Review
Insulin Infusion Therapy in Critically Ill Patients

https://doi.org/10.1016/j.jcjd.2014.01.016Get rights and content

Abstract

While dysglycemia (hyperglycemia, hypoglycemia and glucose variability) is clearly associated with increased mortality in critically ill patients, target range of blood glucose control remains controversial. Standardized insulin infusion protocols constitute the basis of treatment of these patients. The choice of protocol and its implementation is a great challenge. In this article, we review the published data to help define the essential elements that compose a good protocol and apply the right conditions to make it safe and effective.

Résumé

Bien que la dysglycémie (hyperglycémie, hypoglycémie et variabilité glycémique) soit de toute évidence associée à l'augmentation de la mortalité chez les patients gravement atteints, l'intervalle cible de la maîtrise de la glycémie demeure controversé. Les protocoles standardisés de perfusion d'insuline constituent la base du traitement de ces patients. Le choix du protocole et sa mise en œuvre constituent un défi majeur. Dans cet article, nous passons en revue les données publiées pour aider à définir les éléments essentiels qui composent un bon protocole et à appliquer les bonnes conditions pour le rendre sûr et efficace.

Introduction

Dysglycemia, which refers to hyperglycemia, hypoglycemia and glucose variability (GV), frequently occurs in critically ill patients regardless of their diabetes status (1). The association between these glucose dysregulations and morbidity and mortality in intensive care units (ICUs) is well documented, but there is still a great deal of debate about their involvement as biomarkers rather than causative factors of adverse outcomes (2).

The landmark study by van den Berghe and colleagues in 2001 (3), referred to as the surgical Leuven 1 study, brought about a major shift in our approach to hyperglycemia in critically ill patients. It was the first evidence that demonstrated a mortality benefit of tight glycemic control (TGC) in the ICU. New guidelines for treatment of hyperglycemia in the ICU targeting the “normal range” of blood glucose (BG) of 4.4 to 6.1 mmol/L followed that study (4). Prior to that publication, the treatment of hyperglycemia in ICUs was not considered a priority and was often initiated only when BG values exceeded the renal threshold for glycosuria (5).

It is well established that hyperglycemia is associated with increased morbidity and mortality in ICUs (6). The prevalence of diabetes in ICUs is probably 25% or higher, but most cases of hyperglycemia in this setting occur in patients without known diabetes (1). These hyperglycemic patients have stress-induced hyperglycemia or undiagnosed diabetes and have higher in-hospital mortality than patients with known diabetes (7).

Since studies by the Leuven group showed positive results for TGC in surgical (3) (Leuven 1), medical (8) (Leuven 2) and pediatric ICUs (9) (Leuven 3), several attempts to reproduce those results have failed to show any benefits. Hypoglycemia appeared to be problematic, and 2 trials were terminated prematurely due to an unacceptable rate of hypoglycemia 10, 11. The larger of these attempts, the Normoglycemia in Intensive Care Evaluation and Surviving Using Glucose Algorithm Regulation (NICE-SUGAR) Study, is another landmark randomized controlled trial (RCT) that was conducted in 41 multinational mixed medical/surgical ICUs, aiming at the tight target range of 4.5 to 6.0 mmol/L and resulting in an increased mortality rate in the TGC group (12). In a recent analysis of the hypoglycemic events in this study, the authors showed a correlation between the severity of the hypoglycemia and the risk for death, but causality could not be proved (13). The fear of iatrogenic hypoglycemia brought about by this study was instrumental in changing recent guidelines for hyperglycemia in the ICU (14).

Iatrogenic hypoglycemia must be distinguished from spontaneous hypoglycemia because the latter is more clearly associated with mortality (15). Despite the increased rate of hypoglycemia in the TGC groups in the Leuven studies, there was a decrease in mortality. The NICE-SUGAR study was the only TGC RCT in which there was an increase in mortality. Moreover, the greatest hazard ratio for death in this study was observed in patients with severe hypoglycemia in the absence of insulin treatment (13).

