Glucose levels compared with diabetes history in the risk assessment of patients with acute myocardial infarction

Diabetes and Metabolism

Glucose levels compared with diabetes history in the risk assessment of patients with acute myocardial infarction Abhinav Goyal, MD, MHS, a,b Shamir R. Mehta, MD, MSc, a,c Hertzel C. Gerstein, MD, MSc, a,c Rafael Dı´az, MD, d Rizwan Afzal, MSc, a Denis Xavier, MD, e Jun Zhu, MD, f Prem Pais, MD, MSc, e Liu Lisheng, MD, f Khawar A. Kazmi, MD, g Mohammad Zubaid, MD, h Leopoldo S. Piegas, MD, PhD, i Petr Widimsky, MD, DrSC, j Andrzej Budaj, MD, PhD, k Alvaro Avezum, MD, PhD, i and Salim Yusuf, MBBS, DPhil a,c Hamilton, Ontario, Canada; Atlanta, GA; Rosario, Argentina; Bangalore, India; Beijing, China; Karachi, Pakistan; Kuwait City, Kuwait; São Paulo, Brazil; Prague, Czech Republic; and Warsaw, Poland

Background

Both a history of diabetes mellitus and elevated inhospital glucose levels predict death after acute myocardial infarction (AMI). However, only diabetes history (and not glucose levels) is routinely considered in AMI risk assessment.

Methods

We conducted a post hoc analysis of 2 randomized controlled trials of AMI with ST-segment elevation to compare the prognostic value of inhospital glucose levels with diabetes history in 30,536 subjects. Average inhospital glucose (mean of glucose levels at admission, 6 hours, and 24 hours), diabetes history, and death at 30 days (occurring in 2,808 subjects) were documented.

Results

Average glucose predicted 30-day death (OR 1.10 per 1-mmol/L [18-mg/dL] increase, 95% CI 1.09-1.11, P < .0001); this was unchanged after adjusting for diabetes history. In contrast, diabetes history alone predicted 30-day death (OR 1.63, 95% CI 1.48-1.78, P < .0001), but not after adjusting for average glucose (OR 0.98, 95% CI 0.88-1.09, P = .72). The C-indices (areas under the receiver operating characteristic curves) for 30-day death were 0.54 for diabetes history alone, 0.64 for average glucose alone, and 0.64 for glucose plus diabetes. Higher glucose levels predicted death in patients with and without diabetes history, but this relationship was more steep in nondiabetic subjects such that their rate of 30-day death (13.2%) matched that of diabetic patients (13.7%) when average glucose was ≥144 mg/dL (8 mmol/L) (P = .55 after multivariable adjustment).

Conclusions

Although diabetes history is routinely considered in the risk stratification of AMI patients, inhospital glucose levels are a much stronger predictor of death and should be incorporated in their risk assessment. Patients with AMI with inhospital glucose ≥144 mg/dL have a very high risk of death regardless of diabetes history. (Am Heart J 2009;157:763-70.)

In patients admitted with acute myocardial infarction (AMI), a history of diabetes mellitus portends a poor prognosis.1-3 This observation has prompted professional societies to advocate more aggressive risk stratification From the aPopulation Health Research Institute, Hamilton Health Sciences, Hamilton, Ontario, Canada, bEmory Schools of Public Health and Medicine, Atlanta, GA, c Department of Medicine, McMaster University, Hamilton, Ontario, Canada, dEstudios Cardiologicos Latinoamerica, Rosario, Argentina, eSt. John's Medical College, Bangalore, India, fCardiovascular Institute and Fu Wai Hospital, Chinese Hypertension League Institute, Beijing, China, gAga Khan University, Karachi, Pakistan, hDepartment of Medicine, Faculty of Medicine, Kuwait University, Kuwait City, Kuwait, iInstituto Dante Pazzanese de Cardiologia, São Paulo, Brazil, jCardiocenter, Third Faculty of Medicine, Charles University, Prague, Czech Republic, and kPostgraduate Medical School, Department of Cardiology, Grochowski Hospital, Warsaw, Poland. Submitted September 23, 2008; accepted December 6, 2008. Reprint requests: Abhinav Goyal, MD, MHS, 1518 Clifton Road NE, Room 456, Atlanta, GA 30322. E-mail: [email protected] 0002-8703/$ - see front matter © 2009, Mosby, Inc. All rights reserved. doi:10.1016/j.ahj.2008.12.007

