1. Introduction

The exercise test has a long history in cardiovascular medicine. Although Willem Einthoven [Einthoven, 1908] was the first to document ST changes in the ECG with exercise, it was 1932 when Goldhammer and Scherf [Goldhammer and Scherf, 1932] proposed exercise electrocardiography as a diagnostic tool for angina. Since then, exercising patients to stress the cardiovascular system has played a central role in the diagnostic work up of coronary artery disease. However, because of a low reported sensitivity [Froelicher et al., 1998], cardiologists have turned away from the simple exercise test and have added nuclear and echocardiography to stress. This has made the generalist feel that the test is inadequate without these expensive modalities. The future of the exercise test clearly includes these new modalities but in order for it's basic utility to continue in the generalist's office we feel that diagnostic and prognostic scores must be included in all reports and that probability estimates should be used to recommend patient management. Optimisation of the test can produce predictive values equal to that of the best more expensive techniques, and the test should be more widely used. The yield of prognostic and diagnostic information from the test argues for a central position in the work up of cardiac patients. We will discuss some of the concepts that can optimise the exercise test.

2. Optimisation

2.1. Protocol

Experience and history demonstrates that the pervasiveness of a concept, method, or product relies only in part on its intrinsic quality. Rather more, the acceptance of a concept relies on its extrinsic survival potential, a complex attribute with temporal, logistical and experiential dimensions [Dawkins, 1976; Still, 1986; Lynch, 1996]. In applied exercise testing, these features are apparent in relation to the choice of treadmill protocol. When treadmill and cycle ergometer testing was first introduced into clinical practice, practitioners adopted protocols used by major researchers such as Balke [Balke and Ware, 1959; Åstrand and Rodahl, 1986; Bruce, 1971; Ellestad, 1996]. However, over time, the Bruce Protocol came to predominate and helped to expand the use of the exercise test. In fact, a survey of 71 cardiology divisions within the US Veterans Health Care system revealed that the Bruce protocol was used by 82 % [Myers et al., 2000], a number similar to that found by others [Stuart and Ellestad, 1980; Muir et al., 1993]. This use is in contrast with recommendations [American…, 2000; Gibbons et al., 1997] that advocate gradual, individualised protocols. The reasons are clear: work rate increments which are too large or rapid result in lower sensitivity for detecting coronary disease [Myers et al., 1991; Panza et al., 1991; Okin et al., 1989], a less reliable test for studying the effects of therapy [Tamesis et al., 1993; Webster and Sharpe, 1989], and a tendency to overestimate exercise capacity [Myers et al., 1991; Tamesis et al., 1993]. The advantage of the experience and data acquired with the Bruce protocol is outweighed by the large number of patients unable to complete even stage 1 (approximately 5 METS), the difficulty in interpreting submaximal gas exchange measurements with large and unequal increments in work, and the poorer estimates of exercise capacity achieved compared with other protocols. In addition, since many clinical laboratories do not perform gas analysis routinely, a clear relationship between those variables that are measured and oxygen uptake is desirable.

An alternative to the Bruce treadmill protocol, first proposed in 1991 [Myers et al., 1991] and more recently appreciated [Myers and Bellin, 1999; Myers, 1998] is ramp testing (advocated for the cycle ergometer in 1981 [Whipp et al., 1981]). This approach, aided by a pre-test activity questionnaire, aims to bring the patient from rest to maximal exercise through a linear increase in work, over approximately 10 minutes. The rationale for this length of test has been well validated. Studies suggest that tests individualized to last 10 minutes produce the highest values for oxygen uptake [Buchfuhrer et al., 1983], better differentiation of treatment effects in clinical trials [Webster and Sharpe, 1989; Redwood et al., 1971], and a closer relation of work rate to oxygen uptake [Tamesis et al., 1993; Myers et al., 1992]. In addition, a focus on total work performed rather than on exercise-time (a highly variable measure) is facilitated by this approach.

