The reported prevalence of hyperglycaemia among those with acute coronary syndrome (ACS) varies widely. An early series reported 20% of patients with acute myocardial infarction (AMI) presenting with plasma glucose ⩾11.0 mmol/l; a fifth having glycosylated haemoglobin (HbA_1c) suggestive of underlying diabetes.1 A recent Swedish study reported 25% of 1957 patients had blood glucose >9.4 mmol/l on admission; 22% had a history of diabetes.2 A US registry of 141 680 patients aged ⩾65 years reported admission blood glucose below 6.1 mmol/l in only 15.1%, between 6.1 mmol/l and 7.8 mmol/l in 26.8%, between 7.9 mmol/l and 9.5 mmol/l in 17.8%, between 9.6 mmol/l and 13.3 mmol/l in 19.8% and more than 13.3 mmol/l in 20.5%.3 Diabetes previously had been diagnosed in 74% of this last subgroup but in the minority in other subgroups.

Hyperglycaemia remains common even after excluding diabetic individuals. Of 38 864 non-diabetic patients with troponin-positive ACS admitted to hospitals in England and Wales, 9.9% presented with blood glucose ⩾11.1 mmol/l,4 and of 1604 non-diabetic patients admitted to French hospitals, 16.4% presented with blood glucose 7.9–9.3 mmol/l and 16.7% presented with blood glucose >9.3 mmol/l.5

ASSOCIATED RISK

In a systematic review of 15 studies (1966–1998) on AMI, the association of hyperglycaemia with increased in-hospital mortality was stronger in non-diabetic patients than in diabetic patients.6 Non-diabetic patients with glucose concentrations 6.1–8.0 mmol/l or more had a 3.9-fold (95% confidence interval (CI) 2.9 to 5.4) higher risk of death than those with lower concentrations. Admission glucose concentrations higher than values in the range 8.0–10.0 mmol/l were associated with increased risk of congestive heart failure or cardiogenic shock. In diabetic patients with glucose concentrations 10.0–11.0 mmol/l or more the relative risk of death was 1.7 (95% CI 1.2 to 2.4).

This association persists in patients receiving thrombolytic therapy7 and primary percutaneous coronary intervention (PCI)8 and across the spectrum of ACS.2 5 9 10 In non-diabetic patients admission plasma glucose predicts both long-term morbidity (for example, re-infarction, hospitalisation with heart failure, adverse ventricular remodelling)11 12 and mortality.3 5 An increase of 1 mmol/l in admission glucose concentration was associated with a 4% increased mortality risk in 737 non-diabetic patients over 50 months (median) after AMI13 and a 4.3% increased risk of early death in 4408 cases of ACS.14

Non-diabetic patients with admission hyperglycaemia have similar mortality to those with established diabetes, even after risk stratification.13 15 In fact, hyperglycaemic patients without a previous diagnosis of diabetes may have a higher short-term mortality risk than hyperglycaemic patients with known diabetes.2

HYPOGLYCAEMIA

Additionally, hypoglycaemia during hospitalisation is associated with increased mortality at two years in diabetic patients with non-ST-elevation myocardial infarction (NSTEMI)/unstable angina.16 A “U-shaped” relation has been described between admission blood glucose and death/re-infarction at 30 days in both diabetic and non-diabetic patients with ST-elevation myocardial infarction (STEMI). The optimum level was 4.5–5.5 mmol/l, with significantly greater comparative adjusted odds of death in those with glucose <4.5 mmol/l (odds ratio 3.37), 8.3–11.0 mmol/l (odds ratio 2.93) and >11.0 mmol/l (odds ratio 3.09).17 In another study the 30-day mortality was 6.3% for those with admission glucose <4.5 mmol/l and 10.4% for those ⩾11.0 mmol/l, compared with 2.6% for euglycaemic patients.18

OCCULT DIABETES OR STRESS RESPONSE

Even excluding those with “known” diabetes, most patients with ACS have abnormal oral glucose tolerance tests (OGTT). In 923 patients admitted with acute manifestations of coronary disease, OGTT revealed 36% had impaired glucose tolerance and 22% were “newly-diagnosed diabetics”.19 In 181 patients with AMI and admission blood glucose of <11.1 mmol/l normal glucose tolerance was reported in only 34% at discharge and 35% at 3 months.20

However, elevated blood glucose concentration is not simply a marker of pre-existing diabetes or glucose intolerance. Admission blood glucose, unlike HbA_1c and fasting blood glucose concentration, is not an independent predictor of an abnormal OGTT.20 21 Further, even in those with a casual glucose level of ⩾11.1 mmol/l, diabetes was later diagnosed in only 50%22 and abnormal glucose tolerance in 69%.21 Admission blood glucose concentration is therefore an unreliable marker for occult diabetes mellitus. The American Diabetes Association recognises this by defining “hospital-related hyperglycaemia”—distinct from “unrecognised diabetes”—as a fasting blood glucose ⩾6.9 mmol/l (126 mg/dl) or a random blood glucose of ⩾11.1 mmol/l (200 mg/dl) during hospitalisation; later reverting to normal.23

Hyperglycaemia at presentation, while often reflecting undiagnosed and persisting abnormalities of glucose handling, may also represent a transient stress response mediated through the autonomic nervous system with release of catecholamines and adrenal corticosteroids.24 Increased hepatic gluconeogenesis and lipolysis lead to increased circulating glucose, free-fatty acids (FFA) and lactic acid concentrations.

