Implementation of Tight Glucose Control for Critically Ill Surgical Patients: A Process Improvement Analysis

Abstract

Background:

Tight glucose control has been advocated as a method to improve outcomes of surgical critical care. However, continuous infusion of insulin has potential morbidity (e.g., neurologic consequences of hypoglycemia), and it remains unclear to what degree the glucose concentration must be controlled. We examined our performance in instituting a protocol for tight glucose control in our surgical intensive care unit (ICU).

Methods:

Prospective study of 220 consecutive patients (February, 2003–March, 2006) who received an infusion of insulin for glucose control for >24 h by protocol. Data collected included age, acuity (Acute Physiology and Chronic Health Evaluation [APACHE] III) score, sex, history of diabetes mellitus, organ dysfunction (Marshall), and death or survival. Infusion-related data included initial glucose concentration, time to glucose <120 mg/dL, h/day of glucose <110 mg/dL and <140 mg/dL, duration of infusion (days), insulin units/day, year of therapy, and complications. Analysis was performed by χ², analysis of variance, and logistic regression, with p < 0.05 considered significant.

Results:

Insulin drips were required by 10.2% of patients (287/2,804); 29 of these (10.1%) had diabetes mellitus. The mean APACHE III score for the treated patients was 77 ± 2 (standard deviation), and the mortality rate was 24%. Hypoglycemia (<60 mg/dL) occurred in 4.2% of patients. The trigger insulin concentration decreased over time (2003 vs. 2005) from 249 ± 14 to 160 ± 5 mg/dL, and the h/day of glucose <140 increased from 11 ± 1 to 16 ± 1. However, age, acuity, APACHE III, days of insulin, time to achieve glucose <120, h/day of glucose <110, and mortality rate were unchanged. By logistic regression, only the year of treatment (odds ratio [OR] 1.871; 95% confidence interval [CI] 1.177, 2.972; p = 0.008] predicted success in controlling the blood glucose concentration to <140 mg/dL; age, illness severity, diabetes history, and trigger glucose concentration [OR 0.996; 95% CI 0.992, 1.001; p = 0.11] did not.

Conclusions:

Success in implementing tight glucose control was modest, albeit improving, despite a specific protocol for administration. No medical reason could be identified for inability to achieve tight glucose control; therefore, successful implementation must be volitional. Education, particularly regarding hypoglycemia, and possible refinement of our protocol may improve our ability to control blood glucose in our ICU.

Close monitoring and treatment of hyperglycemia has been advocated in critically ill surgical patients to improve survival and decrease perioperative morbidity [1]. A large, randomized, controlled trial in predominantly surgical patients demonstrated better outcomes in patients with tight glycemic control: <110 mg/dL compared with 200 mg/dL [2]. In critically ill patients with renal failure, hyperglycemia may be an ominous portent of death [3].

Many investigators have described the possible mechanisms of the benefit of intensive insulin therapy [4 –14]. The positive effects most likely relate to the direct effect of acute glycemic control rather than the infusion of the insulin hormone [15]. Some authors contend that the therapy minimizes derangements in normal host defense mechanisms, whereas others claim tight glycemic control modulates the release of inflammatory mediators [16 –23].

Because a plethora of data suggest that improved glycemic control is beneficial in critically ill surgical patients, our institution undertook an initiative in February 2003 to maintain tighter blood sugar control via insulin infusion in our surgical intensive care unit (ICU). After three years of the program with prospective data collection, we assessed our practice of this technique. In particular, we determined whether our blood glucose control was becoming more stringent over time, and whether our patients having infusions were receiving larger amounts of insulin. An additional goal was to determine whether the mortality rate was changing. We hypothesized that glycemic control would improve over time, as measured by the proportion of time in a 24-h period the blood glucose concentration was <140 mg/dL as well as <110 mg/dL. We also hypothesized that stricter glycemic control would decrease the mortality rate and the number of nosocomial infections in our critically ill patients.

Patients and Methods

Data were collected prospectively on all patients admitted to the surgical ICU of New York–Presbyterian Hospital/Weill Cornell Medical Center from February, 2003–March, 2006. Included were 287 consecutive patients who received an infusion of insulin for glycemic control for at least 24 h according to an institutional protocol (see Appendix). If patients required an insulin infusion, subcutaneous insulin by “sliding scale” was prohibited. Patients who were receiving intravenous insulin admixed with parenteral nutrition solution also were excluded. The insulin infusion protocol was modified by the institution's Formulary and Therapeutics Committee Subcommittee on Critical Care Therapeutics in early January, 2006, in response to observations made during the initial data analysis for the submission of an abstract on December 1, 2005, for presentation at the 2006 Annual Meeting of the Surgical Infection Society. The trigger glucose concentration was decreased from 150 mg/dL to 110 mg/dL for patients with hyperglycemia, and the “hold” parameter for interruption of the insulin infusion for hypoglycemia was liberalized. The protocol changes were accompanied by an educational initiative to inform critical care nurses and surgical house staff of the changes and the importance of adherence to the protocol.

