Accuracy of Exercise Stress Test for Coronary Artery Disease
Accuracy of Exercise Stress Test for Coronary Artery Disease
Our systematic review discovered five main modalities of exercise stress testing (treadmill ECG, treadmill echocardiography, bicycle ECG, bicycle echocardiography, MPI). There are other test modalities which are used (including cardiac MRI and CT), but did not fit our detailed inclusion criteria. To our knowledge, this is the most comprehensive review of exercise testing in this patient population to date.
The major conclusion is that the diagnostic accuracy of exercise testing varies, depending upon the age, gender and clinical characteristics of the patient, the prevalence of CAD, and the modality of test used. Stress testing seems most useful in low and intermediate prevalence, because in high prevalence, the change in probability will not lead to the exclusion or definite inclusion of CAD. In young patients and in men, a positive exercise test is better at discriminating CAD than in older patients and women. A positive test is more clinically useful from bicycle echocardiography than other test modalities at ruling in CAD. Overall, echocardiographic testing performed better than ECG testing in terms of positive likelihood ratios. Exercise testing, whether using echocardiography or ECG, is probably more useful at confirming CAD than excluding it. In addition, the different test modalities have differing diagnostic performances depending on the degree of coronary stenosis which is being predicted. For example, treadmill ECG has higher positive and lower negative likelihood ratios when predicting> 70% stenosis, whereas the other modalities perform better when predicting< 70% stenosis.
There are several limitations to our study. Firstly, a major weakness of the current review lies in the methodological problems of individual studies (e.g. lack of low prevalence studies, number of studies, size of study populations within studies). Secondly, prior knowledge of test results within studies brings validity of the individual study results into question. Thirdly, there is the possibility that our search missed studies but we made our search as comprehensive as possible. Fourthly, although our review was initially premised on exercise testing in relation to CAD diagnosis, the endpoints of included studies ranged from catheterization data and post mortem exams to ACS. Therefore, exercise testing in practice may well provide a mechanism of prognostic assessment (predicting those who would suffer events and/or who would be referred to catheterisation) as well as diagnostic assessment. Fifthly, the results presented herein do not take into account the cost effectiveness of the different test modalities and the technical, training and resource implications of the different tests, as well as patient and clinician preferences. For example, bicycle echocardiography or MPI are more expensive, take longer and require more training of the concerned health professionals than treadmill ECG testing. Finally, additional factors which have been previously shown to have an impact on choice of test modality, such as ischaemic territory, left ventricular ejection fraction and patient attributes (weight, functional capacities), could not be considered in this meta-analysis.
Understanding how test results impact on individual patients is important for deciding future management options. Consider patient 1, a 63-year-old man with angina, diabetes and smoking history. He has a high risk of CAD therefore treadmill ECG, the most commonly performed test would not be helpful. If he is not mobile enough to exercise, pharmacological stress testing may be an option, but has equally disappointing diagnostic value. Treadmill echocardiography performs a little better, but the most clinically useful test is bicycle echocardiography with an overall LR+ of 11.34 and a LR− of 0.17. If patient 1 was hypertensive, bicycle echocardiography is even more discerning (LR+ 29.38 and LR− 0.11).
Patient 2 is a 49-year-old woman with angina and no other cardiovascular risk factors. She has an intermediate probability of CAD. Treadmill ECG, bicycle ECG and bicycle echocardiography have low diagnostic accuracy in women. However, treadmill echocardiography (LR+ 9.31) and MPI (+LR 13.24) are particularly useful in women, and should be the tests of choice in this situation.
The implications of our findings are important for the choice of test modality in any particular clinical setting, since CAD is very common and the method and accuracy of its diagnosis have far-reaching consequences in terms of human and financial resources, as well as for the management of the individual patient. Clinicians have concentrated on individualising the treatment of CAD, but our study shows that there is great scope for individualising the diagnosis of CAD using exercise testing.
Overall, our meta-analyses show that exercise testing, whether using echocardiography or ECG, is probably more useful at confirming CAD than excluding it. Therefore, its widespread use as the "gatekeeper" to coronary angiography in the clinical setting is useful when the test result is positive. However, a negative exercise test result is not as good at excluding CAD and the clinical challenge of knowing which of these test-negative patients benefit from urgent coronary angiography and treatment remains. Prevalence (and therefore pretest probability) of CAD seems to be a key factor in how well a stress test performs in diagnosis of CAD. When CAD is less likely (or prevalence is low), the diagnostic value of stress testing is higher than situations where CAD is more likely. Therefore, in high to middle prevalence settings, a stress test is useless because it changes the post-test probability of CAD very little in either direction.
Imagine a 30-year-old man with atypical chest pain and no cardiac risk factors. When CAD is unlikely as in this scenario, a positive stress test is most useful and may actually tilt the probability over the threshold for doing coronary angiography. In practical terms, stress testing is only useful in those situations where clinicians would feel comfortable not performing coronary angiography in the case of a negative stress test. A recent analysis from a contemporary US population showed that patients with a positive result on non-invasive stress testing were only 28% more likely to have CAD on coronary angiography than those who did not undergo any testing (41.0% vs. 35.0%; p < 0.001; adjusted odds ratio, 1.28; 95% CI, 1.19–1.37), highlighting the need for improved risk stratification prior to angiography.
Our findings have implications on how results of diagnostic tests are published, which is particularly relevant to newer imaging modalities such as cardiac PET, cardiac MRI and multi-slice CT. Future publications need to provide raw numbers to allow calculations of sensitivity, specificity and likelihood ratios, and must consider modalities side by side, and in sequence if a test has a negative result. In addition, more prospective cohort studies of diagnostic tests in the setting of CAD are required, considering both diagnostic and prognostic implications side by side.
