Professional Documents
Culture Documents
Intensive Insulin Therapy in Intensive Care: An Example of The Struggle To Implement Evidence-Based Medicine
Intensive Insulin Therapy in Intensive Care: An Example of The Struggle To Implement Evidence-Based Medicine
Intensive Insulin Therapy in Intensive Care: An Example of The Struggle To Implement Evidence-Based Medicine
A
long with improving patients’ of ITT (see Text S1). We searched evidence for IIT does not yet support
safety and reducing medical for reasons why IIT had not been a grade-A recommendation (based
errors, one of the main implemented. We compared factors on the highest level of evidence), it
challenges in medicine is implementing that hindered implementation of IIT does appear to be stronger than the
new strategies that have the potential with factors hindering the adoption of evidence in support of a strategy of
to improve health outcomes. After the other recently introduced strategies, tolerating hyperglycemia [10]. Another
process of critically appraising clinical both in ICU medicine and general suggestion is just to target a BGC of
trials has finished, and the results medicine. over 150 mg/dl [9], or to reserve IIT
of this appraisal are used to guide solely for critically ill patients after
changes in clinical practice, it is then Current Recommendations on IIT elective surgery [11].
time to critically appraise the success of and Feared Complications One of the most frequently
implementation. Following publication of the first mentioned and feared complications
In other words, are physicians randomized controlled trial of IIT by of IIT is hypoglycemia. Indeed,
really performing the new strategy van den Berghe and colleagues [4],
in its entirety? If they are not, what several groups have recommended
are the barriers to implementation? IIT as the standard of care for those Funding: Marcus J. Schultz is supported by a
Unfortunately, there is no “golden personal grant from the Netherlands Organization for
who are critically ill. These groups Health Research and Development; NWO-VENI grant
bullet” for successful implementation include the Joint Commission 2004 (project number 016.056.001)
of new strategies in medicine [1,2]. on Accreditation of Healthcare
Competing Interests: The authors have declared
However, common factors in the Organization (http:⁄⁄www.jcaho. that no competing interests exist.
failure of implementation have been org), the Institute for Healthcare
identified, including environmental Citation: Schultz MJ, Royakkers AANM, Levi M,
Improvement (http:⁄⁄www.ihi.org), and Moeniralam HS, Spronk PE (2006) Intensive insulin
factors and factors related to the the Volunteer Hospital Organization therapy in intensive care: An example of the struggle
strategy itself [3]. (http:⁄⁄www.vha.com). In addition, IIT to implement evidence-based medicine. PLoS Med
3(12): e456. doi:10.1371/journal.pmed.0030456
Critically ill patients without diabetes is promoted as a part of a care bundle
often develop hyperglycemia. Until for sepsis by the American Thoracic Copyright: © 2006 Schultz et al. This is an
recently, it was common practice to Society (http:⁄⁄www.thoracic.org) and open-access article distributed under the terms
of the Creative Commons Attribution License,
treat only marked hyperglycemia in experts in the field [8]. Also, IIT has which permits unrestricted use, distribution, and
these patients, since hyperglycemia was become, to some extent, a benchmark reproduction in any medium, provided the original
considered to be an adaptive response author and source are credited.
for the quality of ICU care [9].
to critical illness. But clinical trials have However, over the last few years, Abbreviations: BGC, blood glucose concentrations;
shown that so-called intensive insulin a number of commentators have ICU, intensive care unit; IIT, intensive insulin therapy;
MeSH, medical subject heading; rh-APC, recombinant
therapy (IIT) aiming at normoglycemia expressed concern about the human-activated protein C
(i.e., blood glucose concentrations applicability of van den Berghe and
[BGC] between 80–110 mg/dl) can colleagues’ findings [4,5] to other Marcus J. Schultz is in the Department of Intensive
Care Medicine and the Laboratory of Experimental
significantly decrease mortality and settings [9–11]. These concerns include Intensive Care and Anesthesiology, Academic
morbidity of patients in the surgical the relatively high mortality in relation Medical Center, University of Amsterdam,
and medical intensive care unit (ICU) Amsterdam, Netherlands and the Hermes Critical
to severity of illness among patients in Care Group, Amsterdam, Netherlands. Annick A. N.