and management of diabetic compared with nondiabetic AMI patients.2,3 However, using diabetes history to guide risk assessment after AMI has limitations. Although 15% to 20% of patients admitted with AMI have a self-reported history of diabetes,1,4,5 the true prevalence may be substantially higher based on follow-up oral glucose tolerance testing.6 In addition, several studies have demonstrated that elevated inhospital glucose levels predict higher mortality in both diabetic and nondiabetic AMI patients.7-12 It is thus possible that incorporating only diabetes history in AMI risk estimates might overlook an increased risk of mortality in patients with no diabetes history but who do have inhospital hyperglycemia. However, no study has directly compared the prognostic value of inhospital glucose levels with diabetic history for the purpose of AMI risk stratification. In 2 recently completed randomized controlled trials of AMI with ST-segment elevation, the Clinical Trial of Reviparin and Metabolic Modulation in Acute Myocardial

American Heart Journal April 2009

764 Goyal et al

Figure 1

Study flow diagram.

Infarction Treatment and Evaluation-Estudios Clinicos Latino America (CREATE-ECLA) trial13 and the Organization for the Assessment of Strategies for Ischemic Syndromes-6 (OASIS-6) trial,14 inhospital glucose levels, diabetes history, and 30-day outcomes were documented in 30,536 subjects. These investigations provide an opportunity to assess the prognostic value of inhospital glucose levels, history of diabetes, and both factors together in the prediction of 30-day death. We hypothesized that inhospital glucose levels would add substantial prognostic value to diabetes history in AMI patients but that diabetes history would add little incremental prognostic information to inhospital glucose levels. We also sought to identify a practical inhospital glucose cut point that clinicians could use to identify AMI patients at high risk of dying regardless of diabetic status.

Methods The CREATE-ECLA and OASIS-6 trials The subjects in this study were from the CREATE-ECLA trial13 and the OASIS-6 trial,14 both coordinated by the Population Health Research Institute (Hamilton, Ontario, Canada). CREATEECLA was a randomized controlled trial in which 20 201 patients with AMI with ST-segment elevation were allocated to glucoseinsulin-potassium (GIK) therapy plus standard care or to standard care alone. OASIS-6 was a subsequently initiated

randomized controlled trial of AMI with ST-segment elevation with a 2×2 partial factorial design in which 12,092 patients were allocated to fondaparinux versus placebo, of whom 8,000 were also to be randomized to GIK infusion versus no infusion.4,14 However, the GIK component of the OASIS-6 trial was terminated after 2,748 patients were enrolled, following the announcement of the neutral results of GIK therapy on 30-day mortality in the CREATE-ECLA trial.4 Despite the early termination of the GIK part of OASIS-6, glucose levels continued to be measured in all patients subsequently enrolled in the fondaparinux component of OASIS-6. The combining of the trial data for analyses was prespecified before the unblinding of either study4 and included 32,293 AMI patients from 49 countries. In both trials, baseline characteristics (including diabetes history), inhospital glucose levels, and vital status at 30 days were documented. Ethics approval was obtained for each of the 2 trials separately, and all subjects provided informed consent.

Glucose levels, diabetes history, study population, and study outcomes Of the 32,293 patients in the combined trial population, 30 536 (95%) who had ≥1 documented inhospital glucose levels were included in this study (Figure 1). Venous glucose levels were measured up to 3 times at the participating hospital's local laboratories within the first 24 hours of hospitalization: at the time hospital admission and 6 and 24 hours after admission. Plasma glucose was measured in over 90% of hospital sites; a combination of plasma and whole blood glucose was measured in most of the

American Heart Journal Volume 157, Number 4

Goyal et al 765

Figure 2

Death at 30 days by average inhospital glucose level. Deciles of average glucose have been calculated separately for patients with and without a history of diabetes.

