Intensive Care Medicine

, Volume 38, Issue 4, pp 634–641

Dyslipidemia: a prospective controlled randomized trial of intensive glycemic control in sepsis


  • Sylas B. Cappi
    • Laboratório da Disciplina de Emergências ClínicasFaculdade de Medicina da Universidade de São Paulo
  • Danilo T. Noritomi
    • Laboratório da Disciplina de Emergências ClínicasFaculdade de Medicina da Universidade de São Paulo
  • Irineu T. Velasco
    • Laboratório da Disciplina de Emergências ClínicasFaculdade de Medicina da Universidade de São Paulo
  • Rui Curi
    • Institute of Biomedical SciencesUniversity of São Paulo
  • Tatiana C. A. Loureiro
    • Institute of Biomedical SciencesUniversity of São Paulo
    • Laboratório da Disciplina de Emergências ClínicasFaculdade de Medicina da Universidade de São Paulo
    • Intensive Care UnitUniversity of São Paulo University Hospital

DOI: 10.1007/s00134-011-2458-z

Cite this article as:
Cappi, S.B., Noritomi, D.T., Velasco, I.T. et al. Intensive Care Med (2012) 38: 634. doi:10.1007/s00134-011-2458-z



Metabolic disturbances are quite common in critically ill patients. Glycemic control appears to be an important adjuvant therapy in such patients. In addition, disorders of lipid metabolism are associated with worse prognoses. The purpose of this study was to investigate the effects that two different glycemic control protocols have on lipid profile and metabolism.


We evaluated 63 patients hospitalized for severe sepsis or septic shock, over the first 72 h of intensive care. Patients were randomly allocated to receive conservative glycemic control (target range 140–180 mg/dl) or intensive glycemic control (target range 80–110 mg/dl). Serum levels of low-density lipoprotein, high-density lipoprotein, triglycerides, total cholesterol, free fatty acids, and oxidized low-density lipoprotein were determined.


In both groups, serum levels of low-density lipoprotein, high-density lipoprotein, and total cholesterol were below normal, whereas those of free fatty acids, triglycerides, and oxidized low-density lipoprotein were above normal. At 4 h after admission, free fatty acid levels were higher in the conservative group than in the intensive group, progressively decreasing in both groups until hour 48 and continuing to decrease until hour 72 only in the intensive group. Oxidized low-density lipoprotein levels were elevated in both groups throughout the study period.


Free fatty acids respond to intensive glycemic control and, because of their high toxicity, can be a therapeutic target in patients with sepsis.


Blood glucoseSepsisFatty acidsNonesterifiedLipids


In the last two decades, the incidence of sepsis has increased worldwide [1]. Patients with sepsis often develop severe comorbidities, such as multiple organ failure, which increases mortality rates [2]. The combination of insulin resistance and hyperglycemia, first described in the nineteenth century and also known as “stress diabetes”, is an initial adaptive response to inflammation [3]. Severe hyperglycemia (blood glucose above 10 mM) is associated with increased morbidity and mortality in a variety of diseases [46]. Insulin resistance is one of the metabolic alterations that occur secondary to inflammatory states; insulin regulates not only blood glucose but also lipid homeostasis [7].

Critically ill patients present various lipid disorders [8]: high levels of triglycerides; low levels of low-density lipoprotein (LDL), with a large quantity of atherogenic small dense LDL particles; and low levels of high-density lipoprotein (HDL) [9, 10]. Paraoxonase, a component of HDL, has been demonstrated to prevent the formation of mildly oxidized LDL (oxLDL) [1113]. Patients with sepsis have severe oxidative stress and low levels of HDL, and they are predisposed to increased production of oxLDL. There is evidence that high oxLDL levels correlate with low levels of HDL, high levels of serum glucose, and high levels of triglycerides [14].

