Clinical and economic impact of exercise electrocardiography and exercise echocardiography in clinical practice

Abstract

Background Patients with known or suspected coronary disease are often investigated to facilitate risk assessment. We sought to examine the cost-effectiveness of strategies based on exercise echocardiography and exercise electrocardiography.

Methods and results We studied 7656 patients undergoing exercise testing; of whom half underwent exercise echocardiography. Risk was defined with the Duke treadmill score for those undergoing exercise electrocardiography alone, and by the extent of ischaemia by exercise echocardiography. Cox proportional hazards models, risk adjusted for pretest likelihood of coronary artery disease, were used to estimate time to cardiac death or myocardial infarction. Costs (including diagnostic and revascularisation procedures, hospitalisations, and events) were calculated, inflation-corrected to year 2000 using Medicare trust fund rates and discounted at a rate of 5%. A decision model was employed to assess the marginal cost effectiveness (cost/life year saved) of exercise echo compared with exercise electrocardiography. Exercise echocardiography identified more patients as low-risk (51% vs 24%, p<0.001), and fewer as intermediate- (27% vs 51%, p<0.001) and high-risk (22% vs 4%); survival was greater in low- and intermediate-risk and less in high-risk patients. Although initial procedural costs and revascularisation costs (in intermediate-high risk patients) were greater, exercise echocardiography was associated with a greater incremental life expectancy (0.2 years) and a lower use of additional diagnostic procedures when compared with exercise electrocardiography (especially in lower risk patients). Using decision analysis, exercise echocardiography (∈2615/life year saved) was more cost effective than exercise electrocardiography.

Conclusion Exercise echocardiography may enhance cost-effectiveness for the detection and management of at risk patients with known or suspected coronary disease.

1 Introduction

The exercise electrocardiogram (ExECG) is a reliable prognostic test for identification of patients who are either very unlikely or very likely to have cardiac events. Patients at intermediate risk may be further evaluated using a stress-imaging test.^1,2 The current AHA/ACC guidelines recommend use of the ExECG as the first test for evaluation of known or suspected coronary disease when the patient is able to exercise and the ECG is of diagnostic quality,³ although they justify the use of imaging tests in patients with previous revascularisation. However while exercise capacity and ST segment changes are useful prognostically, they are less accurate than alternative competing stress imaging technologies. This difference between the known differential in the accuracy of the investigations and the current guidelines often leads to some insecurity on the part of physicians ordering stress testing protocols.

One of the major arguments against the early performance of a stress imaging test in patients with known or suspected coronary disease would be the greater cost of such a strategy. However, this argument does not take into account differences in downstream costs that may arise from the use of a less accurate initial test, which may prompt the inappropriate use of more expensive investigations, as well as failing to identify patients who go on to potentially costly complications. Moreover, in contrast to radionuclide imaging, exercise echocardiography (ExEcho) represents a smaller cost increment from the ExECG alone. An alternative approach to this controversy would be to perform a cost analysis for each strategy, taking into account the costs of initial investigation, as well as downstream testing and subsequent management decisions such as revascularisation, as well ascomplications.⁴ The purpose of this study was therefore to address whether lower downstream and event costs over the ensuing 5 years compensated for the greater initial expense of using ExEcho in preference to a ExECG for the diagnosis of coronary artery disease.

2 Methods

2.1 Patient populations

Patients with suspected or known coronary artery disease were referred for the evaluation of chest pain symptoms using ExEcho (n=3860) or ExECG (n=3796) for the initial evaluation. This study was observational rather than randomised. Tests were selected by the patient's physician and varied somewhat according to institutional practice. However, to our knowledge, patients were selected appropriately (i.e. no missing data due to nondiagnostic ECG or technically impossible echocardiograms). All patients who completed the echocardiogram had an attempted diagnosis and were analysed. Each patient gave informed consent for the procedure and the study was approved by the relevant institutional review boards.

2.2 Exercise testing

Patients underwent maximum symptom-limited tests, using treadmill protocols chosen in accordance with the age and functional status of the patient. Patients were continuously monitored clinically and with the ECG, and standard endpoints were used.³ The presence and severity of angina, ST segment changes and exercise capacity were used to calculate a Duke treadmill score⁵ in all patients. In patients undergoing a less vigorous stress than the Bruce protocol, the Duke score was calculated for an exercise time on the standard Bruce protocol that was equivalent to the patient's peak exercise workload.