remaining sites; and very few sites measured only whole blood glucose. The average glucose through the first 24 hours for each patient was calculated, and average glucose was used to assess the relationship between glucose and outcomes in order to use the totality of the available glucose data and to minimize regressiondilution bias. Average glucose has also been shown to be the best glucose metric for predicting short-term mortality in patients with acute coronary syndromes.12 In patients in whom all 3 glucose values were not documented, the average glucose was calculated as the mean of all available (ie, 1 or 2) glucose measures. Patients were classified as having or not having a history of diabetes based on selfreport and medical history. The primary end point of both the CREATE-ECLA and the OASIS-6 trials was 30-day all-cause death, which was also the main outcome of this study. Deaths were adjudicated centrally, and vital status was ascertained in all 30,536 patients (100%) included in this study.

Data analysis and statistical methods Analysis of glucose as a continuous measure. Four logistic regression models for 30-day mortality were constructed to evaluate the prognostic value of diabetes history and average glucose. The first model included average glucose alone; the second included diabetes history alone; the third included glucose, diabetes, and their interaction term to determine if the relationship between glucose and mortality differed among diabetic and nondiabetic subjects; and the fourth was a multivariable model that included glucose and diabetes history adjusted for several covariates, including age, sex, systolic blood pressure, admission heart rate, Killip class, location of infarction on admission electrocardiogram (anterior wall vs other), and administration of reperfusion therapy (yes versus no). These

covariates were chosen because they conferred over 90% of the prognostic value for 30-day mortality in the Global Utilization of Streptokinase and TPA for Occluded coronary arteries-I (GUSTO-I) trial database of over 40,000 patients with AMI.15 For each of the 4 logistic regression models, the model C-statistic (equivalent to the area under the receiver operating characteristic curve16) was derived to determine the model's ability to discriminate survivors from nonsurvivors, and ORs for 30-day death were reported for diabetes history and average glucose (per 1-mmol/L [18-mg/dL] increase). Models for 30-day mortality were repeated using admission glucose instead of average glucose for comparison purposes. Potential differences assessed by statistical interaction tests were explored in the relationship between average glucose and death in important subgroups, including by sex, and by age ≥60 or <60 years (the median age of the population). As the relationship of average glucose with death did not differ between patients allocated to GIK versus no GIK (P value for interaction = .22), both GIK and control patients were pooled in all analyses.

Analysis of glucose levels in patient categories along with diabetes history. We also analyzed data after dividing patients into 4 groups (Figure 1) based on the presence or absence of a history of diabetes and whether glucose levels were <144 or ≥144 mg/dL (8 mmol/L). The cut point of 144 mg/ dL was chosen because a subgroup analysis of the CREATE-ECLA trial indicated the risk of 30-day death was markedly higher among patients with admission glucose ≥144 mg/dL (8 mmol/L) compared with <126 mg/dL (7 mmol/L), or 126 to <144 mg/dL,13 and because this cut point coincided with the average glucose level around which the risk of 30-day death in nondiabetic patients approached that of diabetic patients (Figure 2). For the

American Heart Journal April 2009

766 Goyal et al

Table I. Baseline characteristics and inhospital therapies by comparison group Average glucose Average glucose Average glucose Average glucose _144 mg/dL, _144 mg/dL, > > <144 mg/dL, <144 mg/dL, no diabetes diabetes no diabetes diabetes Number of patients (%) Average glucose level, mg/dL (mean, SD) Baseline characteristics Age (y) (mean, SD) Female sex (%) Weight (kg) (mean, SD) Systolic blood pressure (mm Hg) (mean, SD) Heart rate (beat/min) (mean, SD) Anterior location of infarction (%) Killip class II, III, or IV (%) Inhospital therapies Aspirin Clopidogrel/ticlopidine β-Blocker ACE inhibitors Lipid-lowering drugs Nitrates Glycoprotein IIb/IIIa inhibitors Any reperfusion therapy Primary percutaneous coronary intervention Thrombolytic therapy

All patients with no diabetes

All patients with diabetes

16 708 (54.7) 113 (18)

728 (2.4) 121 (22)

8388 (27.5) 196 (81)

4712 (15.4) 248 (79)

25 096 (82.2) 142 (63)

