Glucose-lowering treatment impacts prognosis in diabetic patients after MI: Post hoc analysis of the DIGAMI 2 trial

Type 2 diabetes mellitus is common and associated with a dismal prognosis in coronary artery disease (CAD) patients.^1,2 While it is unclear exactly what causes the poor prognosis in these patients compared with nondiabetic individuals, it may be attributed to various factors,³ such as uncertainties in optimal glucose-lowering treatment and difficulties in reaching glucose targets. We conducted an epidemiological analysis of the DIGAMI 2 (Diabetes Mellitus Insulin-Glucose Infusion in Acute Myocardial Infarction 2) trial to study the impact of different glucose-lowering agents in patients with type 2 diabetes who experienced a myocardial infarction (MI).

Methods

The DIGAMI 2 trial randomized 1253 diabetic patients with MI to receive 1 of the following 3 treatments: (1) 24-hour insulin-glucose infusion followed by subcutaneous insulin-based long-term glucose control; (2) the same initial treatment followed by standard glucose control; or (3) glucose-lowering treatment according to local practice.⁴ There were no significant differences between the 3 groups in the primary end point of total mortality or in the secondary end points of cardiovascular mortality, MI, and stroke. Furthermore, there was no difference in achieved glucose control after hospital discharge. Our post hoc analysis of the DIGAMI 2 trial examines the impact of various glucose-lowering drugs in all patients discharged from the hospital (n = 1181).

The Cox proportional hazards model was the basis for our main analyses. Covariates adjusted for in the multivariable models included age, smoking habits, previous MI, previous congestive heart failure, creatinine level at randomization, sex, percutaneous coronary intervention or coronary artery bypass graft surgery before randomization or during hospitalization, and blood glucose level at randomization. Insulin and blood glucose levels were used as time-dependent variables. Two-tailed statistical tests were used at a 5% significance level and SAS version 8.02 was used for all statistical analyses.

Results

The 1181 patients discharged were followed for a median of 2.1 years (interquartile range, 1.03-3.00 years) and no patient was lost to follow-up. The mean age was 67.9 years, 67% were men, and the average duration of diabetes was 7.9 ± 8.3 years. Admission blood glucose was 12.5 ± 4.3 mmol/L and glycosylated hemoglobin (HbA_1c) was 7.7% ± 1.8%. Newly detected diabetes (duration <1 year) was seen in 251 patients (21%). About one-third of patients (n = 436) were treated with oral glucose-lowering drugs: sulphonyl-ureas, n = 268; metformin n = 200; and acarbose, n = 9. At hospital discharge, 690 (58%) patients were on insulin and these numbers remained stable during follow-up. A total of 173 (15%) patients were on different combinations of glucose-lowering drugs and 176 (15%) did not receive pharmacological treatment.

A total of 206 patients died, 162 had a nonfatal MI, and 54 had a stroke. Our first analysis compared the different glucose-lowering options, including oral agents, insulin, or no pharmacological treatment (Figure 1). After adjustment, none of the glucose-lowering alternatives had an impact on mortality. Treatment with sulphonylureas did not influence cardiovascular morbidity, whereas patients on metformin had a lower incidence of MI or stroke. Cardiovascular events were more frequent in patients on insulin (Figure 2).

As a next step, patients on newly instituted insulin therapy (n = 317) were compared with those discharged from the hospital without insulin (n = 489). There were no differences in mortality (total or cardiovascular), but an increased number of MIs and strokes were once again observed in those treated with insulin (adjusted hazard ratio [HR], 1.95; 95% confidence interval [CI], 1.35-2.82; P < .004; Figure 3).

Our final analysis compared patients who were randomized to newly instituted insulin (study group 1, n = 245) with patients in study groups 2 and 3 who were not receiving chronic insulin treatment (n = 422). The results were consistent with the previous analyses and insulin treatment was associated with an increased risk for reinfarction or stroke (HR, 2.07; 95% CI, 1.37-3.14; P < .006).

Discussion

Our post hoc analysis of the DIGAMI 2 trial indicates that the choice of glucose-lowering treatment may affect prognosis in diabetic patients who have experienced an MI. Chronic insulin treatment was associated with an increased risk of both MI and stroke, while metformin seemed to be protective and sulphonylurea appeared to have a neutral effect with no impact on outcome.

Our findings that sulphonylureas did not affect outcome is important and reassuring because it suggests that these agents are safe in CAD patients, unlike previous studies that suggested that sulphonylureas might be harmful to the ischemic heart.^5-7

Patients on metformin had a lower rate of MI and stroke, and there was no effect on mortality. In the UKPDS (United Kingdom Prospective Diabetes Study) trial, metformin improved mortality and morbidity in overweight diabetic patients.⁸ Our study expands this finding by showing a beneficial effect of metformin on stroke and MI in patients with established CAD. Dissimilarity in study populations and follow-up time may explain the difference in mortality in our study compared with the UKPDS.

In the DIGAMI 1 trial, insulin-based, glucose-lowering treatment improved glucose control compared with conventional treatment, which probably explains why decreased mortality was observed in this trial.⁹ The DIGAMI 2 trial did not confirm these results and a possible explanation might be that similar glycemic control was achieved in all 3 treatment groups.⁴ Our post hoc analysis of DIGAMI 2 indicates that long-term treatment with insulin may be associated with a dismal outcome. This finding cannot be explained by less effective glucose control or longer duration of diabetes among insulin-treated patients since adjustments for these factors were included in our analyses. Furthermore, when examining only patients on newly instituted insulin treatment, increased mortality was still observed.

The finding that insulin may be harmful in patients with cardiovascular disease (eg, heart failure, CAD) has been suggested previously in both retrospective and registry studies.^10,11 Our epidemiological analyses of DIGAMI 2 were based on prospective follow-up. Moreover, all events were carefully adjudicated. Our analyses demonstrated an insignificant trend toward increased mortality, and a longer period of follow-up is needed to determine if the increased risk of MI and stroke will translate into increased mortality over time.

Combining available information from both DIGAMI trials and studies that examined insulin in the intensive care setting suggest that meticulous glucose control rather than insulin per se has a beneficial effect on mortality and cardiovascular morbidity in CAD patients.^4,9,12 One may surmise that if insulin is used to lower glucose levels, the target should be euglycemia to avoid the potentially negative effects of insulin counteracting the beneficial effects of glycemic control.

We can only speculate on what caused the potentially harmful effect of insulin, but high concentrations of insulin have been associated with endothelial dysfunction, platelet dysfunction, and a proatherogenic effect.^13,14 In 2007, Antoniades and colleagues showed that insulin treatment, compared with metformin and sulphonylurea treatments, resulted in an increased release of inflammatory cytokines, which may affect the thrombotic mechanisms in diabetic patients with CAD.¹⁵

Conclusions

The agent used for long-term glucose control after MI appears to affect patient prognosis. Based on our post hoc analysis of the DIGAMI 2 trial, metformin appears to be protective, sulphonylureas appear to have a neutral effect, and insulin may be potentially harmful. Prospective randomized trials are warranted to further investigate the impact of different glucose-lowering drugs after MI and to elucidate why insulin may be deleterious in CAD patients.

Disclosure

The authors received research funding for this report, but have no financial interests or affiliations that would represent a conflict of interest with the content of this article. Professor Malmberg is employed by both the Karolinska Institutet and Astra-Zeneca; however, this company has not been involved in the trial and has no interest in the topic of this manuscript.