European Journal of Internal Medicine 23 (2012) 564574

Contents lists available at SciVerse ScienceDirect

European Journal of Internal Medicine

journal homepage: www.elsevier.com/locate/ejim

Original article

Intensive versus conventional glucose control in critically ill patients: A meta-analysis of randomized controlled trials
Yan Ling, Xiaomu Li, Xin Gao
Department of Endocrinology and Metabolism, Zhongshan Hospital, Fudan University No. 180 Fenglin Road, Shanghai 200032, China

a r t i c l e

i n f o

a b s t r a c t
Background: Critically ill patients commonly develop hyperglycemia. It remains unclear, however, to what extent correcting hyperglycemia will benet these patients. We performed this meta-analysis to evaluate the benets and risks of intensive glucose control versus conventional glucose control in critically ill adult patients. Methods: A systematic literature search of MEDLINE, PubMed, and Cochrane databases (until June 2011) was conducted using specic search terms. Randomized controlled trials that compared intensive glucose control with a target glucose goal b 6.1 mmol/l (110 mg/dl) to conventional glucose control in adult intensive care patients were included. The random-effect model was used to estimate the pooled risk ratio of the two treatment arms. Results: Twenty two studies that randomized 13,978 participants were included in the meta-analysis. Overall, intensive glucose control did not reduce the short-term mortality (RR = 1.02, 95% CI: 0.951.10, p = 0.51), 90 day or 180 day mortality (RR = 1.06, 95% CI: 0.991.13, p = 0.08), sepsis (RR = 0.96, 95% CI: 0.831.12, p = 0.59) or new need for dialysis (RR = 0.96, 95% CI: 0.831.11, p = 0.57). The incidence of hypoglycemia was signicantly higher in intensive glucose control group compared with conventional glucose control group (RR = 5.01, 95% CI: 3.457.28, p b 0.00001). Conclusions: This meta-analysis found that intensive glucose control in critically ill adults did not reduce mortality but is associated with a signicantly increased risk of hypoglycemia. 2012 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved.

Article history: Received 26 December 2011 Accepted 28 February 2012 Available online 24 March 2012 Keywords: Intensive glucose control Conventional glucose control Critically ill Randomized controlled trials Meta-analysis

1. Introduction Hyperglycemia is an independent risk factor for poor outcomes of numerous surgical and medical conditions [13]. Hyperglycemia may contribute to poor immune function, oxidative stress, increased susceptibility to infectious complications, impaired recovery of organ failure, and endothelial and myocardial dysfunction [4,5]. Patients in the intensive care unit (ICU) commonly develop hyperglycemia [4,6]. It remains unclear, however, to what extent correcting hyperglycemia will benet these critically ill ICU patients. During the past decade, many clinical trials investigated the effects of intensive insulin therapy on clinical outcomes of critically ill patients. Van den Berghe and colleagues found that intensive insulin therapy to maintain blood glucose at or below 6.1 mmol/l reduced morbidity and mortality among critically ill patients in the surgical ICU [7]. Following this study, a similar clinical trial was conducted

by Van den Berghe et al. in the medical ICU [8]. Although no benecial effect on overall mortality was found, this study demonstrated that tight glycemic control reduced the mortality in the subset of patients with an ICU stay of 3 days or more [8]. However, recent clinical trials failed to demonstrate a signicant improvement in mortality with intensive glucose control in critically ill patients [912]. The largest study to date, NICE-SUGAR [13], a multicenter, multinational randomized controlled trial among 6104 critically ill medical and surgical patients even found an increased risk of mortality with intensive glucose control. Moreover, these studies consistently showed that intensive glucose control was associated with an increased risk of severe hypoglycemia. Based on the conicting evidences, we conducted this metaanalysis to evaluate the benets and risks of intensive versus conventional glucose control in critically ill ICU patients. 2. Methods

Abbreviations: ICU, intensive care unit; RCT, randomized controlled trial; CG, conventional glucose control; TG, intensive glucose control; BG, blood glucose; RR, risk ratio; CI, condence interval; SD, standard deviation; AACE, American Association of Clinical Endocrinologists; ADA, American Diabetes Association. Corresponding author. Tel./fax: + 86 21 64439025. E-mail address: gao.xin@zs-hospital.sh.cn (X. Gao).

2.1. Search strategy We searched MEDLINE, PubMed, and Cochrane databases (Central Register of Controlled Trials and Cochrane Database of Systematic Reviews) (until June 2011) to identify the randomized trials examining

0953-6205/$ see front matter 2012 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.ejim.2012.02.013

Y. Ling et al. / European Journal of Internal Medicine 23 (2012) 564574

565

MEDLINE Potentially relevant studies (n=2660)

Cochrane databases Potentially relevant studies (n=404)

PubMed Potentially relevant studies (n=1150)

Excluded (n=2638): After title review (n=2464) After abstract review (n=161) After full-text article review (n=13) 6 Post-hoc/subgroup analysis of RCTs 3 Intervention began at the start of surgery 3 Target glucose in the intervention group over 110 mg/dl 1 full-text not available

Excluded (n=383): After title review (n=315) After abstract review (n=43) After full-text article review (n=25) 10 Post-hoc/subgroup analysis of RCTs 4 Intervention began at the start of surgery 3 Target glucose in the intervention group over 110 mg/dl 2 No outcome reported 2 Cohort studies 1 No target glucose defined 3 Commentary articles

Excluded (n=1128): After title review (n=1002) After abstract review (n=93) After full-text article review (n=33) 14 Post-hoc/subgroup analysis of RCTs 5 Intervention began at the start of surgery 3 Target glucose in the intervention group over 110 mg/dl 4 No outcome reported 2 Cohort studies 1 Non-randomized trial 3 Commentary articles 1 duplicate report

