The initial query gave 356 hits. After screening, 20 GIK and three IGL trials were reviewed. For the GIK studies performed before 1994, a review article was selected that provided a thorough analysis of these trials. From 1994 to 2004, six GIK trials were published that included patients with AMI or unstable angina. These differed from the prior studies in that the included patients received reperfusion therapy. The three IGL studies are discussed separately and in more detail as they are more relevant to current clinical practice. An overview of the study characteristics is given in Table
1. Outcomes and glycaemic parameters are provided in Table
2.
Table 1
Overview of randomised controlled trials
| Meta-analysis of nine RCTs implementing GIK in AMI | 0/1/0/99 | Multicentre, international | 1965–1987 |
Pol-GIK | Chest pain and ECG changes <24 h, IDDM excluded | 0/60/0/40 | Multicentre (16), Poland | 1994–1995 |
| STEMI patients | 0/100/0/0 | Belgrade, Serbia | 2000–2001 |
GIPS I | AMI presenting within 24 h of symptoms | 91/0/4/5 | Zwolle, the Netherlands | 1998–2001 |
GIPS II | STEMI <24 h, heart failure excluded | 93/2/0/5 | Multicentre (7), the Netherlands | 2003–2004 |
Create-ECLA | STEMI presenting within 12 h of symptom onset | 9/74/0/7 | Multicentre (470), international (>10) | 1998–2004 |
OASIS 6 | STEMI presenting within 24 h (later 12 h) | 31/45/0/24 | Multicentre (447) international (41) | 2003–2004 |
DIGAMI I | AMI <24 h and diabetes or glucose of >11.0 mmol/l | 0/50/0/50 | Multicentre (19), Sweden | 1990–1993 |
DIGAMI II | AMI <24 h, type 2 diabetes or >11.0 mmol/l | 42/36/0/22 | Multicentre (44) International (6, Europe) | 1998–2003 |
HI-5 | AMI <24 h with diabetes or >7.8 mmol/l | 35/32/0/33 | Multicentre (6) Australia | 2001–2004 |
Table 2
Study outcomes and glycaemic parameters
| Control | 972 | | | | 21 | | | | | |
GIK | 956 | | | | 16 | 0.76 (0.6–0.9) | 0.004 | – | | |
Pol-GIK | Control | 460 | – | 7 | 6.2 | 4.8 | | | 6.5 | | |
GIK | 494 | <16.8 | 6.9 | 5.9 | 8.9 | 1.85 (1.1–3.1) | 0.01 | 11.1 | 6 | 0.01 |
| Control | 40 | – | | | 10.0b
| | | 4.0 | | |
GIK | 78 | – | | | 3.0b
| 0.30 (0.0–1.4) | 0.08 | 10.0 | 12 | 0.18 |
GIPS I | Control | 464 | – | 8.5 | 8.1 | 5.8 | | | 8.2 | | |
GIK | 476 | 7.0–11.0 | 8.5 | 7.7 | 4.8 | 0.83 (0.5–1.4) | 0.50 | 6.5 | 12 | 0.32 |
GIPS II | Control | 445 | – | 8.3 ± 2.5 | – | 1.8 | | | 3.9 | | |
GIK | 444 | – | 8.5 ± 2.8 | – | 2.9 | 1.61 (0.7–3.8) | 0.27 | 5.3 | 12 | 0.33 |
Create-ECLA | Control | 10,107 | – | 9.0 | 7.5 | 9.7 | | | – | | |
GIK | 10,088 | – | 9.0 | 8.6 | 10 | 1.03 (0.9–1.1) | 0.45 | – | | |
OASIS 6 | Control | 1374 | – | | | 6.7 | | | 10.4 | | |
GIK | 1374 | – | | | 7.6 | 1.13 (0.9–1.5) | 0.36 | 10.8 | 6 | NS |
DIGAMI I | Control | 314 | | 15.7 ± 4.2 | 11.7 ± 4.1 | 15.6 | | | 26.1 | | |
Insulin 24 h + 3 months SC | 306 | 7.0–10 | 15.4 ± 4.1 | 9.6 ± 3.3 | 12.4 | 0.79 (0.5–1.2) | NS | 18.6 | 12 | 0.027 |
DIGAMI II | Standard practice | 306 | | 12.9 ± 4.6 | 10.0 ± 3.6 | 7.5 | | | 17 | | |
Insulin 24 h + 3 months SC | 474 | 7.0–10.0 + fasting 5.0–7.0 | 12.8 ± 4.5 | 9.1 ± 3.0 | 7.5 | 1.00 (0.6–1.7) | NS | 15 | | |
Insulin 24 h | 473 | 7.0–10.0 | 12.5 ± 4.4 | 9.1 ± 2.8 | 7.5 | 1.00 (0.6–1.6) | NS | 12 | 12 | NS |
HI-5 | Control | 114 | | 11.1 ± 3.5 | 9.0 ± 2.8a
| 7.1c
| | | 7.9 | | |
Insulin/dextrose | 126 | 4.0–10.0 | 10.8 ± 4.1 | 8.3 ± 2.2a
| 4.4c
| 0.62 (0.2–1.8) | 0.42 | 6.1 | 6 | 0.62 |
Glucose Normalisation
In the prethrombolytic era, two observational studies using historical controls showed inconsistent results for the beneficial effect of insulin treatment in diabetic AMI patients. Clark et al. reported a reduced incidence of arrhythmias and death in the patients treated with IGL [
