Evaluation of Global Registry of Acute Cardiac Events and Thrombolysis in Myocardial Infarction scores in patients

with suspected acute coronary syndrome^?,??,?

Steve W. Goodacre PhD ^a,?,1, Mike Bradburn MSc ^a,1,

Abdikudus Mohamed MB, ChB ^a,1, Alasdair Gray MB, ChB ^b,1

^aSchool of Health and Related Research, University of Sheffield, Regent Court, 30 Regent Street, Sheffield, S1 4DA,

United Kingdom

^bEmergency Department, Royal Infirmary of Edinburgh, Edinburgh EH164SU, United Kingdom

Received 16 August 2010; revised 2 September 2010; accepted 9 September 2010

Abstract

Purposes: We aimed to evaluate the Global Registry of Acute Cardiac Events (GRACE) and Thrombolysis in Myocardial Infarction scores in patients with suspected but not proven acute coronary syndrome (ACS).

^? Funding sources: The RATPAC trial was funded by the United Kingdom National Institute for Health Research (NIHR) Health Technology Assessment (HTA) Programme (no. 06/302/19). The study funders had no role in study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the paper for publication. The researchers were independent of the study funders. The views and opinions expressed therein

are those of the authors and do not necessarily reflect those of the NIHR HTA.

^?? Prior presentations: none.

^? The RATPAC Research Team: Charlotte Arrowsmith (RATPAC Research Nurse, Derriford Hospital, Plymouth); Julian Barth (Consultant in Chemical

Pathology, Leeds General Infirmary/Co-applicant); Jonathan Benger (Professor of Emergency Care, University of the West of England/Co-applicant); Mike

Bradburn (Senior Medical Statistician, Clinical Trials Research Unit, University of Sheffield); Simon Capewell (Professor of Epidemiology, University of Liverpool/Co-applicant); Tim Chater (Database Manager, Clinical Trials Research Unit, University of Sheffield); Tim Coats (Professor of Emergency Medicine/ Co-applicant); Paul Collinson (Consultant in Chemical Pathology, St George’s Hospital, London/Co-applicant); Cindy Cooper (Director, Clinical Trials Research Unit, University of Sheffield); Mandy Cooper (RATPAC Research Nurse, Leicester Royal Infirmary); Judy Coyle (RATPAC Research Nurse, Edinburgh Royal Infirmary); Liz Cross (Trial Manager, Health Services Research, University of Sheffield); Simon Dixon (Professor of Health Economics, Health Economics & Decision Science/Co-applicant); Patrick Fitzgerald (Research Fellow, Health Economics & Decision Science, University of Sheffield); Emma Gendall (RATPAC Research Nurse, Frenchay Hospital, Bristol); Steve Goodacre (Professor of Emergency Medicine, Health Services Research, University of Sheffield/Chief Investigator); Emma Goodwin (RATPAC Research Nurse, Barnsley Hospital); Alasdair Gray (Consultant in Emergency Medicine, Royal Infirmary of Edinburgh/Co-applicant); Alistair Hall (Professor of Clinical Cardiology, University of Leeds/Co-applicant); Kevin Hall (RATPAC Research Nurse, Barnsley Hospital); Taj Hassan (Consultant in Emergency Medicine, Leeds General Infirmary/Co-applicant); Julian Humphrey (Consultant in Emergency Medicine, Barnsley Hospital); Steven Julious (Senior Lecturer in medical statistics, Medical Statistics Group, Health Services Research, University of Sheffield/Co-applicant); Jason Kendall (Consultant in Emergency Medicine, Frenchay Hospital, Bristol); Vanessa Lawlor (RATPAC Research Nurse, Frenchay Hospital, Bristol); Sue Mackness (RATPAC Research Nurse, Leicester Royal Infirmary); Yvonne Meades (RATPAC Research Nurse, Leeds General Infirmary); David Newby (Professor of Cardiology, University of Edinburgh/Co-applicant); Dawn Newell (RATPAC Research Nurse, Leicester Royal Infirmary); Doris Quartey (RATPAC Research Nurse, Leeds General Infirmary); Karen Robinson (RATPAC Research Nurse, Leicester Royal Infirmary); Glen Sibbick (RATPAC Research Nurse, Leicester Royal Infirmary); Jason Smith (Consultant in Emergency Medicine, Derriford Hospital, Plymouth); Roz Squire (RATPAC Research Nurse, Derriford Hospital, Plymouth).

