already been
applied. There was also no difference in hospital survival be-
tween groups, but the study was not designed to assess differ-
ences in mortality.
In summary, the current study indicates that IV ceftriax-
one is more effective than oral norfloxacin in the prophylaxis
of bacterial infections in patients with cirrhosis, gastrointes-
tinal hemorrhage, and severe liver failure. This is related
to 2 features. First, patients treated with norfloxacin are
predisposed to develop infections because of gram-negative
bacilli and non-enteroccocal streptococci resistant to quino-
lones. By contrast, these bacteria are highly susceptible to
third-generation cephalosporins. Second, intravenous ad-
ministration of prophylactic antibiotics is better than oral
administration in patients with severe hepatic failure, en-
cephalopathy, and active gastrointestinal bleeding. Intrave-
nous ceftriaxone should, therefore, be used instead of oral
norfloxacin in the prophylaxis of bacterial infections in cir-
rhotic patients with advanced cirrhosis and upper gastroin-
testinal bleeding. Because the epidemiology of bacterial in-
fections differs greatly between geographic areas, further
studies in other countries are needed to support this con-
1. Rimola A, Bory F, Teres J, Perez-Ayuso RM, Arroyo V, Rodés J.
Oral, nonabsorbable antibiotics prevent infection in cirrhotics
with gastrointestinal hemorrhage. Hepatology 1985;5:463–467.
2. Soriano G, Guarner C, Tomas A, Villanueva C, Torras X, Gonzalez
D, Sainz S, Anguera A, Cusso X, Balanzo J, Vilardell F. Norfloxacin
prevents bacterial infection in cirrhotics with gastrointestinal
hemorrhage. Gastroenterology 1992;103:1267–1272.
3. Bleichner G, Boulanger R, Squara P, Sollet JP, Parent A. Fre-
quency of infections in cirrhotic patients presenting with acute
gastrointestinal haemorrhage. Br J Surg 1986;73:724–726.
4. Deschenes M, Villeneuve JP. Risk factors for the development of
bacterial infections in hospitalized patients with cirrhosis. Am J
5. Bernard B, Grange JD, Khac EN, Amiot X, Opolon P, Poynard T.
Antibiotic prophylaxis for the prevention of bacterial infections in
cirrhotic patients with gastrointestinal bleeding: a meta-analysis.
6. Hsieh WJ, Lin HC, Hwang SJ, Hou MC, Lee FY, Chang FY, Lee SD.
The effect of ciprofloxacin in the prevention of bacterial infection
in patients with cirrhosis after upper gastrointestinal bleeding.
Am J Gastroenterol 1998;93:962–966.
7. Pauwels A, Mostefa-Kara N, Debenes B, Degoutte E, Levy VG.
Systemic antibiotic prophylaxis after gastrointestinal hemorrhage
in cirrhotic patients with a high risk of infection. Hepatology
8. Blaise M, Pateron D, Trinchet JC, Levacher S, Beaugrand M,
Porriat JL. Systemic antibiotic therapy prevents bacterial infection
in cirrhotic patients with gastrointestinal hemorrhage. Hepatology
9. Hou MC, Lin HC, Liu TT, Kuo BI, Lee FY, Chang FY, Lee SD.
Antibiotic prophylaxis after endoscopic therapy prevents rebleed-
ing in acute variceal hemorrhage: a randomized trial. Hepatology
10. Soriano G, Guarner C, Tomas A, Novella MT, Domenech JM,
Navarro B, Balanzo J, Vilardell F. Predictive factors of bacterial
infections in cirrhotic patients submitted to prophylactic selective
intestinal decontamination (abstract). Gastroenterology 1993;
11. Follo A, Llovet JM, Navasa M, Planas R, Forns X, Francitorra A,
Rimola A, Gassull MA, Arroyo V, Rodés J. Renal impairment after
spontaneous bacterial peritonitis in cirrhosis: incidence, clinical
course, predictive factors and prognosis. Hepatology 1994;20:
12. Navasa M, Follo A, Filella X, Jimenez W, Francitorra A, Planas R,
Rimola A, Arroyo V, Rodés J. Tumor necrosis factor and interleu-
kin-6 in spontaneous bacterial peritonitis in cirrhosis: relation-
ship with the development of renal impairment and mortality.