Glucose variability (GV) is the third aspect of dysglycemia that has been shown to correlate with mortality in the ICU (16). Different glucometrics have been used, such as standard deviation (SD) of the mean BG level (17) and mean absolute glucose change (18), to demonstrate that GV is an independent predictor of mortality in critically ill patients. Two recent large retrospective analyses have shown that high GV was independently associated with higher mortality in the non-diabetes mellitus cohort but not in the diabetes cohort, although severe hypoglycemia correlated with increased mortality in both cohorts of patients 19, 20. Also, an analysis of the Leuven studies revealed that the decline in mortality in the TGC group was associated with decreased mean BG levels, but there was no decrease in GV (21).

The optimal target range for BG in the treatment of hyperglycemia in critically ill patients has shifted markedly over the past 15 years; the changes have been driven mainly by the landmark studies mentioned above. Retrospective studies have shown that the relation between mean BG values and mortality and postoperative infections follows a U-shaped curve in hospitalized patients 22, 23, 24, and a safe range of BG in ICUs has been defined as being between 7 and 9 mmol/L (24). Interestingly, patients with established diabetes have a curve slightly shifting to the right, with a blunted relationship between hyperglycemia and death (5).

Several meta-analyses of the effects of TGC in the ICU have been published over the past 5 years 25, 26, 27, 28, 29, 30, 31, 32. None found a significant benefit in terms of mortality. Of these 8 analyses, 4 found a decreased risk for morbidity in surgical patients 26, 29, 31, 32. However, in all settings, TGC increased hypoglycemia by 3 to 8 times compared with conventional controls. Substantial disparities exist among these studies, and much has been written in an attempt to reconcile the results, especially when comparing the landmark studies of Leuven and NICE-SUGAR 2, 5, 33.

As illustrated in Table 1, several methodological differences are present between the first Leuven study and the NICE-SUGAR study, which resulted in opposite conclusions concerning the benefit of targeting normal values of BG in ICUs. The first major difference concerns the environment of the studies. The Leuven 1 study was performed in a single centre in Leuven (Belgium) with experienced nurses who were well versed in the use of their intuitive paper protocol. In the NICE-SUGAR study, the cases were spread over 41 centres and treated by a heterogeneous group of nurses who had had much less exposure to a new computerized protocol. Surgical patients were studied in the Leuven 1 trial, compared to a mixed population of surgical and medical patients in the NICE-SUGAR study. It is interesting to note that in the Leuven 2 trial, in which medical patients were studied, TGC significantly reduced morbidity but reduced mortality only in a subset of subjects treated for 3 days or more in ICUs (8). Less than 50% of patients in the TGC group in the NICE-SUGAR study reached the target, compared to 70% in the Leuven trial. Also, the target range for the control group was higher in the Leuven trial (10 to 11.1 mmol/L) than in the NICE-SUGAR trial (8 to 10 mmol/L). The resulting gap in BG values for the treated and control groups was narrower in the NICE-SUGAR trial, with >50% overlap between the 2 groups compared to <10% in the Leuven study. Consequently, this study cannot be considered to replicate the Leuven study (5).

The ways in which the BG levels were measured in both studies constitute another major difference. Indeed, a blood-gas analyzer was used with arterial blood samples in Leuven, while various blood glucose meters were used in NICE-SUGAR, with samples from a mixture of arterial, venous and capillary sites. Point-of-care (POC) glucose meters are not accurate enough and are not designed to cope with the numerous interferences affecting the measurement of BG in the ICU 33, 34. Major inaccuracies in BG values in the NICE-SUGAR trial have subsequently been reported at a participating site in Canada that could result in unacceptable glucose fluctuations and hypoglycemia (35). Moreover, BG levels can vary quite significantly if measured by finger-stick capillary specimens compared with arterial or venous specimens, especially during a hypoperfusion state in an ICU (34). A recent systematic review confirmed that BG measures were more accurate using arterial blood than using capillary blood (36).

Feeding strategies in the ICU reveal a third major difference between these 2 broad trials. In the Leuven study, early parenteral nutrition was used in most patients, whereas feeding relied almost exclusively on the enteral route in the NICE-SUGAR study (37). To explore the best feeding strategy in the ICU, van den Berghe's group designed a multicentre trial with patients randomly assigned to either early (<48h) or late initiation (>7 days) of parenteral nutrition and treated with Leuven's usual TGC protocol (38). The results indicated that postponing parenteral nutrition for 1 week did not affect mortality but markedly reduced morbidity and accelerated recovery. Surprisingly, this study suggests that patients were not treated optimally from a nutrition standpoint in the Leuven studies, and the resulting morbidities could have been reduced, causing an even more positive outcome compared to the outcome of the NICE-SUGAR study. On the other hand, early feeding in the NICE-SUGAR context could possibly decrease the risk for hypoglycemia and GV and the associated morbidities. Therefore, the interplay between feeding strategies and BG control is complex and requires further studies to determine the best overall protocol to follow in the ICU.