* Corresponding author. Tel.: +44 114 222 0842; fax: +44 114 222 0749.

E-mail address: [email protected] (S.W. Goodacre).

¹ On behalf of the RATPAC Research Team.

0735-6757/$ - see front matter. Crown Copyright (C) 2012 Published by doi:10.1016/j.ajem.2010.09.013

Basic procedures: We conducted a secondary analysis of data from the RATPAC trial. Standardized data were collected from 2263 patients presenting to 6 emergency departments with suspected but not proven ACS. Patients were followed up by record review and postal questionnaire at 30 and 90 days after recruitment to identify Major adverse events, defined as death, emergency revascularization, life- threatening arrhythmia, hospitalization for ACS, or nonfatal acute myocardial infarction (AMI).

Main findings: Data were available for 2243 patients (mean age, 54.5 years; 58% male). The major Adverse event rate was 43 (2%) of 2243 after 30 days and 62 (3%) of 2243 after 90 days. The c statistics

for 30-day events were 0.717 (95% confidence interval [CI], 0.698-0.735) for GRACE and 0.682 (95% CI, 0.662-0.701) for TIMI. The corresponding 90-day c statistics were 0.726 (95% CI, 0.707-0.745) for GRACE and 0.693 (95% CI, 0.674-0.712) for TIMI. The c statistic for patient age alone was 0.656 for 30-day events and 0.689 for 90-day events.

Principal conclusions: The GRACE and TIMI scores are little better than age alone as predictors of major adverse events in patients with suspected but not proven ACS, and thus add little to Prognostic assessment of such patients.

Crown Copyright (C) 2012 Published by

Introduction

Chest pain is one of the commonest presentations to the emergency department (ED) and accounts for 6% of adult attendances [1]. Assessment of these patients poses a significant challenge to emergency physicians and primarily involves diagnosing acute coronary syndrome (ACS). Initial assessment includes the clinical history, examination, electrocardiogram (ECG), and chest radiography. However, their diagnostic limitations are well recognized [2-4], and many patients have no clear diagnosis after initial ED assessment. Consequently, studies have developed or evaluated methods to accurately stratify the risk of major adverse events among these patients so that those at high risk can be identified and treated optimally [5-8]. These risk stratification tools use elements of clinical history and physical examination, biomarkers, and cardiac Imaging techniques to categorize patients according to their risk of adverse outcome.

The 7-item Thrombolysis In Myocardial Infarction

1. tool was developed using databases of large clinical trials of non-ST-elevation ACS and uses age, risk factors for coronary heart disease , aspirin use, known CHD, rest angina, ST-segment deviation, and raised cardiac markers to predict major adverse events. In contrast, the more recent 8- item Global Registry of Acute coronary events
2. risk score was derived from a multinational register, with a population of patients across the entire spectrum of ACS, and uses age, pulse, systolic blood pressure, creatinine, Killip class, cardiac arrest at admission, elevated cardiac markers, and ST-segment deviation. Both risk scores have been shown to be accurate Prognostic tools for patients with diagnosed ACS in the short term (14 to 30 days, respectively), with slightly modified versions being used to predict medium-term events (6 months to 1 year) [11,12].

diagnosis of ACS is often uncertain in the ED, so risk- stratification methods need to be validated in patients with suspected rather than diagnosed ACS. Studies of the TIMI risk score have shown prognostic value when applied to

unselected patients presenting to the ED with chest pain [13-19]. However, it is not clear whether this prognostic value derives from predicting outcomes in those with definite ACS alone or whether it applies to those with suspected but not diagnosed ACS. We aimed to evaluate the TIMI and GRACE scores specifically in a population of patients with chest pain in which cases with diagnosed ACS were excluded.

