Background In patients with acute coronary syndrome (ACS), increased plasma glucose levels are associated with
worse outcome. Our aim is to ascertain the values of admission and fasting glucose for prediction of death among patients with
ACS; and to compare their predictive capacities.
Methods The relationships of mortality to plasma glucose levels among 811 consecutive patients hospitalized with ACS
were estimated using spline Cox models. Blood samples were obtained upon admission and after overnight fast. The predictive
capacities of fasting and admission glucose were compared using time-dependent receiver operating characteristic curves.
Results
Fasting and admission glucose levels were higher among the 151 patients who died (18.6%) than among
survivors (P < .001). Among the 558 patients with no history of diabetes (68.8%) there was a J-shaped dependence of the alltime mortality hazard ratio on fasting glucose that persisted when adjusted for covariates: hazard was lowest at 110 mg/dL
(6.1 mmol/L), and significantly greater at levels <90 mg/dL (5.0 mmol/L) or >117 mg/dL (6.5 mmol/L). Likewise among nondiabetic patients, the predictive capacities of admission and fasting glucose were similar for forecast times of up to about
1 year, but for later times the area under the receiver operating characteristic curve was larger for fasting glucose than
admission glucose (P < .05). Neither admission nor fasting glucose levels discriminated among diabetic patients in regard to
risk of death.
Conclusions Both admission and fasting glucose may be used for triage of nondiabetic ACS patients; fasting
glucose may additionally be useful for long-term management, for which the relationship with the all-time mortality hazard
ratio is J-shaped. (Am Heart J 2009;158:989-97.)
990 Cid-Alvarez et al
Methods
Patients
The study group comprised the 811 patients admitted to our
coronary care unit between September 2003 and March 2007
with a tentative diagnosis of ACS that was confirmed in
accordance with the universal definition of myocardial infarction by the patient's presenting (i) a troponin I level >99th
percentile of our reference population (0.6 ng/mL) in at least 1
of 3 blood samples taken during the first 6 hours after
admission, (ii) electrocardiographic (ECG) changes indicative of
new ischemia (new ST-T changes, new left bundle branch block,
or the development of pathological Q waves), and/or iii)
imaging evidence of new loss of viable myocardium or new
regional wall motion abnormality.
Glucose measurements
Serum glucose was measured by the glucose oxidase method,
using a Siemens Advia 2400 AutoAnalyzer, in blood samples
taken upon admission and after an overnight fast of >8 hours
within 24 hours of admission. Patients were stratified into tertile
groups defined by fasting plasma glucose.
Ethics
All 811 patients agreed to participate in the study, which was
reviewed and approved by the regional ethics committee.
Statistical analysis
Single-variable Cox models were used to evaluate association
in the whole study group between the all-time risk of death (i.e.,
the risk of death at any time during the study) and each of the
baseline demographic and clinical variables, including admission and fasting glucose. For the latter, given the nonlinear
nature of the association, the models were constructed using
natural cubic splines, HR curves and their asymptotic 95% CIs
being defined relative to minimum risk values16; for the other
variables, linear models were used.17
Multivariable Cox models of all-time risk of death were also
constructed that, together with fasting glucose (which in the
single-variable analyses proved to have greater predictive value
than admission glucose), included variables of known
prognostic value, and variables of unproven prognostic value
that nevertheless emerged as significant predictors of mortality
in 2-covariate analyses with fasting glucose as the other
covariate. In view of the results obtained in the two-covariate
analysis with diabetes, independent multivariate Cox models
were constructed for diabetic and nondiabetic patients. In
both cases, the variables included because they proved
significant in the two-covariate analyses were age, heart
failure (Killip class 2), and ST-segment elevation myocardial
infarction (STEMI), as well as fasting glucose; in addition, the
inclusion of sex, previous coronary artery disease, hypertension, creatinine and anemia was forced. In all the analyses
described in this paragraph, splines were employed to model
glucose dependence.18
Taking data censoring into account, the log hazard ratios of
the unadjusted Cox models were used as criterion variables X to
construct time-dependent ROCs19 via the corresponding sensitivity and specificity functions: for given time t and criterion
threshold c,
sensitivity c; t = PXNcjTVt;
specificity c; t = PXVcjTNt;
where T is survival time. For each time t, the area under the
time-dependent ROC [AUC(t)] was calculated, and 95% CIs for
the AUC(t) curves and for the difference between the admission
and fasting glucose AUC(t) curves were constructed using a nonparametric small-sample bootstrap method in which the
underlying Cox models were recalculated for each resample.
All statistical analyses were carried out in R using the packages
survival (for fitting parametric Cox models), splines (for fitting
nonparametric Cox models), and survival ROC (for timedependent ROC curves). These packages are freely available at
http://cran.r-project.org.
