A randomized controlled trial comparing a computer-assisted insulin protocol with a strict and conventional protocol for glucose control in critically ill patients

Intensive Care Unit, Hospital Israelita Albert Einstein, São Paulo 05652-000, Brazil.
Journal of critical care (Impact Factor: 2). 07/2009; 24(3):371-8. DOI: 10.1016/j.jcrc.2009.05.005
Source: PubMed
ABSTRACT
The objective of this study is to evaluate blood glucose (BG) control efficacy and safety of 3 insulin protocols in medical intensive care unit (MICU) patients.
This was a multicenter randomized controlled trial involving 167 MICU patients with at least one BG measurement >or=150 mg/dL and one or more of the following: mechanical ventilation, systemic inflammatory response syndrome, trauma, or burns. The interventions were computer-assisted insulin protocol (CAIP), with insulin infusion maintaining BG between 100 and 130 mg/dL; Leuven protocol, with insulin maintaining BG between 80 and 110 mg/dL; or conventional treatment-subcutaneous insulin if glucose >150 mg/dL. The main efficacy outcome was the mean of patients' median BG, and the safety outcome was the incidence of hypoglycemia (<or=40 mg/dL).
The mean of patients' median BG was 125.0, 127.1, and 158.5 mg/dL for CAIP, Leuven, and conventional treatment, respectively (P = .34, CAIP vs Leuven; P < .001, CAIP vs conventional). In CAIP, 12 patients (21.4%) had at least one episode of hypoglycemia vs 24 (41.4%) in Leuven and 2 (3.8%) in conventional treatment (P = .02, CAIP vs Leuven; P = .006, CAIP vs conventional).
The CAIP is safer than and as effective as the standard strict protocol for controlling glucose in MICU patients. Hypoglycemia was rare under conventional treatment. However, BG levels were higher than with IV insulin protocols.

Full-text

Available from: José Eluf-Neto
A randomized controlled trial comparing a
computer-assisted insulin infusion protocol with
a strict and a conventional protocol for glucose control
in critically ill patients
Alexandre B. Cavalcanti MD
a,b,c,
, Eliezer Silva PhD
a
, Adriano J. Pereira MD
a
,
Milton Caldeira-Filho MD
d
, Francisca P. Almeida RN
a
, Glauco A. Westphal MD
e
,
Renate Beims RN
f
, Caio C. Fernandes MD
b
, Thiago D. Correa MD
a
,
Marcos R. Gouvea BCS
g
, José Eluf-Neto PhD
c
a
Intensive Care Unit, Hospital Israelita Albert Einstein, São Paulo 05652-000, Brazil
b
Intensive Care Unit, Hospital Estadual Mário Covas, Santo André 09190-615, Brazil
c
Faculdade de Medicina da Universidade de São Paulo, São Paulo 01246-903, Brazil
d
Intensive Care Unit, Hospital Dona Helena, Joinville 89204-250, Brazil
e
Intensive Care Unit, Centro Hospitalar UNIMED, Joinville 89204-060, Brazil
f
Intensive Care Unit, Hospital Municipal São José, Joinville 89202-000, Brazil
g
Education and Research InstituteHospital Israelita Albert Einstein, São Paulo, 05652-000, Brazil
Keywords:
Insulin;
Hyperglycemia;
Hypoglycemia;
Blood glucose;
Critical care
Abstract
Purpose: The objective of this study is to evaluate blood glucose (BG) control efficacy and safety of
3 insulin protocols in medical intensive care unit (MICU) patients.
Methods: This was a multicenter randomized controlled trial involving 167 MICU patients with at least
one BG measurement 150 mg/dL and one or more of the following: mechanical ventilation, systemic
inflammatory response syndrome, trauma, or burns. The interventions were computer-assisted insulin
protocol (CAIP), with insulin infusion maintaining BG between 100 and 130 mg/dL; Leuven protocol,
with insulin maintaining BG between 80 and 110 mg/dL; or conventional treatmentsubcutaneous
insulin if glucose N150 mg/dL. The main efficacy outcome was the mean of patients' median BG, and
the safety outcome was the incidence of hypoglycemia (40 mg/dL).
Results: The mean of patients' median BG was 125.0, 127.1, and 158.5 mg/dL for CAIP, Leuven, and
conventional treatment, respectively (P = .34, CAIP vs Leuven; P b .001, CAIP vs conventional). In
CAIP, 12 patients (21.4%) had at least one episode of hypoglycemia vs 24 (41.4%) in Leuven and
2 (3.8%) in conventional treatment (P = .02, CAIP vs Leuven; P = .006, CAIP vs conventional).
Corresponding author. Hospital Albert EinsteinCTI-Adultos, 05651-901 São Paulo, Brazil. Tel.: +55 11 98829343; fax: +55 11 35547083.
E-mail address: alexandrebiasi@hotmail.com (A.B. Cavalcanti).
0883-9441/$ see front matter © 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcrc.2009.05.005
Journal of Critical Care (2009) 24, 371378
Page 1
Conclusions: The CAIP is safer than and as effective as the standard strict protocol for controlling
glucose in MICU patients. Hypoglycemia was rare under conventional treatment. However, BG levels
were higher than with IV insulin protocols.
© 2009 Elsevier Inc. All rights reserved.
