Glucose-Insulin-Potassium Therapy in Patients With ST-Segment Elevation Myocardial Infarction FREE

Small experimental and clinical studies have supported the use of glucose-insulinpotassium (GIK) infusion in the treatment of acute ST-segment elevation myocardial infarction (STEMI).^1- 4 However, the Clinical Trial of Reviparin and Metabolic Modulation in Acute Myocardial Infarction Treatment and Evaluation— Estudios Clinicos Latino America (CREATE-ECLA)⁵ of 20 201 patients with acute STEMI demonstrated a neutral effect of GIK infusion on 30-day mortality. A separate study, the Organization for the Assessment of Strategies for Ischemic Syndromes-6 (OASIS-6) trial, evaluated the effect of GIK infusion vs no infusion on 6-month clinical outcomes in 2748 patients with acute STEMI. The principal results of the GIK component of the OASIS-6 trial through 6 months are presented herein. The results from the combined analysis of the OASIS-6 and CREATE-ECLA GIK trials showing the association between GIK therapy and 30-day outcomes in a total of 22 943 patients with STEMI also are presented herein. We also examined whether GIK infusion may cause harm in the early postinfusion period and whether this early harm could be explained by infusion-related hyperglycemia, hyperkalemia, or net fluid gain.

The OASIS-6 randomized controlled trial compared fondaparinux with placebo in 12 000 patients with STEMI.⁶ Of the original patients in OASIS-6, 8000 were to be randomized to open-labeled GIK infusion or no infusion in a 2 × 2 partial factorial design.⁶ The steering committee terminated the GIK component of the trial after only 2748 patients had been randomized to GIK infusion or control, following the announcement of the neutral effect of GIK therapy in the CREATE-ECLA study in November 2004.⁵ However, results of the GIK component of OASIS-6 were not analyzed until completion of the fondaparinux arm of the trial. Local ethics committee approvals were obtained separately for each of the OASIS-6 and CREATE-ECLA trials, and participants provided informed consent in both trials.

In the GIK component of the OASIS-6 trial, each center prepared high-dose GIK solution consisting of 25% glucose, 50 U/L of regular insulin, and 80 mEq/L of potassium. Randomization and initiation of GIK infusion occurred as soon as possible after hospital presentation, and GIK was infused at a rate of 1.5 mL/kg per hour for 24 hours. Serum glucose and potassium levels were serially assessed at randomization, 6 hours, and 24 hours. Fluid input and urine output were monitored during the 24-hour infusion period. The primary outcome of the GIK component of the OASIS-6 trial was all-cause mortality at 30 days. Secondary outcomes included heart failure, the composite of death and congestive heart failure, MI, stroke, cardiogenic shock, and cardiac arrest at 30 days, as well as the 6-month rates of all clinical events.

The CREATE-ECLA randomized controlled trial had a 2 × 2 partial factorial design evaluating the effects of open-label 24-hour GIK infusion compared with no infusion in 20 201 patients with acute STEMI, 15 570 of whom also were randomized to a 7-day blinded comparison of reviparin with placebo.^5,7- 8 High-dose GIK infusion identical to that used in OASIS-6 was initiated immediately following randomization and continued for 24 hours. Serum glucose, potassium, and fluid balance were all measured as in OASIS-6. The primary end point (30-day mortality) and the secondary end points were the same as in OASIS-6, except that follow-up beyond 30 days was not performed in CREATE-ECLA.