5440 (17.8) 230 (86)

58.4 (12.7) 20.3 70.0 (13.8) 130 (25)

63.1 (11.0) 30.5 72.4 (15.6) 135 (25)

60.8 (12.4) 26.5 70.5 (14.0) 130 (28)

61.5 (10.7) 33 70.9 (14.1) 133 (27)

59.3 (12.7) 22.6 70.4 (13.9) 130 (26)

61.7 (10.8) 32.6 71.1 (14.3) 133 (27)

77 (16)

80 (16)

79 (19)

82 (19)

78 (17)

81.7 (18.5)

50.3

49.2

50.8

49.1

50.3

49.1

9.6

13.6

15.7

16.8

11.3

16.4

97.1 50.0 77.3 74.6 62.5 74.4 6.8

96.0 59.5 77.5 77.3 65.2 71.8 7.4

96.9 50.5 71.4 72.5 59.3 75.1 7.0

97.2 57.7 72.2 76.0 62.9 71.5 7.7

97.0 50.7 75.7 74.0 62.0 74.7 7.4

97.1 58.0 72.9 76.2 63.2 71.6 7.7

77.9 16.0

79.8 17.7

83.5 17.4

83.8 16.0

79.7 17.4

83.2 16.3

62.2

62.8

66.5

68.1

62.5

67.4

ACE, Angiontensin-converting enzyme.

4 patient comparison groups, baseline characteristics and inhospital therapies were compared as mean (SD) values or as percentages. Rates of 30-day death were then compared among the 4 groups, and logistic regression models for 30-day death (both unadjusted and adjusted for the same covariates as in the models of glucose as a continuous measure) were constructed comparing each group to the referent group (no history of diabetes and glucose <144 mg/dL). In addition, to determine how well average glucose predicted death using different glucose cut points, we reported the sensitivity and specificity for average glucose levels of 108, 126, and 144 mg/dL (6, 7, and 8 mmol/L) and compared these test characteristics with those of diabetes history. SAS version 9.1 (SAS Institute Inc, Cary, NC) and SPSS version 14.0 (SPSS, Chicago, IL) were both used for statistical analysis. All statistical tests were 2-sided, and P < .05 was declared statistically significant.

Funding sources The OASIS-6 trial was funded by Sanofi-Aventis (Cedex, France), Organon (Oss, The Netherlands), and GlaxoSmithKline (Collegeville, PA). The CREATE-ECLA trial had no external funding. Both trials were conducted by their respective steering committees and the Population Health Research Institute. The external funding sources for the trials had no role in design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of this manuscript.

Table II. Logistic regression models for 30-day death: average glucose as a continuous measure Odds Ratio (95% Confidence Interval)

P

Model C-statistic

1.10 (1.09-1.11)

<.0001

0.64

1.63 (1.48-1.78)

<.0001

0.54

1.10 (1.09-1.11)

<.0001

0.64

Known diabetes Average glucose⁎

0.98 (0.88-1.09) 1.07 (1.06-1.08)

.72 <.0001

0.79

Known diabetes

0.98 (0.88-1.10)

.76

Models

Covariates

Model 1

Average glucose only⁎ Known diabetes only Average glucose⁎

Model 2 Model 3† Model 4‡

⁎ Odds ratios for average glucose is shown per 18-mg/dL (1-mmol/L) increment. † Average glucose and known diabetes are adjusted only for each other. ‡ Average glucose and known diabetes are adjusted for each other as well as for age, sex, systolic blood pressure, Killip class, heart rate, location of infarction (ie, anterior wall), and administration of reperfusion therapy.

