ANNALS
OF
SURGERY
Vol.
222,
No.
1,
57-65
©
1995
Lippincott-Raven
Publishers
Controlled
Clinical
Trial
of
Selective
Decontamination
for
the
Treatment
of
Severe
Acute
Pancreatitis
Ernest
J.
T.
Luiten,
M.D.,*
Wim
C.
J.
Hop,
M.Sc.,t
Johan
F.
Lange,
M.D.,
Ph.D.,t
and
Hajo
A.
Bruining,
M.D.,
Ph.D.*
From
the
Department
of
Surgery,
University
Hospital
Rotterdam-Dijkzigt
and
Erasmus
University,*
the
Department
of
Epidemiology
and
Biostatistics,
Erasmus
University,
t
and
the
Department
of
Surgery,
St.
Clara
Hospital,t
Rotterdam,
the
Netherlands
Objective
A
randomized,
controlled,
multicenter
trial
was
undertaken
in
102
patients
with
objective
evidence
of
severe
acute
pancreatitis
to
evaluate
whether
selective
decontamination
reduces
mortality.
Summary
Background
Data
Secondary
pancreatic
infection
is
the
major
cause
of
death
in
patients
with
acute
necrotizing
pancreatitis.
Controlled
clinical
trials
to
study
the
effect
of
selective
decontamination
in
such
patients
are
not
available.
Methods
Between
April
22,
1990
and
April
19,
1993,
102
patients
with
severe
acute
pancreatitis
were
admitted
to
16
participating
hospitals.
Patients
were
entered
into
the
study
if
severe
acute
pancreatitis
was
indicated,
on
admission,
by
multiple
laboratory
criteria
(Imrie
score
2
3)
and/or
computed
tomography
criteria
(Balthazar
grade
D
or
E).
Patients
were
randomly
assigned
to
receive
standard
treatment
(control
group)
or
standard
treatment
plus
selective
decontamination
(norfloxacin,
colistin,
amphotericin;
selective
decontamination
group).
All
patients
received
full
supportive
treatment,
and
surveillance
cultures
were
taken
in
both
groups.
Results
Fifty
patients
were
assigned
to
the
selective
decontamination
group
and
52
were
assigned
to
the
control
group.
There
were
18
deaths
in
the
control
group
(35%),
compared
with
1
1
deaths
(22%)
in
the
selective
decontamination
group.
(adjusted
for
Imrie
score
and
Balthazar
grade:
p
=
0.048).
This
difference
was
mainly
caused
by
a
reduction
of
late
mortality
(>2
weeks)
due
to
significant
reduction
of
gram-negative
pancreatic
infection
(p
=
0.003).
The
average
number
of
laparotomies
per
patient
was
reduced
in
patients
treated
with
selective
decontamination
(p
<
0.05).
Failure
of
selective
decontamination
to
prevent
secondary
gram-negative
pancreatic
infection
with
subsequent
death
was
seen
in
only
three
patients
(6%)
and
transient
gram-negative
pancreatic
infection
was
seen
in
one
(2%).
In
both
groups
of
patients,
all
gram-negative
aerobic
pancreatic
infection
was
preceded
by
colonization
of
the
digestive
tract
by
the
same
bacteria.
57
58
Luiten
and
Others
Conclusion
Reduction
of
gram-negative
colonization
of
the
digestive
tract,
preventing
subsequent
pancreatic
infection
by
means
of
selective
decontamination,
significantly
reduces
morbidity
and
mortality
in
patients
with
severe
acute
necrotizing
pancreatitis.
Despite
improvement
in
surgical
strategies,
the
mor-
tality
of
patients
with
acute
necrotizing
pancreatitis
re-
mains
high,
between
20%
and
70%.1-8
Infection
of
pan-
creatic
necrosis
is
the
most
important
cause
of
late
mor-
tality
in
severe
acute
pancreatitis.37,9-'3
The
value
of
prophylactic
antibiotics
has
not
been
clearly
demon-
strated
in
patients
with
severe
acute
pancreatitis
and
pos-
sibly
is
due
to
patient
selection,
inadequate
spectrum,
in-
sufficient
doses,
or
tissue
penetration.12"14-'7
Intravenous
antibiotics,
which
penetrate
the
pancreas-
blood
barrier,
may
not
protect
the
necrotic
nonperfused
areas
in
and
around
the
inflamed
pancreas
against
infec-
tion.