Despite these potential advantages there have been few direct comparisons of the ramp protocol with others [Myers et al., 1991; Kaminsky and Whaley, 1998; Bhadha et al., 1995; Bader et al., 1999; McInnis et al., 1999]. Of these, three studies used cardiovascular patients [Myers et al., 1991; Bader et al., 1999; Will and Walter, 1999], while two studied healthy women [Bhadha et al., 1995] and obese women [McInnis et al., 1999]. Two used a ramped Bruce protocol (ramping between the stages of the classical Bruce) [Kaminsky and Whaley, 1998; Will and Walter, 1999] while the others used individualised protocols. Taken overall, these studies suggest that a ramped protocol is preferred and better tolerated by patients [Myers et al., 1991; Will and Walter, 1999] produces an oxygen uptake-to-work ratio closer to unity (in some patients [Myers et al., 1991] but not those over 60 years), and may result in higher values for metabolic equivalents [Will and Walter, 1999] and exercise duration [Bader et al., 1999; Will and Walter, 1999]. Although this last finding was reported in only one study [Will and Walter, 1999], such an observation has relevance for the predictive characteristics of the test. The higher workloads achieved by the same patients on the treadmill compared to the cycle ergometer are associated with improved exercise test sensitivity for coronary artery disease [Hambrecht et al., 1992]. It makes intuitive sense that, given two tests in the same patient, the one capable of eliciting a higher oxygen uptake would represent a truer examination of cardiopulmonary function (rather than a reflection of local muscle fatigue).

2.2. Diagnosis

The primary use of exercise testing in clinical medicine is in the diagnosis of coronary artery disease. Meta-analysis of studies has shown a specificity of around 80% and a sensitivity of around 70% for obstructive coronary disease by angiography [Froelicher and Myers, 2000; Gianrossi et al., 1989]. However, many of these studies suffered methodological problems of limited challenge and work up bias. Only considering consecutive patients presenting for evaluation of chest pain can avoid both problems. In the only trial to date to clearly avoid work up bias [Froelicher et al., 1998] values of 85% and 45% were found for specificity and sensitivity respectively. Future studies should avoid these problems.

Computers have enabled the use of multivariate prediction equations that can, by incorporating patient data with the exercise test, improve on this predictive accuracy. Such equations demonstrate similar accuracy to much more expensive tests. Although such equations may seem esoteric and even intimidating, simple scores do as well and automated programs are readily available on desktop computers and the internet to carry out the calculations [Froelicher, 1999].

One problem with such equations is their lack of portability to populations other than those in which they were developed. One way of overcoming this is to use a consensus approach, where a final classification of low or high risk is made depending on the consensus among equations validated in different populations [Do et al., 1997]. Such a method has recently been shown to predict angiographic disease better than cardiologists with a special interest in this field [Lipinski et al., 1999]. Consensus and the use of prognostic scores (such as that from Duke [Mark et al., 1991]) to diagnose CAD has other advantages. Rather than a binary yes/no approach to stratification, separating patients into low, intermediate and high risk groups suggests a management course. To explain, low risk patients would need no further testing at that time, high risk patients would need an invasive study, and intermediate risk patients would require another non-invasive study. Assuming the intermediate group is eventually diagnosed, current data suggests that this test strategy would produce a sensitivity and specificity of around 90%.

Another approach to improving test characteristics is to investigate different ECG criteria. Atwood and colleagues took 100 computed measurements from digitised exercise ECG recordings and related them to angiographic data on the same patients [Atwood et al., 1998]. They found computerised measurements at 3.5 minutes of recovery, from lead V5, 60 milliseconds after the QRS complex (ST60), to be superior to all other single measurements. Further, prediction equations that included clinical and exercise test data exhibited the greatest diagnostic power. In a new approach, Michaelides examined 245 patients who underwent exercise testing with standard 12 leads, right ventricular leads, and thallium-201 scintigraphy. They found sensitivities of 66%, 92% and 93%, and specificities of 88%, 88% and 82% respectively for the detection of "any" coronary artery disease by angiography [Michaelides et al., 1999]. The QRS complex has also been a focus of investigation as a marker of myocardial ischemia consequent on its use in scoring systems to predict infarct size. The Athens score [Michaelides et al., 1995] is calculated from exercise induced changes in the QRS complex and more recently, a Dutch group [vanCampen et al., 1996] have proposed a score based on composite changes in the Q,R and S waves in two leads (aVF, V5). These authors reported a sensitivity of 88% and a specificity of 85% for coronary artery disease, in comparison with values for ST depression of 54.9% and 83% respectively. These findings need to be validated in other populations.

