A simple score for predicting coronary artery disease in patients with chest pain
From the Cardiothoracic Centre, Guy's and St Thomas' Hospitals, London, UK
Address correspondence to Dr E.B. Wu, Department of Medicine and Therapeutics, Prince of Wales Hospital, Ngan Shing Road, Shatin, Hong Kong. email: ebwu{at}netvigator.com
Received 23 July 2004 and in revised form 4 July 2005
 |
Summary |
|---|
 Top Summary IntroductionMethodsResultsDiscussionReferences |
|---|
Background: We have previously derived a chest pain score by comparing those with and without coronary artery disease on angiography, which was subsequently validated in patients attending coronary angiography.
Aim: To test the predictive validity of the score prospectively in a more varied out-patient population, and to determine whether it had predictive validity in addition to exercise testing.
Design: Prospective clinical study.
Methods: The score was applied to 405 out-patients with chest pain who subsequently underwent coronary angiography. Framingham risk analysis and exercise testing were performed in 155.
Results: The score had a sensitivity of 91.4% and specificity of 28% for coronary artery disease, which was found in 31.8%, 51%, 63%, and 82% of those with scores of 0, 1, 2, and 3, respectively. Gender (p < 0.001), age (p < 0.001), and chest pain score (p = 0.009) independently predicted coronary artery disease on multivariate Poisson regression analysis. The chest pain score had additive predictive value with Framingham risk analysis and Duke's score.
Discussion: This simple chest pain score can predict coronary anatomy with similar sensitivity to exercise testing, and can be used in conjunction with exercise testing and other measures. Further validation of the chest pain score in the primary care setting will be useful.
|
Introduction |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Chest pain is reported by 16–20% of the population of the UK1 and USA,2 and usually has a benign non-cardiac aetiology.3 In those referred to a cardiologist, the incidence of cardiac disease may be as low as 11–27%.4,5 In the UK, chest pain clinics have been developed to improve access for patients with acute coronary syndromes, but even these report an incidence of non-cardiac pain of about 50–70%.6–8 Furthermore, the incidence of normal coronary anatomy in patients investigated invasively varies widely, between 11% and 37%, at different cardiac centres,9,10 and also between different physicians at a single centre.11 It is possible that better clinical assessment could lower the rate of referral of cases with non-cardiac pain, and reduce the incidence of normal coronary anatomy.
In routine clinical practice, patients with chest pain are assessed in terms of their chest pain characteristics, coronary risk profile, and the results of non-invasive investigation, usually a treadmill exercise test. Chest pain has traditionally been classified as ‘typical’ or ‘atypical’ of a cardiac origin, which are subjective terms open to wide interpretation, even with the use of standardized questionnaires.12,13 We therefore looked at chest pain characteristics, and found three chest pain characteristics that were significantly different between those with and those without coronary artery disease.14 From this we derived a chest pain score, and confirmed this score on prospective testing15 to be directly related to the likelihood of coronary artery disease in patients undergoing coronary angiography.
The aim of this study was to evaluate the chest pain score further in a larger cohort, combined with results of exercise testing and coronary risk analysis. The end-point was coronary artery disease on angiography.
|
Methods |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Patients
From 1 September 1997 to 31 March 1999, 405 patients who fulfilled the inclusion criteria were recruited from 363 consecutive patients seen as out-patients at our department, and 829 consecutive patients undergoing day-case coronary angiography. Those attending day-case coronary angiography have been described elsewhere.15
Inclusion criteria included chest pain for >1 month without previous history of myocardial infarction, coronary angiography, angioplasty, or coronary artery bypass grafting. Patients were excluded if there were pathological Q waves on electrocardiography or regional wall motion abnormalities on the echocardiogram. None had more than mild valve disease, left ventricular hypertrophy, or left bundle branch block.
Thus in total 405 patients were included, who had a chest pain score assessment and subsequent coronary angiography. Of these, 155 also underwent exercise tolerance testing. The study was approved by the local committee on ethical practice and all subjects gave verbal consent.