Results Glucose levels and baseline characteristics Of the 30,536 study participants with documented glucose data, 25,425 (83%) had ≥2 glucose levels, and

American Heart Journal Volume 157, Number 4

Goyal et al 767

Table III. Logistic regression models for 30-day death by comparison groups of average glucose and diabetes history (data are presented for different average glucose cut points) Comparison group by diabetes status and average glucose cut point (mg/dL)

Adjusted models⁎

Unadjusted models

N

Average glucose cut point = 144 mg/dL Glucose <144, no diabetes 16708 Glucose <144, diabetes 728 Glucose ≥144, no diabetes 8388 Glucose ≥144, diabetes 4712 Average glucose cut point = 126 mg/dL Glucose <126, no diabetes 12002 Glucose <126, diabetes 377 Glucose ≥126, no diabetes 13094 Glucose ≥126, diabetes 5063 Average glucose cut point = 108 mg/dL Glucose <108, no diabetes 5999 Glucose <108, diabetes 148 Glucose ≥108, no diabetes 19097 Glucose ≥108, diabetes 5292

Dead at 30 d (%)

Odds ratio (95% Confidence Interval)

P

Odds ratio (95% Confidence Interval)

P

996 (6.0) 61 (8.4) 1106 (13.2) 645 (13.7)

Reference 1.44 (1.10-1.89) 2.40 (2.19-2.62) 2.50 (2.25-2.78)

– .008 <.0001 <.0001

Reference 1.15 (0.86-1.53) 1.89 (1.71-2.08) 1.96 (1.74-2.20)

– .34 <.0001 <.0001

660 (5.5) 38 (10.0) 1442 (11.0) 668 (13.2)

Reference 1.93 (1.37-2.72) 2.13 (1.93-2.34) 2.61 (2.33-2.92)

– .0002 <.0001 <.0001

Reference 1.38 (0.95-2.01) 1.75 (1.57-1.94) 2.01 (1.78-2.28)

– .088 <.0001 <.0001

342 (5.7) 18 (12.1) 1760 (9.2) 688 (13.0)

Reference 2.29 (1.38-3.80) 1.68 (1.49-1.89) 2.47 (2.16-2.83)

– .0013 <.0001 <.0001

Reference 1.48 (0.85-2.57) 1.40 (1.23-1.60) 1.84 (1.59-2.14)

– .17 <.0001 <.0001

⁎ Adjusted for age, sex, systolic blood pressure, Killip class, heart rate, location of infarction (ie, anterior wall), and administration of reperfusion therapy.

22,871 (75%) had 3 glucose levels recorded in the first 24 hours of hospitalization. The mean average glucose level was 157 ± 76 mg/dL for the entire study population. Table I shows baseline characteristics among comparison groups stratified by average glucose level and diabetes history. Forty-three percent of patients (n = 13,100) had an average glucose ≥144 mg/dL, of whom 64% (n = 8,388) did not have a history of diabetes. In patients with diabetes history, the mean age was slightly higher; a larger proportion comprised female patients, and the percentage of those with Killip class II, III, or IV was greater than in nondiabetic patients. The use of inhospital therapies was similar among comparison groups except that the use of clopidogrel or ticlopidine and the use of thrombolytic therapy were slightly higher, and the use of β-blockers was slightly lower among patients with compared to those without a history of diabetes.

Relationship between glucose as a continuous measure and 30-day mortality At 30 days, 2,808 deaths were documented. In the overall population, admission glucose predicted 30-day death (OR 1.07 per 1-mmol/L [18 mg/dL] increase, 95% CI 1.06-1.08, P < .0001, C-statistic 0.61). However, average glucose was an even stronger mortality predictor (OR 1.10, 95% CI 1.09-1.11, P < .0001, C-statistic 0.64) (Table II). Diabetes history alone also predicted 30-day death (OR 1.63, 95% CI 1.48-1.78, P < .0001). The prognostic value for diabetes was attenuated (yet still statistically significant) after adjusting for only admission glucose (adjusted OR 1.13, 95% CI 1.02-1.25, P = .02) but was completely eliminated after adjusting for only average glucose (adjusted OR 0.98; 95% CI 0.88-1.09, P = .72) (Table II). In contrast, the prognostic value of average