The
route
by
which
sterile
pancreatic
necrosis
be-
comes
infected
is
not
yet
known.
Experimental
studies
and
clinical
observations
have
suggested
that
transloca-
tion
of
bacteria
toward
the
pancreas
occurs
hematoge-
nously,1819
transmurally
through
the
colon,20-22
via
lym-
phogenous
routes,20,23
via
ascites,19,23
and
through
bile24
and
duodenal
chyme
reflux.25
Because
gram-negative
bacteria-predominantly
isolated
from
the
pancreatic
necrosis-are
of
enteric
origin,
the
source
of
the
translo-
cating
bacteria
probably
is
the
intestine.2790422232627
Prevention
of
translocation
by
intraluminal
elimination
of
aerobic
gram-negative
micro-organisms
in
the
intesti-
nal
tract
may
be
an
effective
method
to
prevent
pancre-
atic
necrosis
from
becoming
infected.
In
a
controlled
ex-
perimental
study
on
rats
with
bile-salt-induced
pancre-
atitis,
Lange
et
al.
demonstrated
a
significant
reduction
of
mortality
in
rats
treated
with
intestinal
lavage
and
in-
traluminal
instillation
of
kanamycin.'9
Isaji
et
al.
re-
cently
demonstrated
in
mice
fed
a
choline-deficient,
ethi-
onine-supplemented
diet
to
induce
pancreatitis
that
oral
antibiotics
caused
a
threefold
reduction
of
infected
ne-
crosis
and
a
significantly
improved
survival.28
Several
clinical
studies
have
demonstrated
that
selec-
tive
decontamination
effectively
eliminates aerobic
gram-negative
bacteria
from
the
intestinal
tract
and
sometimes
reduces
gram-negative
septic
complications
in
intensive
care
unit
patients.
However,
results
regard-
29-34
ing
reduction
of
mortality
are
conflicting.
Thi
ra
domized,
controlled
clinical
trial
was
undertaken
to
eval-
uate
whether
selective
decontamination
reduces
mortal-
ity
in
patients
with
objective
evidence
of
severe
acute
pancreatitis.
METHODS
Between
April
22,
1990
and
April
19,
1993,
102
pa-
tients
with
objective
clinical
signs
of
severe
acute
pancre-
atitis
were
admitted
to
16
participating
hospitals.
The
di-
agnosis
of
acute
pancreatitis
had
been
established
on
the
basis
of
clinical
examination
and
elevated
plasma
levels
of
amylase
(>
1000
international
units/L),
or
at
diagnos-
tic
laparotomy
(ten
patients).
All
patients
were
scored
ac-
cording
to
multiple
laboratory
criteria
(Imrie
score)35
and
contrast-enhanced
computed
tomography
(CE-CT)
examinations
were
used
to
classify
disease
severity
(Bal-
thazar
grades)36
within
48
hours
of
hospital
admission
(Table
1).
Patients
were
included
in
the
study
if
the
following
cri-
teria
were
met:
severe
pancreatitis
was
indicated
by
three
or
more
points
according
to
the
Imrie
score
and/or
CT
findings
corresponding
with
Balthazar
grade
D
or
E.
Findings
at
diagnostic
laparotomy
were
not
accepted
as
an
inclusion
criteria.
Exclusion
criteria
were
defined
as
follows:
allergy
to
one
of
the
antibiotics
of
the
selective
decontamination
regimen;
younger
than
18
years
of
age;
postoperative
pancreatitis
after
pancreatic
surgery;
and
bacteriologically
proven
infected
necrosis
at
the
time
of
randomization.
The
attending
clinician
obtained
in-
formed
consent
from
the
patient
or
relatives.
Patients
who
satisfied
the
criteria
were
randomly
as-
signed
to
receive
standard
treatment
(control
group)
or
the
same
standard
treatment
plus
selective
decontami-
nation
(selective
decontamination
group).
A
24-hour
randomization
service
was
available
to
randomize
pa-
tients
with
stratification
per
center.
Follow-up
CT
scans
were
repeated
every
week
until
discharge
or
death.
The
study
was
approved
by
the
ethics
committees
of
the
par-
ticipating
hospitals.