An important but under-appreciated caveat to the usefulness of exercise testing is the fundamental incapacity of ST depression to localise ischemia. Although suggested by the recurrent finding of V5 as the lead of maximum differentiation regardless of coronary anatomy, several direct investigations have corroborated the result [Mark et al., 1987; Abouantoun et al., 1984; Ikeda et al., 1985]. The electrophysiological explanation for this apparent anomaly has only recently come to light [Li et al., 1998].

3. Novel Indications

3.1. Expired Gases and Heart Failure

Although the most common use of the exercise test in clinical cardiology is in the diagnosis of coronary artery disease, several other indications have recently been reported. The rise of exercise training as a treatment for heart failure [Coats, 1999] has led to renewed interest in the assessment of patients with reduced ventricular function. Most of the work in this area has focused on pre-transplant patients [Myers and Gullestad, 2000; Beniaminovitz and Mancini, 1999] and, in particular, the ability of ventilatory and gas exchange measures to stratify risk and predict outcome. The result of one study [Mancini et al., 1991] fostered the concept of a cutpoint value for peak VO2 (14ml/kg/min) with patients who attained values above this point displaying 1 and 2-year survival rates similar to those in transplanted patients, while those below this value had survival rates significantly worse. Although appealing to clinicians, the concept of a cutpoint has been questioned recently. One study found that ventilatory variables (VE/VCO2) and a chronotropic index (see below) fared better than peak VO2 in predicting death, while another [Myers et al., 2000] found similar discriminatory power for each of 7 cutpoints between 10ml/kg/min and 17ml/kg/min. In fact, an earlier publication from the same group [Myers et al., 1998] showed that peak VO2 outperformed clinical variables, right heart catheterization data, exercise time and other exercise test data in predicting outcome.

As a result of these seminal studies, major bodies have recommended the inclusion of gas analysis in exercise tests carried out i) to evaluate transplant patients, and ii) to differentiate cardiac exercise intolerance or dyspnea from pulmonary causes [Gibbons et al., 1997]. Recent data has suggested however, that these techniques are underused. One survey found that less than 3% of exercise tests carried out in the Veterans Health Care system in the United States included gas exchange measurements despite more than 15% of the patients meeting the class I criterion (Table 1).

The promising data from randomised controlled trials of exercise training in heart failure [Belardinelli et al., 1999; Wielenga et al., 1999] suggest a IIb indication for the use of exercise testing with gas analysis in the cardiac patient. The increasing prevalence of heart failure demands an increasing role for this method of assessment for diagnosis, prognosis, or evaluation of therapy indications.

TABLE 1. Indications for the use of gas exchange measurements
in exercise testing (from [Gibbons et al., 1997]

Class I		Class I
Conditions for which there is evidence and/or general agreement that a given procedure or treatment is useful and effective.		1. Evaluation of exercise capacity and response to therapy in patients with heart failure who are being considered for heart transplantation.
		2. Assistance in the differentiation of cardiac versus pulmonary limitations as a cause of exercise-induced dyspnea or impaired exercise capacity when the cause is uncertain.
Class II	IIa	Class IIa
	Weight of evidence/opinion is in favor of usefulness/efficacy.	1. Evaluation of exercise capacity when indicated for medical reasons in patients in whom subjective assessment of maximal exercise is unreliable.
Conditions for which there is conflicting evidence and/or a divergence of opinion about the usefulness/efficacy of a procedure or treatment.
	IIb	Class IIb
	Usefulness/efficacy is less well established by evidence/opinion.	1. Evaluation of the patient's response to specific therapeutic interventions in which improvement of exercise tolerance is an important goal or end point.
		2. Determination of the intensity for exercise training as part of comprehensive cardiac rehabilitation.
Class III		Class III
Conditions for which there is evidence and/or general agreement that the procedure/treatment is not useful/effective and in some cases may be harmful.		1. Routine use to evaluate exercise capacity.