The plan to combine the individual patient data of the OASIS-6 and CREATE-ECLA GIK trials was made prior to the unblinding of either trial. There were 3 prespecified 30-day outcomes in the combined analysis: death, new-onset heart failure (diagnosed by clinical criteria and corroboration by chest x-ray), and the composite of death or heart failure at 30 days. New-onset heart failure was included because prior studies suggested that GIK infusion could enhance the recovery of myocardial pump function following STEMI.^9- 10

Prespecified analyses also were performed to test the hypothesis that GIK infusion might have an early beneficial effect on clinical outcomes (≤3 days after randomization), with little benefit after this period (days 4-30). This hypothesized temporal difference in GIK infusion's effects was based on studies that showed that GIK infusion favorably affected myocardial pump function,^9- 10 free fatty acid levels,¹¹ and arrhythmogenicity¹² for 2 to 3 days after GIK infusion, with no additional improvement thereafter. In the analysis of events from 4 to 30 days, patients who had already experienced the event in question (death, heart failure, or the composite of death or heart failure) within the first 3 days were excluded. Data on time to heart failure were not collected in 3798 patients in CREATE-ECLA, and therefore they were excluded from the analyses of heart failure and the composite of death or heart failure from 0 to 3 days and 4 to 30 days.

On observing in prespecified analyses an unexpected early hazard in the GIK infusion group in the first 3 days, post hoc analyses were performed to explore the mechanisms by which GIK therapy might cause this early harm. In the CREATE-ECLA trial, serum glucose and potassium levels were higher at 6 and 24 hours after randomization in the GIK infusion group compared with the control (usual care) group, and the net fluid balance was almost 600 mL greater at 24 hours in the GIK infusion group.⁵ Therefore, to determine the effect of postrandomization glucose levels on clinical outcomes, the average of the 6- and 24-hour glucose levels for each patient was calculated, and then patients were classified into 1 of 3 categories: patients with an average postrandomization glucose level of less than 126 mg/dL (to convert to mmol/L, multiply by 0.0555); a level of 126 mg/dL or higher but less than 144 mg/dL; and a level of 144 mg/dL or higher. Similarly, patients were stratified by average postrandomization potassium level (ie, the average of 6- and 24-hour potassium levels) of less than 4.0 mEq/L (a level often considered to be hypokalemia in the setting of acute MI; to convert to mmol/L, multiply by 1.0); a level of 4 mEq/L or higher but less than 5 mEq/L (normokalemia); and a level of 5 mEq/L or higher. Finally, patients also were stratified by fluid balance in the first 24 hours: negative fluid balance for patients whose urine output was higher than 110% of the total fluid input; positive fluid balance when urine output was less than 90% of the total fluid input; and neutral fluid balance when the urine output was within 10% of the fluid input. Three-day rates of death, heart failure, and the composite of death or heart failure were compared among strata of postrandomization glucose, potassium, and fluid balance.

Because there is continued interest in whether early administration of GIK therapy may prove beneficial in patients with acute STEMI,¹³ a subgroup analysis comparing the effect of GIK infusion vs control on clinical event rates (death, heart failure, and the composite of death or heart failure) also was performed in patients stratified by time from STEMI symptom onset to randomization (<2, 2-<4, 4-<8, and ≥8 hours).

Analyses of the overall 30-day and 6-month outcomes in the OASIS-6 GIK trial and 30-day outcomes in the combined trial population were performed using the intention-to-treat approach. Time to clinical events between the GIK infusion and control groups were compared using the log-rank statistic. The Cox proportional hazards model was used to estimate hazard ratios (HRs) and 2-sided 95% confidence intervals (CIs) for clinical events from 0 to 30 days, and then again separately for 0 to 3 days and 4 to 30 days. For the analyses of outcomes by strata of postrandomization glucose, potassium, and fluid balance, the Wald χ² test for trend and Cox proportional hazards models were used to compare the proportion of events among the different strata.

All models for death and the composite of death or heart failure were adjusted for potential confounders including age, sex, history of diabetes, GIK infusion allocation, infarction location (eg, anterior vs other walls), reperfusion therapy (yes or no), and time from symptom onset to randomization. Models evaluating the effect of each covariate (glucose, potassium, and fluid balance) on outcomes also were adjusted for the other 2 covariates. The subgroup analysis comparing the effect of GIK infusion with control by strata of time to randomization did not require statistical adjustment because time to randomization was a prerandomization factor. P≤.05 was used to define statistical significance, and all tests were 2-sided. SAS version 9.1 (SAS Institute Inc, Cary, North Carolina) and SPSS version 14.0 (SPSS Inc, Chicago, Illinois) software were used for all statistical analyses.