Sixty-five discrete data elements were collected daily for the first seven days (midnight to 11:59 pm) after the beginning of the infusion (Table 1). Demographic data included age, sex, ICU admission data (elective/emergency, diagnosis), severity of illness on study entry (Acute Physiology and Chronic Health Evaluation [APACHE]-II and –III scores), history of diabetes mellitus, multiple organ dysfunction (MOD) score [24] and the scores of each of its six components, ICU length of stay (LOS), and death or survival.

Table 1.
Data Collected

Demographics

Age

Acuity (Acute Physiology and Chronic Health Evaluation [APACHE II])

Acuity (APACHE III)

Sex

History of diabetes mellitus

Admission diagnosis

Multiple organ dysfunction (MOD) score

MOD component scores

Cardiovascular

Central nervous system

Hematologic

Hepatic

Pulmonary

Renal

Emergency or elective admission

Date of admission

Intensive care unit (ICU) length of stay

Death or survival

Infection data

Infection on admission

Source

Pathogen

Nosocomial infection

Pneumonia

Central venous catheter-related infection

Intra-abdominal infection

Urinary tract infection

Surgical site infection

Other (e.g., sinusitis)

Infusion-related data

Initial glucose concentration (trigger)

Time of day started

Number of hours of insulin infusion on day one

Time (h) to glucose concentration <120 mg/dL

Duration of infusion (days)

Collected every day for first seven days:

H/day of glucose concentration <110 mg/dL

H/day of glucose concentration <140 mg/dL

Insulin units/day

Complications

Calculated variables

Mean no. of insulin units/day

Mean h/day <110 mg/dL

Mean h/day <140 mg/dL

Insulin infusion-related data collected were the initial plasma glucose concentration (trigger), time to achieve a glucose concentration <120 mg/dL, h/day that the blood glucose concentration was maintained <140 mg/dL and <110 mg/dL, the time of day of the start of the infusion, the duration of infusion (days), insulin units/day, and the year of therapy. Hypoglycemia (blood glucose <60 mg/dL) and any complications related thereto were recorded. The development of nosocomial infections, the source of infection, and the specific pathogen(s) isolated were noted as well. Nosocomial infections were defined according to the criteria of the U.S. Centers for Disease Control and Prevention [25] except for hospital-acquired/ventilator-associated pneumonia, the diagnosis of which required the identification of >10⁴ colony-forming units (cfu)/mL of a predominant pathogen by quantitative culture of sputum collected by bronchoscopic bronchoalveolar lavage [26].

Univariable analysis of coordinate variables was performed by contingency table analysis with the Fisher exact test, and univariable analysis of continuous variables was performed using the Mann-Whitney U test. Another non-parametric test (Kolmogorov-Smirnov) was used to examine differences in data that were not normally distributed according to determination of skewness and kurtosis. Changes in continuous variables with respect to time were analyzed by repeated measures analysis of variance (ANOVA) with Bonferroni correction, and correlations were made by the Pearson method. Differences in survival were assessed by Kaplan-Meier analysis. Statistical significance was defined as p = 0.05. Univariable results of possible significance (p < 0.15) were tested in multivariable models for independence of effect on death; the dependent variable under study after auto-correlation was excluded by matrix correlations and the Durbin-Watson statistic. For binary logistic regression analysis, the durations of blood glucose concentrations <140 mg/dL and 110 mg/dL were dichotomized on the basis of the mean values of 15 h/day and 9 h/day, respectively. This time was determined by a post-hoc analysis based on the initial results. Odds ratios (ORs) and the 95% confidence interval (CI) were determined; and the sensitivity, specificity, and goodness of fit (model χ²; Hosmer-Lemeshow test) were determined for each logistic regression model. Data are expressed as mean values ± standard error.