13. Ruiz del Arbol L, Urman J, Fernandez J, Gonzalez M, Navasa M,
Monescillo A, Albillos A, Jimenéz W, Arroyo V. Cardiovascular,
renal and hepatic hemodynamic derangement in cirrhotic pa-
tients with spontaneous bacterial peritonitis. Hepatology 2003;
14. Sort P, Navasa M, Arroyo V, Aldeguer X, Planas R, Ruiz del Arbol
L, Castells L, Vargas V, Soriano G, Guevara M, Gines P, Rodes J.
Effect of intravenous albumin on renal impairment and mortality
in patients with cirrhosis and spontaneous bacterial peritonitis.
N Engl J Med 1999;341:403–409.
15. Goulis J, Armonis A, Patch D, Sabin C, Greenslade L, Burroughs
AK. Bacterial infection is independently associated with failure to
control bleeding in cirrhotic patients with gastrointestinal hemor-
rhage. Hepatology 1998;27:1207–1212.
October 2006 CEFTRIAXONE IN BLEEDING CIRRHOTIC PATIENTS1055
16. Vivas S, Rodríguez M, Palacio MA, Linares A, Alonso JL, Rodrigo
L. Presence of bacterial infection in bleeding cirrhotic patients is
independently associated with early mortality and failure to con-
trol bleeding. Dig Dis Sci 2001;46:2752–2757.
17. Bernard B, Cadranel JF, Valla D, Escolano S, Jarlier V, Opolon P.
Prognostic significance of bacterial infection in bleeding cirrhotic
patients: a prospective study. Gastroenterology 1995;108:
18. Pohl J, Pollmann K, Sauer P, Ring A, Stremmel W, Schlenker T.
Antibiotic prophylaxis after variceal hemorrhage reduces inci-
dence of early rebleeding. Hepatogastroenterology 2004;51:
19. Soares-Weiser K, Brezis M, Tur-Kaspa R, Leibovici L. Antibiotic
prophylaxis for cirrhotic patients with gastrointestinal bleeding.
Cochrane Database Syst Rev 2002;CD002907.
20. Rimola A, Garcia-Tsao G, Navasa M, Piddock LJ, Planas R, Ber-
nard B, Inadomi JM. Diagnosis, treatment and prophylaxis of
spontaneous bacterial peritonitis: a consensus document. Inter-
national Ascites Club. J Hepatol 2000;32:142–153.
21. Wong F, Bernardi M, Balk R, Christman B, Moreau R, Garcia-Tsao
G, Patch D, Soriano G, Hoefs J, Navasa M on behalf of the
International Ascites Club. Sepsis in cirrhosis: report on the 7th
meeting of the International Ascites Club. Gut 2005;54:718–
22. Dupeyron C, Mangeney N, Sedrati L, Campillo B, Fouet P, Leluan
G. Rapid emergence of quinolone resistance in cirrhotic patients
treated with norfloxacin to prevent spontaneous bacterial perito-
nitis. Antimicrob Agents Chemother 1994;38:340–344.
23. Aparicio JR, Such J, Pascual S, Arroyo A, Plazas J, Girona E,
Gutierrez A, de Vera F, Palazon JM, Carnicer F, Perez-Mateo M.
Development of quinolone-resistant strains of Escherichia coli in
stools of patients with cirrhosis undergoing norfloxacin prophy-
laxis: clinical consequences. J Hepatol 1999;31:277–283.
24. Fernandez J, Navasa M, Gomez J, Colmenero J, Vila J, Arroyo V,
Rodés J. Bacterial infections in cirrhosis: epidemiological
changes with invasive procedures and norfloxacin prophylaxis.