1. Introduction
A randomized controlled trial involving 1548 surgical
patients admitted to intensive care observed 42% relative
reduction in death risk and significant morbidity reduction in
the group under strict glycemic control compared with
conventional treatment [1]. Further randomized trials did not
show mortality reduction with intensive insulin therapy [2-4]
or observed a beneficial effect on mortality restricted to the
subgroup with length of intensive care unit (ICU) stay longer
than 3 days [5]. Nevertheless, guidelines from professional
organizations have recommended strict glycemic control
with continuous intravenous insulin infusion for the manage-
ment of critically ill patients [6,7], and it has become routine
practice in many ICUs.
The major concern of strict glucose control is hypoglyce-
mia [2,8]. In addition to a higher risk of hypoglycemia,
implementation of a strict glucose control protocol in ICU
routine may face difficulties because of increased workload
and lack of experience of nursing staff. Therefore, to guarantee
successful implementation of strict glucose control in critical
care, it is essential that the protocol be efficacious, safe (low
risk of hypoglycemia), and practical to use. Algorithms for
glucose control in critical care patients have been evaluated in
several studies [9]. However, most studies were case series or
before-after comparisons, with the lack of an adequate
standard strict glucose control group limiting the interpreta-
tion of results [9]. More recently, randomized trials were
conducted to compare a nonlinear model predictive control
algorithm delivered by a computerized system with conven-
tional strict glucose control protocol [10,11]. Hypoglycemic
episodes were virtually absent, and very strict glucose control
was achieved with model predictive control. However, the
software is not currently available for routine use.
We developed an intravenous insulin infusion algorithm,
which aims to maintain blood glucose (BG) levels between
100 and 130 mg/dL. To make it user friendly, we developed a
computer program for Windows desktops or handhelds to
display the algorithm (computer-assisted insulin protocol
[CAIP]). The desktop and printed version of the program is
available as electronic supplemental material.
The study primary efficacy objective was to compare
glucose control during ICU stay obtained with CAIP vs a
standard strict glycemic control protocol (Leuven protocol)
[1] or a conventional intermittent insulin administration
protocol (conventional treatment) in critically ill patients.
The primary safety objective was to compare the incidence of
hypoglycemia during ICU stay with CAIP vs a standard strict
glycemic control protocol (Leuven protocol) or a conven-
tional intermittent insulin administration protocol (conven-
tional treatment) in critically ill patients.
2. Methods
2.1. Participants
The study was conducted in 5 ICUs from 5 different
Brazilian institutions: Hospital Estadual Mário Covas, Santo
André, Brazil, a 32-bed, teaching, closed ICU in a 321-bed
hospital; Hospital Israelita Albert Einstein, São Paulo, Brazil,
a 30-bed, teaching, open ICU in a 450-bed hospital; Hospital
Municipal São José, Joinville, Brazil, an 8-bed, teaching,
closed ICU in a 200-bed hospital; Hospital Dona Helena, a
7-bed, nonteaching, closed ICU in a 120-bed hospital;
Centro Hospitalar UNIMED, Joinville, Brazil, a 8-bed,
nonteaching, closed ICU in a 140-bed hospital.
Adult medical patients admitted to the ICU were eligible for
the study if they had at least one BG measurement 150 mg/dL
or higher plus one of the following: (1) mechanical ventilation
for an acute process, with expected duration of 24 hours or
longer; (2) trauma; (3) burn; (4) systemic inflammatory
response syndrome (SIRS, modified criteria), with at least
3 of the following: (a) core temperature 38°C or 36°C; (b)
heart rate of 90 beats/min or higher, except in patients with a
medical condition or receiving a medication known to prevent
tachycardia; (c) respiratory rate of 20 breaths/min or higher, or
aPa
CO
2
of 32 mm Hg or lower; (d) white blood cell count
12,000 or 4000/mm
3
or N10% immature neutrophils.
Patients were excluded if they were younger than 21 years,
were surgical patients, were admitted because of diabetic
ketoacidosis or nonketotic hyperosmolar state, or were in a
state in which death was perceived as imminent.
The study protocol and consent form were approved by
the ethics review board of each institution. The study was
performed in accordance with ethical standards stated in the
Declaration of Helsinki. Written consent was obtained from
every patient or the next of kin when the patient was unable
to give it. The study protocol was registered at ClinicalTrials.
gov with the number 00410852.
2.2. Interventions
The following treatments were evaluated:
1. Computer-assisted insulin protocol: the target range
for BG was 100 to 130 mg/dL using continuous
372 A.B. Cavalcanti et al.
Page 2
intravenous insulin infusion. Insulin dose adjustments
were assisted by a computer program running on an
ICU desktop or a handheld (electronic supplemental
material: computer program and printed version).
2. Strict glycemic control protocol (Leuven protocol):
continuous intravenous insulin infusion with adjust-
ments according to a protocol developed and used by
Van den Berghe et al [1,5], which aims to maintain BG
between 80 and 110 mg/dL.
3. Conventional treatment: intermittent subcutaneous
insulin administration according to a sliding scale.
Insulin is given for BG levels higher than 150 mg/dL.
Computer-assisted insulin protocol was developed with
the aim of being easier to use and safer and with a similar
efficacy in controlling BG than the standard protocol for
strict BG control in critically ill patients [1]. The character-
istics of CAIP are as follows:
1. Blood glucose goal between 100 and 130 mg/dL. This
was chosen because there is no evidence that an
intermediate level (between 110 and 180 mg/dL) is
inferior to a very strict range (80 to 110 mg/dL) and the
risk of hypoglycemia may be lowered. Lately, other
studies have suggested that targets up to 140 mg/dL are
associated with the highest survival rates [12,13].