Supported
by
a
grant
from
Merck
Sharp
&
Dohme
B.V.,
the
Nether-
lands,
and
a
grant
from
Roussel
B.V.,
the
Netherlands.
Address
reprint
requests
to
Prof.
Dr.
H.A.
Bruining,
Department
of
Surgery,
University
Hospital
Rotterdam-Dijkzigt,
Dr.
Molewat-
erplein
40,
3015
GD
Rotterdam,
the
Netherlands.
Accepted
for
publication
September
13,
1994.
Control
Group:
Standard
Treatment
A
nasogastric
tube
was
always
inserted.
Intravenous
crystalloid
solutions
were
given
according
to
clinical
re-
quirements.
Oxygen
therapy,
based
on
arterial
blood
gas
analysis,
was
administered
by
face
mask
and
was
re-
Ann.
Surg.
*-July
1995
Selective
Decontamination
for
Severe
Acute
Pancreatitis
59
>55
years
<2.00
mmol/l
>
16
mmol/I
>600
U/I
>
10
mmol/l
>15
109/1
<32
g/l
<60
mm
Hg
(7.5
kPa)
Degree
of
disease
severity
according
to
Balthazar
classificationt
Grade
A
Grade
B
Grade
C
Grade
D
Grade
E
Normal
pancreas.
Focal
or
diffuse
enlargement
of
the
pancreas
(including
contour
irregularities,
nonhomogeneous
attenuation
of
the
gland,
dilatation
of
the
pancreatic
duct,
and
foci
of
small
fluid
collections
within
the
gland,
as
long
as
there
is
no
evidence
of
peripancreatic
disease).
Intrinsic
pancreatic
abnormalities
associated
with
haziness
and
streaky
densities
representing
inflammatory
changes
in
the
peripancreatic
fat.
As
C
plus
single
ill-defined
fluid
collection
(phlegmon)
in
or
adjacent
to
the
pancreas.
As
C
plus
two
or
multiple,
poorly
defined
fluid
collections
or
the
presence
of
gas
in
or
adjacent
to
the
pancreas.
LDH
=
lactate
dehydrogenase;
WBC
=
white
blood
cell
count;
PaO2
=
arterial
oxy-
gen
concentration.
*
The
Imrie
score
equals
the
number
of
separate
criteria
present
(minimum:
0;
maxi-
mum:
8).
t
Computed
tomography
scan
with
use
of
oral
(1/2
hour
before)
and
intravenous
contrast
(rapid
intravenous
drip).
placed
by
assisted
ventilation
if
the
patient
developed
respiratory
insufficiency.
Cultures
from
the
oropharynx,
rectum,
sputum,
gastric
content,
and
urine
were
taken
on
admission
to
the
hospital
and
twice
a
week
until
dis-
charge.
If
fever
(.39
C)
was
present,
blood
cultures
were
taken.
Except
for
urine,
qualitative
semiquantitative
bacteriologic
analysis
was
performed
routinely
on
all
cul-
tures.
Cultures
of
pancreatic
necrosis
and
ascites
were
obtained
at
laparotomy
or
by
means
of
ultrasonography
or
CT-guided
percutaneous
puncture,
as
described
by
Gerzoffet
al.,'0
if
there
was
clinical
suspicion
of
infected
pancreatic
necrosis.
Patients
underwent
surgery
if
an
ul-
trasongographic
or
CT-guided
puncture
showed
pres-
ence
of
bacteria
or
if
the
condition
was
deteriorating
de-
spite
aggressive
supportive
treatment.
Surgery
was
per-
formed
either
by
transverse
or
median
laparotomy.
If
repeated
laparotomies
were
foreseen,
a
laparostomy,
i.e.,
a
ventral
open
packing
of
the
abdominal
cavity,
was
cre-
ated.2
Antibiotics
were
prescribed
according
to
the
anti-
biogram
only
in
the
presence
of
concurrent
infection.
Enteral
feeding
was
replaced
by
total
parenteral
nutrition
only
if
recurrent
gastric
retention
was
present.
Selective
Decontamination
Group:
Standard
Treatment
with
Adjuvant
Selective
Decontamination
Patients
randomized
to
the
selective
decontamination
group
received
the
same
treatment
as
the
control
group
with
the
addition
of
selective
decontamination.