Eligible RCTs (n=22)

Eligible RCTs (n=21)

Eligible RCTs (n=22)

No additional studies identified from references of relevant articles Eligible RCTs included in current meta-analysis (n=22)
Fig. 1. Flow diagram of literature search and study selection. RCT: randomized controlled trial.

the impact of intensive insulin therapy on critically ill patients. The following search formula was used: (intensive care units OR intensive care OR critical care OR critical illness OR postoperative care OR cardiac care facilities OR ICU OR CCU OR coronary care OR recovery room OR burn unit OR critically ill OR cardiac care OR cardiac care units) AND (blood glucose OR insulin OR intensive insulin OR glycemic control) AND (clinical trial OR randomized OR trial). The references of relevant articles were also reviewed to identify any potentially eligible studies. 2.2. Study selection

mortality. The secondary outcomes included 90 or 180-day mortality, hypoglycemia, sepsis and new need for dialysis. We dened hypoglycemia as patients with at least one blood glucose measurement lower than 2.2 mmol/l. We dened sepsis to encompass the terms sepsis, septicemia, bacteremia or a description of positive blood cultures. New need for dialysis was dened as patients without a preexisting dialysis requirement who subsequently developed acute renal failure that required dialysis.

2.4. Statistical analysis Eligible clinical trials were published in English and met the following criteria: (1) trials adopting randomized, controlled design; (2) the setting was an adult ICU; (3) the intervention group received intensive insulin therapy with a target glucose level b 6.1 mmol/l (110 mg/dl), and the control group received conventional insulin therapy with a higher target glucose level; (4) trials reported at least one of the following outcomes: mortality, hypoglycemia, sepsis or new need for dialysis. Studies were excluded if the intervention began before or during the surgery rather than during the ICU stay or if adequate details of the study were not available. We also excluded trials with a sample size less than 20. 2.3. Data abstraction, quality assessment and outcome denition Two investigators (YL and XL) independently assessed and abstracted pertinent data from trials in duplicate. They also assessed the methodological quality of the trials using the Jadad scale [14] to derive a score of 0 to 5, with higher scores indicating higher quality. The investigators resolved discrepancies through discussion, and a third investigator (XG) resolved any disagreement if necessary. Abstracted data included year of publication, each study's methodology, ICU setting, baseline patient characteristics, interventions, mean or median glucose concentration achieved, follow-up duration, and clinical outcomes. The primary outcome was short-term mortality, dened as 28-day mortality. If this outcome was not recorded, we preferentially used the outcomes in the following order: hospital mortality and ICU We used the risk ratio (RR) as the summary measure of association. We obtained the RR with 95% condence interval (CI) by pooling trial-specic cumulative incidence ratios of the included trials using random-effects model. For studies with no event of interest in a treatment group, 1.0 was added to all cells for continuity correction [15]. Statistical heterogeneity was measured using the I 2 statistic (a scale of 0100%, >75% represented very large between-study inconsistency) [16]. We identied the type of ICU (medical, surgical or mixed) as potential sources of heterogeneity and conducted prespecied subgroup analysis based on ICU setting. Sensitivity analyses were conducted by omitting each trial one at a time from analysis, and by deleting trials with high weight, and thereafter computing metaanalysis estimates for the remaining studies. We also did subgroup analysis on the following variables: 1) the mean blood glucose level achieved in the intensive glucose control group; 2) the difference of mean blood glucose level achieved in the intensive and conventional glucose control groups; 3) the standard deviation (SD) of mean blood glucose level (as an index of glycemic variability) achieved in the intensive glucose control group; 4) the percentage of patients who were diabetic; 5) the mean APACHE II score; 6) the percentage of medical patients; 7) the mean daily insulin dose in the intensive glucose control group. We used the median of the above variables as the cut points for subgroup analysis. We quantitatively assessed publication bias using the Begg adjusted-rank correlation test [17] and Egger regression asymmetry test [18] (p value b 0.05 considered signicant). Analyses were performed with the RevMan 5.0 software (The

566

Table 1 Characteristics of randomized controlled trials meeting inclusion criteria. Y. Ling et al. / European Journal of Internal Medicine 23 (2012) 564574 Study/year No. of Reason for patients admission, % Mean Male sex, % age, year 62.8 71 Diabetes, % Mean APACHE II score Follow up duration Intervention Mean daily insulin dose, IU/d CG: 33 TG: 71 Target blood glucose level, mmol/l CG: 1011.1 TG: 4.46.1 Achieved blood glucose level, mmol/l Mean morning BG CG: 8.5 TG: 5.7 NA Mean BG CG: 7.96 TG: 5.13 NA Mean BG CG: 8.06 TG: 6.85 Mean BG CG: 9.9 TG: 5.5 Outcomes included in meta analysis

Surgical ICU Van den Berghe et al. [7]

1548

Cardiac surgery, 63

13

9.0 Hospital (median) stay

Both groups: insulin infusion

Mortality, hypoglycemia, new need for dialysis, sepsis Hypoglycemia Mortality, sepsis

Hoedemaekers et al. [20] Bilotta et al. [21]

20 483

CABG, 100 Neurosurgery, 100

64.2 57.1

90 58

0 10

9.1 NA

Hospital stay 6 months

Both groups: insulin infusion Both groups: insulin infusion

NA CG: 21 TG: 54

CG: b 11.1 TG: 4.46.1 CG: b 11.94 TG: 4.446.11 CG: 1011.1 TG: 4.46.1 CG: b 10 TG: 4.46.1

Meng et al. [22] Coester et al. [23]