2. Adjustments in insulin infusion are delivered according
to the current insulin infusion rate, glucose level, and
variation per unit of time between the last and current
glucose measurements. A table developed by one of the
authors (ABC) based on actual management of patients
in a 10-bed ICU in 2002 guides the adjustments. It was
further refined during a 2-year period of use. The table
is used as follows: (a) The current BG level is identified
in the first column of the table, which contains 10
intervals for current BG. (b) The appropriate interval of
variation in glucose (between current and last measure-
ments) is identified. (c) The guidelines for insulin
adjustments and time to repeat glucose measurement
are available in the same row.
3. In fact, the guidelines on starting and adjusting insulin
infusion and timing of glucose measurements are
delivered by a simple computer program for Windows
PCs or handheld devices. The program is based on the
table described above. The program has a simple layout
and is very user-friendly, needing minimal training.
The aim is to eliminate the need of calculations, thus,
making the use of CAIP easier and faster.
4. Increases in insulin infusion are always additive (eg, if
current BG is 140 mg/dL and it increased 15 mg/dL per
hour, then insulin should be increased by 0.6 U/h),
whereas decreases are proportional (eg, if current BG is
140 mg/dL and it decreased 80 mg/dL per hour since
the last measurement, then insulin infusion should be
reduced by 50%). The objective was to prioritize safety
as well to avoid fluctuations in glucose control when
patients are closer to the target range. For instance, if a
patient is receiving insulin at a high rate (eg, 10 U/h)
and his/her BG dropped fast to a near normal level, then
an absolute decrease in insulin rate may be insufficient
to avoid further decrease in BG, whereas a proportional
reduction (eg, 50%) may be more adequate. The
inverse situation is also true; if a patient is receiving a
low dose of insulin, then an absolute decrease might be
too much, leading to hyperglycemia, whereas a
proportional decrease may be more appropriate.
5. Intervals between measurements are not fixed.
Instead, their sizes depend on both current glucose
level and variation per unit of time between last and
current measurement.
6. Adjustments in insulin infusion and time to make the
next BG measurement are always specific values, not
ranges, to standardize conduct, and decrease reliance
on experience and time spent to make decisions.
In all groups, insulin adjustments were made by the
nursing staff. Randomized treatment was administered until
ICU discharge. Other insulin formulations or oral hypogly-
cemic agents were not used during the ICU stay. When the
patient was discharged to the step-down unit or the ward, BG
was controlled according to his/her physician discretion.
Feeding was prescribed according to each participating
ICU routine practice. However, it was recommended that
patients receive adequate caloric and glucose intake. The
study protocol suggested infusion of 200 to 300 g of glucose
per day for patients without enteral or parenteral nutrition.
Enteral nutrition was initiated as soon as possible. Parenteral
nutrition was recommended when enteral nutrition could not
be administered from the second ICU day. An energy intake
between 105 and 147 kJ/kg per day was suggested, with
adjustments for special situations.
Blood glucose was measured with Advantage glucometer
using Accu-Chek Advantage test strips (Roche Diagnostics,
Mannheim, Germany). Measurements were made using
capillary whole blood obtained from patient's fingertip. For
patients with shock or receiving vasoconstrictors, it was
recommended that arterial sampling be used to measure BG.
Clinical and demographic data were collected at baseline
for all patients, including information necessary to calculate
APACHE (Acute Physiology and Chronic Health Evaluation)
II and Sequential Organ Failure Assessment scores [14,15].
2.3. Outcome measures
The mean of patients' median BG during the ICU stay
was the primary efficacy outcome, and the incidence of
hypoglycemia (40 mg/dL) was the primary safety outcome.
A variable number of BG measurements, in some cases,
hundreds, were obtained from each patient. To deal with the
repetitive measurements, we defined the primary efficacy
end point as the mean of each patient's median; that is, we
calculated a summary measure for each patient [16].
373Computer-assisted insulin infusion protocol for glucose control
Page 3
The secondary outcomes were as follows: (1) hypergly-
cemic index, with a cutoff at 140 mg/dL (HGI 140); (2)
proportion of hypoglycemic episodes (b40 mg/dL) in
relation to total BG measurements per patient; (3) number
of BG measurements obtained per patient; (4) proportion of
time with BG controlled between 60 and 140 mg/dL; (5)
nurse perception about the feasibility of the 3 protocols. The
HGI is the area under the curve and above a predefined BG
cutoff (we used 140 mg/dL) divided by time under
observation [12]. The inde x is calculated after simple
interpolation of each patient's BG measurements. It
represents the burden of BG above normal and is the
best glucose index to predict mortality [12].
A questionnaire was issued to all 60 nurses who applied
the study's protocols to evaluate their perception regarding
the protocols. For each protocol, CAIP, Leuven, or
conventional, the following question was asked: In relation
to the use of the protocol (CAIP, Leuven, or conventional),
considering issues related to time spent to execute protocol
tasks and protocol complexity, do you believe it was (a) very
easy, (b) easy, (c) difficult, or (d) very difficult? It also asked
which of the 3 tested protocols the nurse would like to be
adopted as the standard protocol in his/her ICU.
2.4. Randomization
The randomization list with 3 groups (CAIP, Leuven
protocol, and conventional treatment), with blocks of 6,
stratified by the center, was generated by computer. The system
analyst who generated the randomization list did not take part
in any other aspects of the study. Investigators enrolling a
patient into the study obtained the assigned treatment in the
study Web site only after registering the patient.
There was no blinding of patients or investigators
because it was not feasible to conceal the different glycemic
control protocols.