The
se-
lective
decontamination
regimen
consisted
of
oral
ad-
ministration
of
colistin
sulfate
(200
mg),
amphotericin
(500
mg)
and
norfloxacin
(Noroxin,
Merck
&
Co.,
West
Point,
PA;
50
mg)
every
6
hours.
A
sticky
paste
contain-
ing
2%
of
the
three
selective
decontamination
drugs
was
smeared
along
the
upper
and
lower
gums
every
6
hours
and
at
the
tracheostomy,
if
present.
The
aforementioned
daily
dose
also
was
given
in
a
rectal
enema
every
day.
A
short-term
systemic
prophylaxis
of
cefotaxime
sodium
(Claforan,
Hoechst-Roussel
Pharm.,
Inc.,
Somerville,
NJ;
500
mg)
was
given
every
8
hours
until
gram-negative
bacteria
were
eliminated
from
the
oral
cavity
and
rec-
tum.
Selective
decontamination
was
discontinued
as
soon
as
the
risk
of
acquiring
a
new
infection
was
ab-
sent-i.e.,
the
patient
was
extubated
and
without
supple-
mentary
oxygen
therapy
or
infusions,
on
regular
oral
diet,
and
mobilized
on
the
ward.
Statistical
Analysis
Power
calculations
at
the
phase
of
trial
design,
assum-
ing
a
decrease
in
mortality
from
50%
to
25%,
led
to
a
total
number
of
154
patients
to
be
included
(alpha
=
0.05
[two-sided]
and
beta
=
0.
10).
Because
the
annual
accrual
rate
was
much
less
than
expected,
after
2
years
it
was
decided
to
limit
the
size
of
the
trial
to
100
evaluable
patients,
thereby
reducing
the
power
to
80%
at
one-sided
testing.
This
decision
was
made
without
consideration
of
the
accumulating
out-
comes.
Percentages
were
compared
by
the
Fisher
exact
test
or
the
chi
square
test,
if
appropriate.
Continuous
data
were
compared
by
the
Mann-Whitney
U
test.
For
mortality,
which
was
the
major
endpoint
in
this
study,
multivariate
analysis
(logistic
regression37)
at
entry
into
the
study,
al-
lowing
for
Imrie
score
and
Balthazar
grade,
was
per-
formed
to
obtain
a
higher
level
of
precision
in
comparing
treatment
groups.
Two-sided
p
values
of
0.05
or
less
were
Table
1.
PROGNOSTIC
SYSTEMS
USED
TO
SELECT
PATIENTS
FOR
INCLUSION
IN
THE
TRIAL
Multiple
laboratory
criteria
(Imrie
score)*
Age
Serum
uncorrected
calcium
Serum
urea
LDH
Blood
glucose
(no
diabetes)
WBC
Serum
albumin
PaO2
Vol.
222
-
No.
1
60
Luiten
and
Others
considered
statistically
significant.
Follow-up
was
con-
tinued
until
death
or
discharge
from
the
hospital.
Table
2.
BASELINE
CHARACTERISTICS
OF
PATIENTS
WITH
ACUTE
NECROTIZING
PANCREATITIS
RESULTS
Inclusions,
Exclusions,
and
Withdrawals
Of
the
109
patients
randomized
into
the
study,
2
(se-
lective
decontamination:
n
=
1;
control:
n
=
1)
were
ex-
cluded
because
of
perioperatively
proven
infected
necro-
sis
immediately
(<
1
hour)
after
randomization
and
be-
fore
treatment
was
started.
In
addition,
five
patients
(selective
decontamination:
n
=
3;
control:
n
=
2)
were
withdrawn
from
the
study
because
the
clinical
diagnosis
was
found
to
be
erroneous
(one
patient
with
streptococ-
cal
sepsis,
one
patient
with
an
acute
aortic
occlusion
im-
mediately
after
coronary
bypass
surgery,
one
patient
with
a
ruptured
pancreatic
pseudocyst,
one
patient
with
chronic
pancreatitis,
and
one
patient
with
an
endoscopic
retrograde
cholangiopancreatograpy-induced
choledo-
chus
perforation).
Of
the
remaining
102
patients,
50
had
been
assigned
to
the
selective
decontamination
group
and
52
to
the
control
group.
Inclusion
scores
are
listed
in
Table
2.