Results

Two thousand eight hundred four patients were admitted into the surgical ICU during the study period, of whom 287 (10.2%) required a continuous infusion of insulin for >24 h. Twenty-nine of the subjects (10.1%) had known diabetes mellitus. The mean APACHE III score for treated hyperglycemic patients was 77.4 ± 1.5 points, and the hospital mortality rate was 24.4%. The mean MOD score was 7.8 ± 0.3 points, and nosocomial infections developed in 55.1% of the study patients.

Episodes of hyperglycemia were recorded. Hypoglycemia occurred in 4.2% of patients. Although acute neurologic changes were not captured, no long-term sequelae attributable to hypoglycemia were noted during the study period.

Demographic data overall and as a function of the year of therapy are shown in Table 2. Neither age, severity of illness, organ dysfunction, the incidence of nosocomial infection, nor the mortality rate fluctuated over the study period. The mean ICU day of initiation of the insulin infusion was 3.32 ± 0.27, and the median day of initiation was 2.

Table 2.
Demographic Data Stratified by Year

All (n = 287) (target <150 mg/dL) 2003 (n = 47) (target <150 mg/dL) 2004 (n = 103) (target <150 mg/dL) 2005 (n = 108) (target <150 mg/dL) 2006 (n = 29) (target <110 mg/dL)

Mean age (years) ± SD 66.3 ± 0.9 65.6 ± 2.2 67.2 ± 1.6 65.9 ± 1.6 65.9 ± 3.0

Mean APACHE III score ± SD 77.4 ± 1.5 77.2 ± 3.9 77.4 ± 2.9 77.9 ± 2.4 77.4 ± 1.5

Mean MOD score ± SD 7.8 ± 0.3 7.3 ± 0.6 7.3 ± 0.5 8.4 ± 0.5 8.7 ± 1.0

Nosocomial infection (%) 55.1 57.4 50.5 60.2 48.3

Mortality rate (%) 24.4 23.4 25.2 23.1 27.6

APACHE = Acute Physiologic and Chronic Health Evaluation; MOD = multiple organ dysfunction; SD = standard deviation.

The mean trigger blood glucose concentration decreased over time, from 249.5 ± 13.8 mg/dL in 2003 to 160.1 ± 5.1mg/dL in 2006 (p < 0.001), with a substantial incremental decrease achieved in 2006 after the protocol was modified and reinforced (Table 3). During this same time, the duration of insulin infusion increased significantly, from 8.8 ± 0.8 to 14.8 ± 2.5 days (p < 0.05), increasing by a full five days from 2005 to 2006. Between 2003 and 2006, the lower trigger glucose concentration resulted in a significant reduction in the elapsed time to achieve a glucose concentration <120 mg/dL, with essentially all of the decrease occurring in 2006 compared with 2005. Similarly, the mean h/day of glucose <140 mg dL and <110 mg/dL increased significantly over time, with a dramatic increase in 2006 (Table 3; Fig. 1). The relation between the time of day the insulin infusion was started and the duration of insulin therapy on day one is shown in Figure 2. The insulin dose (units/day) decreased significantly (p < 0.01) with respect to time (Fig. 3), even as control of blood glucose improved.

FIG. 1.
Mean duration of time per day that glucose concentration was kept below 140 mg/dL (light bars) (overall mean 15.0 ± 0.5 hours) or 110 mg/dL dark bars (overall mean 9.4 ± 0.4) hours. Both relationships are statistically significant (p < 0.01) with respect to time.

FIG. 2.
The hour of starting a continuous insulin infusion is shown, stratified by year of therapy. Shown together is the total duration of insulin infusion therapy on day one of the infusion. Earlier start times (p < 0.05) were associated with longer duration of therapy (p < 0.05).

FIG. 3.
The mean daily dose of insulin (units/day) is shown, stratified by year of therapy. The amount of insulin required daily to maintain control of blood glucose concentration decreased significantly (p < 0.01), despite tighter control over time.

Table 3.
Insulin Infusion-Related Demographic Data, Stratified by Year ^a

All (n = 287) 2003 (n = 47) 2004 (n = 103) 2005 (n = 108) 2006 (n = 29) P value

Glucose concentration trigger^b 201.3 ± 4.4 249.5 ± 13.8 202.7 ± 6.3 194.5 ± 7.3 160.1 ± 5.1 <0.001

Days of insulin 8.8 ± 0.8 7.8 ± 1.8 6.5 ± 1.1 9.6 ± 1.5 14.8 ± 2.5 0.022

Hours to ≤120 mg/dL 3.9 ± 0.2 4.8 ± 0.6 3.7 ± 0.3 4.2 ± 0.3 2.6 ± 1.0 0.023

Mean hours <110 mg/dL 9.4 ± 0.4 7.3 ± 0.9 9.1 ± 0.6 9.7 ± 0.6 13.0 ± 1.1 0.005

Mean hours <140 mg/dL 15.1 ± 0.4 11.5 ± 1.2 14.8 ± 0.6 15.9 ± 0.6 18.2 ± 0.9 0.001

a
Mean ± standard deviation.

b
Glucose concentration at which insulin infusion was initiated.