Methods

The RATPAC (Randomized Assessment of Triage using Point-of-care Assay of Cardiac markers) trial was a pragmatic randomized control trial comparing diagnostic assessment with a 90-minute point-of-care triple cardiac marker test to conventional diagnostic assessment without the triple test. The main findings have been published elsewhere [20]. In this article we present findings from a planned secondary analysis of the RATPAC data evaluating the TIMI and GRACE risk scores. Ethical approval for the study was granted by Leeds East Research Ethics Committee.

The trial was undertaken in 6 diverse EDs in the United Kingdom from January 30, 2008, to June 2, 2009. Patients were enrolled into the RATPAC trial if they were an adult older than 18 whose presenting complaint was chest pain and who had a normal or nondiagnostic electrocardiogram. They were excluded if they had an obvious noncardiac cause, diagnostic electrocardiogram changes (N1 mm ST deviation or N3 mm T-wave inversion), suspected serious non-CHD cause (pulmonary embolism), known CHD with prolonged (N1 hour) or Recurrent episodes of typical Ischemic pain, social problems or comorbidities requiring admission, presentation more than 12 hours after their most significant episode of chest pain, previously participated in trial, unable to understand trial information, and if consent was declined. Our definition of suspected but not diagnosed ACS for this study was based on clinical data available at first assessment (ie, before any blood testing), so patients with diagnostic

ECG changes were excluded but those with abnormal troponin levels at presentation were included.

All data were collected using a standardized case report form. Demographic data, components of the TIMI and GRACE risk scores as well as results of initial assessment were documented by the recruiting clinician at the ED. Research nurses then used ED and hospital inpatient notes to record management decisions at initial attendance and admission, extract resource use data, and identify subsequent attendances/admissions and adverse events up to 3 months. In addition, participants were sent questionnaires at 30 and

90 days after recruitment to identify any major adverse events that were not recorded in the hospital notes.

Patients were randomized to receive either (1) diagnostic assessment using a point-of-care panel consisting of troponin I, creatine kinase-MB (mass), and myoglobin measured at baseline and at 90 minutes or (2) standard diagnostic assessment according to local protocols. The point-of-care protocol advised hospital admission and investigation if there was any Troponin elevation, a sustained elevation of creatine kinase-MB (mass), or a myoglobin gradient rise of more than 25% from baseline. The standard diagnostic assessment was applied at the discretion of the treating physician but consisted of a troponin measurement at least 12 hours after last significant symptoms at 5 hospitals and at least 8 hours after last significant symptoms at 1 hospital. All subsequent patient management was at the discretion of the treating physician according to local protocols.

Both GRACE and TIMI scores were assessed for their ability to predict the occurrence of major adverse events at 30 or 90 days. A major adverse event was defined as any of death, emergency revascularization, life-threatening arrhyth- mia, hospitalization for ACS, or nonfatal acute myocardial infarction. Emergency revascularization was defined as a revascularization procedure occurring within 24 hours of the decision to revascularize. We aimed to evaluate the prognostic rather than diagnostic value of the scores, so we did not include hospitalization for ACS or nonfatal acute myocardial infarction occurring at initial presentation in the definition. Hospitalization for ACS was defined as any hospital admission occurring after discharge from the initial presentation for which the physician responsible for hospital care recorded ACS as the final diagnosis. Nonfatal acute myocardial infarction was defined according to the universal definition [21].