3.2. Heart Rate

Exercise chronotropic incompetence or heart rate impairment has been known for some time to predict all-cause mortality in healthy populations [Ellestad and Wan, 1975; Morris et al., 1991; Lauer et al., 1996]. More recently, when combined with a measure of exercise capacity as the "chronotropic index" [Wilkoff and Miller, 1992] (see Glossary) it has been shown to be predictive of all-cause mortality independent of thallium ischemia [Lauer et al., 1999]. In fact, the risk associated with relative bradycardia was equal to that of a perfusion defect, and the effects were additive. The mechanism for these negative associations is not clear. Control of exercise heart rate involves both sympathetic and parasympathetic input. It may be that some form of autonomic dysfunction, which does not necessarily require localised perfusion defects, accompanies the chronic sympathetic overactivation of heart failure. Certainly, we know that beta receptors are down regulated [Colucci et al., 1989] and that this can be reversed by ACE inhibition [Kawai et al., 1999]. Regardless of the mechanism, the finding demands attention to maximal heart rates. Ironically, the tests of many patients who fail to reach their age predicted maximal heart rate are labelled "non-diagnostic". This, together with the finding that including all patients regardless of maximal heart rate, actually improves test characteristics for the diagnosis of coronary artery disease [Gauri et al., 1999], argues strongly for a role for this simple measurement in routine exercise testing.

The heart rate fall after exercise has also been suggested as an important prognostic marker (HR "recovery" [Imai et al., 1994]). A recent study found that a delayed decrease in the heart rate during the first minute after graded exercise was a powerful and independent predictor of all-cause mortality in 2428 patients followed for six years [Cole et al., 1999]. One potential difficulty with the measurement of HR recovery as suggested by these authors, however, is the use of a cool down period which, although generally advised, has been shown to negatively impact the sensitivity of the test.

3.3. Hypertension

The significance of an exaggerated blood pressure rise to exercise has been recognised [Dlin et al., 1983] and debated [Bassett et al., 1998] for over 15 years. Only recently however has there been a clear consensus that such a rise represents a risk factor for new onset hypertension. The Framingham Offspring study [Singh et al., 1999] showed that an exaggerated diastolic BP rise was associated with a 2 to 4 fold risk of new onset hypertension. An elevated recovery SBP was also predictive of hypertension in men alone. Also, Kjeldsen and colleagues demonstrated that systolic blood pressure during a bicycle ergometer exercise test was a stronger predictor of total cardiovascular mortality, and of morbidity and mortality from myocardial infarction, than the resting blood pressure of the same participants [Kjeldsen et al., 1997]. These findings argue for a role for the exercise test in the work up and risk stratification of hypertensive patients.

3.4. Exercise Capacity

The measurement of exercise capacity, made more difficult by the prevalence of non-continuous exercise protocols, is often lost in the focus on ST segment analysis. In fact, exercise capacity turns out to be a strong, independent risk factor for all cause and cardiovascular mortality [Morris et al., 1991; Vanhees et al., 1994; Snader et al., 1997; Blair et al., 1995] and one, moreover, that can be altered by training. Blair and colleagues showed that for every minute of increase in maximal treadmill time, there was a corresponding 7.9% decrease in the risk of mortality [Blair et al., 1995]. Another study that examined a variety of factors including SPECT thallium perfusion scanning, found that the strongest predictor of all cause mortality was estimated fair or poor functional capacity (adjusted RR 3.96) [Snader et al., 1997]. These analyses demonstrate exercise capacity to be a remarkably powerful prognostic indicator, and, in comparison with the measures discussed above in relation to heart rate, one that implies a mechanism for its own amelioration. Goldstein and Holmboe have emphasized the importance of the potential to modify poor prognostic indicators, if we are to argue for measuring them [Goldstein and Holmboe, 1999].

4. Conclusion

The generalist is increasingly called upon to begin the work-up of the cardiac patient. Escalating pressure on specialist time, relentless sub-specialization of the field, and the mounting expense of interventional procedures means that potential enhancement of non-invasive tests should be welcomed [Marcus et al., 1995]. Against this background, the recent developments summarized in this paper emphasize the critical role that the exercise test can play today. The addition of gas analysis for heart failure patients can add important prognostic information. The use of multi-variable scores can enhance the tests diagnostic capabilities in patients presenting with chest pain. The exercise test is inexpensive, brief, requires minimal equipment and space, and can be carried out safely [Franklin et al., 1997; Gibbons et at., 1989].

Glossary

MET	Metabolic unit. 1MET equates to resting metabolic rate (typically taken to be 3.5ml/kg/min)
Work up bias	A biostatistical error caused when study participants are chosen for the gold standard test on the basis of the test in question. To avoid this, participants need to agree to undergo both tests regardless of the outcome of the first.
Chronotropic index	HR reserve divided by metabolic reserve
HR reserve	(HRstage - HRrest / HRpeak - HRrest) x 100
Metabolic reserve	(METstage - METrest / METpeak - METrest) x 100

References