25. Ortiz J, Vila MC, Soriano G, Miñana J, Gana J, Mirelis B, Novella
MT, Coll S, Sabat M, Andreu M, Prats G, Sola R, Guarner C.
Infections caused by Escherichia coli resistant to norfloxacin in
hospitalized cirrhotic patients. Hepatology 1999;29:1064–1069.
26. Novella M, Solà R, Soriano G, Andreu M, Gana J, Ortiz J, Coll S,
Sabat M, Vila MC, Guarner C, Vilardell F. Continuous versus
inpatient prophylaxis of the first episode of spontaneous bacte-
rial peritonitis with norfloxacin. Hepatology 1997;25:532–536.
27. de Franchis R. Updating consensus in portal hypertension: report
of the Baveno III Consensus Workshop on definitions, methodol-
ogy and therapeutic strategies in portal hypertension. J Hepatol
28. Sobel JD, Kaye D. Urinary tract infections. In: Mandell GL, Bennet
JE, Dolin R, eds. Principles and practice of infectious diseases.
4th ed. New York: Churchill Livingstone, 1995:662–690.
29. Caly WR, Strauss E. A prospective study of bacterial infections in
patients with cirrhosis. J Hepatol 1993;18:353–358.
30. Rimola A, Bory F, Planas R, Xaubet A, Bruguera M, Rodés J. Acute
bacterial infections in liver cirrhosis. Gastroenterol Hepatol
31. Andreu M, Barrufet P, Force L, Solá R, Verdaguer A, Panadés A.
Fever in liver cirrhosis: a 6-month follow-up prospective study.
Med Clin 1985;84:433–436.
32. Clemente G, Barajas JM, Serrano MI, Pérez de Ayala MV,
Menchén P, Senent MC. Bacterial infections in liver cirrhosis.
Gastroenterol Hepatol 1986;9:285–290.
33. Suarez C, Pajares JM. Epidemiology of infections in liver cirrho-
sis. Rev Clin Esp 1981;160:299–303.
34. Palazón JM, Garcia A, Gómez A. Nosocomial infections in pa-
tients with cirrhosis. Gastroenterol Hepatol 1984;7:120–122.
35. Levi D, Terg R, Podesta A, Abecasis R, de Gle G, Rafaelli C.
Bacterial infections in patients with liver cirrhosis. Gastroenterol
36. Yoshida H, Hamada T, Inuzuka S, Ueno T, Sata M, Tanikawa K.
Bacterial infection in cirrhosis, with and without hepatocellular
carcinoma. Am J Gastroenterol 1993;88:2067–2071.
37. Llovet JM, Rodriguez-Iglesias P, Moitinho E, Planas R, Bataller R,
Navasa M, Menacho M, Pardo A, Castells A, Cabre E, Arroyo V,
Gasull MA, Rodés J. Spontaneous bacterial peritonitis in patients
with cirrhosis undergoing selective intestinal decontamination.
J Hepatol 1997;26:88–95.
Received February 2, 2006. Accepted June 15, 2006.
Address requests for reprints to: Miguel Navasa, MD, Liver Unit,
Hospital Clínic, Villarroel 170, 08036, Barcelona, Spain. e-mail:
Mnavasa@clinic.ub.es; fax: (34) 93-451 5522.
Supported in part by grants from the Fondo de Investigación Sani-
taria (FIS 00/0921) and the Instituto de Salud Carlos III (BEFI 00/
9380 and C03/02).
The authors thank E. Quintero and A. Pardo, Servicio de Digestivo,
Hospital Universitario de Canarias, Tenerife; J. Such, Servicio de Diges-
tivo, Hospital Universitario de Alicante; and C. Vila, Servicio de Diges-
tivo, Hospital del Mar, Barcelona, for their participation in patient
1056 FERNÁNDEZ ET AL GASTROENTEROLOGY Vol. 131, No. 4