2.5. Statistical methods
We defined the sample size as 165 patients (55 per group)
to detect a difference of 20 mg/dL between the highest and
lowest of group mean BG levels using 1-way analysis of
variance, assuming a SD of 33 mg/dL for each group,
2-tailed type 1 error of 0.05, and power of 80% [1].
Categorical variables were displayed as absolute and
relative frequencies. Numerical variables were presented as
means and SDs or medians and interquartile (IQR) ranges, as
appropriate. Comparisons of proportions were made using
χ
2
test. Comparisons of quantitative variables were carried
out using Mann-Whitney test. We evaluated whether
different participant centers modified the effect of glucose
control protocols on mean glucose and on the risk of
hypoglycemia using 2-way analysis of variance and logistic
regression, respectively. All P values presented are 2-tailed.
3. Results
Between May 4, 2005, and December 4, 2006, we
randomized 168 patients in 5 ICUs: 56 allocated to CAIP, 58
Fig. 1 Study flow diagram.
374 A.B. Cavalcanti et al.
Page 4
to Leuven protocol, and 54 to conventional treatment
(Fig. 1). We followed up all patients until ICU discharge
(primary efficacy and safety outcome analysis) and hospital
discharge, except for one (conventional treatment group)
whose consent to participate in the study was withdrawn by
her next of kin. Ninety-day follow-up was obtained from all
patients except one who had been assigned to conventional
treatment. All patients were analyzed according to the group
they were randomly allocated (intention-to-treat principle),
except for the patient who withdrew consent.
Groups were generally comparable at baseline (Table 1).
Mean age was approximately 60 years, and approximately
half of the patients were female. Patients were very sick as
denoted by around 80% needing mechanical ventilation at
baseline, 68% to 83% having 3 or more SIRS criteria, and
median APACHE II score varied from 20 to 25. Length of
stay in ICU was similar in all groups (median, 7 days).
The mean of patients' median BG was 125.0 ± 17.7,
127.1 ± 32.2, and 158.5 ± 49.6 mg/dL for CAIP, Leuven, and
conventional treatment, respectively (P b .001, CAIP vs
conventional treatment; P = .34, CAIP vs Leuven) (Table 2).
The incidence of hypoglycemia (40 mg/dL) was lower in
CAIP group than in Leuven group, although it was higher in
CAIP than conventional treatment (Table 2).
When episodes of hypoglycemia were considered in
relation to the number of BG measurements done, we have
found that each patient in CAIP protocol had a mean 0.43%
of glucose measurements below 40 mg/dL compared with
Table 1 Baseline demographic and clinical characteristics
Characteristic CAIP (n = 56) Leuven protocol (n = 58) Conventional treatment (n = 53)
Female sex, no. (%) 24 (42.9) 32 (55.2) 27 (50.9)
Age (y)
Mean (SD) 63.6 (17.6) 58.8 (18.4) 61.0 (19.7)
Inclusion criteria, no. (%)
Mechanical ventilation 47 (83.9) 47 (81.0) 41 (77.4)
Trauma 2 (3.6) 3 (5.2) 6 (11.3)
SIRS 46 (82.1) 48 (82.8) 36 (67.9)
Diagnostic category, no. (%)
Respiratory 13 (23.2) 18 (31.0) 18(34.0)
Other sepsis 13 (23.2) 8 (13.8) 4 (7.6)
Cardiovascular 10 (17.9) 10 (17.2) 6 (11.3)
Neurologic 7 (12.5) 7 (12.1) 14 (26.4)
Hematologic or oncologic 6 (10.7) 6 (10.3) 0 (0.0)
Trauma 3 (5.4) 4 (6.9) 8 (15.1)
Gastrointestinal, liver or pancreas 3 (5.4) 4 (6.9) 1 (1.9)
Metabolic or renal 1 (1.8) 1 (1.7) 2 (3.8)
History of diabetes mellitus, no. (%) 17 (30.4) 20 (34.5) 14 (26.4)
Treated with diet only
a
2 (11.8) 4 (20.0) 0 (0.0)
Oral antidiabetic agent (no insulin)
a
10 (58.8) 7 (35.0) 8 (57.1)
Insulin (±oral antidiabetic agent)
a
5 (29.4) 9 (45.0) 6 (42.9)
Body mass index (kg/m
2
)
Mean (SD) 25.5 (5.2) 26.4 (4.6) 25.0 (3.9)
SOFA
Median (IQR) 8 (5-10) 7.5 (5-11) 6 (4-8)
APACHE II score
Median (IQR) 24.5 (17-27) 21 (17-26) 20 (15-27)
Table 2 Primary outcomes: mean of patients' median BG and incidence of hypoglycemia
Outcome CAIP
(n = 56)
Leuven protocol
(n = 58)
Conventional treatment
(n = 53)
P
CAIP vs Leuven CAIP vs conventional treatment
Median BG
a
(mg/dL)
Mean 125.0 127.1 158.5 .34 b.001
Patients with hypoglycemia
b
No. (%) 12 (21.4) 24 (41.4) 2 (3.8) .02 .006
a
Each patient's median BG was used as a summary measure to derive a group mean.
b
Patients with at least one BG of 40 mg/dL or less for all patients.
375Computer-assisted insulin infusion protocol for glucose control
Page 5
0.55% in Leuven group (P = .04) and 0.03% in conventional
group (P = .007) (Table 3). Other secondary outcomes are
also presented in Table 3.
We found no difference of BG control among centers (P =
.71), as assessed by the means of glucose medians, nor was
there modification of the BG control protocols effect by
center (interaction P = .35). Likewise, there was no difference
in the incidence of hypoglycemia between centers (P N.05),
nor did the center influence the risk of hypoglycemia
observed in the 3 groups (all interaction P N .05).