glucose was undiminished after adjusting for diabetes history (Table II). Moreover, the model C-statistics (areas under the receiver operating characteristic curve) for average glucose alone (C = 0.64) was unchanged after including history of diabetes in the models, whereas the C-statistic for history of diabetes alone increased substantially (from a C-index of 0.54-0.64) with the addition of average glucose. After multivariable adjustment, the prognostic value of average glucose remained strong, whereas history of diabetes was no longer prognostic (Table II). There was a statistically significant interaction between average glucose and diabetes history (P = .029) (Figure 2). In diabetic subjects, the risk of death was fairly consistent across the first 6 glucose deciles up to an average glucose level of around 240 mg/dL and then increased beyond this level. In nondiabetic subjects, the risk of death increased starting at a glucose level of 120 mg/dL and intersected with the risk of diabetic patients at a glucose level near 140 mg/dL (Figure 2). Beyond 140 mg/dL, the risk of death for nondiabetic patients was similar to or exceeded that of diabetic patients at any given glucose level. The relationship of average glucose and death differed by sex (P value for interaction = .0002) and was stronger in women (OR 1.11 per 1-mmol/L [18 mg/dL] increase, 95% CI 1.10-1.12, P < .0001) than in men (OR 1.08, 95% CI 1.07-1.09, P < .0001). The relationship of average glucose and death also differed between patients younger and older than the median age of 60 years (P value for interaction = .035); however, the clinical difference was marginal (for age ≥60 years, OR 1.10, 95% CI 1.09-1.11, P < .0001; for age < 60 years, OR 1.09, 95% CI 1.07-1.10, P < .0001).

768 Goyal et al

American Heart Journal April 2009

Figure 3

Death at 30 days among the 4 comparison groups stratified by average glucose <144 or ≥144 mg/dL (8 mmol/L) and diabetic status. P values shown are adjusted for age, sex, systolic blood pressure, Killip class, heart rate, location of infarction (i.e. anterior wall), and administration of reperfusion therapy.

Comparison of 30-day mortality among categories of glucose levels and diabetes history In patients with average glucose <144 mg/dL, those with a history of diabetes had a higher risk of mortality than those without diabetes (8.4% vs 6.0%, unadjusted OR 1.44, 95% CI 1.10-1.89, P = .008). However, this difference in risk was attenuated and was no longer statistically significant after multivariable adjustment (adjusted OR 1.15, 95% CI 0.86-1.53, P = .34) (Table III, Figure 3). In patients with average glucose ≥144 mg/dL, the rates of death among diabetic and nondiabetic subjects were equally high (13.7% versus 13.2% respectively, Table III, Figure 3) in both unadjusted (OR 1.04, 95% CI 0.94-1.16, P = .42) and adjusted (OR 1.04, 95% CI 0.92-1.16, P = .55) models. The analyses were repeated using admission glucose instead of average glucose. For admission glucose <144 mg/dL, 30-day mortality was higher among those with compared to those without diabetes history (8.6% vs 6.5%), although this was not statistically significant after multivariable adjustment (adjusted OR 1.14, 95% CI 0.87-1.50, P = .33). For admission glucose ≥144 mg/dL, 30-day mortality was higher among those with compared to those without diabetes history (13.7% vs

11.8%), and this difference was statistically significant after multivariable adjustment (adjusted OR 1.18, 95% CI 1.06-1.34, P = .0044). Because GIK therapy increased the mean glucose level at 6 and 24 hours compared with no GIK infusion,4 we repeated the above analyses for average glucose and diabetes history after restricting the population to only those 19,113 patients who did not receive GIK therapy. Similar results were obtained. In patients with average glucose <144 mg/dL, there was no difference in the rate of 30-day mortality between patients with (7.0%) and without diabetes history (5.8%) in either unadjusted (OR 1.24, 95% CI 0.88-1.74, P = .23) or adjusted analysis (OR 0.92, 95% CI 0.64-1.33, P = .66). Similarly, in patients with average glucose ≥144 mg/dL, the rates of death among diabetic (13.5%) and nondiabetic subjects (14.6%) were equally high in both unadjusted (OR 0.92, 95% 0.80-1.05, P = .20) and adjusted (OR 0.92, 95% CI 0.79-1.08, P = .31) analyses.

Sensitivity analysis for different glucose cut points At the glucose cut point of 144 mg/dL, the sensitivity and specificity of average glucose for the prediction of death was 62.4% and 59.0% respectively. As the cut point

American Heart Journal Volume 157, Number 4

for average glucose level decreased, the sensitivity of average glucose for the prediction of death increased and the specificity decreased (for a glucose cut point of 126 mg/dL, sensitivity = 75.2% and specificity = 42.1%; for a glucose cut point of 108 mg/dL, sensitivity = 87.2% and specificity = 20.8%). In contrast, for history of diabetes, the specificity was high (82.9%), but the sensitivity (25.1%) was very low. Table III compares the rates of death among the 4 groups when using different cut points of average glucose.