This catecholamine response occurs early, is restricted to the first five days and is proportional to the size of infarction—being associated with faster heart rate, poorer Killip class and lower ejection fraction on discharge.25 26 Hyperglycaemia has been labelled an “epiphenomenon”—a marker of larger infarction and poorer left ventricular function—being associated with larger infarcts,27 28 heart failure on admission5 and elevated brain natriuretic peptide.10

ADVERSE EFFECTS OF HYPERGLYCAEMIA

Regardless of cause, and notwithstanding the effects of co-existing insulin resistance, excess glucose may be directly detrimental during ACS, offering a target for treatment. The molecular mechanisms for this adverse effect have been reviewed elsewhere29 30 and include the promotion of oxidative stress, impairment of endothelial function, promotion of coagulation, non-enzymatic glycation of platelet glycoproteins with abrupt changes in aggregability, amplification of inflammation (with adverse effects most marked in those with co-existing raised inflammatory markers31), suppression of immunity and direct toxicity to myocytes and promotion of apoptosis. Acute hyperglycaemia has been shown to impair ischaemic preconditioning, attenuate the protective effect of pre-infarction angina on microvascular function32 and reduce the effectiveness of collateral blood supply into ischaemic zones.33^–36

During short-term (28 days) follow-up after AMI, admission blood glucose is an independent predictor of mortality (after adjustment for HbA_1c, age, gender and heart failure before the event).37 In-hospital mortality following primary PCI is strongly associated with admission hyperglycaemia irrespective of a previous diagnosis of diabetes; hyperglycaemia is associated with increased in-hospital mortality, while a previous diagnosis of diabetes without admission hyperglycaemia is not.8 Admission hyperglycaemia, but not HbA_1c, remained an important predictor of death following multivariate analysis in 504 consecutive admissions with STEMI (odds ratio 4.91, 95% CI 2.03 to 11.9).38

In STEMI, admission hyperglycaemia independently is associated with incomplete resolution of ST-segment elevation,39 persisting occlusion of the infarct-related artery after thrombolytic treatment18 and (unlike HbA_1c or previous diagnosis of diabetes) predicts poor flow in this artery before,40 and poorer myocardial perfusion and ST-segment resolution after, successful primary PCI.41^–43 Thus, hyperglycaemia appears to be associated with an adverse balance between coagulation and fibrinolysis, and with microvascular dysfunction.

PERSISTING HYPERGLYCAEMIA

Persisting hyperglycaemia is an important issue when attempting to interpret studies on the effects of insulin treatment during AMI.

Hyperglycaemia (blood glucose ⩾8.9 mmol/l) that persists from admission to at least 24 hours after symptom onset is associated both with reduced myocardial perfusion despite patency of the infarct-related artery and with pre-discharge left ventricular impairment.44 This suggests that acute and continuing elevation of blood glucose, rather than an underlying “diabetic state”, may promote microvascular dysfunction, contributing to poorer outcome. In 1219 non-diabetic patients following AMI, higher admission glucose concentrations predicted greater mortality (hazard ratio (HR) 1.12, 95% CI 1.04 to 1.20, per 0.6 mmol/l increase), while larger reductions in glucose over 24 hours predicted lower mortality (HR 0.91, 95% CI 0.86 to 0.96, for every 0.6 mmol/l fall in the first 24 hours) at 30 days. Baseline glucose and the 24-hour change in glucose remained significant predictors of death by 180 days.45 In a smaller observational study (n = 417) of patients undergoing primary PCI, the occurrence of major cardiac adverse events by 30 days was associated with greater blood glucose on admission and with greater mean blood glucose (average of 7.4 determinations per patient) over the first two days.46 Persisting hyperglycaemia was an independent predictor of risk and was a stronger predictor than admission blood glucose.

This was confirmed in a study of 16 871 patients with “biomarker-positive” ACS.47 There was a significant association between in-hospital death and increasing levels of admission blood glucose, mean blood glucose during hospitalisation, “time-averaged-glucose” (the area under the curve) and “hyperglycaemia index” (area under the curve for hyperglycaemic values only). Persisting hyperglycaemia was a more accurate predictor of death than single admission measures; mean glucose concentration was the most practical tool. There was a “J-shaped” relation for this measure, with the optimum average blood glucose during hospitalisation in the range 5.5–6.1 mmol/l.