High serum levels of free fatty acids (FFAs) typically cause necrosis, with a rapid loss of membrane integrity, lysosomal leakage, and cell swelling [15, 16]. These effects seem to be associated with oxidative stress because they can be partially prevented by antioxidants, such as tocopherol [1518]. It has been shown that serum FFAs play a role in heart rate variability [19, 20]. Elevated serum levels of FFAs can disrupt the structure and function of the cardiac plasma membrane, as well as increasing the concentration of intracellular calcium, thereby affecting cardiac activity [21, 22].

Lipid homeostasis can be restored by insulin administration, and recent data on glucose control support the use of this clinical intervention. A randomized clinical trial conducted in a surgical intensive care unit (ICU) [23] showed 40% lower mortality and morbidity rates among patients in the intensive glycemic control. However, in a multicenter prospective randomized controlled study the authors found that intensive glucose control provided no benefits in terms of mortality or morbidity [24].

In patients with sepsis, lipid metabolism, glycemic control, and mortality are relevant issues. The aims of the study were to examine the effects of two different insulin therapy protocols for glycemic control on the lipid profile and lipid metabolism in patients with sepsis, more specifically cholesterol lipoproteins and FFAs plasma levels; to analyze if intensive glucose control correlates with a lower oxLDL formation, vasoactive drug use, lactate clearance, acidosis and renal support.


Study site and design

This was a single-center, prospective, interventional randomized controlled clinical trial, parallel-group study conducted in a 14-bed medical-surgical ICU at a university hospital in São Paulo, Brazil. Consecutive patients were enrolled. All patients were over 18 years of age, diagnosed at ICU admission with severe sepsis or septic shock, in accordance with the criteria established by the American Society of Critical Care [25]. We excluded patients who were pregnant, as were those with HIV, cancer, or leptospirosis, as well as those previously receiving statin therapy and those with do-not-resuscitate orders.

To detect a 20% reduction in FFA and oxLDL levels [19], with a 5% level of significance (two-tailed) and a power of 80%, the required sample size (given an anticipated dropout rate of 10%) was 24 patients per group.

Randomization procedures

The study participants were randomized to the intervention or control group. This process consisted of simple 1:1 randomization [26], through the use of sealed, opaque envelopes. Beginning with a stack of 20 envelopes, we allocated patients until only 4 envelopes remained, then replenishing the stack to 20 and repeating this procedure until all patients had been allocated.


Two separate institutional protocols were created, and the ICU staff was trained in the use of both protocols for 3 months prior to the beginning of the study.

Patients in the intervention group were submitted to stricter glycemic control (glycemia target range 4.4–6.1 mmol/l). In the conservative group, glycemic control was less strict (glycemia target range 7.8–10 mmol/l). In both groups, arterial blood samples were collected on an hourly basis until the continuous insulin dose had stabilized. It means that if three consecutive measures were stable in the desired blood glucose range, then the control changed to every 2 h.

Data collection

We collected the following data: patient age; history of diabetes; Acute Physiology and Chronic Health Evaluation (APACHE) II score at ICU admission; amount of vasoactive drugs administered; site of infection; and culture findings, when available.


Serum samples were collected at baseline, as well as at 24, 48, and 72 h after ICU admission, as shown in Fig. 1. Additional samples (for FFA determination) were collected at 4 h after ICU admission. Serum C-reactive protein (CRP), triglycerides, total cholesterol, and HDL were determined by commercial kits with an automated analyzer (ADVIA 1650; Bayer Diagnostics, Tarrytown, NY, USA). Serum LDL was determined using an ELISA kit (Human LDL Kit; GenWay Biotech, Inc, San Diego, CA, USA). Non-esterified (free) fatty acids were measured with a colorimetric enzymatic technique using a commercial kit (NEFA C; Wako Chemicals, Richmond, VA, USA). We measured oxLDL using an ELISA (Mercodia AB, Uppsala, Sweden). Lactate was measured by spectrophotometric assay in an automated analyzer (COBAS FARA; Roche Molecular Diagnostics, Basel, Switzerland).
Fig. 1