The nurses' perception regarding the protocols' feasibility
is depicted in Fig. 2. All 60 nurses answered the
questionnaires. In terms of complexity and time spent to
execute the protocol tasks, 11.7% found the CAIP difficult or
very difficult, as compared with 38.4% for Leuven protocol
and 13.3% for conventional treatment (P = 0.78 for CAIP vs
conventional treatment; P b .001 for CAIP vs Leuven). Fifty-
six percent of the nurses would like the CAIP to be adopted
as the standard protocol in their ICU, 22% preferred the
Leuven protocol, 15% preferred the conventional protocol,
and 7% believe all the protocols were alike.
4. Discussion
4.1. Summary of findings
The CAIP allows a BG control as effective as the
standard protocol for tight glucose control proposed by Van
den Berghe et al [1,5] . Both maintained BG at normal
nonfasting levels. However, the risk of hypoglycemia was
lower with the computer-assisted protocol, and it was
considered easier to use than the Leuven protocol. The
conventional protocol led to a minimal risk of hypoglycemia,
although it was clearly inferior to the intravenous protocols
in avoiding hyperglycemia.
4.2. Strengths and limitations
Great care was taken to guarantee the integrity of results
of this clinical trial. Randomization was central using a Web
site that assured concealment of the allocation list. Other
items such as careful data collection and intention-to-treat
analysis were used in this study.
This study has also some limitations. We did not collect
data regarding the number of patients assessed for eligibility,
because of a shortage of human resources. Nevertheless,
whenever one of the researchers was in the participant
centers, all potentially eligible patients were systematically
approached to be enrolled in the study. Therefore, we believe
there is no limitation on external generalization of this
study's results.
Adherence to the protocols was suboptimal in one of
the study ICUs because of understaffing. Many hypogly-
cemic episodes might have been avoided if adherence
had been better. Glucometers used to measure BG at the
bedside, such as those used in this study, have been recently
shown to have insufficient accuracy [17-19]. Blood glucose
control in this study may have been more effective and
the risk of hypoglycemia may have been lower if it was
carried out in hemogasometers. However, it is unlikely that
this factor would neutralize t he superiority of CAIP
compared with Leuven protocol. In addition, most ICUs
Fig. 2 Nurses' perceptions about feasibility of protocols.
Table 3 Secondary outcomes: efficacy and safety of the protocols in attaining BG control
Outcome CAIP
(n = 56)
Leuven protocol
(n = 58)
Conventional treatment
(n = 53)
P
CAIP vs Leuven CAIP vs conventional treatment
Episodes of hypoglycemia
a
(%)
Mean 0.43 0.55 0.03 .04 .007
HGI 140 (mg/dL per hour)
Median (IQR) 4.2 (2.0-9.6) 8.7 (2.5-20.2) 20.5 (5.1-42.8) .10 b.001
No. of BG measurements
Median (IQR) 100 (33-192) 105 (35-312) 49 (39-77) .52 .01
Proportion of time BG controlled between 60 and 140 mg/dL (%)
Mean (SD) 71.8 (18.0) 67.9 (20.8) 47.1 (30.2) .50 b.001
a
Percentage of hypoglycemic episodes per patient = ([BG 40 mg/dL for all BG measurements] * 100).
376 A.B. Cavalcanti et al.
Page 6
use point-of-care glucometers and do not have hemogas-
ometers. Therefore, the results of this study may be more
applicable in most settings.
4.3. Comparison with previous literature
Patients from CAIP group had mean BG median level
similar to those of Leuven group, approximately 125 mg/dL,
even though the target BG range of the latter is 80 to
110 mg/dL. The failure of Leuven protocol to achieve the
target range in this study, as opposed to the trials of Van den
Berghe et al [1,5], might be due to the fact that it is difficult to
implement this protocol into the routine of different ICUs,
because it depends mainly on the experience and motivation
of the nursing team. Instructions for insulin dose adjustments
are not always precise in the Leuven protocol, demanding
decisions to be made by nurses. Another explanation for the
mean glucose level being higher than the Leuven protocol's
target might be that the patients included in this study had
higher baseline glucose levels than those evaluated in the 2
Leuven trials [1,5]. This is because, as opposed to the
Belgium studies, hyperglycemia (150 mg/dL) was a
necessary inclusion criterion in the present trial.
In spite of the lower risk of hypoglycemia in CAIP group
compared with Leuven group, the risk was still considerable.
In comparison with other studies [1,5,20,21] , the frequency
of hypoglycemia observed in this study was higher. The
incidence of hypoglycemia was 21.4% in the CAIP group
and 41.4% in the Leuven group, whereas in other studies,
which used the Leuven protocol, the incidence varied
between 5.1% and 18.7%. We believe that the higher
incidence of hypoglycemia observed in this study was
attributable to (1) the longer ICU stay of patients enrolled in
this study (median, 7 vs 3 days in the first study by Van den
Berghe et al [1]). Assuming that the risk of hypoglycemia is
relatively constant while patients are on an insulin drip,
patients in this study would have approximately double the
chance of experiencing an episode of hypoglycemia. In fact,
the mean proportion of hypoglycemic measurements per
patient was low in CAIP group (0.43%). (2) The eligibility
criteria in this study, with the need of one episode of
hyperglycemia and presence of SIRS or mechanical ventila-
tion, leads to the selection of patients with a higher risk of
hypoglycemia [8,22]. In fact, different samples will have a
remarkably different risk as evidenced by the 5.1% and
18.7% incidence of hypoglycemia in the intensive insulin
arms of the surgical and medical Leuven trials, respectively.