In the OASIS-6 trial, 2748 patients were randomized to GIK therapy (Figure 1). Baseline characteristics were well balanced between the groups (Table 1). There were no differences between the GIK infusion and control groups in the 30-day outcomes of death (P = .36), heart failure (P = .31), or the composite of death or heart failure (P = .41) (Table 2). There also were no differences in any secondary 30-day end points or in 6-month clinical event rates (Table 2).

Overall, 22 949 patients were randomized in the OASIS-6 and the CREATE-ECLA GIK trials. Six patients (3 from each of the GIK and control groups) were lost to follow-up at 7 days in the CREATE-ECLA trial, and therefore 22 943 patients were included in the combined analysis. Compared with patients in CREATE-ECLA, patients in OASIS-6 were older, had a higher prevalence of previously diagnosed hypertension and congestive heart failure, and had a slightly lower prevalence of previously diagnosed type 2 diabetes mellitus (Table 1). Although rates of appropriate use of medical therapy were high in both trials, clopidogrel or ticlopidine, β-blockers, and glycoprotein IIb/IIIa inhibitors were used and primary percutaneous coronary intervention was performed more often in OASIS-6 than in CREATE-ECLA.

In the combined OASIS-6 and CREATE-ECLA GIK trial results (Table 3 and Figure 2), there were no differences between the GIK infusion and control groups in the 30-day rate of death (P = .33), heart failure (P = .82), or the composite of death or heart failure (P>.99). In the analyses from days 0 to 3 (Figure 2), the risks of death and the composite of death or heart failure were higher in the GIK group compared with the control group, with 712 deaths (6.2%) in the GIK group and 632 deaths (5.5%) in the control group (HR, 1.13; 95% CI, 1.02-1.26; P = .03); and 1509 death or heart failure events in the GIK group (15.8%) and 1388 events in the control group (14.5%) (HR, 1.09; 95% CI, 1.02-1.18; P = .02). There was a consistent but nonstatistically significant increase in heart failure in the GIK group (1233 events or 12.9%) compared with the control group (1153 events or 12.0%) at 3 days (HR, 1.08; 95% CI, 0.99-1.17; P = .07).

From days 4 to 30 (Figure 2), GIK infusion was associated with a statistically significantly lower occurrence of heart failure (2.4% vs 3.1%; HR, 0.78 [95% CI, 0.65-0.94]; P = .10) and the composite outcome of death or heart failure (4.1% vs 5.0%; HR, 0.81 [95% CI, 0.70-0.93]; P = .004), as well as a nonstatistically significantly lower rate of death (3.7% vs 4.0%; HR, 0.91 [95% CI, 0.80-1.05]; P = .20).

The mean (SD) serum glucose level was 162 (85) mg/dL at randomization in both groups. Mean (SD) glucose levels were higher in the GIK infusion group compared with the control group at both 6 hours (186 [103] vs 148 [70] mg/dL) and at 24 hours (153 [79] vs 135 [63] mg/dL), respectively. The mean (SD) baseline serum potassium concentration was 4.0 (0.6) mEq/L in both groups. Potassium concentrations were higher in the GIK infusion group compared with control group at both 6 hours (4.2 [0.6] vs 4.1 [0.6] mEq/L) and 24 hours (4.4 [0.7] vs 4.0 [0.6] mEq/L), respectively. In the GIK infusion group, the median total fluid input was 3000 mL, median urine output was 1850 mL, and the median net fluid gain was 950 mL. In the control group, the median fluid input was 1700 mL, median output was 1300 mL, and the median net fluid gain was 400 mL.