As we can see from these analyses, it is difficult to determine optimal targets of treatment in view of such disparities. Targeting normal values of 4.4 to 6.1 mmol/L was proven to be effective in reducing morbidity and mortality in the unique context of the Leuven studies. However, these conditions have been very difficult to reproduce, for the various reasons explained. The NICE-SUGAR settings correspond more closely to real-life experience in our institutions, with practical shortcomings related to inaccurate tools for monitoring BG and deficiencies in the design and implementation of the intensive insulin protocol (IIT). In this context, it seems more reasonable to aim for a moderate target range of BG similar to the range for the conventional group in the NICE-SUGAR trial (8 to 10 mmol/L) as a compromise, to avoid the deleterious effects of hypoglycemia and GV (5).

Most medical organizations have adopted this target for glucose control in ICUs in their most recent guidelines 14, 39, 40, 41, 42. Two of these guidelines also propose more stringent BG control in certain subpopulations. The Society of Thoracic Surgeons proposes a more stringent upper limit of 8.3 mmol/L following 3 days in an ICU after cardiac surgery for patients with comorbidities (40). For the American Diabetes Association, lower BG targets (6.1 to 7.8 mmol/L) may be appropriate in selected patients, but strong evidence is lacking (14). Guidelines from the Society of Critical Care Medicine also reached a consensus glycemic goal range of 5.6 to 8.3 mmol/L to avoid excursion of BG >10 mmol/L, which has been clearly associated with an increased risk for death, length of stay and infection in an ICU 16, 34. They set their upper limit at 8.3 mmol/L, particularly for patients with cardiac surgery and patients with trauma, based on evidence of morbidities in these patients (34). In a recent international multicentre cohort study of patients admitted to 23 ICUs, BG >7.8 mmol/L was associated with an increased risk for mortality in patients without diabetes but not in patients with diabetes, where even values of BG >10 mmol/L were not related to increased mortality (19). Moreover, diabetes was independently associated with decreased mortality in the entire cohort. Therefore, patients with diabetes seem to behave differently from patients without diabetes and may benefit from higher target ranges of BG. Despite the controversy, there is general consensus that a BG level of 10 mmol/L constitutes the absolute upper limit and that hypoglycemia (BG <4 mmol/L) should be avoided. The upper limit could be lowered in certain subpopulations if the adjustments are made safely without hypoglycemia.

The disparities in the results of the TGC trials have demonstrated abundantly that an effective and safe IIT is much more than a simple order set. There are many practical elements involved, including existing local technical and human resources that can make or break an insulin infusion protocol and its implementation. Failure to implement the protocol and replicate the results of the Leuven group at other medical centres is a good example. Because the Leuven protocol is based on simple rules with a high level of intuitive decision making by users, it is difficult to reproduce. Recently, a group from the Netherlands (8 ICUs) undertook to reproduce the Leuven protocol closely by investing heavily in training their staffs, with visits to Leuven and gradual implementation (43). They succeeded in obtaining a median BG level of 5.8 mmol/L with an initial increase in the hypoglycemia rate that improved over time, confirming the importance of the environment and well-trained staff when implementing a protocol.