The TIMI and GRACE risk score for each patient was calculated using the algorithms in Table 1. Missing data were handled in 2 ways: firstly by imputation of missing data (ie, including all patients) and secondly by using complete case analysis (ie, including only patients who had sufficiently complete data for their respective scores). The imputation assumed that the risk factor was “normal” unless documented otherwise as outlined in Table 2. The troponin results can be used in 2 ways. If the score is used at patient presentation, then only the first available troponin result will be relevant, whereas if the score is used after patient

Table 1 Risk scores

Characteristic

TIMI (0-7)

GRACE (0-258)

Age >=65

>=3 risk factors for CHD Use of aspirin in last 7 d Known CHD (stenosis >=50)

N1 episode rest angina in b24 h ST-segment deviation

Elevated cardiac markers Age (y)

b40 40-49

50-59

60-69

70-79

>=80

Pulse (beats/min)

b70 70-89

90-109

110-149

150-199

N200

Systolic blood pressure (mm Hg)

b80 80-99

100-119

120-139

140-159

160-199

N200

Creatinine (umol/L) 0-35.3

35-70

71-105

106-140

141-176

177-353

>=354

Killip class Class I Class II Class III Class IV

Elevated cardiac markers ST-segment deviation

Cardiac arrest on admission ^a

Score

1

1

1

1

1

1

1

0

18

36

55

73

91

0

7

13

23

36

46

63

58

47

37

26

11

0

2

5

8

11

14

23

31

0

21

43

64

15

30

-

^a Because of the study design, patients were automatically excluded they had an arrest on admission.

admission, then all troponin results can be used, with the test being considered positive if any measurement was positive. Each scoring system therefore gives rise to 4 potential scores: complete case using only first troponin result, complete case using any troponin result, imputed using only first troponin result, and imputed using any troponin result. The main analysis was based on using all troponin results and imputation of missing data.

90 days. Some 1750 (76%) of 2243 patients had sufficient data for a complete case analysis when calculating GRACE risk score and 1798 (80%) of 2243 patients for TIMI risk

Table 2 Missing and imputed data

Characteristic

No. patients missing data

0

222

Rule

Imputed score

TIMI GRACE

Age

Risk factors for CHD

Not applicable (NA) NA

Assume no (modal value)

0

NA NA

p d

Table 3 Baseline characteristics

n (%) or mean +- SD

Normal	1544 (69)
T-wave inversion	153 (7)
ST depression	35 (2)
ST elevation	16 (1)
Bundle-branch block	54 (2)
Other	384 (17)
Unknown/missing	55 (2)
Pulse (beats/min) 76.4 +- 15.1 Systolic blood pressure (mm Hg) 140.4 +- 22.7 Diastolic blood pressure (mm Hg) 80.2 +- 14.3 Killip class (%) Class I 2040 (91) Class II 13 (1) Class III 0 Class IV 1 (b1) Missing 189 (8) Diagnosis after initial assessment (%)
Nonspecific chest pain	452 (20)
Anxiety	100 (4)
Angina, no ACS	351 (16)
Acute coronary syndrome	666 (30)
Gastroesophageal	232 (10)
Musculoskeletal	210 (9)
Other	179 (8)
Unknown/missing	53 (2)

Aspirin use Known CHD	39 64	Assume no (modal value) Assume no	0 0	NA NA		Age (y) Sex (%) Male	54.5 +- 14.1 1307 (58)
Rest angina	122	(modal value) Assume no	0	NA		Female Past history of CHD (%)	936 (42)
		(modal value)			Previous myocardial infarction		125 (6)
ST-segment	55	Assume normal	0	0	Angina with positive diagnostic test		99 (4)
deviation		(modal value)			Previous coronary artery bypass graft		27 (1)
Elevated	151	Assume normal	0	0	Previous angioplasty		71 (3)
cardiac		(modal value)			Stenosis N50% on angiography		26 (1)
markers					Unproven clinical label of CHD		67 (3)
Pulse Systolic blood	3 9	Impute group mean (76 beats/min) Impute group mean	NA NA	7 37	Other Risk factors (%) Diabetes		22 (1) 178 (8)
pressure		(140 mm Hg)			Hypertension		737 (33)
Serum	226	Impute modal value	NA	8	Hyperlipidemia		553 (27)
creatinine					Present smoker		626 (28)
Killip class	189	Assume class I	NA	0	Ex-smoker (last 10 y)		273 (13)
		(modal value)			Cocaine abuse		16 (1)
Cardiac arrest	0	NA	NA	NA	First degree with angina/MI, onset b60 y		696 (33)
at admission					Use of aspirin in previous 7 d		422 (19)
					Duration of worst onset of chest pain		230.2
and arrival at ED (min) Discrete variables were presented as frequencies or Duration of longest episode of 97.9 ercentages, whereas continuous variables with normal worst pain (min) istribution were expressed as a mean value. For illustrative First ECG in ED (%)