[4–7]. the control group in one study [4]; the M. Royakkers is in the Department of Intensive Care
We questioned whether IIT truly frequent administration of parenteral Medicine, Tergooi Hospitals, Blaricum, Netherlands.
Marcel Levi is in the Department of Internal Medicine,
has become part of standard therapy calories (which is unusual among most Academic Medical Center, University of Amsterdam,
in ICU patients and, if it is applied, ICUs); the single-center design of the Amsterdam, Netherlands. Hazra S. Moeniralam is
to what extent? We performed a two studies [4,5]; and the fact that the in the Department of Intensive Care Medicine, St.
Antonius Hospital, Nieuwegein, Netherlands. Peter
systematic search of the medical investigators could hardly be blinded. E. Spronk is in the Department of Intensive Care
literature, in which we focused on The results of two much larger trials Medicine, Academic Medical Center, University of
surveys and reports on the practice Amsterdam, Amsterdam, Netherlands; the Hermes
are awaited (the GLUControl trial Critical Care Group, Amsterdam, Netherlands; and
[12] and the NICE-SUGAR trial [13]). the Department of Intensive Care Medicine, Gelre
In the mean time, different experts Hopsitals, Lukas, Apeldoorn, Netherlands.
The Essay section contains opinion pieces on topics
of broad interest to a general medical audience.
give different recommendations: * To whom correspondence should be addressed.
some argue that although the E-mail: m.j.schultz@amc.uva.nl
Krinsley et al. [6,41] 2004/2005 Before–after cohort 1,600 Less than 140 “Not changed” Safe
Kanji et al [21] 2004 Before–after cohort 100 Nurses 80–110 16%
Grey et al. [42] 2004 Randomized 61 80–120 32%
controlled trial
Zimmerman et al. [43] 2004 Prospective cohort 342 Nurses 80–150 7%
Laver et al. [44] 2004 Prospective cohort
Goldberg et al. [45] 2004 Prospective cohort 118 Nurses 100–140 0.2% Safe
Goldberg et al. [46] 2004 Prospective cohort 52 Nurses 100–140 0.3% Safe
Ku et al. [47] 2005 Before–after cohort 156 Nurses Safe
Thomas et al. [48] 2005 Before–after cohort 891 Safe
Chant et al. [49] 2005 Before–after cohort 86 Nurses 90–140 0.2%–0.4% Safe
Bland et al. [50] 2005 Randomized 10 Nurses “Rare” Safe
controlled trial
Moeniralam et al. [51] 2005 Before–after cohort 7,327 Nurses and 80–140 3.3%–4.0% Safe
physicians
Taylor et al. [22] 2006 Before–after cohort 281 Physicians and 120–150, 80–110 1.1%–3.4% Safe
nurses
5.1% of patients treated with IIT in range [IQR] 162–216 mg/dl). The (MJS, PES, and HSM) [19]. Over 100
surgical intensive care versus 0.7% reported median clinically important participants of the annual meeting
of control patients developed severe threshold for hypoglycemia was 72 of the Dutch Society of Intensive
hypoglycemia (BGC is defined as mg/dl (IQR 54–72 mg/dl). ICU nurses Care were surveyed, most of them
less than 40 mg/dl) [4]. In medical acted on slightly but significantly ICU physicians. Of the participants,
ICU patients, severe hypoglycemia higher thresholds than ICU physicians 69% stated that IIT was already
occurred even more often with IIT: (a difference of 9 mg/dl). being applied in their ICU, while 7%
18.7% of study patients versus 3.1% Avoidance of hyperglycemia was mentioned they would start with this
in the conventionally treated group judged most important for patients intervention shortly. Of those that said
encountered severe hypoglycemia with diabetes, a recent seizure, they applied IIT in their ICU, 62%
[5]. Of note, the recent multicenter advanced liver disease, or acute used some sort of intensive insulin
VISEP trial in Germany by the SepNet myocardial infarction. Surprisingly, protocol with sliding scales. Twenty-
group was discontinued prematurely avoiding hyperglycemia was judged six percent stated that their ICU used