Selective
decontamination
was
started
within
24
hours
of
randomization.
Ten
patients
(selective
de-
contamination:
n
=
8;
control:
n
=
2)
with
severe
acute
pancreatitis
had
to
be
randomized
only
on
the
basis
of
the
multiple
laboratory
criteria
(Imrie
score
2
3)
because
their
condition
did
not
permit
transport
from
the
inten-
sive
care
unit
to
the
CT
scanner
at
that
time.
Of
these
patients,
fluid
collections
in
or
adjacent
to
the
severely
inflamed
pancreas
(personal
communication
with
the
at-
tending
surgeon
immediately
postoperatively)
were
demonstrated
on
the
first
day
of
the
study
during
lapa-
rotomy
in
eight
patients
and
with
abdominal
ultrasound
in
one
patient.
Because
of
these
results,
the
Balthazar
grade
was
classified
as
grade
E.
In
the
other
patient
(con-
trol
group;
Imrie
score
=
3),
a
CT
scan
was
performed
only
after
5
days
of
treatment,
and
it
demonstrated
a
peripancreatic
fluid
collection.
The
latter
patient
also
un-
derwent
surgery
on
the
first
day
after
randomization;
however,
the
pancreatic
loge
was
left
untouched.
The
Balthazar
grade
at
the
time
of
randomization
was
un-
available
for
this
patient.
Comparability
of
Control
and
Selective
Decontamination
Group
Both
treatment
groups
appeared
well
matched
for
age,
sex,
etiologic
factors,
Imrie
score,
and
Balthazar
grade.
Characteristics
for
both
groups
are
listed
in
Table
2.
The
mean
Imrie
score
was
3.2
for
both
groups.
Patients
with
an
Imrie
score
of
8
were
not
encountered
in
this
study.
Mean
age
(years)
Sex
Male
Female
Etiology
Alcohol
Gallstones
Hyperparathyroidism
Blunt
abdominal
trauma
Postoperative
ERCP-induced
Unknown
Imrie
score
0
1
2
3
4
5
6
7
8
Balthazar
degree
of
disease
severity
Grade
A
Grade
B
Grade
C
Grade
D
Grade
E
Day
1
unavailable
Selective
Decontamination
Group
(n
=
50)
56
(26-91)
31
19
19
17
0
1
2
1
10
5
8
2
10
12
9
3
1
0
0
0
3
21
26
0
Control
Group
(n
=
52)
55
(20-88)
29
23
12
19
2
0
2
3
14
4
7
10
6
13
6
2
4
0
0
0
4
20
27
1*
ERCP
=
endoscopic
retrograde
cholangiopancreatography.
*
Computed
tomography
scan
was
performed
on
day
5:
grade
D.
Mortality
Eleven
patients
(22%)
in
the
selective
decontamina-
tion
group
died
as
compared
with
18
patients
(35%)
in
the
control
group.
This
difference
is
not
significant
(p
=
0.19).
The
95%
confidence
limits
of
the
difference
(con-
trol
group
minus
selective
decontamination
group)
in
mortality
ranges
from
less
than
4%
to
more
than
30%.
Survival
according
to
treatment
group
is
shown
in
Figure
1.
All
deaths
occurred
within
80
days.
In
each
of
both
groups,
six
patients
died
of
multiple-organ
failure
with
documented
sterile
pancreatic
necrosis.
Ten
patients
in
the
control
group
died
of
a
gram-negative
pancreatic
sep-
sis
syndrome
compared
with
only
three
such
patients
in
the
selective
decontamination
group
(p
=
0.07).
In
each
Ann.
Surg.
.
July
1995
Selective
Decontamination
for
Severe
Acute
Pancreatitis
61
aS_~~~
~~~~~~~~~-
-
-
L
-
~~~~~~~__--
-
- -
-
--
L--
-
Selective
Decontamination
group
-----
Control
group
0.00
-
30
day
60
90
Figure
1.
Overall
survival
according
to
treatment.
Overall
mortality
rates
at
90
days:
selective
decontamination
group
=
22%;
control
group
=
35%.
Adjusted
for
lmrie
score
and
Balthazar
grade,
p
=
0.048.
Difference
in
mortality
rates
equals
13%
(95%
confidence
limits:
-4%,
+30%).
group,
one
patient
died
of
sepsis
due
to
a
solitary
gram-
positive
pancreatic
infection.
