Review of the DIGAMI trial

Randomized trial of insulin-glucose infusion followed by subcutaneous insulin treatment in diabetic patients with acute myocardial infarction (DIGAMI Study): effects on mortality at 1 year

J Am Coll Cardiol 1995;26:57-65.

Background Patients with diabetes have higher short- and long-term rates of mortality following acute myocardial infarction (AMI). Possible explanations for this include increased fatty acid metabolism, compromising glycolysis in ischemic and nonischemic areas as well as impairment of platelet and fibrinolytic function. This led to the theory that both processes could be improved with insulin infusion. Small studies at the time provided conflicting results.

The Diabetes Mellitus Insulin-Glucose Infusion in Acute Myocardial Infarction (DIGAMI) trial sought to test the hypothesis that rapid improvement of metabolic control in diabetic patients with AMI by means of insulin-glucose infusion would decrease early mortality and that continued good metabolic control would improve subsequent prognosis.

Patients Patients admitted to the CCUs of 19 Swedish hospitals with suspected AMI within the preceding 24 hours with a blood glucose level >11 mmol/l (198 mg/dl) with or without a previous history of diabetes mellitus. Exclusion criteria included inability to participate for reasons of health, refusal to give consent, residence outside the catchment area or enrollment in other studies.

Baseline characteristics There were 1,240 patients who met inclusion criteria and 50% were excluded, mainly due to inability or unwillingness to participate. Compared to those enrolled, excluded patients were older (72 years of age) and more were women. The majority of those enrolled were men (62%) at an average age of 68 years. Nearly 40% had a history of previous MI and 22% had congestive heart failure. The mean time from the onset of symptoms to randomization was 13 hours. More than 80% of patients were non-insulin dependent. The average HbA1c at randomization was 8% and the blood glucose was 279 mg/dl.

Procedures Patients randomized to insulin therapy were started on an insulin-glucose infusion at 30 ml/h and blood glucose was checked after 1 hour. The infusion rate was adjusted according to protocol. The infusion was continued until stable normoglycemia was attained for ≥24 hours. Subcutaneous administration of insulin was given immediately after cessation of the infusion, according to a multidose regimen, with the aim of maintaining normoglycemia. Serum potassium was measured immediately before the infusion and then after 6, 12, and 24 hr and was checked immediately in patients who developed any kind of clinically significant arrhythmia. Control patients were treated according to standard coronary care unit practice and did not receive insulin unless it was deemed clinically indicated.

Endpoints The primary endpoint was all-cause mortality at 3 months. The investigators hypothesized that the insulin-glucose infusion followed by multidose subcutaneous insulin for 3 months would reduce the mortality rate by 30%, from a 35% mortality rate in the control group (35/100 to 24.5/100). Based on that assumption, a sample size of 600 was needed to demonstrate the expected mortality reduction with a 5% significance level and power of 80%.

Results 620 patients were randomized, 306 to the intervention group and 314 to the control group. The blood glucose level was significantly lower in the insulin group 24 hours after randomization (173 mg/dl vs 211 mg/dl; p<0.0001) and at hospital discharge (148 mg/dl vs 162 mg/dl; p<0.01) but not at 3 months (153 mg/dl vs 162 mg/dl; p=NS) or 1 year (averages not provided). HbA1c was significantly lower at 3 months (7.0% vs 7.5%; p<0.01). Hypoglycemia during insulin infusion occurred in 15% of patients in the intervention group and none in the control group and hospital length of stay was significantly longer in the intervention group (11.3 vs 9.5 day; p=0.04).

At 3 months, insulin treatment did not significantly reduce the primary endpoint of mortality compared to controls (12.4% vs 15.6%; p=NS); however, the 1 year mortality (not the primary endpoint) was significantly lower (18.6% vs 26.1%; p=0.03).

Results are provided for subgroups of patients based on risk groups classified according to whether they were insulin dependent or not at baseline, and whether they were at high cardiovascular risk or not at baseline. Interestingly, a very large and statistically significant difference was observed for patients in the lowest risk stratum (no insulin, low cardiovascular risk). The average mortality in this group was the lowest (6.5% vs 13.5% at 3 months) and the risk reduction was more than 3x greater (55%) than in all other groups who experienced progressively higher mortality rates (risk reduction ranged from -11% to 15% at 3 months). In contrast the average mortality in the highest risk group was unchanged at 3 months (24.4% vs 26.0%; p=NS) and 1 year (41.0% vs 46.0%; p=NS).

Conclusions In the DIGAMI trial, insulin-infusion followed by multidose subcutaneous insulin did not significantly reduce the primary endpoint of all-cause mortality at 3 months compared to control. Insulin infusion was associated with increased length of hospital stay and hypoglycemia.

Overall, the mortality rate in DIGAMI was lower than anticipated and the trial was underpowered. Furthermore, the anticipated risk reduction, which formed the basis of sample size estimates was unrealistically optimistic at 30%. Had it been 20% the sample size would have had to been 4 times larger at 2,572 patient. If the authors had chosen a 10% mortality reduction as their minimal clinically important difference (an historically realistic estimate for mortality reduction), the sample size would have needed to be 10,722. Thus, the 1-year results should be viewed with extreme skepticism and at best are hypothesis generating only.

For multiple interventions reviewed thus far (nitrates and magnesium), we have reviewed false positive signals generated from small, underpowered studies that were reversed when appropriately powered trials were undertaken.

Finally, the results of the risk-based subgroups in DIGAMI suggest important treatment effect heterogeneity but should also be viewed with extreme caution due to low power in the overall trial.