Discussion
Our systematic review discovered five main modalities of exercise stress testing (treadmill ECG, treadmill echocardiography, bicycle ECG, bicycle echocardiography, MPI). There are other test modalities which are used (including cardiac MRI and CT), but did not fit our detailed inclusion criteria. To our knowledge, this is the most comprehensive review of exercise testing in this patient population to date.
The major conclusion is that the diagnostic accuracy of exercise testing varies, depending upon the age, gender and clinical characteristics of the patient, the prevalence of CAD, and the modality of test used. Stress testing seems most useful in low and intermediate prevalence, because in high prevalence, the change in probability will not lead to the exclusion or definite inclusion of CAD. In young patients and in men, a positive exercise test is better at discriminating CAD than in older patients and women. A positive test is more clinically useful from bicycle echocardiography than other test modalities at ruling in CAD. Overall, echocardiographic testing performed better than ECG testing in terms of positive likelihood ratios. Exercise testing, whether using echocardiography or ECG, is probably more useful at confirming CAD than excluding it. In addition, the different test modalities have differing diagnostic performances depending on the degree of coronary stenosis which is being predicted. For example, treadmill ECG has higher positive and lower negative likelihood ratios when predicting> 70% stenosis, whereas the other modalities perform better when predicting< 70% stenosis.
There are several limitations to our study. Firstly, a major weakness of the current review lies in the methodological problems of individual studies (e.g. lack of low prevalence studies, number of studies, size of study populations within studies). Secondly, prior knowledge of test results within studies brings validity of the individual study results into question. Thirdly, there is the possibility that our search missed studies but we made our search as comprehensive as possible. Fourthly, although our review was initially premised on exercise testing in relation to CAD diagnosis, the endpoints of included studies ranged from catheterization data and post mortem exams to ACS. Therefore, exercise testing in practice may well provide a mechanism of prognostic assessment (predicting those who would suffer events and/or who would be referred to catheterisation) as well as diagnostic assessment. Fifthly, the results presented herein do not take into account the cost effectiveness of the different test modalities and the technical, training and resource implications of the different tests, as well as patient and clinician preferences. For example, bicycle echocardiography or MPI are more expensive, take longer and require more training of the concerned health professionals than treadmill ECG testing. Finally, additional factors which have been previously shown to have an impact on choice of test modality, such as ischaemic territory, left ventricular ejection fraction and patient attributes (weight, functional capacities), could not be considered in this meta-analysis.
Understanding how test results impact on individual patients is important for deciding future management options. Consider patient 1, a 63-year-old man with angina, diabetes and smoking history. He has a high risk of CAD therefore treadmill ECG, the most commonly performed test would not be helpful. If he is not mobile enough to exercise, pharmacological stress testing may be an option, but has equally disappointing diagnostic value. Treadmill echocardiography performs a little better, but the most clinically useful test is bicycle echocardiography with an overall LR+ of 11.34 and a LR− of 0.17. If patient 1 was hypertensive, bicycle echocardiography is even more discerning (LR+ 29.38 and LR− 0.11).
Patient 2 is a 49-year-old woman with angina and no other cardiovascular risk factors. She has an intermediate probability of CAD. Treadmill ECG, bicycle ECG and bicycle echocardiography have low diagnostic accuracy in women. However, treadmill echocardiography (LR+ 9.31) and MPI (+LR 13.24) are particularly useful in women, and should be the tests of choice in this situation.
The implications of our findings are important for the choice of test modality in any particular clinical setting, since CAD is very common and the method and accuracy of its diagnosis have far-reaching consequences in terms of human and financial resources, as well as for the management of the individual patient. Clinicians have concentrated on individualising the treatment of CAD, but our study shows that there is great scope for individualising the diagnosis of CAD using exercise testing.
Overall, our meta-analyses show that exercise testing, whether using echocardiography or ECG, is probably more useful at confirming CAD than excluding it. Therefore, its widespread use as the "gatekeeper" to coronary angiography in the clinical setting is useful when the test result is positive. However, a negative exercise test result is not as good at excluding CAD and the clinical challenge of knowing which of these test-negative patients benefit from urgent coronary angiography and treatment remains. Prevalence (and therefore pretest probability) of CAD seems to be a key factor in how well a stress test performs in diagnosis of CAD. When CAD is less likely (or prevalence is low), the diagnostic value of stress testing is higher than situations where CAD is more likely. Therefore, in high to middle prevalence settings, a stress test is useless because it changes the post-test probability of CAD very little in either direction.
Imagine a 30-year-old man with atypical chest pain and no cardiac risk factors. When CAD is unlikely as in this scenario, a positive stress test is most useful and may actually tilt the probability over the threshold for doing coronary angiography. In practical terms, stress testing is only useful in those situations where clinicians would feel comfortable not performing coronary angiography in the case of a negative stress test. A recent analysis from a contemporary US population showed that patients with a positive result on non-invasive stress testing were only 28% more likely to have CAD on coronary angiography than those who did not undergo any testing (41.0% vs. 35.0%; p < 0.001; adjusted odds ratio, 1.28; 95% CI, 1.19–1.37), highlighting the need for improved risk stratification prior to angiography.
Our findings have implications on how results of diagnostic tests are published, which is particularly relevant to newer imaging modalities such as cardiac PET, cardiac MRI and multi-slice CT. Future publications need to provide raw numbers to allow calculations of sensitivity, specificity and likelihood ratios, and must consider modalities side by side, and in sequence if a test has a negative result. In addition, more prospective cohort studies of diagnostic tests in the setting of CAD are required, considering both diagnostic and prognostic implications side by side.