240 88

Severe traumatic brain injury, 100 Severe traumatic brain injury, 100

45.5 38.5

66 86

10 1.2

9.5 6 months (median) 15.5 6 months (median)

Cao et al. [24]

179

Gastric cancer, 100

58.8

32

100

NA

28 days

Both groups: NA insulin infusion NA CG: subcutaneous insulin injection or insulin infusion TG: insulin infusion Both groups: NA insulin infusion

Mortality, hypoglycemia Mortality, hypoglycemia, sepsis

CG: 1011.0 TG: 4.46.1

Mortality, hypoglycemia, sepsis

Medical ICU Van den Berghe et al. [8]

1200

Respiratory, 42.7; gastrointestinal, liver, 25.5 Out of hospital ventricular brillation, 100 Respiratory, 32; sepsis, cardiovascular, neurologic, 44 Acute ischemic stroke, 100

63.5

62

17

23

90 days

Both groups: insulin infusion

CG: 10 TG: 59

CG: 1011.1 TG: 4.46.1

Oksanen et al. [25]

90

63.7

79

12

24.9 30 days (median) 20.5 90 days (median) NA 120 days

Both groups: insulin infusion CG: subcutaneous insulin injection TG: insulin infusion CG: subcutaneous insulin injection TG: insulin infusion CG: subcutaneous insulin injection or insulin infusion TG: insulin infusion

CG: 12.5 TG: 22 NA

CG: 6.08.0 TG: 4.06.0 CG: b 8.3 TG: 4.46.1 CG:b11.1 TG: 4.446.11 CG: 8.3 TG: 4.46.1

Cavalcanti et al. [26] 112

59.9

46

30

Kreisel et al. [27]

40

71.6

60

33

CG: 5.4 TG: 13.3 Mean insulin infusion rate CG: 1.4 IU/h TG: 2.39 IU/h

Green et al. [28]

81

Ischemic stroke, intracerebral 51 hemorrhage, subarachnoid hemorrhage, 35; traumatic brain injury, 49

69

NA

NA

90 days

Mean morning BG CG: 8.49 TG: 6.16 Median BG CG: 6.4 TG: 5.0 Median BG CG: 8.8 TG: 7.1 Mean BG CG: 8.01 TG: 6.49 Mean BG CG: 7.9 TG: 6.2

Mortality, hypoglycemia, new need for dialysis, sepsis

Mortality, hypoglycemia

Hypoglycemia

Mortality

Mortality, hypoglycemia, sepsis

Mixed ICU Mitchell et al. [30]

70

Medical: 61 Surgical: 3 9 Medical: 83 Surgical: 17 Sepsis Medical: 47 Surgical: 53 Medical: 49 Surgical: 16 Trauma: 35 Sepsis Medical: 64 Surgical: 32 Medical: 62 Surgical: 38 Medical: 40 Surgical: 47 Trauma: 13 Medical: 75 Surgical: 25 Medical: 62 Surgical: 38 Medical: 75 Surgical: 11

65.4

60

14

20.5 Hospital (median) stay 22.8 Hospital stay 90 days

Both groups: insulin infusion Both groups: insulin infusion Both groups: insulin infusion Both groups: insulin infusion Both groups: insulin infusion Both groups: insulin infusion Both groups: insulin infusion

Arabi et al. [9]

523

52.4

75

40

CG: 0 TG: 35.7 (median) CG: 31.4 TG: 71.2 CG: 5 TG: 32 (median) CG: 12.5 TG: 52.4 CG: 38.8 TG: 74.5 CG: 16.9 TG: 50.2

CG: 1011.1 TG: 4.46.1 CG: 1011.1 TG: 4.46.1 CG: 1011.1 TG: 4.46.1 CG: 1011.1 TG: 4.46.1 CG: 1011.1 TG: 4.46.1 CG: b 10 TG: 4.56.0

Brunkhorst [10]

et

al. 537

64.6

60

30

20.3

De La Rosa et al. [11] 504

46.6

60

12

15.7

Hospital stay 90 days

Iapichino et al. [29]

90

62.3

65

17

NA

NICE-SUGAR study [13] Glucontrol [12]

6104

62.2

63

20

21.1

90 days

study 1101

64.6

62

18

15 Hospital (median) stay

Taslimi et al. [31] Savioli et al. [32]

129 90

55.5 61

52 62

53 13

10 NA

Hospital stay 90 days

Both groups: insulin infusion Both groups: insulin infusion CG: insulin infusion or subcutaneous insulin injection TG: insulin infusion Both groups: insulin infusion

Median insulin CG: 7.810.0 TG: 4.46.1 infusion rate CG: 0.32 IU/h TG: 1.30 IU/h NA CG: 6.912.5 TG: 4.46.1 CG: 36 CG: 1011.1 TG: 57 TG: 4.46.1 CG: 46 TG: 71 (median)

Median BG CG: 7.9 TG: 5.4 Mean BG CG: 9.5 TG: 6.4 Mean morning BG CG: 8.4 TG: 6.2 Median morning BG CG: 8.2 TG: 6.5 Mean BG CG: 9.0 TG: 6.1 Mean BG CG: 8.0 TG: 6.4 Median BG CG: 8.0 TG: 6.5 NA

Mortality, hypoglycemia

Mortality, hypoglycemia, new need for dialysis, sepsis Mortality, hypoglycemia, new need for dialysis Mortality, hypoglycemia, new need for dialysis Mortality, hypoglycemia

Mortality, hypoglycemia, new need for dialysis, sepsis Y. Ling et al. / European Journal of Internal Medicine 23 (2012) 564574 Mortality, hypoglycemia

COIITSS study [33]

509

64

65

NA

NA

180 days

Mean BG CG: 8.8 TG: 6.2 CG: not dened NA TG: 4.46.1

Mortality, new need for dialysis Mortality

Mortality, hypoglycemia

Arabi et al. [34]

240

Medical: m83 Surgical: m17

51.1

68

40

25.3

180 days

CG: 23 TG: 62.8

CG: 1011.1 TG: 4.46.1

Mean BG CG: 8.6 TG: 6.2

Mortality, hypoglycemia, new need for dialysis, sepsis

ICU: intensive care unit; CG: conventional glucose control; TG: intensive glucose control; BG: blood glucose.