Discussion Diabetes mellitus predicts adverse outcomes after AMI,1-3 and professional society guidelines recommend diabetes history as a component of AMI risk stratification.2,3 However, whether diabetes history remains as useful in AMI risk assessment after accounting for inhospital glucose levels is unclear. In this study, average inhospital glucose level was a much stronger mortality predictor than diabetes history: diabetes history was no longer prognostic of adverse outcome after adjusting for average inhospital glucose, whereas the prognostic value of glucose was undiminished after adjusting for diabetes history. Among patients with average glucose ≥144 mg/dL (8 mmol/L), almost two thirds did not have a history of diabetes, and their adjusted risk of death was just as high as patients with diabetes. These hyperglycemic patients with no diabetes history would have been overlooked as high risk if diabetic status alone were used for risk assessment.

Glucose and prognosis in AMI patients The prognostic value of inhospital glucose values in AMI patients has been previously reported in both diabetic and nondiabetic subjects.7-12 Our study extends the knowledge from these prior studies by demonstrating that the risk of death in nondiabetic patients with hyperglycemia (glucose ≥144 mg/dL) is equivalent to that of diabetic patients. Our study also demonstrates the prognostic superiority of average glucose over admission glucose in the risk stratification of AMI patients, a finding consistent with a prior study.12 Most studies reporting an association between glucose and mortality after AMI have measured only one glucose level (usually on admission).7,8,10 Using a single glucose value dilutes the estimated effect between a parameter and the outcome of interest (ie, regression-dilution bias), which is minimized by using repeated measures.17 A recent scientific statement on hyperglycemia in acute coronary syndromes18 identified several gaps in knowledge, including defining what should be considered an abnormal inhospital glucose level. Although the relationship between glucose and mortality after AMI is continuous, our data provide data to support their expert opinion18 that an inhospital glucose level ≥140 mg/dL

Goyal et al 769

(∼8 mmol/L) can be used to identify high-risk AMI patients regardless of their diabetes history.

Limitations Hemoglobin A1c was not measured in this study. Future studies are warranted to determine the additive prognostic utility of short-term (ie, glucose) and long-term (ie, hemoglobin A1c) markers of glycemia. In addition, random (and not fasting) glucose levels were collected. Although fasting levels may be superior to nonfasting levels in AMI risk assessment,11 random glucose levels are more practical to obtain during AMI hospitalization and are still highly prognostic. Our study does not determine whether glucose is simply a marker of risk or an actual mediator of adverse outcomes. This is an active area of research and debate, and clinical trials addressing this question are ongoing.19 Nevertheless, the prognostic value of inhospital glucose levels and its utility in AMI risk stratification are clear. Finally, one might suggest that some hyperglycemic patients with no diabetes history may in fact have undiagnosed diabetes. However, this suggestion is inconsistent with the steeper relationship between glucose and mortality observed in nondiabetic compared with diabetic patients (Figure 2). Furthermore, admission hyperglycemia is an extremely poor predictor of undiagnosed diabetes in AMI subjects based on predischarge oral glucose tolerance testing.20

Conclusions Patients with no diabetes history with elevated inhospital glucose levels (≥144 mg/dL) have the same high risk for short-term death after AMI as patients with diabetes history. The use of inhospital glucose levels in addition to diabetes history greatly enhances the identification of high-risk patients. Glucose levels are easy and inexpensive to measure and are already obtained routinely several times during AMI hospitalization. Therefore, elevated glucose levels should be integrated along with diabetes history into the risk stratification of AMI patients.

Acknowledgements We thank Dr Peter W.F. Wilson who provided helpful suggestions for this manuscript.