Clinical assessment
All patients were interviewed by one author (EBW). In a subgroup of 100 patients, the chest pain score was also assessed by questionnaire format, randomized to either before or after the interview. Significant bias was observed, with patients inevitably filling out the questionnaire as they had answered in the interview, but this reproducibility was not seen when the patient took the questionnaire first. The simplicity of the chest pain score makes for difficult comparison between observers, due to patient recall. Coronary risk was assessed using the Framingham database.16 Chest pain was assessed with a modification of the Master17 questionnaire, with the addition of the three specific questions (Figure 1) constituting the previously validated scoring system14: (i) ‘If you go up a hill (or another individually appropriate stressor) on ten separate occasions, on how many of these do you experience chest pain?’ The answer to this question was called the reproducibility index. (ii) ‘If you have the pain ten times in a row, how many happen when you are resting or sitting quietly?’ The answer to this question was called the rest index. (iii) ‘How long does the pain usually last?’ The answer to this question was called the duration index. The answers to these questions were dichotomized as either ‘typical’ or ‘atypical’ based on the previous logistic regression analysis.14 For question 1, a reproducibility index of 10/10 was defined as ‘typical’ and a score of 1 to 9 out of 10 was defined as ‘atypical’. For question 2, a rest index of 0 or 1 out of 10 was defined as ‘typical’ and 2 or more out of 10 was defined as ‘atypical’. Pain duration of 5 min or shorter was ‘typical’ and longer than 5 min was ‘atypical’. One point was given for every answer defined as ‘typical’ to give a chest pain score ranging between 0 and 3. Thus a chest pain score of 0 was unequivocally ‘atypical’, 3 unequivocally ‘typical’ and 1 to 2 were intermediate.
|
Exercise test
A standard Bruce protocol exercise test was performed on a semi-automated treadmill (Quinton Q55xt or Marquette Mach IV). The 12-lead ECG and heart rate were monitored throughout exercise and during recovery, and blood pressure was measured at the end of every stage of exercise and at the end of the recovery phase. The time of first occurrence of chest pain and the symptom leading to stopping exercise were noted. A test was considered positive if there was abnormal ST segment depression, defined as >1 mm of planar or down-sloping depression measured 80 ms after the J point (or 60 ms for heart rates >140 bpm). The exercise test was also scored using the Duke system,18 which includes exercise time, the degree of pain, and the magnitude of ST segment depression. This has been validated in both in-patients and out-patients.19,20
Coronary angiography
Coronary angiography was requested according to standard clinical criteria by the clinicians in charge of each case, and was not part of the protocol of this study. The standard Sones or Judkins techniques were used. The angiogram was recorded as abnormal if a stenosis judged by eye to be 50% or more of the luminal diameter was shown in at least one main epicardial vessel or a major branch. The angiogram was recorded as normal or near normal if the most severe stenosis was <50% of the luminal diameter as assessed by eye. Patients with near-normal angiograms were grouped as normal, an oversimplification that was permitted because it was likely to underestimate rather than exaggerate the validity of our results.
Statistics
Mean and standard deviation values were calculated. A student's t-test was used for comparison of continuous variables and the Mann-Whitney U test for non-parametric data. The 
2 test was used for the comparison of chest pain scores between groups. The univariate relative risk (RR) and 95%CIs between those with and without an end-point were calculated. Factors with a significant univariate relative risk, which included the chest pain score, coronary risk and age, were included in a multivariate Poisson regression analysis21 for adjusted relative risk and 95%CIs. A receiver operating characteristic (ROC) curve was plotted to access the discrimination of the model. Areas under the curve (AUC) were compared to identify the discrimination capacity among chest pain score, Framingham risk, and Dukes risk score. A p value of <0.05 was considered to be significant. Statistical analyses were performed using Stata 8.0 and Medcalc.
|
Results |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Patients
There were 405 patients, mean (SD) age 60.6 (9.5) years: 268 (66%) males aged 60.5 (9.1) years and 137 (34%) females aged 60.8 (10.2) years. Of these patients, 244 (60%) had significant coronary artery disease and 161 (40%) had normal coronary anatomy.