2.3 Exercise echocardiography

Echocardiography was performed at rest using commercially available equipment. Images in the same planes were obtained immediately post-exercise,⁶ and rest and stress images were compared side-by-side, by echocardiographers trained in ExEcho. Interpretations were made independent of clinical, exercise or angiographic data and results were made available to the physicians responsible for the care of the patient. The original interpretations of the tests, made by a number of cardiologists at each institution, was used in preference to a core laboratory analysis, as the intention was to assess the performance of the test in standard practice.

Based on the extent of abnormal wall motion, resting LV function was evaluated as normal, mild, moderate or severely reduced.⁷ Standard criteria were used for the interpretation of ExEcho;^8,9akinesis or dyskinesis at rest were used as markers of infarction, and new or worsening wall motion abnormalities were used to identify ischaemia. Myocardial segments were combined into vascular territories for the purpose of expressing the extent of ischaemia as one, two or three vessel coronary artery disease. The apex, anteroseptal, septal and anterior walls were attributed to the left anterior descending, the lateral to the left circumflex, and the inferior and basal septal to the right coronary. The posterior wall was attributed to the circumflex or right coronary if either was abnormal; in patients with isolated posterior wall abnormalities, these were ascribed to the left circumflex.

2.4 Follow-up

All patients were followed for the occurrence and date of major adverse events including all-cause and ischaemic heart disease deaths, myocardial infarction, and the occurrence of percutaneous coronary intervention or coronary artery bypass surgery. Follow-up data were gathered by clinic review or telephone contact with the patient or physician after 2.5±2.0 years (range 0.5 to 5 years) in the ExECG and after 3.2±2.0 years (range 0.5 to 5 years) in the ExEcho group.

2.5 Statistical analysis

Continuous variables were expressed as mean and standard deviation, and compared by t tests and analysis of variance. Categorical variables were recorded as frequency and percentage, and compared by chi-square analysis. A p value <0.05 was considered statistically significant. A multivariable Cox proportional hazards model, including clinical and exercise test variables, was developed for the estimation of time to cardiac death or myocardial infarction; patients undergoing coronary revascularisation were censored at the time of their procedure. To prevent model overfitting, we included one variable for every ten clinical outcomes. The final multivariable model included all significant estimators of time to cardiac death or infarction with a p<0.20.

2.5.1 Defining pretest clinical risk

Pretest clinical risk was defined using an estimated predicted probability of cardiac death or myocardial infarction, derived from a Cox proportional hazards model that included age, gender, diabetes, angina class, cigarette smoking, hypertension, diabetes, hyperlipidaemia, prior revascularisation and previous myocardial infarction.^10–12 Based upon their predicted annualised risk of death or myocardial infarction (hin.nhlbi.nih.gov/atpiii/calculator.asp?usertype=prof), patients were classified as being at low (<0.6%), intermediate (0.6–2%), and high (>2%) risk. After comparing the final multivariable model in each group for similarities in event classification (i.e. C-index), we compared clinical characteristics and pretest risk groups for the ExEcho and ExECG patient groups (Table 1).

A risk-adjusted Cox proportional hazards model (controlling for pretest clinical risk using the previously described score including risk factors, symptoms, and prior coronary disease history) was calculated for the (1) ExEcho extent of ischaemia (0, 1, 2, and 3 vascular territories) and (2) low, intermediate, and high risk Duke treadmill score. A stratified Cox model was employed to assess the relative changes in event-free survival for intermediate to high risk patients undergoing coronary revascularisation.

2.5.2 Propensity score

For each of the cost and outcome risk-adjusted analysis, a propensity score was developed to control for referral bias to coronary angiography and revascularisation procedures as well as other variations in practice and referral pattern across sites.¹² The propensity score was developed based upon multivariable logistic regression predictors of referral to coronary angiography or revascularisation. The propensity score was added to each multivariable cost and clinical outcome models as a covariate.

2.6 Cost analysis

Detailed resource utilisation was obtained through epidemiologic tracking of clinical outcomes and major cardiovascular procedure use. Standard cost estimations were calculated using a median of charges (adjusted by a national median cost-charge ratio (http://www.hcfa.gov/stats/pufiles.htm) as well as supplemented by available published cost data.^12–17 Diagnostic costs were accumulated through the first 90 days after the initial testing date; total diagnostic cost included the cost of initial testing (either exercise treadmill testing or echocardiography) as well as the use of a diagnostic angiography. Follow-up costs included the use of coronary revascularisation procedures and cardiac-related hospitalisations. Societal economic cost for cardiac death was based upon cost estimates (http://www.americanheart.org/statistics/economic.html). Total costs were calculated as the sum of diagnostic and follow-up costs. All costs were inflation-corrected to year 2000 and discounted at a rate of 5%. The original cost data were obtained in US$ and expressed in ∈ at an exchange rate of 1.00.