Three binary logistic regression models were created for the dependent variables duration of blood glucose < 140 mg/dL (Table 4), development of a nosocomial infection (Table 5), and death (Table 6). By logistic regression, the year of treatment (OR 1.871; 95% CI 1.192, 2.717; p = 0.005), trigger glucose concentration (OR 0.994; 95% CI 0.988, 0.999; p = 0.025), the start time of insulin infusion (OR 0.929; 95% CI 0.887, 0.972; p = 0.001), and the ICU day of start of insulin infusion (OR 1.145; 95% CI 1.009, 1.299; p = 0.036) predicted success in controlling blood glucose <140 mg/dL (dichotomized at the mean value of ±15 h/day). Therefore, requiring insulin later in the ICU stay rather than nearer to admission, and starting insulin as soon as possible after the trigger blood glucose criterion is met, are associated with better glucose control. Also, by logistic regression, the development of a nosocomial infection was independently associated with insulin therapy later in the ICU stay (OR 1.223; 95% CI 0.987, 0.998; p = 0.007), a larger insulin dose requirement (units/day)(OR 1.015; 95% CI 1.005, 1.025; p = 0.004), but paradoxically a lower trigger glucose concentration (OR 0.993; 95% CI 0.987, 0.998; p = 0.029).

Table 4.
Binary Logistic Regression (Dependent Variable: Glucose Concentration <140 mg/dL for >15 h/day)

P value Odds ratio 95.0% Confidence interval

Year 0.005 1.8 1.192, 2.717

Age 0.083 0.981 0.961, 1.002

PMH 0.827 1.131 0.376, 3.400

Trigger 0.025 0.994 0.988, 0.999

Time start 0.001 0.929 0.887, 0.972

ICU Day 0.036 1.145 1.009, 1.299

Model χ² 49.220, p value 0.0001; Nagelkerke R² 0.284; Hosmer-Lemeshow test χ² 11.051, p value 0.199.

ICU = intensive care unit; PMH = positive medical history; Trigger = glucose concentration at which insulin infusion was initiated.

Table 5.
Binary Logistic Regression (Dependent Variable: Nosocomial Infection)

P value Odds ratio 95.0% Confidence interval

Age 0.520 0.992 0.969, 1.016

Intensive care unit day 0.007 1.223 1.057, 1.415

Trigger 0.029 0.993 0.987, 0.998

Mean no. units/day 0.004 1.015 1.005, 1.025

Mean time < 110/day 0.374 1.046 0.947, 1.156

Mean time < 140/day 0.501 1.034 0.938, 1.14

Model χ² 39.366, p value 0.0001; Nagelkerke R² 0.276; Hosmer-Lemeshow test χ² 6.565, p value 0.584.

Trigger = glucose concentration at which insulin infusion was initiated.

Table 6.
Binary Logistic Regression (Dependent Variable: Death)

P value Odds ratio 95.0% Confidence interval

Age 0.751 1.004 0.981, 1.027

APACHE III score 0.17 0.986 0.966, 1.006

MOD score 0.0001 1.56 1.35, 1.804

Mean time

<140/day for 15 h/day 0.0001 0.72 0.623, 0.831

Model χ² 118.490, p value 0.0001; Nagelkerke R² 0.664; Hosmer-Lemeshow test χ² 7.464, p value 0.488.

APACHE = Acute Physiology and Chronic Health Evaluation; MOD = multiple organ dysfunction.

The binary logistic regression for death is shown in Table 6. The Kaplan-Meier survival curve is shown in Figure 4. The mortality rate was decreased significantly by a mean duration of glucose concentration <140 mg/dL of more than 15 h/day (OR 0.78; 95% CI 0.623, 0.831; p < 0.0001).

FIG. 4.
Kaplan Meier survival plot of glucose concentration times day <140 mg/dL. Time was dichotomized at the mean value of 15 hours/day. The curves are significantly different (p < 0.01).

