Influence of Diabetes Mellitus on Clinical Outcome in the Thrombolytic Era of Acute Myocardial Infarction

1.1 Patient Population

A total of 41,021 patients presenting within 6 h of chest pain consistent with a myocardial infarction and ST segment elevation were enrolled in the GUSTO-I trial from December 27, 1990 to February 22, 1993. Data were collected as to whether patients had a previous or current diagnosis of diabetes, and if so, whether they were treated with insulin injections. The specifics of study design, patient eligibility, data collection and management and quality assurance have been previously detailed ([23]).

1.2 Thrombolytic Strategies

Patients were randomized to 1) streptokinase, 1.5 million U over 60 min, with subcutaneous heparin; 2) streptokinase, 1.5 million U over 60 min, with intravenous heparin bolus and infusion, adjusting the dose to maintain an activated partial thromboplastin time of (aPTT) of 60 to 85 s; 3) accelerated t-PA, 15-mg bolus and then infusion of 0.75 mg/kg body weight (up to 50 mg) over 30 min and 0.5 mg/kg (up to 35 mg) over another 60 min, plus intravenous heparin; or 4) the combination of intravenous t-PA (1.0 mg/kg over 60 min) [up to 90 mg], with 10% administered as a bolus) and streptokinase (1.0 million U over 60 min) infused through separate catheters with intravenous heparin. Patients were to receive chewable aspirin as soon as possible, followed by a daily dose of 160 or 325 mg. Intravenous bolus injection of 10 mg of atenolol, in two divided doses, followed by a daily oral dose of 50 to 100 mg, was also recommended for patients without contraindications. The use of other medications, such as nitrates, calcium channel blockers, angiotensin-converting enzyme inhibitors and digitalis, coronary angiography and revascularization procedures were at the discretion of the attending physician.

1.3 End Points

The primary end point was all-cause mortality at 30 days. Other important clinical outcomes, including reinfarction, stroke and bleeding, were also collected prospectively at 30 days, and mortality was assessed at 1 year. The cause of focal neurologic deficit was determined with the aid of computed axial tomography, magnetic resonance imaging or postmortem examination, and classified as primary intracranial hemorrhage, cerebral infarction with conversion to hemorrhage, nonhemorrhagic cerebral infarction or unknown (when such examinations were not performed). The stroke data were interpreted by an independent Stroke Review Committee that had no knowledge of the treatment assignment. Bleeding was classified as severe when it was intracerebral or resulted in substantial hemodynamic compromise that required intervention; moderate when blood transfusion alone was required; or mild when no transfusion was required.

1.4 Coronary Angiography and Revascularization

Coronary angiography was performed in patients during the in-hospital period according to the protocol of the angiographic study and at the discretion of the attending physicians. Among those who underwent coronary angiography, a major epicardial vessel was considered diseased if lumen diameter narrowing ≥50% was present, with at least ≥70% stenosis ([26]). Left ventriculography was performed in the right anterior oblique projection.

1.5 Statistical Analysis

Demographic characteristics of patients with and without diabetes were compared using the unpaired Student ttest and Wilcoxon test for continuous variables and the Pearson chi-square test for categoric variables. Continuous variables are described as the median and interquartile range, and categoric variables are described as the percentage of nonmissing data. Tests for synergism between diabetes and other single risk factors was performed using the Breslow-Day test for homogeneity. Multiple risk factors and possible interactions were considered using logistic regression models. The cohort of all randomized patients and the cohort of 30-day survivors were analyzed separately using multivariable Cox regression analyses. Stepwise models, including the statistically significant variables, and models including a complete set of prespecified risk factors were both fit. Kaplan-Meier product limit estimates and log-rank tests were used for analysis of late death (with 1 SE for 1-year mortality).