purposes, GRACE was categorized into 5 categories (b60, 60-79, 80-89, 90-109, N110) based on the approximate 20th, 40th, 60th, and 80th centiles of the score in the data set. Areas under the receiver operating characteristic curves (AUCs or c statistic) were derived by treating the GRACE and TIMI as continuous scores, and 95% confidence intervals (CIs) were calculated using the exact binomial method. Formal statistical comparison of the c statistics was performed using the methodology of DeLong et al [22]. Analyses were performed using the Stata statistical software, version SE 10.1 [23].

Results

In total, 2263 patients were enrolled into the trial at the 6 participating centers, but 20 were excluded for the following reasons: inadequate consent (n = 12), consent withdrawn (n = 2), and recruited in error (n = 6). Therefore, there were 2243 (99%) patients eligible for data analysis. Hospital record review for major adverse events was achieved for 2215 (98.8%) of 2243. Questionnaire responses were received from 1635 (72.3%) of 2243 at 30 days and 1514 (67.5%) at

Any event Death

nonfatal myocardial infarction Hospitalization for ACS without

myocardial infarction Life threatening arrhythmia

Emergency revascularization

43 (2)

2 (b1)

7 (b1)

17 (b1)

62 (3)

8 (b1)

10 (b1)

27 (1)

3 (b1)

23 (1)

8 (b1)

24 (1)

Table 4 Major adverse events

n (%)

30 d 90 d

(n = 2243) (n = 2243)

score. Most of the missing data were due to lack of documentation of risk factors of CHD (n = 222), rest angina (n = 122), and elevated cardiac markers (n = 151) in the case of the TIMI risk score, whereas most of those in the GRACE risk score were due to serum creatinine (n = 226) and Killip class (n = 189). Details are provided in Table 2.

Baseline characteristics of the study population are as shown in Table 3. Mean patient age was 54.5 years and they were predominantly male (58%). Approximately 19% of patients had a history of proven or suspected CHD. The most common risk factors were hypertension (33%), first-degree family member with CHD onset younger than 60 years (33%), current smoker (28%), and hyperlipidemia (27%). The mean duration between onset of worst pain and arrival at the ED was 230 minutes and mean duration of the longest episode of worst pain was 98 minutes. The initial ECG was interpreted as normal in most of the cases (69%), however, a small proportion of patients had ST deviation (3%) and T- wave inversion (7%) but were deemed not to be diagnostic of ACS by the recruiting doctor and research nurses. Ulti- mately, 1744 (78%) patients were admitted into hospital.

By 30 days, 43 (2%) patients had experienced a major adverse event: 2 (b1%) died, 3 (b1%) had life-threatening arrhythmias, 23 (1%) underwent emergency revasculariza-

Fig. 1 Proportion with a major adverse event (MAE) according to GRACE score.