The Society of Critical Care Medicine did an extensive review of the literature to draw up recommendations concerning the essential components of an insulin infusion system (34). Likewise, the Society of Hospital Medicine offers its recommendations with examples of insulin infusion protocols on its website (http://www.hospitalmedicine.org/ResourceRoomRedesign/GlycemicControl.cfm). A proper protocol should make it possible to set different BG target ranges. As discussed previously, a range of 8 to 10 mmol/L may be appropriate in most situations in an ICU, but tighter ranges may be considered for certain situations, such as after cardiac surgery. If the protocol is to be used outside an ICU, different target ranges can be used for pregnancy, diabetic ketoacidosis or other settings, according to the pathology or the local resources available. There should be clear instructions about the BG threshold for initiating insulin infusion and the initial rate. A validated algorithm for subsequently adjusting the rate should be based not only on the current BG level but also on the dynamic change of BG levels over time. This makes it possible to reach the target faster if correction of hyperglycemia is too slow or to avoid hypoglycemia if the correction is too fast. A concentration of 1 unit/mL of insulin for the infusion is recommended to limit volume intake in an ICU along with priming new tubing with at least 20 mL of waste volume to avoid adsorption of insulin to the IV tubing (34). Accurate insulin administration requires a reliable infusion pump that can deliver insulin dose in increments of 0.1 unit per h.

Monitoring BG during IIT is crucial, especially with tight BG target ranges. It is suggested that BG levels be monitored every 1 to 2 h, with more frequent readings in unstable patients. Many protocols allow testing every 4 h once BG values have been stabilized in the target range, but routine monitoring at a frequency of 4 h is discouraged because the rates of hypoglycemia increase above 10% in many of these protocols (34). Studies have shown the superiority of blood gas analyzers compared to POC glucose meters for measuring BG levels in the ICU 33, 34, 44. The 1987 Clarke error grid, used to describe the clinical accuracy of glucometers in outpatient settings, is unsuitable in an ICU. The US Food and Drug Administration is currently establishing tighter performance requirements for BG monitors used in hospitals (33). However, POC glucose meters remain the most popular instruments for monitoring BG levels in ICUs in terms of time and money. Additional consideration should also be given to the sampling site of the blood. As mentioned previously, arterial blood from indwelling catheters should be the primary source of samples for measuring glucose levels, and venous blood should be used as an alternative source. Capillary samples should be used only in patients whose severity of illness does not justify invasive vascular monitoring 33, 34, 44.

Frequent BG monitoring according to the recommendations is time-consuming for nurses and constitutes a major barrier in the proper implementation of an insulin infusion protocol. For this reason, continuous glucose monitoring (CGM) has been explored to reduce the burden and increase the safety of IIT (44). Initial reports of continuous monitoring with interstitial glucose sensors have shown acceptable accuracy 33, 44. Intravascular BG monitoring by devices using various techniques based on microdialysis, absorption spectroscopy, optical scattering or fluorescence is currently being studied in ICUs. The first studies involving the Eirus microdialysis system (Maquet Critical Care, Solna, Sweden) and the OptiScanner (Optiscan Biomedical, Hayward, CA), based on infrared spectroscopy, have shown very promising results, with superior accuracy compared to interstitial CGM 45, 46. These CGM systems have potential for avoiding hypoglycemia by trend detection, optimizing insulin titration, reducing nurses' workloads and making it possible to analyze new parameters of glucose regulation such as GV and glucose complexity.

Procedures for treating hypoglycemia are an important component of insulin infusion protocols. Clear indications should be given to stop the infusion when BG <4 mmol/L and to administer 50% dextrose intravenously, titrated according to the hypoglycemic value, in order to avoid rebound hyperglycemia and a rise in GV (34). BG testing should be repeated every 15 m, with further dextrose administration until hypoglycemia is corrected. Nutritional considerations can have an important effect on glycemic control and can have a major influence in terms of insulin requirements. Providing constant carbohydrate calories simplifies glycemic management. Guidance for handling situations in which caloric intake could be interrupted, such as “field trips” to the operating room or for imaging studies, should be incorporated into the order set to avoid hypoglycemia. The transition from insulin infusion when patients are stable and start eating deserves special instructions in the protocol. It is suggested that all patients with type 1 diabetes, patients with type 2 diabetes who have an insulin infusion rate >0.5 unit/h and patients without diabetes who have an insulin infusion rate >1 unit/h would benefit from transition to a basal-nutritional-correction subcutaneous insulin regimen (34). Insulin infusion should be stopped only after the initiation of subcutaneous insulin to avoid rebound hyperglycemia. There are various ways to determine the total daily dose of insulin necessary for the transition based on the last hours of IV insulin infusion, and guidance should be given in the protocol to this effect (47).