Study flow diagram

Statistical analysis

Statistical analyses were performed with the Statistical Package for the Social Sciences, version 10.0 (SPSS Inc., Chicago, IL, USA). Data are presented as mean and standard deviation (SD) for variables with Gaussian distribution and as median and interquartile (IQR) range (25th–75th percentile) for variables with non-Gaussian distribution. We used the Kolmogorov–Smirnov test for normality test analysis. The level of statistical significance was set at P < 0.05. For comparison between groups, we used the Mann–Whitney U test or Student’s t test for nonparametric and parametric data, respectively. For analyses of repeated measures, we used Friedman’s test with Bonferroni correction or repeated measures analysis of variance followed by the Tukey–Kramer post hoc test. For some analyses, we categorized patients as survivors and nonsurvivors. In order to assess the significance of the observed correlations, we used Pearson’s correlation coefficient.


The study was approved by the Research Ethics Committee of the University of São Paulo University Hospital. All participating patients or their legal guardians gave written informed consent prior to their inclusion in the study.


Of the 69 eligible patients, 6 were excluded: 2 for presenting with strong evidence of leptospirosis and 4 for declining to participate. Therefore, the final study sample consisted of 63 patients (Fig. 1). All participating patients were initially admitted to the emergency room. Table 1 shows the baseline data, by group. The two groups did not differ significantly at baseline.
Table 1

Baseline comparison of key characteristics


Conservative glycemic control (n = 35)

Intensive glycemic control (n = 28)

Age (years)

53 ± 19

53 ± 19

Male gender

19 (54)

17 (61)


19 ± 7

19 ± 7

History of dyslipidemia

6 (17)

5 (18)

History of diabetes

8 (23)

7 (25)

Site of infection


17 (49)

15 (53)


11 (31)

5 (18)


4 (11)

3 (11)

 Urinary tract

1 (3)

3 (11)


2 (6)

2 (7)

Isolated germs

 Gram negative

9 (26)

6 (21)

 Gram positive

10 (28)

13 (46)


1 (3)

1 (4)

Glycemia (mM)

7.8 (5.6–8.9)

8.0 (5.4–10.1)

Total cholesterol (mg/dl)

88 (63–111)

95 (71–125)

HDL cholesterol (mg/dl)

17 (11–30)

18 (14–24)

LDL cholesterol (mg/dl)

38 (20–46)

36 (28–51)

Triglycerides (mg/dl)

115 (78–215)

153 (92–232)

oxLDL (mg/dl)

13 (6–23)

21 (12–28)

FFAs (µmol/l)

0.08 (0.05–0.10)

0.07 (0.05–0.12)

CRP (mg/dl)

220 (168–279)

293 (170–347)

Data are presented as mean ± SD, n (%), or median (IQR) as appropriate

APACHE acute physiology and chronic health evaluation, IQR interquartile range (25th–75th percentile), HDL high-density lipoprotein, LDL low-density lipoprotein, oxLDL oxidized LDL, CRP C-reactive protein

The appropriateness of the intervention was evidenced by the statistically significant differences between the two groups at 24, 48, and 72 h post-admission, in terms of the level of glycemia and the number of units of insulin administered per day. The values of blood glucose at 48 h were 8.6 ± 2.2 versus 5.5 ± 1.0; and at 72 h they were 9.2 ± 1.9; the amount of insulin at 48 h was 30 ± 49.2 UI/d versus 126.8 ± 175 UI/d; and at 72 h it was 29.8 ± 42.1 UI/d versus 103 ± 99.3 UI/d.