2.1 Baseline Characteristics

Of the 41,021 patients enrolled into the study, diabetic and insulin treatment status were complete in 99.7% and 99.8%, respectively. This included 34,888 nondiabetic, 1,643 insulin-treated and 4,301 non–insulin-treated diabetic patients. Patients with diabetes were older, more frequently women, nonwhite, enrolled in the United States and more likely to have a previous history of vascular disease, anterior location of infarction and worse Killip class at presentation (Table 1). In addition, patients with diabetes presented, on average, ∼15 min later for treatment than nondiabetic patients. The diurnal variation regarding onset of myocardial infarction was fairly similar among patients with and without diabetes mellitus, with a primary peak at ∼9 to 10 am, and a secondary peak at ∼2 to 3 pm. Similar trends were also observed among insulin and non–insulin-treated diabetic patients.

2.2 Study Medications

A total of 5,796 (96.4%) patients with diabetes and 33,871 (97.1%) patients without diabetes completed their assigned thrombolytic regimen. On admission, aspirin was administered to 96.6% and 97.4% of patients with and without diabetes, respectively, and in 80.2% and 78.7% this dose was ≤160 mg. Nearly all patients (99.5%) who were assigned to receive intravenous heparin, had an infusion for 24 h, and at least half achieved target therapeutic anticoagulation. However, 11% of patients assigned to the subcutaneous heparin group did not receive this therapy. Instead, 36% of these patients received intravenous heparin, with recurrent ischemia reported as the principal reason. An intravenous beta-adrenergic blocking agent was administered to 45.7% and 44.2% of patients with and without diabetes, respectively. In contrast, 68.5% of patients with diabetes received oral beta-blockers, compared with 71.6% of patients without diabetes.

2.3 Major Clinical Outcomes

2.3.1 Thirty-Day Mortality (Table 2and Table 3)

View this table:

• View inline
• View popup

Table 2

Thirty-Day Outcomes and Diabetic Status

View this table:

• View inline
• View popup

Table 3

Thirty-Day Outcome According to Thrombolytic Strategy

The 30-day mortality for patients with diabetes was significantly higher (10.5%) than that for nondiabetic patients (6.2%) (odds ratio [OR] 1.77, 95% confidence interval [CI] 1.61 to 1.95). Moreover, patients with diabetes treated with insulin had an even higher mortality (12.5%) than that of those not treated with insulin (9.7%) (OR 1.32, 95% CI 1.10 to 1.57). Duration of diabetes was not significantly associated with 30-day mortality after adjusting for other baseline characteristics (p = 0.687) ([17]). We also found that the mortality rates associated with different thrombolytic regimens in patients with and without diabetes mellitus were similar to the entire cohort of patients in GUSTO-I, in that patients with diabetes receiving accelerated t-PA had the lowest mortality (Fig. 1). As with the overall patient cohort, several demographic features, including older age (≥65 years), female gender, hypertension and a previous myocardial infarction, were found to be univariable predictors for poor 30-day outcome among patients with and without diabetes. However, adjusting for baseline risk factors ([17]), the size of the effect for each of these demographic features was similar among patients with and without diabetes.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1

ORs and 95% CIs for 30-day mortality in patients with and without diabetes. SK = streptokinase.

2.3.5 One-Year Mortality

The higher mortality among patients with diabetes compared with those without it continued to increase after 30 days. At 30 days, mortality among patients with diabetes was ∼40% higher than that among nondiabetics. However, by the end of the first year, mortality rates were over 60% higher among patients with diabetes (14.5 ± 0.5%) than among nondiabetics (8.9 ± 0.2%, p = 0.0001) (Fig. 2). Indeed, for survival after 30 days, an additional 4.3% of patients with diabetes died compared with 2.7% of patients without diabetes. Multivariable models show that diabetes remained an independent predictor of mortality throughout the first year of follow-up. Likewise, the higher mortality among insulin-treated diabetic patients (17.8 ± 1.0%) than among non–insulin-treated diabetic patients (13.1 ± 0.5%) remained evident (p = 0.0001).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2

Kaplan-Meier estimates of cumulative mortality by diabetic status.