An excellent validated insulin titration program is no guarantee of effective and safe control of glycemia in the ICU unless implementation is carefully planned. The insulin infusion system must be tailored to local resources and the subset of patients treated. The availability of appropriate hospital staff, accurate monitoring technology and continuous assessment of glucometrics are essential aspects of successful implementation (34). Team education is of paramount importance, and great efforts should be invested in teaching the nurses, who bear the burden of carrying out the protocol (43). Explaining the rationale behind the importance of glycemic control, the practicalities of the protocol, the situations that demand drastic changes in insulin requirements, and when alerting a physician is appropriate should help in overcoming barriers to executing the protocol. It is also essential to get feedback from the staff and organize regular sessions during the early phase of implementation so as to clarify and adjust the protocol accordingly.

There are many published insulin infusion protocols, and the adoption of protocols with demonstrated safety and efficacy is highly recommended (14). However, there is not a single proven ideal protocol that would fit every ICU. As explained above, a validated protocol should be chosen and customized according to local resources and patient characteristics. Protocols vary in insulin dosing intensity and complexity, and comparison is made more difficult by differing glucometrics, BG target ranges, monitoring methodologies and the populations studied. There are currently no data from a prospective RCT comparing the impact of various types of glucose control protocols on morbidity and mortality. However, available evidence indicates that static control algorithms dictating insulin delivery rate based solely on the last glucose measurement should be discouraged (48). Dynamic-control algorithms, based on the last 2 BG levels and rate-of-change of BG levels, can be classified in 2 broad categories: infusion rate (IR) algorithms and maintenance rate (MR) algorithms 49, 50.

In IR algorithms, the new insulin infusion rate is the result of a direct adjustment of the previous IR and sometimes requires additional calculations or qualitative assessments by the nursing staff. These are often referred to as learning algorithms and are possibly more flexible (50). The Leuven 3, 8, Portland (51) and Yale (52) protocols belong to this class of algorithms. Table 2 shows characteristics and some glucose metrics from these studies. The Leuven protocol, discussed above, includes a high level of intuitive decision making by users. The Portland protocol, used for patients after cardiac surgery and revised many times since 1992, is a complex protocol with a combination of fixed and relative adjustments of the IR. The most recent versions with various target ranges can be downloaded from the Portland Diabetic Project site. The Yale protocol has been published, showing results in both medical and surgical ICUs, and has also been revised to adapt the BG target range as the guidelines continue to evolve (53). It uses 2 tables to help the nurse identify the desired change of IR and convert the change into a new IR.

The MR algorithms seek to define a maintenance rate, also referred as a multiplier or column, according to the patient's insulin sensitivity. The paper-based protocols use multiple columns (sliding scales) corresponding to different multipliers or levels of sensitivity to insulin. The designation of the multiplier (column) used to calculate the new IR is determined according to the rate of change in BG levels (49). These algorithms have potential for improved adaptation to insulin resistance, making it possible to achieve target BG levels rapidly for all levels of insulin sensitivity (50). Moreover, the burden of interpretation of these protocols by nursing staff may be less than that of the IR protocols (49). The first published paper based on this concept used 5 columns of increasing insulin resistance in cardiac surgery patients (54). In the North Carolina protocol, Braithwaite et al pushed the concept with a 6-columns algorithm for trauma patients (55). Details of these 2 protocols are shown in Table 2.

These paper-based protocols are quite complex, and they are nurse-dependent for interpretation and application. Both types of algorithms have been translated into mathematical equations, with several benefits. Indeed, they reduce the number of parameters needed to determine the IR of insulin, from more than 50 in certain paper-based protocols to 5 or fewer in computer-based protocols (50). With fewer parameters, they are more adaptable to differing subsets of patients in an ICU and to local resources. Simplicity of use makes them less prone to error. In a crossover study in which a computer-based protocol was compared to the same paper-based protocol in simulated scenarios for ICU nurses, there were significantly fewer errors in the titration and transition phases of the protocol (56).