Cid-Alvarez et al 991
Non-survivors
Survivors
151 (19)
76 10
105 (69)
100 (66)
69 (46)
77 (51)
60 (40)
60 (40)
52 (34)
79 (52)
78 (52)
32 (21)
56 (57)
75 (50)
19 (3-45)
1.2 (1.0-1.6)
148 (111-217)
660 (81)
65 13
500 (76)
363 (55)
184 (28)
325 (49)
334 (51)
176 (27)
261 (39)
565 (86)
161 (24)
72 (11)
163 (30)
128 (19)
10 (2-40)
1.0 (0.8-1.1)
124 (106-175)
<.001
.138
.014
<.001
.916
.466
.003
.510
<.001
<.001
<.001
<.001
<.001
.061
<.001
<.001
135 (102-181)
114 (99-144)
<.001
CAD, Coronary artery disease. LVEF, left ventricular ejection fraction. Values are
means SDs, or medians with the corresponding interquartile range in parentheses,
as appropriate.
Coronary angiography results were available for 645 patients.
the design and conduct of this study, all study analyses, the
drafting and editing of the manuscript, and its final contents.
Results
Baseline and general
The 811 patients enrolled in the study were aged 67
13 years (mean SD), 206 (25%) were women, and 253
(31%) had a history of diabetes. The median duration of
follow-up was 18 months (range, 0-48 months), and 151
patients (18.6%) died. Nonsurvivors were older on
admission than survivors (76 10 vs 65 13 years, P <
.001), and were more likely to have a history of
hypertension (P = .014), diabetes (P < .001) and previous
myocardial infarction (P = .016), but there were no
significant differences between survivors and non-survivors as regards sex, smoking habit or dyslipemia. The
nonsurvivor group had higher incidences of three-vessel
disease and atrial fibrillation, smaller left ventricular
ejection fractions, higher serum creatinine levels at
admission, higher Killip classes, and were less likely to
receive coronary angiography and percutaneous intervention (P < .001 in all cases) (Table I).
Table II summarizes the characteristics of the patient
groups defined by fasting glucose tertiles. Among the
three patient groups defined by fasting glucose tertiles
there were significant differences in age (P < .001),
creatinine on admission (P = .052), peak troponin in
the first 6 hours (P < .001), and prevalences of heart
failure, diabetes and 3-vessel disease (P < .001 in all
three cases) (see Table II).
<106
(<5.8)
106-135
(5.8-7.5)
>135
(>7.5)
Age, y
65 14
68 13
70 11
Male sex, %
76
77
71
Hypertension, %
52
58
62
Diabetes, %
12
22
60
Dyslipaemia, %
49
44
55
Smoking, %
45
42
38
Prior CAD, %
32
25
30
STEMI, %
29
25
30
Killip class 2, %
15
18
29
Atrial fibrillation, %
9
16
16
Three-vessel disease, %
31
27
44
Anemia, %
25
24
26
Troponin I, ng/dL
5 (1-21)
15 (3-53)
17 (4-56)
Creatinine, mg/dL
1.09 0.60 1.11 0.57 1.21 0.70
P
<.001
.240
.080
<.001
.062
.358
.193
.193
<.001
.057
<.001
.954
<.001
.052
Values are means SDs, or medians with the corresponding interquartile range in
parentheses, as appropriate. CAD: coronary artery disease.
Coronary angiography results were available for 645 patients.
992 Cid-Alvarez et al
Figure 1
Nonparametric estimates of the dependence of all-time risk of death on fasting glucose among ACS patients without (A) and with (B) a prior
diagnosis of diabetes mellitus (log hazard ratio, with 95% confidence intervals; unadjusted analyses). Ref, reference values. To convert mg/dL of
glucose to mmol/L, divide by 18.
Multivariate analysis
The J-shaped relationship between fasting glucose
levels and log hazard ratio among non-diabetic patients
persisted after controlling for confounding factors: least
hazard was associated with fasting glucose levels of
Discussion
The above results show that in our population neither
admission nor fasting glucose is predictive of all-cause
mortality among ACS patients with known diabetes.
However, among ACS patients without a prior diagnosis
of diabetes unadjusted risk is significantly greater than
minimum for admission glucose levels >150 mg/dL
(8.3 mmol/L); and as fasting glucose deviates in either
Cid-Alvarez et al 993
Figure 2
Nonparametric estimates of the dependence of all-time risk of death on admission glucose among ACS patients without (A) and with (B) a prior
diagnosis of diabetes mellitus (log hazard ratio, with 95% confidence bands; unadjusted analyses).
reported that among 462 diabetic and 1101 nondiabetic patients followed up for a median 24 months,
fasting glucose was only associated with increased risk
of death in the non-diabetic group, in which its
prognostic value was additional to that of the GRACE
risk score and left ventricular ejection fraction.15
Almost all the studies cited above distinguished only
between normal glucose levels and 1 levels of hyperglycemia. However, Svensson et al11 reported that among
diabetic patients a hypoglycemic lowest inhospital glucose
value was also associated with worse adjusted all-cause 2year mortality risk, and Kosiborod et al10 that hypoglycemic mean glucose was likewise associated with increased
inhospital mortality, while Pinto et al. obtained similar
results for admission glucose as regards 30-day mortality.20
The present study found that both hypoglycemic and
Forecast time dependence of the AUC for glucose-based prediction of all-cause mortality among ACS patients without (A) and with (B) a prior diagnosis of diabetes mellitus. Continuous lines, based
on admission glucose; dotted lines, based on fasting glucose. Panels C (for nondiabetic patients) and D (for diabetic patients) show the difference between AUCs based on fasting and admission
glucose, with the corresponding 95% confidence bands.