Therefore, estimates of hypoglycemic risk may not be
comparable between studies; instead, di fferent insulin
protocols should be submitted to a randomized comparison,
as we did. (3) The ICU that enrolled most of the patients in
this study had a low nurse-to-patient ratio and high staff
turnover (Hospital Mário Covas, Santo André, SP, Brazil).
Compliance with all protocols was sometimes inadequate, in
special at night (eg, omission of glucose measurements with
unchanged insulin infusion). Therefore, many hypoglycemic
episodes arose because of protocol noncompliance and might
have been avoided.
4.4. Interpretation of study findings and
clinical implications
The CAIP group had approximately half the incidence of
hypoglycemia (40 mg/dL) compared with the Leuven
group. We thin k t he following features o f CAIP reduc ed the
risk of hypoglycemia: (1) higher target range (100-130
mg/dL); (2) insulin infusion and next BG measurement are
indicated by the microcomputer software, which may
have facilitated its implementation; (3) insulin is adjusted
not only according to current BG level, but also according
to the rate of change since the last measurement, which
allows a smoother BG control. We believe that the
incidence of hypogl ycem ia observed with CAIP might be
further lowered by the following: (1) strict adherence to
protocol orders, (2) use o f arterial blood for glucose
measurement instead of fingerstick blood, and (3) use of
hemogasometers within ICU to mea sure BG inste ad of
point-of-care glucometers.
Good acceptance of the insulin protocol by the nursing
staff is critical for a smooth implementation of strict glucose
control [9]. We evaluated the nurses' perception regarding
the feasibility of the insulin protocols and also which of those
the nurses would prefer to use in their ICU. Between the 2
continuous insulin infusion protocols, the CAIP was
considered easier to use than the Leuven protocol. Also,
CAIP was considered as easy as the conventional protocol.
Most nurses chose CAIP as the protocol they would like to
see implemented in their ICUs. We believe that the easy use
of the computer program was an important determinant for
turning CAIP into a more practical protocol than Leuven.
Also, CAIP displays the exact dose of insulin to be infused;
therefore, as opposed to Leuven protocol, CAIP did not
require decisions to be made by the nurses, allowing them to
concentrate on other aspects of patient care.
Recently, several randomized controlled trials [2-5] have
had results conflicting with those obtained in the very
influential study of Van den Berghe et al [1]. Most of the
newer studies found a similar mortality with the strict or the
liberal glucose control strategies [2,3,5]. More remarkably,
NICE-SUGAR, the largest mu lti cen ter trial evaluatin g
intensive vs conventional glucose control in critically ill
patients, observed an increase in the risk of death at day 90th
associated with the i ntensive glucose control strategy
(81-108 mg/dL) compared with a conventional strategy
(b180 mg/dL) [4]. Our study did not aim and was not
powered to evaluate the effect of different BG goals on
patients' outcomes. However, based on the available
evidence, it is sensible to conclude that BG ranges below
180 mg/dL, but not very strict targets, are probably safe and
effective goals for critically ill patients.
We conclude that CAIP protocol may be used to achieve
glucose control within 100 to 130 mg/dL in critically ill
377Computer-assisted insulin infusion protocol for glucose control
Page 7
patients with a lower risk of hypoglycemia than the Leuven
protocol and may have better acceptance by the nursing staff.
4.5. Future research
In line with the available evidence showing that BG levels
just below 180 mg/dL are probably superior to very strict
levels, protocols such as CAIP may be adapted and tested to
target higher BG levels. On the other side, development of
glucose control devices, which integrate continuous or very
frequent glucose measurement with insulin infusion (closed
loop), may greatly improve glucose control in critically ill
patients while virtually avoiding hypoglycemia.
Acknowledgments
This study was conducted with financial resources from
research support no. 2005/50557-5 of Fundação de Amparo
a Pesquisa do Estado de São Paulo. Fundação de Amparo a
Pesquisa do Estado de São Paulo had no role on the study
design, conduction, data analysis, drafting the manuscript, or
decision to submit to publication.
Roche Diagnóstica Brasil kindly donated the glucometers
and test stripes used in this study.
We are indebted to the nurses for collecting BG and
complying with insulin protocols.
Appendix A. Supplementary data
Supplementary data associated with this article can be
found, in the online version, at doi:10.1016/j.jcrc.2009.05.005.
References
[1] Van den Berghe G, Wouters P, Weekers F, et al. Intensive insulin
therapy in the critically ill patients. N Engl J Med 2001;345:
1359-67.
[2] Brunkhorst FM, Engel C, Bloos F, et al. Intensive insulin therapy and
pentastarch resuscitation in severe sepsis. N Engl J Med 2008;358:
125-39.
[3] Preiser JC. Intensive glycemic control in med-surg patients (European
Glucontrol trial). Society of Critical Care Medicine's 36th Critical
Care Congress; 2007.
[4] Finfer S, Chittock DR, Su SY, et al. Intensive versus conventional
glucose control in critically ill patients. N Engl J Med 2009;360:
1283-97.
[5] Van den Berghe G, Wilmer A, Hermans G, et al. Intensive insulin
therapy in the medical ICU. N Engl J Med 2006;354:449-61.
[6] American College of Endocrinology and American Diabetes Associa-
tion consensus statement on inpatient diabetes and glycemic control.
Endocr Pract 2006;12:458-68.