Fig. 2 Proportion with a major adverse event (MAE) according to TIMI score.

tion, 7 (b1%) had nonfatal acute myocardial infarction, and 17 (b1%) were hospitalized for ACS without diagnosis of myocardial infarction . By 90 days, 62 (3%) patients had experienced a major adverse event: 8 (b1%) died, 8 (b1%) had life-threatening arrhythmias, 24 (1%) underwent emergency revascularization, 10 (b1%) had nonfatal acute myocardial infarction, and 27 (1%) were hospitalized for ACS without acute myocardial infarction. This is summarized in Table 4. Some patients had more than one event in the Composite outcome. The causes of death in patients who died during the study were myocardial infarction (n = 2), cancer (n = 2), pneumonia, multiorgan failure, and a further 2 died of unknown causes. Hospital record review identified all of the major adverse events except one nonfatal myocardial infarction that was reported by the patient after attending another hospital.

The relationship between the 30- and 90-day major adverse event rate and the different risk groups in each risk score is summarized in Figs. 1 and 2. This indicates that risk of a major adverse event increases with increasing score.

As shown in Fig. 3, the AUC or c statistic of GRACE and TIMI risk scores for 30-day events were 0.717 (95% CI,

Fig. 3 Receiver operating characteristic curves.

Fig. 4 Receiver operating characteristic curves.

0.698-0.735) and 0.682 (95% CI, 0.662-0.701), respectively.

The c statistics of GRACE and TIMI risk scores for 90-day events were 0.726 (95% CI, 0.707-0.745) and 0.693 (95%

CI, 0.674-0.712). In contrast, the c statistic for age alone as a predictor of major adverse events was 0.656 for 30-day events and 0.689 for 90-day events. There was no statistically significant difference between the 2 risk scores’ ability to predict 30- or 90-day events (?2 test, P = .29 and P = .26, respectively) (Fig. 4).

The c statistics for all analyses are shown in Table 5. predictive accuracy of the risk scores was slightly lower when only first troponin result was used, whereas results for complete case analysis were similar to the results for imputed analysis.

Discussion

	TIMI score, complete case, first troponin positive	0.664	0.642-0.686
The GRACE and TIMI risk scores were developed in	TIMI score, imputed,	0.682	0.662-0.701
patients with diagnosed ACS and have been evaluated in	any troponin positive a
several studies of unselected patients in the ED setting. Conway Morris et al [16] showed an association between TIMI risk score or “front door” score (TIMI minus	TIMI score, imputed, first troponin positive 90 d	0.662	0.642-0.682

but diagnostic parameters that were inadequate as a sole means of supporting decision-making.

These ED studies included patients with unstable angina, significant ECG changes [14,18], and in some circum- stances, ST-elevation myocardial infarction [13,16]. The predictive value of the TIMI and GRACE scores in patients with diagnosed ACS is well established, whereas their value in undiagnosed chest pain is less clear. We intentionally excluded patients with a clear diagnosis of ACS, so our study evaluated score performance in a specific patient group who present a major management challenge. The participation of 6 diverse hospitals means that findings are likely to be generalizable to a range of settings. Our study, like previous studies, demonstrated an association between each risk score and major adverse events. However, simply demonstrating an association does not mean that a score is clinically useful. AUC analysis showed that neither score accurately discriminated between those who will and will

GRACE score, complete case,	0.701	0.679-0.722
first troponin positive
GRACE score, imputed, any troponin positive a	0.717	0.698-0.735
GRACE score, imputed,	0.701	0.681-0.720
first troponin positive
TIMI score, complete case,	0.678	0.656-0.700
any troponin positive

troponin component) and adverse outcomes in patients with Undifferentiated chest pain. Lyon et al [17] showed that the TIMI risk score and the GRACE risk score had similar discriminant value (AUC 0.79 vs 0.80) for adverse outcomes in patients with undifferentiated chest pain. Hess et al [18] evaluated the original TIMI score and a modified version that classified patients with ST-segment deviation or troponin elevation as high risk in patients with acute chest pain. The modified TIMI score out-performed the original (AUC, 0.83 vs 0.79), but the authors concluded