Gram-positive
sepsis
of
un-
known
origin,
without
pancreatic
infection,
was
the
cause
of
death
in
one
patient
in
each
of
both
groups.
The
Imrie
score
at
entry
into
the
study
appeared
to
correlate
very
strongly
with
mortality
(Fig.
2).
Mortality
was
0%,
for
an
Imrie
score
of
0
or
1,
and
it
gradually
increased
to
100%,
for
patients
with
an
Imrie
score
of
7
(Ptrend
<
0.001).
Mortality
also
increased
with
increasing
Balthazar
grade,
although
these
differences
were
less
pro-
nounced
(Ptrend
=
0.04)
(Fig.
2).
The
worsening
of
prog-
nosis
with
increasing
Imrie
score
and
Balthazar
grade
was
apparent
in
each
separate
treatment
group.
Overall
mortality
in
the
selective
decontamination
group
versus
the
mortality
in
the
control
group
appeared
to
be
signifi-
cantly
lower
(p
=
0.048),
using
multivariate
analysis
al-
lowing
for
Imrie
score
and
Balthazar
grade
(Table
3).
This
analysis
also
demonstrates
the
importance
of
the
Imrie
score
in
predicting
mortality.
There
was
no
sig-
nificant
relation
between
mortality
and
the
Balthazar
grade.
1.00
-
-
0.75-
.,,
DU
n
n
a
0.50-
.c
uz
0.25
-
0.00
-
s
30
day
60
90
t__-----
---------
Balthazar
grade
P
(trend)
=
0.04
30
day
60
90
Figure
2.
Survival
according
to
an
Imrie
score
of
0/1
(n
=
24),
2
(n
=
12),
3
(n
=
16),
4
(n
=
25),
5
(n
=
15),
6
(n
=
5),
and
7
(n
=
5),
respectively
(upper
panel).
Survival
according
to
Balthazar
grade
C
(n
=
7),
D
(n
=
41),
or
E
(n
=
53),
respectively
(lower
panel).
Both
as
assessed
at
entry
into
the
study,
for
both
treatment
groups
combined.
Severe
acute
pancreatitis
was
defined
according
to
Imrie
score
2
3
points
and/or
CT
findings
ac-
cording
to
Balthazar's
degree
of
disease
severity
grade
D
or
E.
Table
3.
MULTIVARIATE
ANALYSIS
OF
MORTALITY
IN
RELATION
TO
TREATMENT,
IMRIE
SCORE,
AND
BALTHAZAR
GRADE
Odds
Factor
Ratio
p
Value
Bacteriologic
Analysis
Secondary
pancreatic
infection
occurred
in
20
pa-
tients
(38%)
in
the
control
group
and
in
9
patients
(18%)
of
the
selective
decontamination
group
(p
=
0.03).
Gram-negative
pancreatic
infection
occurred
in
17
pa-
tients
(33%)
in
the
control
group
and
in
only
4
patients
(8%)
in
the
selective
decontamination
group
(p
=
0.003).
Pancreatic
necrosis
was
not
infected
in
11
of
16
patients
who
died
early,
in
contrast
to
only
3
of
13
patients
who
died
after
2
weeks
(p
=
0.03).
This
difference
is
similar
for
both
groups.
Treatment
Control
Selective
decontamination
lmrie
score
Balthazar
grade
C/D
E
1*
0.3
(0.3)
3.7t
(3.9)
1*
1.8
(-)
0.048
(0.049)
<
0.001
(<
0.001)
0.354
(-)
*
Reference
category.
t
Relative
to
patients
who
have
an
lmrie
score
of
1
point
less.
Data
given
are
odds-ratios
for
mortality.
(Odds-ratios
>
1
indicate
an
increased
mor-
tality;
<
1
indicate
a
decreased
mortality.)
Data
between
parentheses
denote
results
when
only
treatment
and
lmrie
score
are
analysed
regarding
mortality.
1.00
-
'
0.75-
.0
.,,
n
0
L
0
O.
0.50
-
.
4
C-
c:'
0.25-
1.00
-
>
0.75-
.,4
.,
n
co
aL
0.50-
-'
n
0.25-
0.00
-
Vol.
222
-
No.
1