Professional Documents
Culture Documents
doi:10.1093/eurheartj/ehi002
Clinical research
KEYWORDS Aims Acute myocardial infarction (AMI) is categorized, according to the presenting
Myocardial infarction; electrocardiogram, into non-ST-elevation myocardial infarction (non-STEMI), ST-
Prognosis; elevation myocardial infarction (STEMI), or bundle branch block myocardial infarction
Cohort study; (BBBMI). Data on the prognostic significance of these categories mainly originate
Selection bias;
from voluntary based registries or large-scale clinical trials and may be hampered
Information bias
by selection and information bias. The aim of this historical cohort study was to evalu-
ate the prognostic significance of different categories of AMI in an unselected cohort.
Methods and results From 1 November 1999 to 31 October 2001, patient records were
reviewed from all admissions to hospitals serving a study region with 139 000 inhabi-
tants. An Endpoint Committee determined whether patients fulfilled the European
Society of Cardiology criteria of AMI. A total of 654 patients with AMI were identified.
The proportion having non-STEMI, STEMI, and BBBMI was 54, 39 and 6%, and the associ-
ated 1 year mortality was 31, 21, and 55%, respectively (log rank 54, P , 0.001). The
more favourable outcome observed in patients with STEMI remained significant
according to multivariable analysis (P ¼ 0.044).
Conclusion In an unselected cohort of patients admitted with AMI, the mortality was
considerably higher than expected from voluntary-based registries and large-scale
clinical trials. The most favourable outcome is observed in patients with STEMI.
European Heart Journal vol. 26 no. 1 & The European Society of Cardiology 2004; all rights reserved.
Mortality in ST-elevation acute myocardial infarction 19
statistics. The final model (Model I) included variables with patients had a median (IQR) age of 67 (58–76) years
P , 0.05. The importance of the categories of AMI in the final and presented with a median (IQR) pre-hospital delay
model was tested using a Likelihood Ratio test (LR test) compar- of 4.0 (1.8–7.6) h.
ing the full model with a sub-model not including the categories The unadjusted 1 year mortality was 28.1% in the com-
of AMI. The final model was refitted in the whole population by
plete cohort, being 30.5%, 20.5%, and 54.8% in patients
replacing missing values with their conditional means (continu-
ous variables) or conditional probabilities (categorical variables)
with non-STEMI, STEMI, and BBBMI, respectively (log
(Model II). These imputed values were obtained as predictions rank 54, P , 0.001) (Table 3) (Figure 1). Lower 1 year
from a regression model or logistic model using all non-missing mortality was observed in patients admitted within 12 h
covariates in each subject.12,13 A third multivariable model of symptom onset (n ¼ 433) compared with patients
(Model III) was presented implementing a fourth category of admitted later after the onset of symptoms (n ¼ 221)
AMI (non-STEMI with pre-existing BBB). Statistical significance (22.9% vs. 38.5%, log rank 30, P , 0.001). Among patients
level was P , 0.05 (two-sided test). The software packages with non-STEMI, lower 1 year mortality was observed in
SPSS 10.0 and STATA 8.0 were used for statistical analyses. patients scheduled for an early interventional strategy
(n ¼ 170) compared with patients treated conservatively
(n ¼ 184) (10.0% vs. 49.5%, log rank 95, P , 0.001).
Results Among patients with STEMI admitted within 12 h of
symptom onset, lower 1 year mortality was observed in
In the 2 year study period, hospital records and labora- those treated with reperfusion therapy (n ¼ 143) com-
tory data were reviewed from 4815 patient admissions. pared with those treated conservatively (n ¼ 62) (10.5%
Table 1 Baseline characteristics, in-hospital and post-discharge data according to different categories of acute myocardial
infarction
Categorical data are presented as % (n) and continuous data as median values (inter-quartile range).
a
At time of follow-up. CAG: coronary angiography. CABG: coronary artery by pass grafting. CK: creatinine kinase. CK(M)B þ BB: creatinine kinase
myocardial band þ brain band. CKMB: creatinine kinase myocardial band. CX: circumflex artery. LAD: left anterior descending artery. n: number of
valid cases. On-scene delay: time from ambulance arrival at the scene of event to ambulance departure. PCI: percutanous coronary intervention.