After adjusting for confounders (including GIK infusion allocation), glucose, potassium, and fluid balance all independently predicted death at 3 days, and glucose and potassium predicted the combined outcome of death or heart failure at 3 days (Table 4). Even though GIK therapy caused a greater 3-day death rate than control (HR, 1.13 [95% CI, 1.02-1.26]; P = .03), GIK infusion allocation was no longer a predictor of 3-day death after adjusting separately for glucose (adjusted P = .86), potassium (adjusted P = .15), and fluid balance (adjusted P = .47).

In subgroup analyses for 30-day mortality (the primary outcome of the individual CREATE-ECLA and OASIS-6 GIK trials), GIK therapy provided no benefit in any stratum of time from symptom onset to randomization (Table 5), including in the 9388 patients randomized within 4 hours (HR, 1.08; 95% CI, 0.94-1.24). In analyses of 8 other end points (death at 3 days and 180 days; heart failure at 3, 30, and 180 days; and the combination of death or heart failure at 3, 30, and 180 days), a similar lack of effect of GIK therapy was observed for all time-to-randomization subgroups (data available from authors on request).

Despite its early termination, the OASIS-6 GIK trial still represents the second-largest GIK trial in patients with acute STEMI published to date, and suggests that GIK therapy has no net clinical effect through 6 months. The combined OASIS-6 and CREATE-ECLA trial analysis of almost 23 000 patients with STEMI (the largest global experience with GIK therapy) demonstrates that GIK infusion has no effect on any important clinical end point through 30 days following STEMI.

However, contrary to our prespecified hypothesis, we observed a higher rate of death and the composite of death or heart failure at 3 days in patients allocated to GIK therapy compared with control. Between 4 and 30 days, there were lower rates of death and the composite of death or heart failure in the GIK infusion group than in the control group, and the overall effect of GIK therapy on 30-day outcomes was neutral. It is possible that despite its early harmful effects, GIK therapy may have delayed benefits that neutralize its early hazard, but a more likely explanation for the observed “late benefit” is postrandomization (survivor) bias.

Patients receiving GIK therapy had higher levels of postrandomization glucose, potassium, and fluid balance than in the control group. Even though GIK infusion had a harmful effect at 3 days, GIK infusion allocation was no longer prognostic of 3-day events after adjusting separately for glucose, potassium, and fluid balance. Moreover, in models adjusted for GIK infusion allocation and other confounders, increasing levels of each of the 3 covariates remained significant predictors of worse outcomes. Therefore, the early harmful effects of GIK therapy largely may be explained by its propensity to increase glucose, potassium, and net fluid gain.

In the combined analysis of almost 23 000 patients, GIK infusion had no effect compared with control on clinical outcomes. Even in subgroup analyses among patients who were randomized within 2 hours (>2800 patients) or within 4 hours (>9300 patients) of symptom onset, there was no reduction in mortality with GIK therapy. Although this subgroup analysis of more than 9300 patients may be underpowered, it is likely that either there is no benefit from early administration of GIK infusion, or there is at best a small effect of no greater than a 6% relative risk reduction (considering the 95% CI of 0.94-1.24 in the stratum randomized within 4 hours).

There are limitations. The OASIS-6 GIK trial was terminated early and therefore lacked power to detect actual differences in 6-month outcomes after GIK infusion. The analyses by strata of postrandomization glucose, potassium, and fluid balance were not prespecified; therefore, these post hoc analyses that suggest an increased mortality related to elevated glucose, potassium, and net fluid gain do not imply that lowering these factors will reduce mortality.

In conclusion, the combined OASIS-6 and CREATE-ECLA trials have demonstrated that GIK infusion does not produce a favorable clinical effect in patients with STEMI. Moreover, GIK therapy appears to initially increase the rate of death postinfarction, and this may be due to its propensity to increase glucose and potassium levels and net fluid gain. Early administration of GIK therapy (within 4 hours) after symptom onset does not appear to have benefit compared with later initiation of GIK therapy. Avoidance of infusion-related hyperglycemia, hyperkalemia, and net fluid gain may be advisable in future studies of metabolic modulation in patients with STEMI.