Discussion

Multiple mechanisms can contribute to the potential benefits of insulin infusion in critically ill patients [27 –30]. Most of the benefits appear to be related to prevention of hyperglycemia, which affects outcomes adversely. In the trauma population, Yendamuri et al. have found that hyperglycemia independently predicts longer ICU and hospital LOS as well as the mortality rate. These authors substantiated greater infectious morbidity for both mild (glucose concentration >135 mg/dL) and moderate (glucose concentration >200 mg/dL) hyperglycemia [31]. Other investigators have observed trauma patients to fare worse when the initial glucose concentration is elevated [32]. Importantly, glycemic variability also appears to predict death in critically ill patients [33].

The anti-inflammatory effects of insulin infusion may contribute to its benefits in critically ill patients. Several cellular mechanisms may be beneficial. Part of the improvement may be related to protection of mitochondrial respiration and prevention of excessive inflammation. Hyperglycemia without antagonism may reduce neutrophil activity and alter specific concentrations of early pro-inflammatory cytokines, including tumor necrosis factor-alpha and interleukin-6 [34].

Intensive insulin therapy may work at the endothelial level by suppressing inducible nitric oxide synthase (iNOS) expression in liver and skeletal muscle, possibly in part via reduced nuclear factor-κB activation. Other actions of insulin may include lowering circulating NO concentrations in both survivors and non-survivors [35]. The metabolic actions of insulin may also benefit other aspects of the physiology of critically ill patients [36]. For example, insulin may contribute to a partial correction of the abnormal serum lipid profile that characterizes critical illness and counteract the associated catabolism [37,38]. Nonetheless, other authors found that increased insulin administration was positively associated with death in critically ill patients regardless of the prevailing blood glucose concentration [39].

In particular, surgical patients may have a better response to tight glycemic control than do medical patients [40]. Whereas insulin infusion was beneficial in an 800-patient study in a combined medical and surgical ICU [41], a meta-analysis examining 35 trials employing insulin infusion in critically ill patients found contrary results. In this meta-analysis, Pittas et al. showed that insulin therapy decreased the mortality rate significantly in the surgical ICU when the aim of therapy was glucose control regardless of whether the patient had diabetes mellitus [42]. These authors also observed a near-significant trend toward fewer deaths in critically ill surgical patients with acute myocardial infarction who did not receive reperfusion therapy. In this meta-analysis, the authors found no randomized trials of insulin in the medical ICU setting. More recently, a large randomized, controlled trial demonstrated that intensive insulin therapy significantly reduced morbidity, but not the mortality rate, among medical ICU patients [43].

The analysis of our initial experience with intensive insulin infusions yielded notable results from a process improvement standpoint. First, our glucose trigger decreased substantially over the study period. This decrease was accompanied by better control of blood glucose to <140 mg/dL, as measured during a 24-h period. Concurrently, we noted that the hours of infusion required to achieve a blood glucose concentration <120 mg/dL decreased. Although outcomes appeared to be more favorable if a lower glucose threshold initiated the insulin therapy, we do not have data regarding any delay definitively causing an adverse outcome. These results were demonstrated most prominently after the revision of our protocol in early 2006, and were accompanied by reinforcement to the house staff and nursing teams by the surgical intensivists at our institution. Successful protocol implementation required intensive teaching for the ordering and treating personnel. Directed feedback also may be necessary to evoke a beneficial effect of this therapy because interventions such as insulin infusion may yield only subtle benefits while being notably labor intensive.

Our results also demonstrated that an essential feature of successful euglycemia is early recognition and treatment of “borderline” hyperglycemia before sustained and difficult-to-treat hyperglycemia can ensue. We noted that in our more recent study period, patients were receiving lower cumulative insulin doses to control their glucose concentration even though the duration of their drips was longer. This observation may also convey that our personnel may have been attempting to wean patients from insulin infusions prematurely during the earlier portion of the study period. Another finding was that the mean start time of infusion changed to earlier in the day during the study (Fig. 2). This may have occurred because nurses and residents may have felt more independent in the later phases of the study, and the de facto trigger was not the presence of the attending physician during morning ICU rounds. Later in the study, we found that despite tighter control, the total insulin dose was less. This result suggests that earlier glucose control may decrease subsequent insulin resistance.

Perhaps the most important feature of our study centers on the difficulty inherent in creating an insulin infusion protocol in a locale where the physicians and nurses are not familiar with the initiative. In the beginning phases of this protocol, the trigger was a mean of about 250 mg/dL despite aggressive education of all staff on the merits of the protocol. We would have thought that because the protocol-specified trigger was 120 mg/dL for two consecutive values, a much lower mean trigger would have been observed. We did see this in the later phases of the study but not initially.