567

568

Y. Ling et al. / European Journal of Internal Medicine 23 (2012) 564574

Cochrane Collaboration, Copenhagen, Denmark). The meta-analysis was prepared according to the PRISMA guideline [19]. 3. Results 3.1. Literature search Detailed search steps were described in Fig. 1. Briey, from the initial literature search we identied 4214 potentially eligible studies. After screening of study titles and abstracts, 136 articles were considered of interest and full text was retrieved for detailed evaluation. Due to the inconsistency with the predened inclusion criteria, 114 of the 136 articles were subsequently excluded. Meanwhile, references of the remaining 22 articles were reviewed, and no additional relevant study was identied. Thus, a total of 22 [713,2034] eligible randomized controlled trials were included in the meta-analysis. 3.2. Study characteristics The baseline characteristics of the enrolled studies were shown in Table 1. Among 22 included trials, 6 trials were conducted in surgical ICUs, 5 in medical ICUs, and 11 in mixed ICUs. A total of 13,978 patients were included for analysis with 6966 allocated to intensive glucose control group and 7012 to conventional glucose control group. The target glucose level of intensive insulin therapy group for all trials was lower than 6.1 mmol/l. However, the target glucose levels of conventional insulin therapy groups were not consistent between trials. Mean achieved glucose levels also varied between trials in both intensive control and conventional care groups. No trials could receive a Jadad score higher than 3 out of 5 due to the lack of double-blind study design (Table 2). 3.3. Meta-analysis for clinical outcomes 3.3.1. Mortality Seventeen trials provided usable data on short-term mortality. Pooled analysis of these trials showed that intensive glucose control did not signicantly reduce short-term mortality compared with conventional glucose control (RR = 1.02, 95% CI: 0.951.10, p = 0.51) (Fig. 2). Subgroup analysis stratied by ICU setting also did not demonstrate a statistically signicant inter-group difference (surgical ICU: RR = 0.82, 95% CI: 0.631.07, p = 0.15; medical ICU: RR = 0.99, 95% CI: 0.841.17, p = 0.91; mixed ICU: RR = 1.06, 95% CI: 0.991.14, p = 0.10) (Fig. 2). There was no evidence for heterogeneity across the individual trials in the primary analysis (I 2 = 8%) or subgroup analysis (I 2 = 012%) (Fig. 2). Eleven trials reported 90-day or 180-day mortality. Pooled analysis of these trials showed that intensive glucose control was not associated with a reduced 90-day or 180-day mortality compared with conventional glucose control (RR = 1.06, 95% CI: 0.991.13, p = 0.08) (Fig. 3). Moreover, we found that patients in the intensive insulin therapy group had an increased risk of 90-day or 180-day mortality compared with those in the control group in the mixed ICU (RR = 1.10, 95% CI: 1.021.19, p = 0.01) (Fig. 3). However, there was no signicant difference in 90-day or 180-day mortality between intensive and conventional glucose control in the surgical and medical ICU (surgical ICU: RR= 0.96, 95% CI: 0.801.15, p = 0.67; medical ICU: RR= 0.98, 95% CI: 0.841.16, p = 0.84) (Fig. 3). No heterogeneity was found across the individual trials in the primary analysis (I2 = 0%) or subgroup analysis (I2 = 02%) (Fig. 3). 3.3.2. Hypoglycemia Hypoglycemia dened as blood glucose level lower than 2.2 mmol/l was reported in eighteen trials. In pooled analysis of these trials, intensive glucose control signicantly increased the risk of hypoglycemia compared with conventional glucose control