References 1. Donahoe SM, Stewart GC, McCabe CH, et al. Diabetes and mortality following acute coronary syndromes. JAMA 2007;298:765-75. 2. Antman EM, Anbe DT, Armstrong PW, et al. ACC/AHA guidelines for the management of patients with ST-elevation myocardial infarction—executive summary. A report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol 2004;44:671-719. 3. The Task Force on the Management of Acute Myocardial Infarction of the European Society of Cardiology. Management of acute myocardial infarction in patients presenting with ST-segment elevation. Eur Heart J 2003;24:28-66.

770 Goyal et al

4. Diaz R, Goyal A, Mehta SR, et al. Glucose-insulin-potassium therapy in patients with ST-segment elevation myocardial infarction. JAMA 2007;298:2399-405. 5. Nallamothu BK, Blaney ME, Morris SM, et al. Acute reperfusion therapy in ST-elevation myocardial infarction from 1994-2003. Am J Med 2007;120:693-9. 6. Bartnik M, Ryden L, Ferrari R, et al. The prevalence of abnormal glucose regulation in patients with coronary artery disease across Europe. Eur Heart J 2004;25:1880-90. 7. Kosiborod M, Rathore SS, Inzucchi SE, et al. Admission glucose and mortality in elderly patients hospitalized with acute myocardial infarction: implications for patients with and without recognized diabetes. Circulation 2005;2005:3078-86. 8. Capes SE, Hunt D, Malmberg K, Gerstein HC. Stress hyperglycaemia and increased risk of death after myocardial infarction in patients with and without diabetes: a systematic overview. Lancet 2000;355: 773-8. 9. Goyal A, Mahaffey KW, Garg J, et al. Prognostic significance of the change in glucose level in the first 24 hours after acute myocardial infarction: results from the CARDINAL study. Eur Heart J 2006;27: 1289-97. 10. Stranders I, Diamant M, van Gelder RE, et al. Admission blood glucose level as risk indicator of death after myocardial infarction in patients with and without diabetes mellitus. Arch Intern Med 2004; 164:982-8. 11. Suleiman M, Hammerman H, Boulos M, et al. Fasting glucose is an important independent risk factor for 30-day mortality in patients with acute myocardial infarction: a prospective study. Circulation 2005; 111:754-60. 12. Kosiborod M, Inzucchi SE, Krumholz HM, et al. Glucometrics in patients hospitalized with acute myocardial infarction. Defining the optimal outcomes-based measure of risk. Circulation 2008;117:1018-27.

American Heart Journal April 2009

13. Mehta SR, Yusuf S, Diaz R, et al. Effect of glucose-insulin-potassium infusion on mortality in patients with acute ST-segment elevation myocardial infarction: the CREATE-ECLA randomized controlled trial. JAMA 2005;293:437-46. 14. Yusuf S, Mehta SR, Chrolavicius S, et al. Effects of fondaparinux on mortality and reinfarction in patients with acute ST-segment elevation myocardial infarction: the OASIS-6 randomized trial. JAMA 2006; 295:1519-30. 15. Lee KL, Woodlief LH, Topol EJ, et al. Predictors of 30-day mortality in the era of reperfusion for acute myocardial infarction. Results from an international trial of 41,021 patients. Circulation 1995;91: 1659-68. 16. Harrell Jr FE, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 1996;15:361-87. 17. MacMahon S, Peto R, Cutler J, et al. Blood pressure, stroke, and coronary heart disease. Part 1, Prolonged differences in blood pressure: prospective observational studies corrected for the regression dilution bias. Lancet 1990;335:765-74. 18. Deedwania P, Kosiborod M, Barrett E, et al. Hyperglycemia and acute coronary syndrome. A scientific statement from the american heart association diabetes committee of the council on nutrition, physical activity, and metabolism. Circulation 2008;117: 1610-9. 19. Goyal A, Nerenberg K, Gerstein HC, et al. Insulin therapy in acute coronary syndromes: an appraisal of completed and ongoing randomised trials with important clinical end points. Diab Vasc Dis Res 2008;5:276-84. 20. Ishihara M, Inoue I, Kawagoe T, et al. Is admission hyperglycaemia in non-diabetic patients with acute myocardial infarction a surrogate for previously undiagnosed abnormal glucose tolerance? Eur Heart J 2006;27:2413-9.