As shown in Fig. 2, the two groups did not differ in terms of caloric intake (total, enteral, or parenteral). Throughout the study period, all patients received a standard enteral diet; that is, protein and lipid contents did not differ.
Fig. 2

Total caloric intake during the study period. There were no statistically significant differences between the two groups at any of the time points evaluated

Total cholesterol and triglycerides remained unchanged throughout the study period in both groups (Table 2). As can be seen in Table 2, HDL also remained unchanged (consistently low) throughout the study period in both groups. Levels of LDL increased significantly over the course of the study, being statistically different from baseline values in both groups at 48 and 72 h post-admission (P < 0.05). There were no statistically significant differences between the two groups in terms of LDL levels (Table 2). Although levels of oxLDL tended to increase over time in both groups, the differences compared with baseline values did not achieve statistical significance at any of the time points evaluated. There were no statistically significant differences between the two groups in terms of oxLDL levels (Table 2).
Table 2

Comparison of key characteristics


Conservative glycemic control (n = 35)

Intensive glycemic control (n = 28)


Total cholesterol 24 h (mg/dl)

76.5 (52.8–104.3)

97 (72–115)


Total cholesterol 48 h (mg/dl)

86 (64–113)

97 (77.5–118.3)


Total cholesterol 72 h (mg/dl)

95.5 (69.5–123.5)

105 (92.5–122.5)


 Triglycerides 24 h (mg/dl)

120 (80–160.8)

117 (91–192)


 Triglycerides 48 h (mg/dl)

137 (104–157)

109 (75.8–162.5)


 Triglycerides 72 h (mg/dl)

114 (83.8–160)

108 (89.5–174)


  HDL 24 h (mg/dl)

12.5 (8.8–22)

17 (12–21)


  HDL 48 h (mg/dl)

12 (8–18)

15.5 (11–22.3)


  HDL 72 h (mg/dl)

16 (10.3–25.3)

17 (11–20)


  LDL 24 h (mg/dl)

35 (20.5–50)

44 (26.5–61)


  LDL 48 h (mg/dl)

43 (28–59)

42 (36.3–64.3)*


  LDL 72 h (mg/dl)

53.5 (32.8–71.8)

62 (47.3–70.8)*


  ox-LDL 24 h (mg/dl)

18.8 (14.1–24.6)

18.5 (15.7–27.4)


  ox-LDL 48 h (mg/dl)

15.2 (14.2–18.1)

22.1 (17.2–27.6)


  ox-LDL 72 h (mg/dl)

17.8 (13.3–22.9)

23.5 (20–28.4)


Data expressed as median and IQR (25th–75th percentile). Non-Gaussian data

NS not significant

P < 0.05 in comparison to baseline values

* denotes P < 0.05 conservative compared to intensive group

Figure 3 shows that, at 4 h post-admission, FFA levels were significantly higher in the conservative glycemic control group than in the intensive glycemic control group (P < 0.05). At hour 48, FFA levels were significantly lower in comparison with baseline values in both groups (P < 0.05). However, by hour 72, FFA levels had increased slightly in the conservative group, whereas the progressive decrease had continued in the intensive group, resulting in a significant difference between groups (P < 0.05). There were also significant differences between survivors and nonsurvivors in the time course of FFA levels over the course of the study (P < 0.05). The intensive and conservative groups differed significantly in terms of lactate clearance and base excess (P < 0.05). There were no differences in mean arterial pressure or in fluid balance between the groups for any period (Figs. 1, 2 ESM).
Fig. 3

Comparison of free fatty acids (FFAs) between the two groups (a), as well as between survivors and nonsurvivors (b); comparison of serum lactate between the two groups (c) and base excess (d). a At 4 h post-admission, FFAs were significantly lower in the intensive glycemic control group, as shown in (a). In both groups, FFAs were significantly lower at hour 24 than at baseline, and there was a significant difference between the two groups at hour 72. b There were significant differences between survivors and nonsurvivors, in terms of FFAs, at 24, 48, and 72 h post-admission. c At 24 and 48 h post-admission, lactate was significantly lower in the intensive group. d Similarly, the recovery of base excess was significantly better in the intensive group at hours 48 and 72. *P < 0.05 between intensive and conservative groups; P < 0.05 versus baseline; P < 0.05 between survivors and nonsurvivors