Table 5 Receiver operating characteristic analysis of GRACE and TIMI

30 d

GRACE score, complete case, any troponin positive

0.722

0.700-0.743

any troponin positive ^a TIMI score, imputed, first troponin positive

0.673

0.653-0.692

^a Main analysis.

c statistic 95% CI

GRACE score, complete case,	0.731	0.709-0.752
any troponin positive
GRACE score, complete case,	0.717	0.695-0.738
first troponin positive
GRACE score, imputed, any troponin positive a	0.726	0.707-0.745
GRACE score, imputed,	0.712	0.693-0.731
first troponin positive
TIMI score, complete case,	0.680	0.658-0.702
any troponin positive
TIMI score, complete case,	0.665	0.643-0.687
first troponin positive
TIMI score, imputed,	0.693	0.674-0.712

that both scores were insufficiently sensitive and specific

to recommend as the sole means of determining disposition in ED chest pain patients. A meta-analysis of prospective cohort studies of the TIMI score [19] in ED chest pain patients came to a similar conclusion based on a strong association between TIMI score and risk of adverse events,

not have a major adverse event. Indeed, the TIMI and GRACE were little better at predicting adverse outcome than patient age.

Data for complete case analysis were available for only 76% and 80% of patients when calculating GRACE and TIMI risk scores, respectively. For the imputed analysis, we assumed that these results (eg, angina, Killip class, history of CHD, elevated cardiac markers) were negative or normal for our cohort. In addition, one of the items (cardiac arrest on admission) in the GRACE risk score was absent because the design of the study. This could potentially lead our study to underestimate the discriminatory accuracy of the risk score. The timing of troponin sampling was not determined by a strict protocol: the first troponin sample was not always taken immediately at presentation and not all patients had repeat troponin sampling. Thus, the study may have underestimated the potential difference between using a score based on presentation data and a score using all available data. Although hospital records were checked for major adverse events in more than 98% of cases, response rates to the questionnaire were 72% at 30 days and 67% at

90 days. If any nonresponders attended another hospital with an adverse event, then they would have been misclassified as having no adverse event. However, among those who responded, the questionnaire identified only one major adverse event that had not already been identified by hospital record review. It is therefore unlikely that significant numbers of events were missed. Finally, it should be reemphasized that the study cohort consisted of selected patients with suspected but not diagnosed ACS. Our findings do not attempt to challenge the established data showing that both scores accurately predict major adverse event in cohorts of patients with diagnosed ACS.

Conclusions

Our study demonstrated an association between the GRACE and TIMI scores and an increased risk of adverse outcome in patients with suspected but not proven ACS. However, neither score accurately discriminated between those who will and will not have major adverse events.

Acknowledgments

We would like to thank Margaret Jane for her invaluable clerical assistance, the staff involved at the NHS sites for all their hard work and participants for their involvement, and Francis Morris for advice on ED data.