because of identical mortality rates in unimportant for surgical patients—the blood-glucose-concentration limits of
the treatment group and in the control targeted patients in van den Berghe 80–110 mg/dl, 73% stated that their
groups but a higher incidence of and colleagues’ pivotal study on IIT in ICU used limits of 80–145 mg/dl, and
hypoglycemia in the IIT group (12.1% patients in the ICU [4]. In McMullin 2% stated that their ICU used limits
versus 2.1%) [14]. Patients in the ICU and colleagues’ paper [16], the of 80–180 mg/dl. Eighteen percent
who were sedated and patients with authors gave no information regarding of respondents said that glycemic
disturbances in the counter-regulatory presumed risks of IIT, in particular the control was applied solely by ICU
responses to hypoglycemia are at risk risk for hypoglycemia and the impact of nurses, 16% said that it was applied by
for neuroglycopenia because of the this risk on the chosen BGC thresholds. ICU physicians alone, and 65% said
absence of clinical symptoms of severe Mackenzie and colleagues recently that it was applied by ICU nurses and
hypoglycemia. Neuroglycopenia may reported a survey on the use of IIT ICU physicians as a team (1% did not
cause cerebral damage, epileptic in large English hospitals [17]. Only answer the question).
insults, or even coma [15]. 25% of ICUs reported blood-glucose- Recently, the Australian and New
concentration targets to be similar Zealand Intensive Care Society
Current Practice of IIT to those used in the study by van Clinical Trials Group (ANZICS-CTG)
Surveys. McMullin and colleagues den Bergh and colleagues [4]. Most conducted a practice survey [20].
surveyed ICU nurses and ICU ICUs in which IIT was performed There were 45 affiliated ICUs that
physicians on the blood-glucose- reported higher normal blood-glucose- were E-mailed a blood-glucose survey,
concentration thresholds that they concentration limits. Interestingly, enquiring as to their familiarity with
acted upon in five university-affiliated most ICU nurses (82%) reported being the van den Berghe and colleagues’
multidisciplinary ICUs in Canada [16]. afraid of hypoglycemia in the patients studies [4,5] and whether IIT had been
The reported clinically important receiving IIT [18]. adopted. If IIT had been adopted,
threshold for hyperglycemia was Mackenzie and colleagues’ findings respondents were asked to which
remarkably high. Indeed, median are partly in line with a recent survey groups of patients IIT was applied and
threshold was 180 mg/dl (interquartile in the Netherlands by three of us the reasons for such selection. If IIT
Intensive insulin therapy Patients in the ICU Poor recognition of target groups, uncertainties on who is Risk for (neuro)-hypoglycemia
to apply the strategy in daily practice, concerns about the
external validity of studies
rh-APC Patients in the ICU Poor recognition of target groups, costs of the strategy, Risk for bleeding
concerns about the validity of the study
Lung-protective mechanical Patients in the ICU Poor recognition of target groups, incorrect translation of Risk for hemodynamic side effects or increased
ventilation using lower tidal calculation of ideal tidal volume for individual patients need for sedation
volumes
Vitamin-K antagonists Patients with atrial fibrillation Risk for bleeding
Lipid-lowering drugs Patients with an Insufficient knowledge regarding recent insights in the
atherothrombotic event effectiveness of this treatment
had not been adopted, respondents than 72 mg/dl as a threshold for Discussion
were asked their reasons for failure to hypoglycemia, incidences were 32% A systemic approach to the
adopt this strategy. Sixty-four percent and 29%, respectively. However, since implementation of research evidence
of ICUs responded to this survey; all the reliability of capillary blood-glucose in daily practice is recommended.