Pre-hospital delay: time from onset of symptoms to arrival at hospital. RCA: right coronary artery. Transport delay: time from ambulance call to
arrival at hospital.
fulfilling the ESC/ACC criteria of AMI (with minor modi- observed in NRMI-3,9 were lower than the 14% observed
fications), and the three categories of AMI (non-STEMI, in the present study. These discrepancies may be
STEMI, and BBBMI) were accepted as independent explained by the fact that all three registries relied on
entities, thus STEMI and BBBMI were not combined. voluntarily reported cases. Thus, it may be that large
Thereby, potential selection and information bias was myocardial infarctions (instantly detectable on admis-
limited. sion) were more likely to be registered, resulting in a
The present study was performed during the same higher proportion of patients with STEMI being included.
period as the Global Registry of Acute Coronary Events Moreover, no methods ensured that consecutive patients
(GRACE), the Euro Heart Survey of Acute Coronary were included in these registries. In GRACE and EHS-ACS,
Syndromes (EHS-ACS), and the third National Registry of some patients were required to give consent to join the
Myocardial Infarction (NRMI-3).8,9,15,16 Nonetheless, the registries, a strategy potentially leading to exclusion of
proportion of patients with STEMI was 52% in GRACE15 high-risk patients who could not receive information or
and 63% in EHS-ACS,8 hence considerably higher than give written consent.8,15 Patients dying within 24 h of
the 39% observed in the present study. Moreover, the admission tended to be excluded from GRACE, which
in-hospital mortality of 5% observed in both GRACE and also in part explains the low mortality observed.17,18
EHS-ACS,8,15 as well as the in-hospital mortality of 9% In EHS-ACS a high proportion of patients were registered
22 C.J. Terkelsen et al.
Table 2 Medication on admission and at discharge according to different categories of acute myocardial infarction
at academic centres with revascularization facilities, scenario.7 Moreover, an inherent selection bias is
hence potentially resulting in selection bias and contri- always present in large-scale clinical trials as evident
buting to the lower mortality observed.8 In GRACE, from the fact that higher mortality is observed among
patients with BBBMI were categorized as STEMI, thus those eligible for inclusion but not included in the
any improved outcome in patients with STEMI as com- trials.25,26 The latter is consistent with the findings in
pared to non-STEMI may have been underestimated.19 the present study, in which the subgroup of patients
Finally, in none of the three registries was an Endpoint with STEMI treated with reperfusion therapy had signifi-
Committee consulted to reach agreement upon diagnosis, cantly lower mortality when compared with remaining
thus potentially leading to information bias.8,9 The STEMI patients admitted within 12 h of symptom onset.
mortality observed in the present study was also The observation in the present study that patients with
considerably higher than expected from numerous non-STEMI have higher unadjusted mortality compared
large-scale clinical trials, in which 1 year mortality con- with patients with STEMI has previously been explained
sistently is reported as ,10%.5,6,20–24 Possible expla- by the fact that the former patients have more pro-
nations are that numerous inclusion and exclusion nounced co-morbidity as well as more frequent multi-
criteria are implemented in the latter trials, patients vessel disease.16 Indeed, we found several risk factors
included in the latter trials are admitted in a favourable to be more frequent among non-STEMI patients.
condition enabling them to give written informed However, there were no differences in the number of
consent, and the proportion of patients receiving diseased vessels, and ejection fraction was higher, and
optimal reperfusion therapy is considerably higher in coronary biochemical marker release lower, among
the latter trials when compared with the real world patients with non-STEMI compared with patients with
Mortality in ST-elevation acute myocardial infarction 23
Table 4 Prognostic significance of selected variables concerning mortality at follow-up according to univariable Cox regression
analyses
is expected in patients with STEMI because of a greater patients should have been identified. However, a small
deterioration in left ventricular function.32,33 In this number of patients initially admitted to the emergency
setting inclusion of Killip class would result in over- care unit may have died immediately following admission
adjustment and thus underestimate the real prognostic before being assigned to a ward. Any such patients would
information achieved from the AMI categorization. be missed in the present study, but they would be few in
Selection bias was limited. In Denmark, all AMI patients number and, if anything, result in an underestimation of
are admitted to public hospitals, and only two hospitals mortality. Potential information bias was also limited
served the study region, hence all hospital-admitted because the AMI diagnosis was only accepted if confirmed
Mortality in ST-elevation acute myocardial infarction 25
Model I, n ¼ 421
Age (years) ,0.001 1.062 (1.042–1.083)
Creatinine level (mmol/L) ,0.001 1.003 (1.002–1.004)
History of heart failure ,0.001 2.448 (1.527–3.925)
Pre-hospital delay (h) ,0.001 1.006 (1.003–1.010)
Ejection fraction 0.003 0.978 (0.964–0.992)
Category of AMI (reference: STEMI) 0.044
Non-STEMI 1.612 (1.036–2.506)
BBBMI 2.071 (1.062–4.038)
Model II, n ¼ 654 (model based on imputed missing values)
Age (years) ,0.001 1.061 (1.047–1.075)
Creatinine level (mmol/L) ,0.001 1.002 (1.001–1.003)
History of heart failure 0.001 1.801 (1.264–2.567)
Ejection fraction 0.003 0.983 (0.972–0.994)
Category of AMI (reference: STEMI) 0.018
Non-STEMI 1.337 (0.988–1.810)
Perspectives
Acknowledgements
The high mortality observed in the present study com-
pared with consistently lower mortality reported in Professor Werner Vach, Department of Statistics, Uni-
voluntary-based registries and large-scale clinical trials versity of Southern Denmark is thanked for statistical
calls for care when extrapolating results and out- support. The study was supported by the Danish Heart
comes of voluntary registries and clinical trials to the Foundation (Grant 01-2-3-28A-22925 and 02-2-3-58-
real world clinical scenario. In planning future trials, 22026), the Laerdal Foundation of Acute Medicine,
one should implement this knowledge and consider limit- Karl G Andersons Foundation, The A.P. Møller Foundation
ing the number of inclusion and exclusion criteria for the Advancement of Medical Science, Elin Holms
to achieve higher event rates and improve the external Foundation and Kirsten Anthonius’ Foundation.