Chest pain scores
One patient failed to complete the chest pain score. The chest pain score was 0 in 66 (16.3%) patients, 1 in 106 (26.2%), 2 in 111 (27.5%), and 3 in 121 (30%). An end-point occurred in 21/66 (31.8%) patients with a score of 0, 54/106 (51%) patients with a score of 1, 70/111 (63%) patients with a score of 2, and 99/121 (82%) patients with a score of 3. Using a chest pain score of zero as negative and a chest pain score of 1–3 as positive predictor for coronary artery disease, the sensitivity of the chest pain score was 91.4% and the specificity 28%. This compares to the Duke score's sensitivity of 82.4% and specificity of 31%.
Differentiating factors
On univariate relative risk analysis of all the chest pain characteristics included in the Master's questionnaire, the three chest pain characteristics that forms the chest pain score (i.e. the exercise score, the rest score and the duration index of <5 min) and the overall chest pain score all showed significant difference between those with and without subsequent coronary artery disease on coronary angiography. Other characteristics that reached significance included pain relief with rest and sublingual nitrates (occurring less frequently in those with normal coronary anatomy), radiation of the pain to the back, dizziness, tingling, palpitations (occurring more frequently in those with normal coronary anatomy), sex, diabetes, and hypertension (Table 1
).
|
|
|
|
|
|
|
Combined risk stratification
Multivariate Poisson regression analysis showed that gender (p < 0.001), age (p < 0.001), and chest pain score (p = 0.009) independently differentiated those with and without coronary artery disease. Multivariate Poisson regression analysis was performed with age, sex, chest pain score and conventional coronary risk factors that were significant on univariate analysis. Relief with rest (p = 0.046), dizziness (p = 0.030), smoking (p = 0.006), hypertension (p = 0.016), and diabetes (p = 0.016) were also found to be significant in addition to age, sex, and the chest pain score in the model (Table 5).
|
|
|
|
Chest pain score in relation to exercise testing and Framingham risk
The chest pain score of zero identified a low-risk group of 26 patients with risk of coronary artery disease of 27% only. When used in conjunction with the Framingham risk analysis, the chest pain score can further differentiate patients into high and low risk groups for coronary artery disease (Table 2). Using a chest pain score of zero together with a Duke's exercise test score of >4, we identified a group of 12 patients without coronary artery disease (Table 3). A combination of chest pain score, Framingham risk profile, and exercise test increased the number of cells with a low risk of end-point. Thus a coronary risk <15%, Duke's score of >4, and chest pain score 0–2 gave a low risk (20%) in 6/30 patients (Table 4). AUC for chest pain score was 0.713 (95%CI 0.635–0.783); while AUC for Framingham risk profile and Duke's exercise test score were 0.641 (95%CI 0.56–0.716) and 0.668 (95%CI 0.588–0.741), respectively. Pair-wise comparisons of ROC curves (Figure 3) were not significant (Chest pain score vs. Framingham risk profile, p = 0.208; chest pain score vs. Duke's exercise test score, p = 00.418; Framingham risk profile vs. Duke's exercise test score p = 0.656).
|
|
Discussion |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
This study shows that a simple chest pain score can stratify patients in terms of the likelihood of having significant coronary artery disease on coronary angiography. The chest pain score had similar sensitivity and specificity to Duke's score on exercise testing.
The chest pain score was derived from a comparative case control study of chest pain characteristics between those with and without coronary artery disease on coronary angiogram.14 Subsequently, the chest pain score was validated in 250 patients with chest pain on the day of attending day-case coronary angiography.15 However, for the chest pain score to be clinically useful, it needs to be able to stratify patients not only on the day of their angiogram, but also at their initial clinical visit, and in conjuncture with exercise testing and Framingham risk analysis. Therefore, this present study extended our previous work by including patients who were recruited at the time when they initially presented as out-patients, and also included their exercise test data in the analysis.