INTERVENTIONAL TRIALS

The original concept that insulin infusions (with glucose and potassium) might be beneficial during AMI was advanced 45 years ago.48 The “glucose hypothesis” advocated the use of glucose-potassium-insulin solution (GIK) to shift myocardial metabolic substrate from FFAs to glucose, thereby reducing the toxic effect of these fatty acids on the myocardium and reducing oxygen demand in ischaemic zones.49 50 This intervention had an “insulin focus”—the restoration of euglycaemia being less important than the correction of relative insulin deficiency, with glucose being infused to allow safe administration of insulin and to provide sufficient alternative substrate for metabolism. Interventions that simply reduce (to normal levels) elevated glucose concentrations have a “glycaemic focus”—the infusion of insulin being used for restoration of euglycaemia. The former strategy reflects a belief that high-dose glucose infusion in the presence of exogenous insulin can improve myocardial cell metabolism during ischaemia; the latter seeks to reverse any direct adverse effects of hyperglycaemia. While both strategies are characterised by insulin infusion, differences in emphasis can lead to important differences in blood glucose levels.

We performed a systematic review for randomised controlled trials of insulin in the early management of AMI (see appendix for search strategy). Tables 1 and 2 present data from nine large (>200 patients) open-label randomised controlled trials; six with “insulin focus”,51^–56 three with “glycaemic focus”.57^–59

To determine whether glucose reduction in ACS is beneficial, evidence should be obtained from an adequately powered randomised trial comparing outcomes in both diabetic and non-diabetic patients allocated either to “usual care” or to more intensive treatment that reduces blood glucose concentration quickly and consistently—preferably before reperfusion of the infarct-related artery. Most importantly, an early and significant difference in blood glucose should be achieved between the groups. No such trial has been published.

In general, large trials with an insulin focus have not shown any significant benefit, while those with a glycaemic focus, albeit including subjects with higher admission blood glucose levels and achieving greater reductions in blood glucose within the first 24 hours, have been inconclusive. The existing trials failed to achieve a substantial difference in blood glucose between comparator groups, did not start treatment early after admission, and, for the “glycaemic focus” trials, failed to recruit the planned number of patients, failed to achieve the target blood glucose levels and recruited few non-diabetic patients.

Paucity of non-diabetic patients

Patients enrolled in the two DIGAMI (Diabetes Insulin-Glucose in Acute Myocardial Infarction study) trials were predominantly diabetic. Only 11% (n = 66) of patients in DIGAMI 1 were non-diabetic with admission blood glucose >11.0 mmol/l,57 while in DIGAMI 2 newly detected diabetes (defined as <1 year in duration) was reported in between 21% and 24% enrolled into the treatment groups.58 Of 240 subjects recruited into HI-5 there were 124 non-diabetics (admission blood glucose ⩾11.1 mmol/l in 26 and 7.8–11.0 mmol/l in 98).59

Lack of difference in blood glucose between comparator groups

Neither the DIGAMI 2 and HI-5 trials nor the CREATE-ECLA (Clinical Trial of Reviparin and Metabolic Modulation in Acute Myocardial Infarction Treatment and Evaluation-Estudios Clinicos Latino America) trial of GIK infusion52 showed benefit of early intensive insulin infusion compared with “standard therapy”. Yet in none of these trials did active treatment lead to early, substantial, reductions in blood glucose compared with control treatment. In DIGAMI 2, despite a stated aim to reduce blood glucose as fast as possible to between 7.0 mmol/l and 10.0 mmol/l, and to achieve a fasting blood glucose 5–7 mmol/l after 24-hour treatment the difference in blood glucose between groups was only 0.9 mmol/l, and the target fasting blood glucose was not reached consistently (9.1 mmol/l in the treatment group). In HI-5 the target blood glucose was 4–10 mmol/l yet there was no statistically significant difference in the mean (of eight readings) glucose level over 24 hours between the interventional and control groups (8.3 mmol/l vs 9.0 mmol/l). In CREATE-ECLA mean glucose levels rose with the GIK infusion, from 9.0 mmol/l to 10.4 mmol/l after 6 hours, before falling to 8.6 mmol/l at 24 hours, compared with an initial fall to 8.2 mmol/l and then to 7.5 mmol/l at 24 hours in the control group. The effects of higher blood glucose in the treatment group may have diluted any beneficial effect of insulin. A combined analysis of 22 943 patients included in the CREATE-ECLA and OASIS-6 (Organization for the Assessment of Strategies for Ischemic Syndromes-6) GIK trials suggested that the detrimental effect of GIK, with respect to early (3-day) death, was explained by the propensity for the infusion to cause elevated glucose (and additionally potassium and fluid gain).60 In fact, the combination of hyperglycaemia with hyperinsulinaemia has been shown to lead to a pro-coagulant and pro-inflammatory state in healthy volunteers.61

In contrast, in the first DIGAMI trial—the only trial to report important mortality benefit—the intervention was associated with a significant 2.1 mmol/l difference at 24 hours compared with the control group.