20]. In contrast, Gwilt et al. did not find a difference in mortality in diabetics treated with an insulin infusion protocol, though they did find a higher mortality rate in diabetic vs. non-diabetic AMI patients [
21].
Between 1990 and 2004 three randomised controlled trials were performed including patients with AMI and known diabetes or hyperglycaemia at admission [
11,
16,
17]. Patients from Europe [
11,
16] and Australia [
17] were included; study size varied from 240 to 1253 patients. IGL was administered via an intravenous insulin regime for at least 24 h to achieve glucose levels of <10.0 mmol/l. The DIGAMI studies included a 3 month subcutaneous insulin regimen as well. Admission glucose levels ranged from 10.8 to 15.7 mmol/l. At 24 h, glucose levels were reduced in both IGL and control groups, but more so in the IGL groups. The difference in glucose levels between patients allocated to IGL or conventional treatment ranged from 2.1 mmol/l (DIGAMI-1) to 0.7 mmol/l (the average over 24 h in HI-5). Long-term mortality was lower in two of the three studies, but this was statistically significant in DIGAMI-1 only. However, mortality was lower in the control group of DIGAMI-2.
DIGAMI-1 [
11] was the first randomised trial for IGL, and showed a 29% relative reduction (18.6% vs. 26.1%) in 1-year mortality among hyperglycaemic or diabetic AMI patients treated with IGL during the first 12 h and continued subcutaneously for 3 months. Interestingly, about half of this difference in mortality was achieved in the first 3 months, while additional benefit occurred at longer follow-up. In this trial, about half of the patients received thrombolytic therapy. The target for IGL in DIGAMI was a glucose level of 7–10 mmol/l.
The second DIGAMI [
16] trial was designed to verify whether normalisation of serum glucose (target 5–7 mmol/l for fasting glucose) would further improve outcome. Three treatment regimens were compared: IGL 24 h, IGL 24 h continued for 3 months subcutaneously and a control group. No significant difference was seen in survival in the IGL groups compared with conventional management (
P = 0.203) while, unexpectedly, the highest survival rate was observed in the control group.
The discrepancy between the first and second DIGAMI can be attributed to several factors. Most importantly, patients were less ill and admission glucose levels were lower in DIGAMI-2 than in DIGAMI-1 and the investigators did not succeed in normalising glucose levels in the IGL groups. The difference in glucose levels vs. controls was 2.1 mmol/l in DIGAMI-1 and only 0.9 mmol/l in DIGAMI-2. Additionally, in DIGAMI-2, 14% of the patients in the control group also received insulin infusion during hospital admission, the overall admission glucose levels were lower and reperfusion treatment was given more often (78% vs. 50%).
The more recent HI-5 study [
17] did find a lower mortality at 3 and 6 months in favour of IGL, which is consistent with DIGAMI-1, although this was not statistically significant. There was also a lower incidence of heart failure during admission (12.7% vs. 22.8%;
P = 0.04) and reinfarction within 3 months (2.4% vs. 6.1%;
P = 0.05) in the insulin-treated group. The mortality rates in the HI-5 study were markedly lower than those in the DIGAMI studies. This can be explained by the younger population (62 vs. 68 years), by inclusion of non-diabetic subjects (48%), increased use of reperfusion therapy (67% vs. 50%) and overall improved care in the more recent time period (2001–2004 vs. 1990–1993) of the HI-5 vs. the DIGAMI study.