Severe episodes of hypoglycemia (blood glucose below 2.2 mM) were seen in two patients in each group (6 and 7% in the conservative and intensive groups, respectively; only four episodes in approximately 100 measurements). We considered hypoglycemia to be one hourly sample below 2.2 mM, and to exclude measurement error these values were confirmed twice. Those hypoglycemic episodes did not provoke any severe complications, such as neurologic disorders or significant cardiovascular alterations. Severe complications as defined in the current literature were acute neurologic complications as cognitive impairment, seizures, agitation, and acute cardiologic complications such as sudden hemodynamic instability during hypoglycemia episodes, new onset arrhythmias, or cardiac arrest. There were no differences between the intensive and conservative groups in terms of the need for dialysis (25% in both) or the in-hospital mortality rate (18 vs. 28%). However, the 28-day mortality rate did differ significantly between the intensive and conservative groups (11 vs. 28%, P < 0.05). In terms of the need for norepinephrine to maintain blood pressure, there was a significant difference between the two groups for the 24- to 48-h period, as well as a trend toward a difference for the 48- to 72-h period (Table 3).
Table 3

Comparison of patient evolution and other key variables


Conservative glycemic control (n = 35)

Intensive glycemic control (n = 28)


Norepinephrine 0–24 h (µg/kg/min)

0.41 ± 0.48

0.22 ± 0.20


Norepinephrine 24–48 h (µg/kg/min)

0.56 ± 0.63

0.20 ± 0.11


Norepinephrine 48–72 h (µg/kg/min)

0.25 ± 0.38

0.05 ± 0.10


Need for hemodialysis, n (%)

9 (25)

7 (25)


Severe hypoglycemia, n (%)

2 (6)

2 (7)


28-day mortality, n (%)

10 (28)

3 (11)


In-hospital mortality, n (%)

10 (28)

5 (18)


Data expressed as mean ± SD, except where otherwise indicated

Student’s t test for intergroup comparisons; mortality analyzed by log-rank test


In our sample of patients with sepsis, we found that the serum lipid profile was characterized by low levels of cholesterol, together with high levels of triglycerides, FFAs, and oxLDL compared to published data in the literature [27, 28]. At 4 h after admission, serum levels of FFAs were reduced in the intense glycemic control group, to levels significantly lower than those observed in the conservative glycemic control group. At hours 24 and 48, there were improvements in FFA levels in both groups. It is of note that, between hour 48 and hour 72, there was no further improvement in the conservative group, whereas the progressive improvement continued in the intensive group, in which FFA levels had normalized by hour 72. In addition, the comparison between survivors and nonsurvivors showed that FFAs remained elevated throughout the study period in the nonsurvivors. Lactic acid metabolism was also altered by the more intensive glycemic control, patients in the intensive group showing more rapid clearance of lactic acid and correction of base excess.

Our primary hypothesis was that glycemic control has a protective effect on serum oxLDL levels in patients with sepsis. High levels of oxLDL have been shown to correlate with an insulin resistant state [14]. Over the course of the study, there were improvements in the serum levels of blood glucose, triglycerides, and FFAs but not in those of oxLDL. Although treatment for hyperglycemia with extra insulin was able to reduce the metabolic changes, it did not change the oxidative stress. Thomas et al. [29] showed that LDL cannot be oxidized without being exposed to reactive oxygen species. Septic patients have severe oxidative stress and low levels of HDL, and then they are predisposed to present an increased production of oxLDL [9, 10]. Receptor uptake of oxLDL amplifies the oxidative stress by inducing reactive oxygen species, reducing nitric oxide, and activating nuclear factor kappa B [30]. The regular clearance of oxLDL is quite rapid with more than 90% reportedly being removed from serum over a 5-min period [31, 32]; there are studies that show a fast LDL clearance in humans [33]. However, in our patients, high levels of oxLDL persisted for the entire 72 h of the study period. Taken together, these indicate continuous production or altered clearance of oxLDL.