The IR algorithms have been translated into various equations. The proportional integral derivative computer protocol has only 4 parameters to define basic responses. It was used recently in an RCT targeting normoglycemia in a pediatric cardiac ICU compared to standard care (57). TGC was achieved with low hypoglycemia, but neither morbidity nor mortality changed significantly (Table 2). The enhanced Model Predictive Control (eMPC) is another computerized protocol of the IR type that uses advanced control algorithms designed by the Closed Loop Insulin Infusion for Critically Ill Patients (CLINICIP) consortium. That protocol was compared to the Leuven protocol in a recent RCT that also targeted normoglycemia in a mixed ICU. The eMPC improved efficacy of TGC without increasing the hypoglycemia rate (58). There are also 2 commercial electronic protocols based on IR algorithms: the EndoTool from Monarch HealthCare (Irvine, CA) and the GlucoCare from Pronia Medical Systems (Louisville, KY). In an RCT including 300 patients in an ICU, the EndoTool protocol was compared with the Portland protocol and resulted in a higher percentage of BG levels in the target range, reduced GV and produced higher satisfaction rates for nurses (59). The computerization of the Yale protocol within the GlucoCare system produced a very low incidence of hypoglycemia, and those episodes that were reported were caused mainly by the decreased frequency of BG testing and by staff failing to comply with the protocol (60).

Paul Davidson translated the MR algorithm into a simple equation: Insulin/Hour = Multiplier × (BG–60), where BG is in mg/dL and Multiplier is the factor of insulin resistance. He devised the Glucommander computer program, which adjusts the multiplier to reach target glucose. This program has been tested extensively in differing environments with excellent results, and many centres have adapted the original equation in their hospital informatics systems to achieve BG control. As shown in Table 2, the implementation of the Glucommander in a cardiac surgery ICU was highly effective in normalizing glucose without hypoglycemia (61). However, the adaptation of this protocol at the Sunnybrook Health Sciences Centre in Toronto produced a reduction in mean BG levels with an increase in hypoglycemic episodes compared to a paper-based protocol (62). Two commercial computer programs based on this algorithm are available. Since 2006, the Glucommander protocol has been marketed by Glytec (Greenville, SC), which has extended the use of the protocol far beyond the ICU to include many areas where IV insulin infusion is advised. The GlucoStabilizer program from Alere Informatics Solution (Waltham, MA) like the Glytec solution, covers not only IV insulin infusion but also subcutaneous insulin treatments. Using this IV protocol, Juneja et al were able to reach the normoglycemia target in nearly all patients and with a low incidence of hypoglycemia (Table 2). The delay in measuring BG was a contributing factor in 67% of episodes of hypoglycemia (63).

There is currently no direct comparison study of the effectiveness, nursing errors or hypoglycemic risks involved in the use of either type of algorithm. However, systematic reviews of the computerized protocols indicate that their implementation improves BG control while keeping the rate of hypoglycemia stable or reducing the rate 64, 65. Although staff adherence to the protocol is easier with computerized programs, delays in BG measurements remain a potential problem. The recent use of accurate CGM and the development of sophisticated mathematical algorithms for delivering IV insulin make the fully automated closed-loop glucose control a reality. Leelarathna et al demonstrated the feasibility of such a system, based on a subcutaneous CGM and the MPC algorithm, in an RCT of 24 patients in the ICU, comparing the closed-loop system with the local paper-based protocol. They proved that the closed-loop system, without nurse intervention, was efficacious and hypoglycemia free, improving on the results of the usual therapy (66).

Complexity is unavoidable in flexible and efficient paper-based IV insulin protocols, making them difficult to use in the busy environment of the ICU. Computerization of these protocols, combined with the use of CGM, appears to be the solution for optimal management of hyperglycemia in ICUs in the future.

Section snippets

Conclusion

There is no doubt that dysglycemia is detrimental for patients in ICUs or that it is necessary to treat it. However, there clearly are few data on which to base firm conclusions concerning glucose levels to target and algorithms to use for insulin infusion therapy. Despite controversies concerning the target ranges of blood glucose control in ICUs, treatment of hyperglycemia in a safe fashion is essential, and insulin infusion therapy should be tailored to match local resources and subtypes of

Acknowledgements

JMB received speaker fees and consulting fees from AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, Eli Lilly, NovoNordisk and Sanofi Aventis and consulting fees from Janssen. LG received honoraria, speaker fees and consulting fees from Eli Lilly and NovoNordisk and speaker fees from AstraZeneca, Bayer, Boehringer Ingelheim, Bristol-Myers Squibb and Sanofi Aventis.