So, insulin infusion in the absence of significant lowering of blood glucose appears to have no effect on outcomes. Yet even in these “negative studies” the positive association between high admission blood glucose and mortality remains. For example, in the HI-5 study the average 24-hour blood glucose was positively associated with mortality (2% vs 11% at 6 months in those ⩽8 mmol/l vs those >8 mmol/l).

Delays to treatment

In all three “glycaemic focus” trials the start of insulin treatment occurred more than 12 hours after the onset of symptoms; delay to start of infusion after symptom onset by a mean of 13 (SD 7) hours in DIGAMI 1 and 12.6–13.5 hours in DIGAMI 2. In HI-5 the infusion was started a mean of 13.2 (8.4) hours after initiation of thrombolytic therapy or PCI. In the CREATE trial the median time from symptom onset to randomisation was 4.7 hours, with the GIK infusion starting within 1 hour of randomisation in more than 90% of patients. Yet, given that the median time from symptom onset to reperfusion was 3.9 hours the likelihood is that GIK was started after reperfusion.

Failure to recruit sufficient patients

The planned recruitment into the DIGAMI 2 trial was 3000 patients, but because of “slow recruitment” the trial was completed with 1253 subjects. The planned recruitment into HI-5 was 850 patients, yet only 244 were included (four of whom were withdrawn soon after).

Safety

Given that hypoglycaemia may also be detrimental (see above), its occurrence in some individuals might reduce the apparent effectiveness of an intervention within a group of patients. Hypoglycaemia was seen (capillary blood glucose <3.5 mmol/l) in 13 of 126 patients with intensive insulin therapy in HI-5 and was more common (blood glucose <3 mmol/l, regardless of symptoms) in the two treatment groups (9.6–12.7%) than in controls (1.0%) in DIGAMI-2. It has also been reported during GIK infusions—for example during the POL-GIK (Polish Glucose-Insulin-Potassium) trial the concentration of insulin in the infusion was reduced (from 32 U/l to 20 U/l) after 7.6% of patients developed hypoglycaemia.55

IMPLICATIONS FOR CLINICAL PRACTICE

Regardless of the aetiology, or the use of reperfusion strategies, hyperglycaemia remains an independent risk marker for poor prognosis in ACS. Its addition to risk stratification tools may increase the discriminatory value of risk score,16 though its utility in this situation needs further exploration.

We believe that blood glucose should be measured at admission in all cases of ACS, though this is less important if no specific therapy was then offered to restore euglycaemia where necessary. Given that this is an unreliable marker of pre-existing or subsequent abnormal glucose metabolism, all non-diabetic patients with ACS should later undergo OGTT to identify those with undiagnosed diabetes or impaired glucose homeostasis.

There is great variation in the use of insulin treatment in those presenting with elevated blood glucose.3 4 The enthusiasm for insulin infusion in managing hyperglycaemia, fostered by the findings of the first DIGAMI trial, has, in our experience, waned since the publication of DIGAMI 2 and CREATE-ECLA. Notwithstanding a failure to appreciate the strengths and weaknesses of these two trials, the speed with which their results appear to have influenced practice suggests a pre-existing dissatisfaction with the intervention. This may reflect recognition that the early control of glycaemia in AMI is problematic62 and may lead to treatment-induced hypoglycaemia. Insulin infusion in the absence of a swift reduction of blood glucose has no effect on outcome, and the necessary trial to quantify the effect of rapid glucose reduction in non-diabetic patients with ACS has not been performed.

The European Society of Cardiology and the European Association for the Study of Diabetes recommend strict blood glucose control with intensive insulin therapy in adults undergoing cardiac surgery (grade IA recommendation) or those with “critical illness” (grade IB) rather than in those with ACS.63 The American Heart Association suggests intensive glucose control with insulin in patients with significant hyperglycaemia (>9.9 mmol/l), regardless of prior diabetes history (level of evidence B).64 The Scottish Intercollegiate Guidelines Network state that “Patients with clinical myocardial infarction and diabetes mellitus or marked hyperglycaemia (>11.0 mmol/l) should have immediate intensive blood glucose control. This should be continued for at least 24 hours”.65

Based upon existing observational data we would support the use of treatments that aim to restore euglycaemia. Efforts should be made to move from “sliding scale” insulin infusions—which are reactive, treating hyperglycaemia after it has occurred and with no allowance for meals—to a more predictive (weight-adjusted) and reliable infusion (utilising basal, prandial and correction doses). There is an urgent need for a large well-designed trial to explore this practice.