were familiar with the studies on IIT, measurements (blood obtained from a
Indeed, before an intervention
but only 10.3% had adopted IIT in all finger stick) are unsatisfactory (there
is implemented, different phases
their patients. In 31% of responding is a high degree of imprecision and a
of accumulating evidence with
centers, IIT was applied in selected high percentage of discordance [23])
respect to the intervention should
patient groups, predominantly those in many of the reviewed studies, the
be followed. A framework for the
that stayed in the ICU for over three incidence of hypoglycemia may have
implementation of research evidence
days, those with sepsis, and postsurgical been higher or lower than reported. In
that leads to understanding of barriers
patients. The reasons for not applying most reports, IIT was considered a safe
and opportunities involved in the
IIT were due to concerns about the risk strategy.
implementation of protocols or
of hypoglycemia and concerns about Personnel involved in intensive
guidelines in health care has been
the external validity of the two studies insulin therapy. Although it was not
proposed [2,24].
by van den Berghe and colleagues. always clearly stated in the papers,
Identifying barriers for IIT is an
Although these surveys may only be it seems that ICU nurses were the
an incomplete reflection of practice primary health-care workers involved important part of the process of
throughout the world, the striking in the application of the IIT protocol its implementation [1]. Fears and
similarities between their results at least (Table 1). Only one study compared barriers should be catalogued and
suggest that IIT is far from being part of an ICU nurse-driven IIT protocol with rationalized. In fact, implementation
the standard care of critically ill patients. a protocol applied by ICU physicians of complex strategies requires a
Targets of intensive insulin therapy. alone [22]. In this cohort study, three thorough social investigation before
Only two of the identified studies consecutive regimens were compared: such strategies will be applied in daily
[21,22] used BGC targets identical IIT applied by ICU physicians with practice. Considering IIT, we now
to those used in the two studies by no specific targets, IIT applied by recognize several hampering factors:
van den Berghe et al. [4,5] (Table ICU nurses aiming at BGC between concerns about the external validity
1). All other studies used different 120–150 mg/dl, and IIT applied by of the two studies by van den Berghe
BGC thresholds, most of them with a ICU nurses aiming at the BGC used by and colleagues [4,5], the potential
higher upper limit (up to 150 mg/dl). van den Berghe and colleagues [4,5]. increased risk of hypoglycemia with
Of interest, most studies found that There was a significant decrease in associated neurological damage, and
higher BGC limits were deliberately average daily BGC, from 190 to 163 uncertainties on how (and who is) to
chosen to facilitate acceptance of the to 131 mg/dl in the three consecutive apply and monitor IIT.
protocol (i.e., because it was suspected phases of the study. The incidence of Factors hampering uptake of other
that there would be an unacceptably severe hypoglycemia (defined as blood ICU treatments. The hampering factors
high incidence of hypoglycemia when glucose concentration less than 40 for the implementation of IIT in
applying the limits used by van den mg/dl) was similar between the groups, intensive care medicine are not unique,
Berghe and colleagues [4,5]). ranging between 1.1% and 3.4%. but are comparable to other strategies,
Incidence of hypoglycemia. The Remarkably, protocol compliance was both in intensive care medicine and in
incidence of hypoglycemia varied reported to be low (only about 50% of other medical specialties. This is nicely
from as low as 0.5% to as high as orders were followed), and blood for illustrated by the implementation
18.7%, when using the threshold BGC monitoring was at times obtained processes of several strategies in the
of 40 mg/dl (Table 1). When using from a finger stick, which may be ICU in the last decade, such as the
BGC less than 60 mg/dl and less unreliable, as explained above. use of recombinant human-activated