[7] Dellinger RP, Carlet JM, Masur H, et al. Surviving Sepsis Campaign
guidelines for management of severe sepsis and septic shock. Intensive
Care Med 2004;30:536-55.
[8] Krinsley JS, Grover A. Severe hypoglycemia in critically ill patients:
risk factors and outcomes. Crit Care Med 2007;35:2262-7.
[9] Meijering S, Corstjens AM, Tulleken JE, et al. Towards a feasible
algorithm for tight glycaemic control in critically ill patients:
a systematic review of the literature. Crit Care 2006;10:R19.
[10] Pachler C, Plank J, Weinhandl H, et al. Tight glycaemic control by an
automated algorithm with time-variant sampling in medical ICU
patients. Intensive Care Med 2008;34:1224-30.
[11] Plank J, Blaha J, Cordingley J, et al. Multicentric, randomized,
controlled trial to evaluate blood glucose control by the model
predictive control algorithm versus routine glucose management
protocols in intensive care unit patients. Diabetes Care 2006;29:271-6.
[12] Vogelzang M, van der Horst I, Nijsten MW. Hyperglycaemic index as a
tool to assess glucose control: a retrospective study. Crit Care 2004;8:
R122-7.
[13] Finney SJ, Zekveld C, Elia A, et al. Glucose control and mortality in
critically ill patients. JAMA 2003;290:2041-7.
[14] Knaus WA, Draper EA, Wagner DP, et al. APACHE II: a severity of
disease classification system. Crit Care Med 1985;13:818-29.
[15] Vincent JL, Moreno R, Takala J, et al. The SOFA (Sepsis-related Organ
Failure Assessment) score to describe organ dysfunction/failure. On
behalf of the Working Group on Sepsis-Related Problems of the
European Society of Intensive Care Medicine. Intensive Care Med
1996;22:707-10.
[16] Matthews JN, Altman DG, Campbell MJ, et al. Analysis of serial
measurements in medical research. BMJ 1990;300:230-5.
[17] Finkielman JD, Oyen LJ, Afessa B. Agreement between bedside blood
and plasma glucose measurement in the ICU setting. Chest 2005;127:
1749-51.
[18] Kanji S, Buffie J, Hutton B, et al. Reliability of point-of-care testing for
glucose measurement in critically ill adults. Crit Care Med 2005;33:
2778-85.
[19] Pereira AJ, Cavalcanti AB, Correa TD, et al. Reliability of arterial,
capillary and venous point-of-care glucose measurements in the
intensive care unit setting: evaluation of two glucometers. Crit Care
2007;11:s56 [abstr].
[20] Cheung NW, Wong VW, McLean M. The hyperglycemia: intensive
insulin infusion in infarction (HI-5) studya randomized controlled
trial of insulin infusion therapy for myocardial infarction. Diabetes
Care 2006;29:765-70.
[21] Devos P, Preiser JC. Current controversies around tight glucose control
in critically ill patients. Curr Opin Clin Nutr Metab Care 2007;10:206-9.
[22] Vriesendorp TM, van Santen S, de Vries JH, et al. Predisposing factors
for hypoglycemia in the intensive care unit. Crit Care Med 2006;34:
96-101.
378 A.B. Cavalcanti et al.
Page 8
  • Source
    • "The majority of these systems focus on vitamin K antagonist dosing to be used in hospitals and community clinics in order to establish and maintain International Normalized Ratio (INR) in the therapeutic range for the prevention of thromboprophylaxis [7] [8] [9]. Also, many CDS systems for insulin therapy exist that give insulin dosing advice or support time of dosing to achieve glucose control in patients in intensive care units [10] [11] [12]. Although most of these systems have been shown to improve glycaemic control, in general the effects of CDS systems in drug monitoring and dosing on patient outcomes are heterogeneous and therefore inconclusive [6]. "
    [Show abstract] [Hide abstract] ABSTRACT: Objective: The rising incidence of type 2 diabetes mellitus (T2DM) induces severe challenges for the health care system. Our research group developed a web-based system named PANDIT that provides T2DM patients with insulin dosing advice using state of the art clinical decision support technology. The PANDIT interface resembles a glucose diary and provides advice through pop-up messages. Diabetes nurses (DNs) also have access to the system, allowing them to intervene when needed. The objective of this study was to establish whether T2DM patients can safely use PANDIT at home. To this end, we assessed whether patients experience usability problems with a high risk of compromising patient safety when interacting with the system, and whether PANDIT's insulin dosing advice are clinically safe. Research design and methods: The study population consisted of patients with T2DM (aged 18-80) who used a once daily basal insulin as well as DNs from a university hospital. The usability evaluation consisted of think-aloud sessions with four patients and three DNs. Video data, audio data and verbal utterances were analyzed for usability problems encountered during PANDIT interactions. Usability problems were rated by a physician and a usability expert according to their potential impact on patient safety. The usability evaluation was followed by an implementation with a duration of four weeks. This implementation took place at the patients' homes with ten patients to evaluate clinical safety of PANDIT advice. PANDIT advice were systematically compared with DN advice. Deviating advice were evaluated with respect to patient safety by a panel of experienced physicians, which specialized in diabetes care. Results: We detected seventeen unique usability problems, none of which was judged to have a high risk of compromising patient safety. Most usability problems concerned the lay-out of the diary, which did not clearly indicate which data entry fields had to be entered in order to obtain an advice. 27 out of 74 (36.5%) PANDIT advice differed from those provided by DNs. However, only one of these (1.4%) was considered unsafe by the panel. Conclusion: T2DM patients with no prior experience with the web-based self-management system were capable of consulting the system without encountering significant usability problems. Furthermore, the large majority of PANDIT advice were considered clinically safe according to the expert panel. One advice was considered unsafe. This could however easily be corrected by implementing a small modification to the system's knowledge base.