Several theories may explain why our study failed to show a stronger discernible benefit in terms of mortality rate or nosocomial infections in patients treated with intensive insulin infusion. Notably, a major meta-analysis by Weiner et al. in 2008 had findings that paralleled ours in that a mortality difference was not observed in patients receiving insulin infusions [44]. A type II error may have been present, or perhaps the effectiveness of our infection diagnoses and treatment protocols masked an effect. Perhaps our study population differed from prior studies with this modality owing to a higher severity of illness in our patients prior to the initiation of insulin therapy. Consequently, our patients may already have developed substantial organ dysfunction, and aggressive maintenance of euglycemia has never been shown to be superior to conventional therapy at reversing organ dysfunction.

Another explanation is that the insulin infusions were not employed frequently or aggressively enough in our entire group of critically ill surgical patients; our incidence of insulin infusions was only 10.3%. Other authors likewise have also that the implementation of insulin infusion protocols may not always be associated with initial complete compliance [45]. The theory that a lack of proper glucose control masked more beneficial results is supported by the Kaplan-Meier analysis, which showed a significantly higher survival rate in patients who were controlled <140 mg/dL. Only in 2006 did we finally improve our insulin infusion process such that patients are controlled <140 mg/dL for the majority of the day. Perhaps if we had been more aggressive with blood sugar control with more patients, we would have seen an unequivocal mortality benefit. On the other hand, a practice targeting blood glucose control <140 mg/dL may be sufficient to prevent the ill consequences of hyperglycemia and mitigate the potential damage of hypoglycemia.

We concluded that insulin therapy in the ICU setting is difficult to implement successfully. The practice requires persistence and vigilance to maintain constant euglycemia in critically ill patients. However, we also demonstrated superior survival for patients who were properly controlled for the majority of the 24-h period, with a blood glucose concentration <140 mg/dL. Further studies will demonstrate whether we can maintain stringency in our protocol and whether the provision of successful intensive insulin therapy to a larger cohort of patients will produce demonstrably decreased mortality and nosocomial infection rates.

	All (n = 287) (target <150 mg/dL)	2003 (n = 47) (target <150 mg/dL)	2004 (n = 103) (target <150 mg/dL)	2005 (n = 108) (target <150 mg/dL)	2006 (n = 29) (target <110 mg/dL)
Mean age (years) ± SD	66.3 ± 0.9	65.6 ± 2.2	67.2 ± 1.6	65.9 ± 1.6	65.9 ± 3.0
Mean APACHE III score ± SD	77.4 ± 1.5	77.2 ± 3.9	77.4 ± 2.9	77.9 ± 2.4	77.4 ± 1.5
Mean MOD score ± SD	7.8 ± 0.3	7.3 ± 0.6	7.3 ± 0.5	8.4 ± 0.5	8.7 ± 1.0
Nosocomial infection (%)	55.1	57.4	50.5	60.2	48.3
Mortality rate (%)	24.4	23.4	25.2	23.1	27.6

	All (n = 287)	2003 (n = 47)	2004 (n = 103)	2005 (n = 108)	2006 (n = 29)	P value
Glucose concentration trigger^b	201.3 ± 4.4	249.5 ± 13.8	202.7 ± 6.3	194.5 ± 7.3	160.1 ± 5.1	<0.001
Days of insulin	8.8 ± 0.8	7.8 ± 1.8	6.5 ± 1.1	9.6 ± 1.5	14.8 ± 2.5	0.022
Hours to ≤120 mg/dL	3.9 ± 0.2	4.8 ± 0.6	3.7 ± 0.3	4.2 ± 0.3	2.6 ± 1.0	0.023
Mean hours <110 mg/dL	9.4 ± 0.4	7.3 ± 0.9	9.1 ± 0.6	9.7 ± 0.6	13.0 ± 1.1	0.005
Mean hours <140 mg/dL	15.1 ± 0.4	11.5 ± 1.2	14.8 ± 0.6	15.9 ± 0.6	18.2 ± 0.9	0.001

	P value	Odds ratio	95.0% Confidence interval
Year	0.005	1.8	1.192, 2.717
Age	0.083	0.981	0.961, 1.002
PMH	0.827	1.131	0.376, 3.400
Trigger	0.025	0.994	0.988, 0.999
Time start	0.001	0.929	0.887, 0.972
ICU Day	0.036	1.145	1.009, 1.299