994 Cid-Alvarez et al
Figure 3
Cid-Alvarez et al 995
Figure 4
Nonparametric estimates of the dependence of all-time risk of death on fasting glucose among ACS patients without a prior diagnosis of diabetes
mellitus, after controlling for potential confounding variables (log hazard ratio, with 95% confidence band; adjusted analysis).
Table III. Cox adjusted hazard ratios for all-cause death during
the 4-year duration of the study among patients with no prior
diagnosis of diabetes
Age, y
Female sex
Hypertension
Smoking
Prior CAD
STEMI
Coronary angiography
Killip 2
Creatinine, mg/dL
Anemia
Fasting glucose, mg/dL (mmol/L)
66 (3.6)
74 (4.1)
90 (5.0)
110 (6.1) (reference level)
119 (6.6)
150 (8.3)
175 (9.7)
200 (11.1)
HR
95% CI
1.07
1.28
0.90
2.01
1.51
2.37
0.65
2.34
1.23
1.76
1.04-1.10
0.70-2.36
0.53-1.50
1.13-3.57
0.90-2.53
1.42-3.96
0.40-1.06
1.42-3.86
0.82-1.84
1.03-2.53
<.001
.420
.630
.018
.112
<.001
.087
<.001
.310
.037
<.001
2.61
2.87
1.79
1.00
1.11
2.21
3.52
5.21
0.55-12.4
1.18-6.98
1.00-3.20
1.02-1.20
1.49-3.28
1.97-6.31
2.57-10.5
The final model included all variables listed above, risk from fasting glucose being
modelled by means of natural splines (see Methods; by way of illustration, results are
listed here for eight fasting glucose values). CAD: coronary artery disease.
996 Cid-Alvarez et al
Study limitations
This study concerned the risk of all-cause mortality
conditional on serum glucose levels measured in a sample
taken at admission and a second sample obtained, within
24 hours of the first, after an overnight fast of >8 hours.
We were not in a position to determine how many
patients had persistent hyper- or hypoglycemia during
hospitalization, or whether deaths were the consequence
of ACS. The observational nature of the study also means
that it does not show whether hyper- and/or hypoglycemia were causally linked to adverse outcome, were
simply markers of greater disease severity, or both.
Moreover, it does not allow the possibility of confounding
due to unmeasured factors to be ruled out. However, as
Conclusions
Among patients who, at admission for ACS, are not
known to suffer diabetes, admission, and fasting plasma
glucose are both strong predictors of mortality. Likewise,
among nondiabetic patients, the predictive values of
admission and fasting glucose are similar for forecast
times of up to about 1 year, but the better predictor of longterm mortality is fasting glucose, for which the relationship
with the log hazard ratio of all-time mortality is J-shaped.
Neither admission nor fasting plasma glucose are useful in
our population for discriminating among known diabetic
patients as regards the risk of death after ACS.
According to our results, fasting glucose concentrations
in nondiabetic patients hospitalized for ACS may serve as
a simple marker to help clinicians stratify risk for optimal
triage and long-term surveillance and management.
References
1. Kosuge M, Kimura K, Kojima S, et al. Japanese Acute Coronary
Syndrome Study (JACSS) Investigators. Effects of glucose abnormalities on in-hospital outcome after coronary intervention for acute
myocardial infarction. Circ J 2005;69:375-9.
2. Capes SE, Hunt D, Malmberg K, et al. Stress hyperglycaemia and
increased risk of death after myocardial infarction in patients with
and without diabetes: a systematic overview. Lancet 2000;355:
773-8.
3. Timmer JR, Van der Horst IC, Ottervanger JP, et al. Myocardial
Infarction Study Group. Prognostic value of admission glucose in
nondiabetic patients with myocardial infarction. Am Heart J 2004;
148:399-404.
4. Wahab NN, Cowden EA, Pearce NJ, et al, on behalf of the ICONS
investigators. Is blood glucose an independent predictor of mortality
in acute myocardial infarction in the thrombolytic era? J Am Coll
Cardiol 2002;40:1748-54.
5. Zeller M, Cottin Y, Brindisi MC, et al. The RICO survey working group.
Impaired fasting glucose and cardiogenic shock in patients with acute
myocardial infarction. Eur Heart J 2004;25:308-12.
6. Vis MM, Sjauw KD, Van der Schaaf RJ, et al. In patients with STelevation myocardial infarction with cardiogenic shock treated with
percutaneous coronary intervention, admission glucose level is a
strong independent predictor for 1-year mortality in patients without a
prior diagnosis of diabetes. Am Heart J 2007;154:1184-90.
7. 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.
Cid-Alvarez et al 997
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.