References (66)

  • A.E. Wagstaff et al.

    Diabetes and hyperglycemia in the critical care setting: has the evidence for glycemic control vanished? (Or…is it going away?)

    Curr Diab Rep

    (2014)
  • G. van den Berghe et al.

    Intensive insulin therapy in critically ill patients

    N Engl J Med

    (2001)
  • G. van den Berghe

    Intensive insulin therapy in the ICU: reconciling the evidence

    Nat Rev Endocrinol

    (2012)
  • M. Falciglia et al.

    Hyperglycemia-related mortality in critically ill patients varies with admission diagnosis

    Crit Care Med

    (2009)
  • M. Egi et al.

    Blood glucose concentration and outcome of critical illness: the impact of diabetes

    Crit Care Med

    (2008)
  • G. van den Berghe et al.

    Intensive insulin therapy in the medical ICU

    N Engl J Med

    (2006)
  • J.C. Preiser et al.

    A prospective randomised multi-centre controlled trial on tight glucose control by intensive insulin therapy in adult intensive care units: the Glucontrol study

    Intensive Care Med

    (2009)
  • F.M. Brunkhorst et al.

    Intensive insulin therapy and pentastarch resuscitation in severe sepsis

    N Engl J Med

    (2008)
  • S. Finfer et al.

    Intensive versus conventional glucose control in critically ill patients

    N Engl J Med

    (2009)
  • S. Finfer et al.

    Hypoglycemia and risk of death in critically ill patients

    N Engl J Med

    (2012)
  • E.S. Moghissi et al.

    American Association of Clinical Endocrinologists and American Diabetes Association consensus statement on inpatient glycemic control

    Diabetes Care

    (2009)
  • M. Carey et al.

    Impact of hypoglycemia in hospitalized patients

    Curr Diab Rep

    (2013)
  • F. Bilotta et al.

    Glycemia management in critical care patients

    World J Diabetes

    (2012)
  • J.S. Krinsley

    Glycemic variability: a strong independent predictor of mortality in critically ill patients

    Crit Care Med

    (2008)
  • J. Hermanides et al.

    Glucose variability is associated with intensive care unit mortality

    Crit Care Med

    (2010)
  • J.S. Krinsley et al.

    Diabetic status and the relation of the three domains of glycemic control to mortality in critically ill patients: an international multicenter cohort study

    Crit Care

    (2013)
  • M.K. Sechterberger et al.

    The effect of diabetes mellitus on the association between measures of glycaemic control and ICU mortality: a retrospective cohort study

    Crit Care

    (2013)
  • G. Meyfroidt et al.

    Dynamic characteristics of blood glucose time series during the course of critical illness: effects of intensive insulin therapy and relative association with mortality

    Crit Care Med

    (2010)
  • J.T. King et al.

    Glycemic control and infections in patients with diabetes undergoing noncardiac surgery

    Ann Surg

    (2011)
  • M. Kosiborod et al.

    Glucose normalization and outcomes in patients with acute myocardial infarction

    Arch Intern Med

    (2009)
  • S.E. Siegelaar et al.

    Mean glucose during ICU admission is related to mortality by a U-shaped curve in surgical and medical patients: a retrospective cohort study

    Crit Care

    (2010)
  • J.O. Friedrich et al.

    Does intensive insulin therapy really reduce mortality in critically ill surgical patients? A reanalysis of meta-analytic data

    Crit Care

    (2010)
  • D.E. Griesdale et al.

    Intensive insulin therapy and mortality among critically ill patients: a meta-analysis including NICE-SUGAR study data

    CMAJ

    (2009)
  • Cited by (11)

    • Inpatient Diabetes Management in the Twenty-First Century

      2016, Endocrinology and Metabolism Clinics of North America
      Citation Excerpt :

      It is important that the hospital’s chosen protocol is validated and has demonstrated safety and efficacy.67 Key elements of an IV insulin infusion protocol are listed in Box 4.66–68 In general, a potential hypoglycemic or hyperglycemic scenario should be anticipated and proactively addressed with clear guidelines in the protocol.

    • In-hospital diabetes management

      2014, Canadian Journal of Diabetes
    View all citing articles on Scopus
    View full text