	P value	Odds ratio	95.0% Confidence interval
Age	0.520	0.992	0.969, 1.016
Intensive care unit day	0.007	1.223	1.057, 1.415
Trigger	0.029	0.993	0.987, 0.998
Mean no. units/day	0.004	1.015	1.005, 1.025
Mean time < 110/day	0.374	1.046	0.947, 1.156
Mean time < 140/day	0.501	1.034	0.938, 1.14

	P value	Odds ratio	95.0% Confidence interval
Age	0.751	1.004	0.981, 1.027
APACHE III score	0.17	0.986	0.966, 1.006
MOD score	0.0001	1.56	1.35, 1.804
Mean time
<140/day for 15 h/day	0.0001	0.72	0.623, 0.831

Footnotes

Author Disclosure Statement

No conflicting financial interests exist.

Appendix

Presented in part at the 26th Annual Meeting of the Surgical Infection Society, San Diego, California, April 2005.

References

Langouche

, van den Berghe

. Glucose metabolism and insulin therapy. Crit Care Clin, 2006; 22:119–129.

van den Berghe

, Wouters

, Weekers

et al. Intensive insulin therapy in the surgical intensive care unit. N Engl J Med, 2001; 345:1359–1367.

Basi

, Pupim

, Simmons

et al. Insulin resistance in critically ill patients with acute renal failure. Am J Physiol, 2005; 289:F259–F264.

Hovorka

, Kremen

, Blaha

et al. Blood glucose control by a model predictive control algorithm with variable sampling rate versus a routine glucose management protocol in cardiac surgery patients: A randomized controlled trial. J Clin Endocrinol Metab, 2007; 92:2960–2964.

Treggiari

, Karir

, Yanez

et al. Intensive insulin therapy and mortality in critically ill patients. Crit Care, 2008; 12:R29.

Krinsley

. Perioperative glucose control. Curr Opin Anaesth, 2006; 19:111–116.

Hoedemaekers

, Pickkers

, Netea

et al.

Intensive insulin therapy does not alter the inflammatory response in patients undergoing coronary artery bypass grafting: A randomized controlled trial [ISRCTN95608630]

Crit Care, 2005; 9:R790–R797.

Zerr

, Furnary

, Grunkemeier

et al. Glucose control lowers the risk of wound infection in diabetics after open heart operations. Ann Thorac Surg, 1997; 63:356–361.

Rassias

, Marrin

, Arruda

et al. Insulin infusion improves neutrophil function in diabetic cardiac surgery patients. Anesth Analg, 1999; 88:1011–1016.

10.

Gale

, Sicoutris

, Reilly

et al. Glycemic control improves outcome in cardiac surgical patients and after myocardial infarction or stroke. Am Surg, 2007; 73:454–460.

11.

Scalea

, Bochicchio

et al. Tight glycemic control in critically injured trauma patients. Ann Surg, 2007; 246:605–610.

12.

Bochicchio

, Joshi

, Bochicchio

et al. Early hyperglycemic control is important in critically injured trauma patients. J Trauma, 2007; 63:1353–1359.

13.

Grey

, Perdrizet

. Reduction of nosocomial infections in the surgical intensive-care unit by strict glycemic control. Endocr Pract, 2004; 10,Suppl 2:46–52.

14.

Bochicchio

, Sung

, Joshi

et al. Persistent hyperglycemia is predictive of outcome in critically ill trauma patients. J Trauma, 2005; 58:921–924.

15.

Lipsett

. The importance of insulin administration in the critical care unit. Adv Surg, 2006; 40:47–57.

16.

Butler

, Btaiche

, Alaniz

. Relationship between hyperglycemia and infection in critically ill patients. Pharmacotherapy, 2005; 25:963–976.

17.

Vanhorebeek

, Ingels

, van den Berghe

. Intensive insulin therapy in high-risk cardiac surgery patients: Evidence from the Leuven randomized study. Semin Thorac Cardiovasc Surg, 2006; 18:309–316.

18.

Dandona

, Chaudhuri

, Ghanim

et al. Anti-inflammatory effects of insulin and the pro-inflammatory effects of glucose. Semin Thorac Cardiovasc Surg, 2006; 18:293–301.

19.

Latham

, Lancaster

, Covington

et al. The association of diabetes and glucose control with surgical-site infections among cardiothoracic surgery patients. Infect Control Hosp Epidemiol, 2001; 22:607–612.

20.

Dunkelgrun

, Schreiner

, Schockman

et al. Usefulness of preoperative oral glucose tolerance testing for perioperative risk stratification in patients scheduled for elective vascular surgery. Am J Cardiol, 2008; 101:526–529.