The study confirmed the prospective validity of the chest pain score even at initial clinical visit, and this risk stratification is useful in an additive manner when used in conjuncture with the Duke exercise test score and Framingham risk analysis. Therefore, in the clinic setting, the chest pain score can be used as a stand-alone scoring system for the diagnosis of coronary artery disease, as 89% of patients with a chest pain score of 3/3 have coronary artery disease (Table 2), and they do not require further non-invasive testing. Similarly, 80% of those with intermediate chest pain scores of 1–2 with high Framingham risk for coronary artery disease will have coronary artery disease, and could therefore proceed straight to angiography. Using this method, 40% of all patients referred for a cardiological opinion could proceed directly to coronary angiography, resulting in a normal coronary angiography rate of only 14% (Table 2), considerably lower than most published series of around 20–30%. The value of exercise testing is reduced by the use of the chest pain score. However, a normal exercise test in patients with chest pain score of 0 is highly predictive of normal coronary anatomy, and these patients can be discharged after reassurance (Table 3). The remaining 53% of patients with intermediate chest pain scores and low Framingham risks have a post-exercise-testing risk of coronary artery disease of around 50%, and will require further non-invasive testing (Table 4).
In this high-risk group, the chest pain score allows the physician to make a positive diagnosis of coronary artery disease in a substantial proportion of patients, reducing the amount of unnecessary non-invasive investigation. However, in a lower-risk group in the community where more patients have a chest pain score of 0, the chest pain score may be more useful in preventing unnecessary referral for investigations. Further research into lower-risk cohorts with longer period of clinical follow-up is necessary to establish the usefulness of the chest pain score in this context.
Limitations
We did not use quantitative analysis of coronary lesions nor intravascular ultrasound, but confined ourselves to the same method of analysis as used in routine clinical practice. Furthermore, we used an imperfect cut-off between our two clinical groups. Some patients with coronary stenoses at around 50% have no evidence of objective myocardial ischaemia, while others may have significant ischaemia. It is therefore possible that some patients in our normal group may have had myocardial ischaemia and others in the coronary group may have had no ischaemia. This dichotomous classification was allowed, as it reflects the common clinical practice where only a very small minority of patients with borderline lesions undergo Doppler flow wire measurements to establish the flow limitation of the lesion, and the majority of lesions that are treated with angioplasty are done so on the basis of eyeballing alone. Also, these potential errors of classification are few, and would tend to underestimate rather than exaggerate the significance of our results.
The study recruited patients who did not have exercise testing before coronary angiography, as this reflects clinical practice, where not all patients inevitably undergo exercise testing. Had we not included these patients, our study might have missed out the highest risk group of patients and invalidated the findings. The results of this study suggest that clinical assessment alone can predict a 85% risk of coronary artery disease in up to 40% of patients, which validates the approach of angiography without exercise testing in high-risk patient groups.
Patients with acute coronary syndromes may have atypical pain, in the sense that it may originate at rest and last longer than 5 min. Our scoring system can only be used for patients with chronic pain of at least 1 month's duration, and must be interpreted carefully in individual cases.
The questionnaire is no more than semi-objective. However, this is not as major a limitation as it first seems, since the statistically determined cut-off between ‘typical’ and ‘atypical’ is at extremes of the scoring scales. A patient is likely to be able to recall whether he or she experiences pain every time he is subjected to a particular stressor, or whether pain almost never occurs at rest. Had the statistically determined threshold been at 7 or 8 times out of 10, its accuracy might have been more in question. A small number of patients were unable to answer the questions adequately. The chest pain score is simple to use, as it is consistent with our clinical experience, and the cut-off points are easy to recall.
One of the uses of the scoring system may be to aid general practitioners to risk stratify patients and to judge the need for referral to a cardiologist. The chest pain score can be used to identify those at low risk of coronary artery disease. Based on chest pain score alone, our low-risk group (chest pain score of 0) had a 21/66 (31.8%) risk of coronary artery disease. This risk would be even lower in the community, where the pre-test risk of coronary artery disease is considerably lower than that for patients on an angiogram waiting list.
Conclusions
Our chest pain score can be used alone to identify patients at high risk (85%) of coronary artery disease. In patients with an intermediate score, the score can be used in conjunction with coronary risk analysis and exercise testing to further risk stratify patients into high and low risk of coronary artery disease. This score may be useful for clinical assessment of chest pain, guiding referral to a cardiologist, in aiding the decision for non-invasive and invasive investigation, and defining ‘typical’ chest pain for future research.
|
Acknowledgments |
|---|
Dr E.B. Wu is funded by a grant from the special Trustee's of Guy's and St Thomas' NHS Trust.
|
References |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
1. Hotopf M, Mayou R, Wadsworth M, Wessely S. Psychosocial and developmental antecedents of chest pain in young adults. Psychosom Med 1999; 61:861–7.