(RR = 5.01, 95% CI: 3.457.28, p b 0.00001) (Fig. 4). In addition, intensive glucose control consistently increased the risk of hypoglycemia compared with conventional glucose control in all ICU settings (surgical ICU: RR = 3.90, 95% CI: 1.609.49, p = 0.003; medical ICU: RR = 6.03, 95% CI: 3.899.34, p b 0.00001; mixed ICU: RR = 5.17, 95% CI: 3.018.89, p b 0.00001) (Fig. 4). Test for heterogeneity found that the COIITSS study [33] and the NICE-SUGAR study [13] had outlying results in subgroup analysis of the mixed ICU (I 2 = 77%) (Fig. 4). Exclusion of the two outlying trials resolved this heterogeneity (I 2 = 16%) but did not signicantly change the nding in the mixed ICU (RR = 4.87, 95% CI: 3.436.93, p b 0.00001). 3.3.3. Sepsis and new need for dialysis Sepsis and new need for dialysis were reported in 10 trials and 8 trials respectively. Pooled analysis showed that intensive glucose control did not signicantly reduce rates of sepsis (RR = 0.96, 95% CI: 0.831.12, p = 0.59) or new need for dialysis (RR = 0.96, 95% CI: 0.831.11, p = 0.57) compared with conventional glucose control (Fig. 56). Subgroup analysis stratied by ICU setting also did not demonstrate any benecial effect of intensive glucose control on sepsis or new need for dialysis (Fig. 56). Test for heterogeneity found that the study by Coester et al. [23] had outlying result in subgroup analysis for sepsis in the surgical ICU (I 2 = 78%) (Fig. 5). Exclusion of this outlying trial resolved the heterogeneity (I 2 = 0%) and also signicantly changed the nding. Intensive glucose control was found to be associated with a reduced risk of sepsis compared with conventional glucose control in the surgical ICU (RR = 0.58, 95% CI: 0.40 0.84, p = 0.004). 3.3.4. Subgroup analysis Stratied trials according to variables prespecied also showed that intensive glucose control did not reduce mortality signicantly (Table 3). 3.3.5. Sensitivity analysis We did sensitivity analysis by omitting each trial once a time from analysis or by deleting trials with high weight, which did not change any nding signicantly. 3.3.6. Publication bias Neither the Egger's nor the Begg's test based on the data of shortterm mortality and hypoglycemia showed statistical signicances (Egger's test: p = 0.96, 0.46, respectively; Begg's test: p = 0.77, 0.55, respectively), suggesting that there was no signicant publication bias existing among the included trials. 4. Discussion In this meta-analysis of randomized controlled trials of intensive glucose control versus conventional glucose control in critically ill adult patients, we found that intensive glucose control had no impact on the overall risk of short-term mortality, 90 day or 180 day mortality, sepsis or new need for dialysis. Subgroup analysis stratied by ICU setting showed that intensive glucose control even increased the risk of 90-day or 180-day mortality in patients from the mixed ICU. Although intensive glucose control may benet patients in the surgical ICU and decrease the risk of sepsis by 42% compared with conventional glucose control when excluding the outlying trial, it had no impact on the overall risk of sepsis. On the other hand, intensive insulin therapy was associated with a 5-fold increased risk of severe hypoglycemia compared with the conventional therapy. Moreover, the risk of hypoglycemia consistently increased in all ICU settings. Our meta-analysis reached a similar conclusion of lack of benet and increased risk of hypoglycemia associated with intensive glucose control in ICU patients as in the previous studies [3537]. The study by Wiener et al. also found that intensive glucose control had a

Y. Ling et al. / European Journal of Internal Medicine 23 (2012) 564574 Table 2 Jadad quality score and other quality criteria of randomized controlled trials included in the meta-analysis. Study Jadad quality score Randomized Method to generate randomization sequence described and appropriate Surgical ICU Van den Berghe et al. [7] Hoedemaekers et al. [20] Bilotta et al. [21] Meng et al. [22] Coester et al. [23] Cao et al. [24] Medical ICU Van den Berghe et al. [8] Oksanen et al. [25] Cavalcanti et al. [26] Kreisel et al. [27] Green et al. [28] Mixed ICU Mitchell et al. [30] Arabi et al. [9] Brunkhorst et al. [10] De La Rosa et al. [11] Iapichino et al. [29] NICE-SUGAR study [13] Glucontrol study [12] Taslimi et al. [31] Savioli et al. [32] COIITSS study [33] Arabi et al. [34] Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Withdrawals Total Method of Method to generate Double- Method of blind double-blinding double-blinding and dropouts Jadad randomization score described described but described and sequence described inappropriate appropriate but inappropriate No No No No No No No No No No No No No No No No No No No No No No No No Yes Yes Yes Yes Yes Yes 3 3 3 3 3 3 Other quality criteria Description of adequate allocation concealment Yes Yes Yes Yes Yes Yes

569

Analysis on basis of intentionto-treat Yes No Yes Yes Yes No

Yes Yes Yes Yes Yes

Yes Yes Yes Yes Yes

No No No No No

No No No No No

No No No No No

No No No No No

Yes Yes Yes Yes Yes

3 3 3 3 3

Yes Yes Yes Yes Yes

Yes Yes Yes Yes Yes

Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Yes Yes Yes Yes Yes Yes Yes No Yes Yes Yes

No No No No No No No No No No No

No No No No No No No No No No No

No No No No No No No No No No No

No No No No No No No No No No No

Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

3 3 3 3 3 3 3 2 3 3 3

Yes No Yes Yes No Yes Yes No No Yes Yes

Yes Yes Yes Yes Yes Yes Yes No Yes Yes Yes

ICU: intensive care unit.

benecial effect on the risk of sepsis in the surgical ICU patients which is similar to our nding [35]. However, this benecial effect is overweighed by the neutral effect on mortality and the harmful effect of increased hypoglycemia. The study by Griesdale et al. found that intensive glucose control decreased the risk of mortality in the surgical ICU, although it had no effect on the overall pooled estimate on mortality [36]. Both Wiener's and our studies did not demonstrate such a benecial effect on mortality in the surgical ICU, suggesting that more evidences are needed to conrm this potential benet. The deleterious effects associated with hyperglycemia during critically ill conditions are well studied, including increased production of reactive oxygen species, exacerbation of inammatory pathways, decreased complement activity, modication in the innate immune system, impairment in endothelial function, abolishment of ischemic preconditioning, and protein glycosylation [4,5,38,39]. Multiple observational studies have shown that hyperglycemia is associated with poor clinical outcomes in multiple patient populations and effective glucose control can improve clinical outcomes [13,6,40,41]. Based on the above evidences, the lack of benecial effects of intensive glucose control on critically ill ICU patients in our study seems puzzling. Blood glucose lowering strategy should achieve dual goals: efcacy and safety. However, intensive insulin therapy was associated with a 5-fold increased risk of severe hypoglycemia compared with the conventional therapy in our study. Many studies