Glucose Target Range and Hypoglycaemia
The results of DIGAMI-2 illustrate the difficulty in regulating glucose levels and achieving ‘optimal’ target ranges. DIGAMI started insulin infusion in patients with a glucose level of 11 mmol/l or higher, and aimed for values between 7 and 10 mmol/l. After the landmark trial from Leuven [
13], in an intensive care population, found a 10.4% absolute difference in mortality in favour of IGL targeted to 4.4–6.1 mmol/l, many subsequent trials used similar targets. However the later trials were unable to reproduce the results [
22]. The most recent NICE-SUGAR trial even reported a higher incidence of death, similar to DIGAMI-2. Since hypoglycaemic episodes occurred more frequently in the IGL group, this resulted in a modification in the ACC/AHA guidelines [
23] to use 10 mmol/l as a threshold for initiating treatment in STEMI patients. The ESC guidelines [
24] still mention a target range of 5 to 7.8 mmol/l; however, these were established before the NICE-SUGAR results appeared.
The fear for hypoglycaemia in certain IGL protocols seems grounded, and prevention requires frequent measurement or a wider glucose target range. Meijering et al. [
25] evaluated insulin protocols in 24 studies (including six with AMI patients). The best results were found using a dynamic scale protocol for continuous intravenous insulin infusion, combined with frequent blood glucose measurement and taking into account changes in glucose levels rather than single values.
Compliance with any insulin protocol is difficult to achieve. Computerised insulin protocols exist [
26‐
28] and can improve protocol compliance and glycaemic regulation; however, frequent measurements are required. More recently, a closed loop system for glycaemic regulation in the intensive care unit was developed using a continuous glucose sensor [
29]. This, however, requires further testing. Also, the reliability of continuous glucose sensors needs further improvement [
30] and validation is necessary in the AMI patient population, particularly in patients with heart failure and hypoperfusion of the skin and subcutaneous tissues.
Glucose–Insulin–Potassium Therapy
A meta-analysis by Fath et al. [
31] of GIK trials done in the pre-reperfusion therapy era (before 1988) showed a lower mortality in the GIK-treated group than in controls (16% vs. 21%;
P = 0.004). However, in later trials that included patients receiving reperfusion therapy (either by thrombolysis or mechanical), GIK did not show beneficial effects. The early randomised trials which were performed between 1994 and 2004 varied in size from 118 [
32] to 20,195 [
33] and included patients in Europe [
32,
34,
35] and other continents [
36,
37]. Reperfusion therapy was given as thrombolysis or primary percutaneous coronary intervention in less than 1% [
31] to 100% of patients enrolled [
32]. Admission glucose values, when reported, varied from 6.9 to 9 mmol/l and were similar across control and GIK groups. All-cause mortality at 30–40 days was higher in the GIK group in four of the six studies (reaching statistical significance in Pol-GIK). In three studies that reported longer follow-up (6–12 months) the mortality difference in favour of the control group remained (also reaching statistical significance in Pol-GIK). Two studies showed a trend in cardiac [
32] and all-cause (GIPS) mortality in favour of GIK, though these did not reach statistical significance. Because in GIPS a subgroup of patients without heart failure had a lower 30-day mortality in the GIK-treated group (1.2% vs. 4.2%,
P = 0.01), GIPS-2 was set up excluding patients with symptoms of heart failure, but this study did not show any beneficial effect of GIK [
34].
The differences in effect of GIK between the earlier and more recent studies can be explained by improvements in treatment (including reperfusion, antiplatelet and β-blocker therapy), which is reflected in the lower mortality rates of the studies performed after 1987. In the more recent studies all 30–40 day mortality rates were below 16%, which is the mortality rate of GIK-treated patients in the meta-analysis of trials from 1965 to 1987.
A parallel could be drawn between GIK and the administration of intravenous magnesium in AMI patients; promising beneficial effects in smaller studies were not confirmed in the large randomised MAGIC trial and the concept was abandoned [
38].