3.1 Early Mortality

The association of increased mortality with diabetes mellitus has been observed in several thrombolytic trials. Taken together (Table 6), with more than 80,000 pooled patients, mortality was at least 60% higher among patients with diabetes than without. Despite differences in baseline characteristics, diabetes mellitus is an important independent predictor of 30-day mortality ([17]) that is not attributable to a larger infarct size ([2, 13, 33–35]). Indeed, in the present study, we found that the median peak creatine kinase level was lower among diabetic patients than nondiabetics (1,281 vs. 1,445 IU/liter), despite the higher mortality associated with diabetes. Similarly, Stone et al. ([13]) reported that early mortality was higher among patients with diabetes, even though their peak creatine kinase levels were lower and their global left ventricular ejection fractions similar to those of nondiabetics. The discrepancy between global left ventricular systolic function and survival after thrombolysis has been observed by several investigators ([36, 37]). Impairment of regional left ventricular function in non–infarct-related areas (which was more common among diabetic patients) ([14, 38]) and other factors intrinsic to diabetics, such as diastolic dysfunction and myocardial fibrosis, may contribute to the higher mortality.

Mortality is higher among diabetic patients treated with insulin than among those who were not treated with insulin (Table 6). The reason for this observation remains uncertain. It may be partly accounted for by differences in metabolic derangements, beyond glucose regulation, between non–insulin-treated and insulin-treated diabetic patients. Furthermore, patients with more severe diabetes were probably more likely to receive insulin treatment. Although animal models of myocardial infarction have shown differences in infarct size ([39]), metabolic activities ([40]) and possibly myocyte function ([41, 42]) between control animals and those with experimentally induced diabetes, limited clinical data suggest that stringent metabolic control may improve cardiovascular outcomes ([9, 43]).

Interaction with other coronary risk factors, such as age and gender, may also affect mortality. Studies in both the prethrombolytic ([1, 11, 13]) and thrombolytic ([14, 16]) eras have shown that mortality is considerably higher in women than in diabetic men. In addition, duration of diabetes has also been found to be associated with mortality ([15, 44]). However, both gender and duration of diabetes were not associated with increased 30-day mortality after adjusting for other baseline risk factors ([17]).

3.2 Other Early Clinical Outcomes

The occurrence of stroke in patients who have had a myocardial infarction not only limits meaningful functional capacity, but also increases the risk for subsequent early mortality ([45]). In contrast to previous trials ([14, 15, 46]), patients with diabetes in the GUSTO-I cohort were found more likely to have cerebrovascular events than nondiabetics. The occurrence of stroke among patients with diabetes was 1.9%, compared with 1.4% among patients without diabetes (p < 0.001). This, too, is consistent with the prethrombolytic era, where strokes, principally due to cerebral infarction, occurred several-fold more frequently among diabetic (7.4%) than nondiabetic (2.7%) patients ([47]). Pullicino et al. ([47]) attributed this increase to concomitant cerebrovascular disease, co-existing risk factors, autonomic neuropathy leading to postural hypotension and cardiogenic shock. It is interesting that although the cerebrovascular event rate for nondiabetics in the prethrombolytic era was comparable to that of the GUSTO-I population, it was considerably higher among patients with diabetes. An important finding in our study is that the incidence of primary intracranial hemorrhage or hemorrhagic conversion after cerebral infarction was not increased among diabetic patients. This is corroborated by the report of the TIMI-II trial ([48]). Therefore, the risk of intracerebral hemorrhage is not increased in patients with diabetes.

Consistent with previous reports ([14, 16]), the rate of myocardial reinfarction after thrombolysis was similar among patients with and without diabetes. Although Barbash et al. ([15]) found that the in-hospital reinfarction rate was higher in patients with diabetes (4.9% vs. 3.8%) and at 6-month follow-up (6.2% vs. 4.2%), in both instances, among their 883 patients, the association was lost with multivariable analysis.

Congestive heart failure after myocardial infarction is common in diabetic patients, particularly in those treated with insulin ([11]). Despite a similar ([35]) or smaller ([8]) infarct size, patients with diabetes are more likely to have heart failure than nondiabetics. Interestingly, it is only among patients with a moderate extent of myocardial injury that the association with heart failure has been evident ([8]). These observations may be related to factors associated with diabetes, such as impairment of regional left ventricular function ([14, 38]), maladaptation of the autonomic system during hemodynamic stress, thickening of the basement membranes leading to local tissue hypoxia ([49]) and other metabolic derangements.