have demonstrated that there is an independent relationship between hypoglycemia and the risk of mortality in critically ill patients [42,43]. Hypoglycemia is associated with irreversible neuronal damage, cardiac arrhythmias and myocardial compromise [44]. It is biologically plausible that hypoglycemia contributes to mortality in critically ill patients with intensive insulin therapy, although the specic mechanisms remain to be elucidated [44]. On the other hand, glycemic variability has been shown to be independently associated with the risk of mortality among various populations of critically ill patients [4548]. Oxidative stress induced by glycemic variability may be responsible for an excess of mortality in these patients [49]. In the present study, we examined the effect of SD of mean blood glucose achieved in the intensive glucose control group on mortality, and did not nd a signicant inuence. Obviously, SD of mean blood glucose is not an optimal measurement for glycemic variability. It is important to develop and implement continuous glucose monitoring devices and closed-loop glucose control system in the future studies. By using the new technologies, studies can not only minimize the risk of hypoglycemia, but also prospectively evaluate the potential impact of glycemic variability on clinical outcomes. Our meta-analysis provided the most current estimate of the effect of intensive insulin therapy on critically ill adults. As in previous studies, our study did not replicate the benecial effect of intensive insulin therapy observed in the rst large landmark trial by Van den

570

Y. Ling et al. / European Journal of Internal Medicine 23 (2012) 564574

Fig. 2. Forest plot for short-term mortality comparing intensive glucose control to conventional glucose control stratied by type of intensive care unit. ICU: intensive care unit; CI: condence interval.

Berghe et al. [7]. It is possible that local features of the population of patients and distinctive approach of critical care in this single center trial may have inuenced the outcomes which cannot be replicated in other studies. The very different results of this study compared with other studies clearly indicate that the optimal level of glycemic control may be different among different populations of patients. Therefore, additional studies are needed to discriminate subgroups of critically ill patients who may benet from intensive insulin therapy. Our study had some limitations. Studies included in our meta-analysis varied with regard to baseline patients' characteristics. However, we only found heterogeneity in some subgroup analysis, which did not inuence our results greatly. Subgroup analysis stratied by ICU setting and other factors proved our results to be robust. On the other hand, the generalizability of our results may be limited by the diverse patient characteristics, different protocol and implementation of intervention, different concomitant therapy, different blood glucose monitoring methodology, and different event rates in the included trials. In summary, our meta-analysis found that intensive glucose control in critically ill adults did not reduce mortality but is associated with a signicantly increased risk of hypoglycemia. Our metaanalysis supports the guideline of American Association of Clinical Endocrinologists (AACE) and American Diabetes Association (ADA), which doesn't recommend intensive insulin therapy targeting a blood glucose concentration below 6.1 mmol/l in critically ill patients,

and a glucose range of 7.810 mmol/l is recommended for the majority of critically ill patients based on the current knowledge [50]. Pending more evidences to guide the development of optimal glucose levels, this less aggressive glucose target should be applied to critically ill patients currently. Learning points Critically ill patients commonly develop hyperglycemia, which is an independent risk factor for poor outcomes of numerous surgical and medical conditions. This meta-analysis of randomized controlled trials in critically ill adults found that intensive insulin therapy with a target glucose level b 6.1 mmol/l did not reduce mortality but is associated with a signicantly increased risk of hypoglycemia compared with the conventional therapy. Intensive glucose control in critically ill patients is not recommended currently, and more evidences are warranted to guide the development of optimal glucose levels in these patients.

Conict of interest statement The authors state that they have no conicts of interest.

Y. Ling et al. / European Journal of Internal Medicine 23 (2012) 564574

571

Fig. 3. Forest plot for 90 d or 180 d mortality comparing intensive glucose control to conventional glucose control stratied by type of intensive care unit. ICU: intensive care unit; CI: condence interval.

Acknowledgements This work was supported by the National Key Technologies R&D Program (2008 BAI52B03) and the National Basic Research Program of China (2011CB504004; http://www.973.gov.cn/Default_3.aspx). References
[1] Whitcomb BW, Pradhan EK, Pittas AG, Roghmann MC, Perencevich EN. Impact of admission hyperglycemia on hospital mortality in various intensive care unit populations. Crit Care Med 2005;33:27727. [2] Krinsley JS. Association between hyperglycemia and increased hospital mortality in a heterogeneous population of critically ill patients. Mayo Clin Proc 2003;78: 14718. [3] Capes SE, Hunt D, Malmberg K, Gerstein HC. Stress hyperglycaemia and increased risk of death after myocardial infarction in patients with and without diabetes: a systematic overview. Lancet 2000;355:7738. [4] Dungan KM, Braithwaite SS, Preiser JC. Stress hyperglycaemia. Lancet 2009;373: 1798807. [5] McCowen KC, Malhotra A, Bistrian BR. Stress-induced hyperglycemia. Crit Care Clin 2001;17:10724. [6] Umpierrez GE, Isaacs SD, Bazargan N, You X, Thaler LM, Kitabchi AE. Hyperglycemia: an independent marker of in-hospital mortality in patients with undiagnosed diabetes. J Clin Endocrinol Metab 2002;87:97882. [7] van den Berghe G, Wouters P, Weekers F, Verwaest C, Bruyninckx F, Schetz M, et al. Intensive insulin therapy in the critically ill patients. N Engl J Med 2001;345:135967. [8] Van den Berghe G, Wilmer A, Hermans G, Meersseman W, Wouters PJ, Milants I, et al. Intensive insulin therapy in the medical ICU. N Engl J Med 2006;354:44961. [9] Arabi YM, Dabbagh OC, Tamim HM, Al-Shimemeri AA, Memish ZA, Haddad SH, et al. Intensive versus conventional insulin therapy: a randomized controlled trial in medical and surgical critically ill patients. Crit Care Med 2008;36:31907. [10] Brunkhorst FM, Engel C, Bloos F, Meier-Hellmann A, Ragaller M, Weiler N, et al. Intensive insulin therapy and pentastarch resuscitation in severe sepsis. N Engl J Med 2008;358:12539. [11] De La Rosa Gdel C, Donado JH, Restrepo AH, Quintero AM, Gonzalez LG, Saldarriaga NE, et al. Strict glycaemic control in patients hospitalised in a mixed medical and surgical intensive care unit: a randomised clinical trial. Crit Care 2008;12:R120.