    Full-text · Article · Jun 2013 · Artificial intelligence in medicine
  • Source
    • "Fourteen studies evaluated medication dosing assistants, providing recommendations specific to drug dosing adjustments, such as insulin dosing or dosing advice for warfarin initiation [14,16-18,21,22,24,27,28,30,32,36,49,50]. These CCDSSs showed improvements in process of care outcomes in 9 of 14 studies (64%) [14,21,22,27,28,32,36,49,50], improvements in patient outcomes in 3 of 10 studies (30%) [32,49,50], and a negative effect on patient outcomes in 1 of 10 (10%) studies [49]. Many studies in the Medication Dosing Assistants section overlap with studies in the therapeutic drug monitoring and thus, are not the primary focus of this review. "
    [Show abstract] [Hide abstract] ABSTRACT: Acute medical care often demands timely, accurate decisions in complex situations. Computerized clinical decision support systems (CCDSSs) have many features that could help. However, as for any medical intervention, claims that CCDSSs improve care processes and patient outcomes need to be rigorously assessed. The objective of this review was to systematically review the effects of CCDSSs on process of care and patient outcomes for acute medical care. We conducted a decision-maker-researcher partnership systematic review. MEDLINE, EMBASE, Evidence-Based Medicine Reviews databases (Cochrane Database of Systematic Reviews, DARE, ACP Journal Club, and others), and the Inspec bibliographic database were searched to January 2010, in all languages, for randomized controlled trials (RCTs) of CCDSSs in all clinical areas. We included RCTs that evaluated the effect on process of care or patient outcomes of a CCDSS used for acute medical care compared with care provided without a CCDSS. A study was considered to have a positive effect (i.e., CCDSS showed improvement) if at least 50% of the relevant study outcomes were statistically significantly positive. Thirty-six studies met our inclusion criteria for acute medical care. The CCDSS improved process of care in 63% (22/35) of studies, including 64% (9/14) of medication dosing assistants, 82% (9/11) of management assistants using alerts/reminders, 38% (3/8) of management assistants using guidelines/algorithms, and 67% (2/3) of diagnostic assistants. Twenty studies evaluated patient outcomes, of which three (15%) reported improvements, all of which were medication dosing assistants. The majority of CCDSSs demonstrated improvements in process of care, but patient outcomes were less likely to be evaluated and far less likely to show positive results.
    Full-text · Article · Aug 2011 · Implementation Science
  • Source
    • "The number of clinics within studies varied from 1 to 66, with the majority being performed at a single centre (63%) [4-9,13-18,21-23, 29,30,32,38], and most involved academic centres (73%)45679,10,12,14,15,1819202122232426,29,31,32,343536373839. Financial support was provided by public funding in 16 studies [4,6,8,9,14,19,21,22,24,25,31,32,343536373839, private funding in eight studies [8,12,13,16,19,27,28,35363739] (four had both), and 13 studies [5,7,10,11,15,17,18,20,23,26, 29,30,33] did not report a funding source.Table 1 summarizes the effectiveness of all CCDSSs on TDMD and Additional file 4,Table S4 provides extensive outcome details. Overall, 60% of studies (18/30)45671011121319,21,24,26,29,30,33,3536373839 showed an improvement for process of care, and 21% (4/19) for patient outcomes [10,33,38,39]. "
    [Show abstract] [Hide abstract] ABSTRACT: Some drugs have a narrow therapeutic range and require monitoring and dose adjustments to optimize their efficacy and safety. Computerized clinical decision support systems (CCDSSs) may improve the net benefit of these drugs. The objective of this review was to determine if CCDSSs improve processes of care or patient outcomes for therapeutic drug monitoring and dosing. We conducted a decision-maker-researcher partnership systematic review. Studies from our previous review were included, and new studies were sought until January 2010 in MEDLINE, EMBASE, Evidence-Based Medicine Reviews, and Inspec databases. Randomized controlled trials assessing the effect of a CCDSS on process of care or patient outcomes were selected by pairs of independent reviewers. A study was considered to have a positive effect (i.e., CCDSS showed improvement) if at least 50% of the relevant study outcomes were statistically significantly positive. Thirty-three randomized controlled trials were identified, assessing the effect of a CCDSS on management of vitamin K antagonists (14), insulin (6), theophylline/aminophylline (4), aminoglycosides (3), digoxin (2), lidocaine (1), or as part of a multifaceted approach (3). Cluster randomization was rarely used (18%) and CCDSSs were usually stand-alone systems (76%) primarily used by physicians (85%). Overall, 18 of 30 studies (60%) showed an improvement in the process of care and 4 of 19 (21%) an improvement in patient outcomes. All evaluable studies assessing insulin dosing for glycaemic control showed an improvement. In meta-analysis, CCDSSs for vitamin K antagonist dosing significantly improved time in therapeutic range. CCDSSs have potential for improving process of care for therapeutic drug monitoring and dosing, specifically insulin and vitamin K antagonist dosing. However, studies were small and generally of modest quality, and effects on patient outcomes were uncertain, with no convincing benefit in the largest studies. At present, no firm recommendation for specific systems can be given. More potent CCDSSs need to be developed and should be evaluated by independent researchers using cluster randomization and primarily assess patient outcomes related to drug efficacy and safety.
    Full-text · Article · Aug 2011 · Implementation Science
Show more