21.

Hermayer

, Neal

, Hushion

et al. Outcomes of a cardiothoracic intensive care web-based online intravenous insulin infusion calculator study at a medical university hospital. Diabetes Technol Ther, 2007; 9:523–534.

22.

D'Alessandro

, Leprince

, Golmard

et al. Strict glycemic control reduces EuroSCORE expected mortality in diabetic patients undergoing myocardial revascularization. J Thorac Cardiovasc Surg, 2007; 134:29–37.

23.

Hafidh

, Reuter

, Chassels

et al. Effect of intravenous insulin therapy on clinical outcomes in critically ill patients. Am J Med Sci, 2007; 333:354–361.

24.

Marshall

, Cook

, Christou

et al. The multiple organ dysfunction (MOD) score: A reliable descriptor of a complex clinical outcome. Crit Care Med, 1995; 23:1638–1652.

25.

http://www.cdc.gov/ncidod/dhqp/healthDis.html. 2008 May 15.

26.

Chastre

, Combes

, Luyt

. The invasive (quantitative) diagnosis of ventilator-associated pneumonia. Respir Care, 2005; 50:797–807.

27.

Vanhorebeek

, Langouche

, van den Berghe

. Glycemic and nonglycemic effects of insulin: How do they contribute to a better outcome of critical illness? Curr Opin Crit Care, 2005; 11:304–311.

28.

Taylor

, Beilman

. Hyperglycemia in the intensive care unit: No longer just a marker of illness severity. Surg Infect, 2005; 6:233–245.

29.

Laird

, Miller

, Kilgo

et al. Relationship of early hyperglycemia to mortality in trauma patients. J Trauma, 2004; 56:1058–1062.

30.

Goldberg

, Inzucchi

. Critical issues in endocrinology. Clin Chest Med, 2003; 24:583–606.

31.

Yendamuri

, Fulda

, Tinkoff

. Admission hyperglycemia as a prognostic indicator in trauma. J Trauma, 2003; 55:33–38.

32.

Sung

, Bochicchio

, Joshi

et al. Admission hyperglycemia is predictive of outcome in critically ill trauma patients. J Trauma, 2005; 59:80–83.

33.

Krinsley

. Glycemic variability: A strong independent predictor of mortality. Crit Care Med, 2008; 36:3008–3013.

34.

Turina

, Fry

, Polk

Jr.

Acute hyperglycemia and the innate immune system: Clinical, cellular, and molecular aspects. Crit Care Med, 2005; 33:1624–1633.

35.

Langouche

, Vanhorebeek

, Vlasselaers

et al. Intensive insulin therapy protects the endothelium of critically ill patients. J Clin Invest, 2005; 115:2277–2286.

36.

Solano

, Totaro

. Insulin therapy in critically ill patients. Curr Opin Clin Nutr Metabol Care, 2004; 7:199–205.

37.

Nylen

, Muller

. Endocrine changes in critical illness. J Intensive Care Med, 2004; 19:67–82.

38.

Bonville

, Parker

, Levine

et al. The relationships of hypocholesterolemia to cytokine concentrations and mortality in critically ill patients with systemic inflammatory response syndrome. Surg Infect, 2004; 5:39–49.

39.

Finney

, Zekveld

, Elia

et al. Glucose control and mortality in critically ill patients. JAMA, 2003; 290:2041–2047.

40.

Khoury

, Klausner

, Ben-Abraham

et al. Glucose control by insulin for critically ill surgical patients. J Trauma, 2004; 57:1132–1138.

41.

Krinsley

. Effect of an intensive glucose management protocol on the mortality of critically ill adult patients. Mayo Clin Proc, 2004; 79:992–1000.

42.

Pittas

, Siegel

, Lau

. Insulin therapy for critically ill hospitalized patients: A meta-analysis of randomized controlled trials. Arch Intern Med, 2004; 164:2005–2011.

43.

van den Berghe

, Wilmer

, Hermans

et al. Intensive insulin therapy in the medical ICU. N Engl J Med, 2006; 354:449–461.

44.

Wiener

, Wiener

, Larson

. Benefits and risks of tight glucose control in critically ill adults: A meta-analysis. JAMA, 2008; 300:933–944.

45.

Schultz

, Wolthuis

, Moeniralam

et al. Struggle for implementation of new strategies in intensive care medicine: Anticoagulation, insulin, and lower tidal volumes. J Crit Care, 2005; 20:199–204.