[12] Preiser JC, Devos P, Ruiz-Santana S, Melot C, Annane D, Groeneveld J, 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;35:173848. [13] Finfer S, Chittock DR, Su SY, Blair D, Foster D, Dhingra V, et al. Intensive versus conventional glucose control in critically ill patients. N Engl J Med 2009;360:128397. [14] Jadad AR, Moore RA, Carroll D, Jenkinson C, Reynolds DJ, Gavaghan DJ, et al. Assessing the quality of reports of randomized clinical trials: is blinding necessary? Control Clin Trials 1996;17:112. [15] Kalil AC, Murthy MH, Hermsen ED, Neto FK, Sun J, Rupp ME. Linezolid versus vancomycin or teicoplanin for nosocomial pneumonia: a systematic review and meta-analysis. Crit Care Med 2010;38:18028. [16] Ioannidis JP, Patsopoulos NA, Evangelou E. Uncertainty in heterogeneity estimates in meta-analyses. BMJ 2007;335:9146. [17] Begg CB, Mazumdar M. Operating characteristics of a rank correlation test for publication bias. Biometrics 1994;50:1088101. [18] Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315:62934. [19] Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gotzsche PC, Ioannidis JP, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate healthcare interventions: explanation and elaboration. BMJ 2009;339: b2700. [20] Hoedemaekers CW, Pickkers P, Netea MG, van Deuren M, Van der Hoeven JG. Intensive insulin therapy does not alter the inammatory response in patients undergoing coronary artery bypass grafting: a randomized controlled trial [ISRCTN95608630]. Crit Care 2005;9:R7907. [21] Bilotta F, Caramia R, Paoloni FP, Delni R, Rosa G. Safety and efcacy of intensive insulin therapy in critical neurosurgical patients. Anesthesiology 2009;110:6119. [22] Yang M, Guo Q, Zhang X, Sun S, Wang Y, Zhao L, et al. Intensive insulin therapy on infection rate, days in NICU, in-hospital mortality and neurological outcome in severe traumatic brain injury patients: a randomized controlled trial. Int J Nurs Stud 2009;46:7538. [23] Coester A, Neumann CR, Schmidt MI. Intensive insulin therapy in severe traumatic brain injury: a randomized trial. J Trauma 2010;68:90411. [24] Cao SG, Ren JA, Shen B, Chen D, Zhou YB, Li JS. Intensive versus conventional insulin therapy in type 2 diabetes patients undergoing D2 gastrectomy for gastric cancer: a randomized controlled trial. World J Surg 2011;35:8592. [25] Oksanen T, Skrifvars MB, Varpula T, Kuitunen A, Pettila V, Nurmi J, et al. Strict versus moderate glucose control after resuscitation from ventricular brillation. Intensive Care Med 2007;33:2093100. [26] Cavalcanti AB, Silva E, Pereira AJ, Caldeira-Filho M, Almeida FP, Westphal GA, et al. A randomized controlled trial comparing a computer-assisted insulin infusion