In studies of patients with sepsis or septic shock, our group has previously demonstrated that, on the first and sixth day of sepsis, nonsurvivors present higher serum levels of FFAs than do survivors [19, 34]. We have also shown that FFA levels are elevated in patients with cardiac mitochondrial damage and changes in heart rate variability [19, 34], as well as that oxidative stress activates poly(adenosine diphosphate-ribose) polymerase in the hearts of nonsurvivors of sepsis [34].

In critically ill patients, serum levels of FFAs are elevated, resulting of higher rates of adipose tissue lipolysis [35]. The increased availability of FFAs is mediated by counter-regulatory hormones and tumor necrosis factor alpha. Tumor necrosis factor alpha and interleukin-1 also promote de novo lipogenesis in the liver and might be responsible for impaired triglyceride removal in peripheral tissues [35]. We were surprised to find that FFA levels, after decreasing progressively in the first 48 h after ICU admission, increased in the conservative glycemic control group between hour 48 and the end of the study period (hour 72), whereas they continued to decrease in the intensive glycemic control group. Insulin release has been shown to be normal during the first 48 h but inadequate thereafter [36, 37]. Therefore, after 72 h, inadequate insulin secretion of septic patients can affect the regular control of FFAs.

Glucose–insulin–potassium (GIK) therapy has been used in myocardial infarction, based on the assumption that GIK lowers circulating FFA levels, which are detrimental during cardiac reperfusion [38]. In experimental models of myocardial infarction and in patients having undergone heart surgery, acute stress promotes elevated serum levels of FFAs within 4–6 h after the insult, and the use of insulin has been shown to reduced those levels [38, 39]. In the present study, we measured FFAs at 4 h after ICU admission and found that intensive glycemic control was more effective than conservative glycemic control in reducing their serum levels. FFAs are toxic to the ischemic myocardium and can lead to cardiac cell membrane damage, calcium overload, and arrhythmias, particularly in cases of insulin deficiency [40].

Elevated levels of FFAs contribute to metabolic acidosis, as well as to cell and organ dysfunction [40]. In surgical patients, the use of an insulin clamp in order to reduce FFA levels has also been shown to reduce levels of acetoacetate and beta-hydroxybutyrate [38]. In the present study, we found that the clearance of lactic acid was more rapid and the improvement in base excess was greater in the intensive glycemic control group. Patients in the intensive glycemic control group required lower doses of norepinephrine in the period between hours 48 and 72. These metabolic and hemodynamic differences may explain our result of reduced mortality in the intensive glycemic control. Other studies had shown severe hypoglycemia episodes more frequent among intensive glycemic control groups. These studies correlated these hypoglycemic episodes with mortality. We had only four episodes of hypoglycemia; the few hypoglicemia episodes was probably a result of the training program with the ICU team performed during 2 months before this study. Our hypothesis based on these data is that metabolic control should target plasma levels of FFAs rather than those of blood glucose.

On the other hand, we have to highlight the limitations of our study. First it was a unicenter trial and we judge that our sample of patients is small for mortality analysis. Nevertheless the patients were maintained in the same study arm during the whole period of ICU stay in order to allow coherent analysis of mortality data, and the project was designed for a metabolic study. In that way we need to be cautious in the interpretation of mortality data. In addition, that kind of study cannot be blinded, which is another limitation.

In conclusion, intensive glycemic control can reduce FFA, but not oxLDL. Therefore, FFAs might represent a better therapeutic target in patients with sepsis.


The São Paulo Research Foundation (FAPESP)- 04/02161-2.

Supplementary material

134_2011_2458_MOESM1_ESM.ppt (119 kb)
Supplementary material 1 (PPT 119 kb)
134_2011_2458_MOESM2_ESM.ppt (118 kb)
Supplementary material 2 (PPT 118 kb)

Copyright information

© Copyright jointly held by Springer and ESICM 2012