2. Savage DD, Devereux RB, Garrison RJ, Castelli WP, Anderson SJ. Mitral valve prolapse in the general population. 2. Clinical features: the Framingham Study. Am Heart J 1983; 106:577–81.[CrossRef][Web of Science][Medline]
3. Chambers J, Bass C. Chest pain with normal coronary anatomy: a review of natural history and possible etiologic factors. Prog Cardiovasc Dis 1990; 33:161–84.[CrossRef][Web of Science][Medline]
4. Kroenke K, Mangelsdorff AD. Common symptoms in ambulatory care: incidence, evaluation, therapy, and outcome. Am J Med 1989; 86:262–6.[CrossRef][Web of Science][Medline]
5. Jain D, Fluck D, Sayer JW, Ray S, Paul EA, Timmis AD. One-stop chest pain clinic can identify high cardiac risk. J Roy Coll Physicians 1997; 31:401–4.
6. Newby DE, Fox KAA, Flint LL, Boon NA. A ‘same day’ direct-access chest pain clinic: improved management and reduced hospitalization. Q J Med 1998; 91:333–7.
7. Sutcliffe SJ, Fox KF, Wood DA. How to set up and run a rapid access chest pain clinic. Br J Cardiol 2000; 7:692–702.
8. Davie AP, Caesar D, Caruana L, et al. Outcome from a rapid-assessment chest pain clinic. Q J Med 1998; 91:339–43.
9. Proudfit WL, Shirey EK, Sones FM. Selective cine coronary arteriography: correlation with clinical findings in 1,000 patients. Circulation 1966; 33:901–10.
10. Kemp HG, Kronmal RA, Vliestra RE, Frye RL. Seven year survival of patients with normal or near normal coronary arteriograms: a CASS registry study. J Am Coll Cardiol 1986; 7:479–83.[Abstract]
11. Dans PE, Keruly JC, Brinker JA, Tabatznik B, Pristoop AS, Buchanan SG. Analysis of physician and hospital differences in ‘negative’ coronary angiogram rates. Arch Int Med 1988; 148:2633–7.
12. Bulpitt CJ, Shipley MJ, Demirovic J, Ebi-Kryston KL, Markowe HL, Rose G. Predicting death from coronary heart disease using a questionnaire. Int J Epidemiol 1990; 19:899–904.
13. Day LJ, Sowton E. Clinical features and follow-up of patients with angina and normal coronary arteries. Lancet 1976; 2:334–7.[CrossRef][Web of Science][Medline]
14. Cooke, RA, Smeeton, N, Chambers, JB. Comparative study of chest pain characteristics in patients with normal and abnormal coronary angiograms. Heart 1997; 78:142–6.
15. Wu EB, Smeeton N, Chambers JB. A chest pain score for stratifying the risk of coronary artery disease in patients having day case coronary angiography. Int J Cardiol 2001; 78:257–64.[CrossRef][Web of Science][Medline]
16. Murabito JM, Anderson KM, Kannel WB, Evans JC, Levy D. Risk of coronary heart disease in subjects with chest discomfort: the Framingham heart study. Am J Med 1990; 89:297–302.[CrossRef][Web of Science][Medline]
17. Master AM. The spectrum of anginal and non cardiac pain. JAMA 1964; 187:894–9.
18. Mark DB, Shaw L, Harrell FE Jr, et al. Prognostic value of a treadmill exercise score in outpatients with suspected coronary artery disease. New Engl J Med 1991; 325:849–53.[Abstract]
19. Iskandrian AS, Ghods M, Helfeld H, Iskandrian B, Cave V, Heo J. The treadmill exercise score revisited: coronary arteriographic and thallium perfusion correlates. Am Heart J 1992; 124:1581–6.[CrossRef][Web of Science][Medline]
20. Shaw LJ, Peterson ED, Shaw LK, et al. Use of a prognostic treadmill score in identifying diagnostic coronary disease subgroups. Circulation 1998; 98:1622–30.
21. Zou G. A Modified Poisson Regression Approach to Prospective Studies with Binary Data. Am J Epidemiol 2004; 159:702–6.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||