study validity. Assuming that the strategy of acute
reperfusion therapy and adjunctive medication con-
tributes to the more favourable outcome observed in
patients with STEMI, further clarification of the role of
References
catheter-based and adjunctive pharmacological inter- 1. Van de Werf F, Ardissino D, Betriu A, Cokkinos DV, Falk E, Fox KA,
ventions in patients with BBBMI and non-STEMI are Julian D, Lengyel M, Neumann FJ, Ruzyllo W, Thygesen C, Underwood SR,
needed. Vahanian A, Verheugt FW, Wijns W. Management of acute myocardial
26 C.J. Terkelsen et al.
infarction in patients presenting with ST-segment elevation. 19. Steg PG, Goldberg RJ, Gore JM, Fox KA, Eagle KA, Flather MD, Sadiq I,
Eur Heart J 2003;24:28–66. Kasper R, Rushton-Mellor SK, Anderson FA. Baseline characteristics,
2. Braunwald E, Antman EM, Beasley JW, Califf RM, Cheitlin MD, management practices, and in-hospital outcomes of patients
Hochman JS, Jones RH, Kereiakes D, Kupersmith J, Levin TN, hospitalized with acute coronary syndromes in the Global Registry
Pepine CJ, Schaeffer JW, Smith EE 3rd, Steward DE, Theroux P, of Acute Coronary Events (GRACE). Am J Cardiol 2002;90:358–363.
Gibbons RJ, Alpert JS, Faxon DP, Fuster V, Gregoratos G, Hiratzka LF, 20. Andersen HR, Nielsen TT, Rasmussen K, Thuesen L, Kelbaek H,
Jacobs AK, Smith SC Jr. ACC/AHA guideline update for the manage- Thayssen P, Abildgaard U, Pedersen F, Madsen JK, Grande P, Villadsen
ment of patients with unstable angina and non-ST-segment elevation AB, Krusell LR, Haghfelt T, Lomholt P, Husted SE, Vigholt E,
myocardial infarction—2002: summary article: a report of the Kjaergard HK, Mortensen LS. A Comparison of coronary angioplasty
American College of Cardiology/American Heart Association Task with fibrinolytic therapy in acute myocardial infarction. N Engl J
Force on Practice Guidelines (Committee on the Management of Med 2003;349:733–742.
Patients with Unstable Angina). Circulation 2002;106:1893–1900. 21. The GUSTO V Investigators. Reperfusion therapy for acute myocardial
3. Braunwald E. Application of current guidelines to the management infarction with fibrinolytic therapy or combination reduced fibrino-
of unstable angina and non-ST-elevation myocardial infarction. lytic therapy and platelet glycoprotein IIb/IIIa inhibition: the
Circulation 2003;108:28111–28137. GUSTO V randomised trial. Lancet 2001;357:1905–1914.
4. Boersma E, Maas AC, Deckers JW, Simoons ML. Early thrombolytic 22. Assessment of the Safety and Efficacy of a New Thrombolytic
treatment in acute myocardial infarction: reappraisal of the golden (ASSENT-2) Investigators. Single-bolus tenecteplase compared with
hour. Lancet 1996;348:771–775. front-loaded alteplase in acute myocardial infarction: the ASSENT-2
5. Fibrinolytic Therapy Trialists’ (FTT) Collaborative Group. Indications double blind randomised trial. Lancet 1999;354:716–722.
for fibrinolytic therapy in suspected acute myocardial infarction: 23. Sinnaeve P, Alexander J, Belmans A, Bogaerts K, Langer A, Diaz R,
collaborative overview of early mortality and morbidity results Ardissino D, Vahanian A, Pehrsson K, Armstrong P, Van de Werf F.
from all randomised trials of more than 1000 patients. Lancet One-year follow-up of the ASSENT-2 trial: a double-blind, randomized
1994;343:311–322. comparison of single-bolus tenecteplase and front-loaded alteplase