572

Y. Ling et al. / European Journal of Internal Medicine 23 (2012) 564574

Fig. 4. Forest plot for hypoglycemia comparing intensive glucose control to conventional glucose control stratied by type of intensive care unit. ICU: intensive care unit; CI: condence interval. protocol with a strict and a conventional protocol for glucose control in critically ill patients. J Crit Care 2009;24:3718. Kreisel SH, Berschin UM, Hammes HP, Leweling H, Bertsch T, Hennerici MG, et al. Pragmatic management of hyperglycaemia in acute ischaemic stroke: safety and feasibility of intensive intravenous insulin treatment. Cerebrovasc Dis 2009;27:16775. Green DM, O'Phelan KH, Bassin SL, Chang CW, Stern TS, Asai SM. Intensive versus conventional insulin therapy in critically ill neurologic patients. Neurocrit Care 2010;13:299306. Iapichino G, Albicini M, Umbrello M, Sacconi F, Fermo I, Pavlovich R, et al. Tight glycemic control does not affect asymmetric-dimethylarginine in septic patients. Intensive Care Med 2008;34:184350. Mitchell I, Knight E, Gissane J, Tamhane R, Kolli R, Leditschke IA, et al. A phase II randomised controlled trial of intensive insulin therapy in general intensive care patients. Crit Care Resusc 2006;8:28993. Taslimi RAR, Talebian MH, Abtahi HR, Jalili M, Nejati A, Labbaf A. The efcacy of intensive glucose management on hospitalized critically ill patients associated mortality rate in intensive care unit. DARU 2009;17:15762. Savioli M, Cugno M, Polli F, Taccone P, Bellani G, Spanu P, et al. Tight glycemic control may favor brinolysis in patients with sepsis. Crit Care Med 2009;37:42431. Annane D, Cariou A, Maxime V, Azoulay E, D'Honneur G, Timsit JF, et al. Corticosteroid treatment and intensive insulin therapy for septic shock in adults: a randomized controlled trial. JAMA 2010;303:3418. Arabi YM, Tamim HM, Dhar GS, Al-Dawood A, Al-Sultan M, Sakkijha MH, et al. Permissive underfeeding and intensive insulin therapy in critically ill patients: a randomized controlled trial. Am J Clin Nutr 2011;93:56977. Wiener RS, Wiener DC, Larson RJ. Benets and risks of tight glucose control in critically ill adults: a meta-analysis. JAMA 2008;300:93344. Griesdale DE, de Souza RJ, van Dam RM, Heyland DK, Cook DJ, Malhotra A, et al. Intensive insulin therapy and mortality among critically ill patients: a meta-analysis including NICE-SUGAR study data. CMAJ 2009;180:8217. Kansagara D, Fu R, Freeman M, Wolf F, Helfand M. Intensive insulin therapy in hospitalized patients: a systematic review. Ann Intern Med 2011;154:26882. Shepherd PR, Kahn BB. Glucose transporters and insulin action implications for insulin resistance and diabetes mellitus. N Engl J Med 1999;341:24857. [39] Amour J, Brzezinska AK, Jager Z, Sullivan C, Weihrauch D, Du J, et al. Hyperglycemia adversely modulates endothelial nitric oxide synthase during anesthetic preconditioning through tetrahydrobiopterin- and heat shock protein 90-mediated mechanisms. Anesthesiology 2010;112:57685. [40] Capes SE, Hunt D, Malmberg K, Pathak P, Gerstein HC. Stress hyperglycemia and prognosis of stroke in nondiabetic and diabetic patients: a systematic overview. Stroke 2001;32:242632. [41] Marik PE, Raghavan M. Stress-hyperglycemia, insulin and immunomodulation in sepsis. Intensive Care Med 2004;30:74856. [42] Krinsley JS, Grover A. Severe hypoglycemia in critically ill patients: risk factors and outcomes. Crit Care Med 2007;35:22627. [43] Egi M, Bellomo R, Stachowski E, French CJ, Hart GK, Taori G, et al. Hypoglycemia and outcome in critically ill patients. Mayo Clin Proc 2010;85:21724. [44] Lacherade JC, Jacqueminet S, Preiser JC. An overview of hypoglycemia in the critically ill. J Diabetes Sci Technol 2009;3:12429. [45] Egi M, Bellomo R, Stachowski E, French CJ, Hart G. Variability of blood glucose concentration and short-term mortality in critically ill patients. Anesthesiology 2006;105:24452. [46] Krinsley JS. Glycemic variability: a strong independent predictor of mortality in critically ill patients. Crit Care Med 2008;36:300813. [47] Hermanides J, Vriesendorp TM, Bosman RJ, Zandstra DF, Hoekstra JB, Devries JH. Glucose variability is associated with intensive care unit mortality. Crit Care Med 2010;38:83842. [48] Meyfroidt G, Keenan DM, Wang X, Wouters PJ, Veldhuis JD, Van den Berghe G. 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;38:10219. [49] Monnier L, Mas E, Ginet C, Michel F, Villon L, Cristol JP, et al. Activation of oxidative stress by acute glucose uctuations compared with sustained chronic hyperglycemia in patients with type 2 diabetes. JAMA 2006;295:16817. [50] Moghissi ES, Korytkowski MT, DiNardo M, Einhorn D, Hellman R, Hirsch IB, et al. American Association of Clinical Endocrinologists and American Diabetes Association consensus statement on inpatient glycemic control. Endocr Pract 2009;15: 35369.

[27]

[28]

[29]

[30]

[31]

[32] [33]

[34]

[35] [36]

[37] [38]

Y. Ling et al. / European Journal of Internal Medicine 23 (2012) 564574

573

Fig. 5. Forest plot for sepsis comparing intensive glucose control to conventional glucose control stratied by type of intensive care unit. ICU: intensive care unit; CI: condence interval.

Fig. 6. Forest plot for new dialysis comparing intensive glucose control to conventional glucose control stratied by type of intensive care unit. ICU: intensive care unit; CI: condence interval.

574

Y. Ling et al. / European Journal of Internal Medicine 23 (2012) 564574

Table 3 Subgroup analysis of short-term mortality comparing intensive glucose control to conventional glucose control stratied by median of variables. Variables Studies Participants Risk ratio 95% CI 0.971.16 0.881.09 0.951.10 0.751.07 0.991.16 0.911.10 0.771.19 0.951.13 0.891.09 0.961.13 0.841.16 0.931.10 0.961.11 0.831.25 0.931.10 0.891.11 0.911.17 0.951.10 0.791.06 0.981.16 0.911.09

Mean BG achieved in treatment group >6.2 mmol/l 5 8320 1.06 6.2 mmol/l 12 4922 0.98 Total 17 13,242 1.02 Difference of mean BG in treatment and control groups >2.27 mmol/l 8 3940 0.90 2.27 mmol/l 6 8424 1.07 Total 14 12,346 1.00 SD of mean BG achieved in treatment group >1.03 mmol/l 5 4029 0.96 1.03 mmol/l 6 7224 1.03 Total 11 11,253 0.99 Percentage of diabetic patients >17% 7 8813 1.04 17% 9 3920 0.99 Total 16 12,733 1.01 Mean APACHE II score >20.5 7 8756 1.03 20.5 7 4088 1.02 Total 14 12,844 1.01 Percentage of medical patients >62% 7 2781 1.00 62% 10 10,461 1.03 Total 17 13,242 1.02 Mean daily insulin dose in treatment group >53.2 IU/d 6 4110 0.92 53.2 IU/d 6 7395 1.07 Total 12 11,505 0.99 BG: blood glucose; SD: standard deviation.