Although most studies ([2, 5, 9, 16]) concur that atrioventricular block is more frequent among diabetic than nondiabetic patients, the occurrence of ventricular tachycardia or fibrillation has been less consistent. Among our patients, onset of ventricular fibrillation was not associated with diabetic status. However, some reports ([15, 50]) suggest that this arrhythmia is more common among patients with diabetes. However, in a recent observational study ([51]), ventricular fibrillation occurred substantially less frequently among diabetic patients treated with glibenclamide compared with diet or other oral hypoglycemic agents.

3.3 Coronary Angiography and Revascularization

Patients with diabetes in GUSTO were more likely to have multivessel disease than patients without diabetes, which is consistent with previous studies ([14]). Coronary artery disease in patients with diabetes is generally more diffuse than in patients without diabetes ([14, 52]). These unfavorable anatomic features may contribute to the increased morbidity and mortality among diabetic patients. Preliminary results from the angiographic substudy of the GUSTO-I population ([53]) showed that the presence of TIMI flow grade 3 in the infarct-related artery at 90 min was similar among patients with and without diabetes (41% vs. 38%). Likewise, the Thrombolysis and Angioplasty in Myocardial Infarction (TAMI) study group ([14]) found that there was no difference in 90-min patency rates (TIMI flow grade 2 or 3) between patients with and without diabetes (71% vs. 70%). Therefore, adverse outcome associated with patients with diabetes cannot be purely related to early infarct artery patency.

Despite the fact that patients with and without diabetes had similar infarct-related artery patency rates, it is interesting to note in the light of the recent preliminary report of the Bypass Angioplasty Revascularization Investigation (BARI) ([54]) that patients with diabetes in our study were more likely to undergo bypass surgery than nondiabetics. This may be attributed to the more frequent occurrence of diffuse, multivessel coronary artery disease and marginally lower left ventricular ejection fraction among those with diabetes.

3.4 Late Mortality

Unlike most patients receiving thrombolytic therapy, who demonstrate no survival benefit after the first month ([55–57]), patients with diabetes continued to show increased mortality compared with nondiabetics at late follow-up (1-year). Although the development of thrombolytic strategies and advancement of therapies for myocardial infarction appear to reduce early and late mortality for patients with diabetes, their mortality remains 40% to 60% higher than that for patients without diabetes (Fig. 3). Compared with nondiabetics and non–insulin-treated diabetics, mortality among insulin-treated patients remained highest during late follow-up. These observations have been noted in previous studies at intermediate follow-up (6-month) ([15, 16]) (Fig. 3), and the operating factors are probably intrinsic to diabetes.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3

Early (30 to 35 days) and late (6 to 12 months) mortality among patients with and without diabetes. DM = patients with diabetes; nonDM = patients without diabetes. ISG = International Tissue Plasminogen Activator/Streptokinase Mortality trial (15).

3.5 Conclusions

In agreement with observations from previous studies ([14–16]), including TAMI and Gruppo Italiano per lo Studio della Sopravvivenza nell’Infarto Miocardico (GISSI-2), our results from the largest reported cohort of diabetic patients (n = 5,944) confirm that patients with diabetes have a higher mortality at 30 days and 1 year, especially among those treated with insulin. The OR for 30-day mortality for patients with diabetes was 1.77 that of patients without diabetes. In addition, we found that survival was improved in patients with diabetes receiving accelerated t-PA compared with other thrombolytic agents, which is consistent with the main results from GUSTO-I. The rates for reinfarction were similar among the different thrombolytic regimens. Unlike previous studies, there was little interaction between diabetes and other risk factors, and stroke occurred more frequently among diabetic than nondiabetic patients. The incidence of intracranial hemorrhage was similar between patients with and without diabetes. Nonetheless, morbidity and mortality remain higher in patients with diabetes than in those without it, even in the contemporary thrombolytic era, despite correcting for differences in baseline clinical characteristics. To further improve outcome after myocardial infarction and thrombolysis among patients with diabetes, newer strategies such as peri-infarction metabolic control ([58]) and primary angioplasty ([59]) should be investigated.