References

1. Goodacre S, Cross E, Arnold J, et al. The health care burden of acute chest pain. Heart 2005;91:229-30.
2. Panju AA, Hemmelgam BR, Guyatt GH, et al. The rational clinical examination. Is this patient having myocardial infarction? JAMA 1998;280:1256-63.
3. Sanchis J, Bodi V, Nunez J, et al. Limitations of clinical history for evaluation of patients with acute chest pain, non-diagnostic electro- cardiogram, and normal troponin. Am J Cardiol 2008;101:613-7.
4. Goodacre S, Pett P, Arnold J, et al. Clinical diagnosis of acute coronary syndrome in patients with chest pain and a normal or non-diagnostic electrocardiogram. Emerg Med J 2009;26:866-70.
5. Reilly BM, Evans T, et al. Impact of a clinical decision rule on the hospital triage of patients with suspected acute cardiac ischaemia in the emergency department. JAMA 2002;288:342-50.
6. Georgiou d, Budoff MJ, Kaufer E, et al. Screening patients with chest in the emergency department using electron beam tomography: a follow-up study. J Am Coll Cardiol 2001;38:105-10.
7. Limkakeng Jr A, Gibler WB, Pollack C, et al. Combination of Goldman risk and initial cardiac troponin I for emergency department chest pain patient risk stratification. Acad Emerg Med 2001;8: 696-702.
8. Jacobs Jr DR, Kroenke C, Crow R, et al. PREDICT: a simple risk score for clinical severity and long-term prognosis after hospitalization for acute myocardial infarction or unstable angina: the Minnesota Heart Survey. Circulation 1999;100:599-607.
9. Antman EM, Cohen M, Bernink PJ, et al. The TIMI risk score for unstable angina/non-ST elevation MI: a method for prognostication and therapeutic decision making. JAMA 2000;284:835-42.
10. The GRACE investigators. Predictors of hospital mortality in the global registry of acute coronary events. Arch Intern Med 2003;163: 2345-53.
11. Eagle KA, Lim MJ, Dabbous OH, et al. A validated prediction model for all forms of acute coronary syndrome: estimating the risk of 6-month postdischarge death in an international registry. JAMA 2004;291:2727-33.
12. Scirica BM, Cannon CP, Antman EM, et al. Validation of the Thrombolysis in Myocardial Infarction risk score for unstable angina pectoris and non-ST elevation myocardial infarction in the TIMI III registry. Am J Cardiol 2002;90:303-5.
13. Chase M, Robey J, Zogby K, et al. Prospective validation of the Thrombolysis in Myocardial Infarction risk score in the emergency department chest pain population. Ann Emerg Med 2006;48:252-9.
14. Pollack CV, Sites FD, Shofer FS, et al. Application of the TIMI risk score for unstable angina and non-ST elevation acute coronary syndrome to an unselected emergency department chest pain population. Acad Emerg Med 2006;13:13-8.
15. Bartholomew BA, Sheps DS, Monroe S, et al. A population based evaluation of the thrombolyisis in myocardial infarction risk score for unstable angina and non-ST elevation myocardial infarction. Clin Cardiol 2004;27:74-8.
16. Conway Morris A, Caesar D, Gray S, Gray A. TIMI risk score accurately risk stratifies patients with undifferentiated chest pain presenting to an emergency department. Heart 2006;92:1333-4.
17. Lyon R, Conway Morris A, Caesar D, Gray A. Chest pain presenting to the emergency department-to stratify risk with GRACE or TIMI? Resuscitation 2007;74:90-3.
18. Hess EP, Perry JJ, Calder LA, Thiruganasambandamoorthy V, Body R, Jaffe A, et al. Prospective validation of a modified thrombolysis in myocardial infarction risk score in emergency department patients with chest pain and possible acute coronary syndrome. Acad Emerg Med 2010;17(4):368-75.
19. Hess EP, Agarwal D, Chandra S, Murad MH, Erwin PJ, Hollander JE, et al. Diagnostic accuracy of the TIMI risk score in patients with chest pain in the emergency department: a meta-analysis. CMAJ 2010;182: 1039-44.
20. Goodacre SW, Bradburn M, Cross E, Collinson P, Gray A, Hall AS on behalf of the RATPAC Research Team. The Randomised Assessment of Treatment using Panel Assay of Cardiac Markers (RATPAC) trial: a randomised controlled trial of point-of-care cardiac markers in the

emergency department. Heart 2010; published online first doi: 10.1136/hrt.2010.203166.

1. Thygesen K, Alpert JS, White HD. Joint ESC/ACCF/AHA/WHF task force for the redefinition of myocardial infarction. Universal Definition of Myocardial Infarction. J Am Coll Cardiol 2007;50(22):2173-95.
2. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated Receiver operating curves: a nonpara- metric approach. Biometrics 1988;44:837-45.
3. StataCorp. Stata Statistical Software: Release 10. College Station (TX): StataCorp LP; 2007.