Comparisons of cost data by treadmill or echocardiography risk groupings were performed using univariate and multivariate analysis of variance or general linear model techniques, controlling for the pretest clinical risk index as described above. For the comparison of costs by pretest clinical risk groups, a general linear model using weighted least squares regression was performed, controlling for the extent of ischaemia or Duke treadmill score. Linear regression techniques were also used in order to calculate predicted costs controlling for pretest clinical risk covariates and for comparison of important patient subsets (e.g. prior history of coronary disease).

2.6.1 Cost effectiveness decision analytic model

A decision analytic model to determine clinical outcome and economic data of ExEcho and electrocardiography using Treeage (version 2.6, Williamston, MA) and Answer Tree (version 3.1, SPSS Inc., Chicago, IL) software. For this analysis, marginal cost effectiveness was defined as cost per life year saved. The calculation of incremental cost effectiveness was defined as ((Total cost for ExEcho)–(Total cost for ExECG))/((Total life years for ExEcho)–(Total life years for ExECG)). For this analysis, costs and clinical outcomes were discounted over the lifetime of each patient. Lifetime costs were derived using a future value estimate based upon observed 3-year costs and predicting lifetime costs of care by published estimates of utilisation and event rates (http://www.aihw.gov.au/publications/health/hsccdda93-4, http://www.hta.nhsweb.nhs.uk).¹⁸ These estimates were compared to results from a Markov model for lifetime costs. Gender-specific life expectancy estimates were derived from published estimates from the National Center for Health Statistics (www.cdc.gov/nchs/fastats/lifeexpec.htm).Estimates of life expectancy were corrected for patients with observed death rates by calculating life years remaining as age at testing+time to follow-up before death.

3 Results

3.1 Clinical characteristics (Table 1)

The groups were comparable apart from a slightly lower average age in those undergoing ExEcho (61 vs 63 years, p<0.001). Approximately 40% of the study cohort were women. Cardiac risk factors were prevalent-nearly half of the patients were hypertensive, 17–18% were diabetic, and 28% were current smokers. A prior history of coronary disease was noted in 25% and 21% of those initially referred for ExEcho and ExECG (p=0.42).

Predicted annual rates of cardiac death or myocardial infarction were generated from a Cox model for both groups (hin.nhlbi.nih.gov/atpiii/calculator.asp?usertype=prof). The groups showed a similar prevalence of low risk (11% vs 12%), intermediate risk (58% vs 60%) and high-risk (30% vs 28%). Of the high risk group, 72% had a history of coronary disease; a rate similar to that for both echocardiography and electrocardiography (p>0.20).

Exercise test findings were similar between the groups. Approximately 40% showed ST depression >̄1.00mm, about 35% had exertional chest pain (limiting or nonlimiting) and most patients completed stage II of the modified Bruce protocol.

3.2 Characteristics of risk groups

Results of stress testing in clinically-defined low-, intermediate- and high-risk groups are summarised in Tables 2 and 3. Low pretest risk patients had better functional capacity than intermediate or high-risk patients, and there was a similargradation in the frequency of chest pain and ST depression.

In the ExEcho group (Table 2), resting left ventricular function was more often normal in low risk as compared to other patients (p<0.0001), but severely depressed left ventricular function was uncommon. The prevalence and extent of inducible ischaemia increased proportionally in higher risk patient subsets (p<0.0001). In those undergoing ExECG (Table 3), higher risk Duke treadmill scores increased proportionate to the increasing clinical risk of the patient (p<0.0001).

3.3 Procedure utilisation rates

The management decisions and outcomes of patients based on their post-test assessment of ischaemic risk are summarised in Table 4. In a risk-adjusted comparison for cardiac catheterisation rates (adjusting for pretest clinical risk and including the propensity score), 12% of patients without exertional ischaemia on ExEcho had angiography performed during follow-up. This rate of angiography increased to 51% of patients with 2–3 vascular territories with ischaemia (p<0.0001). Interestingly, the same linear increase in angiography rates were not observed for those individuals undergoing initial testing with ExECG. Of those patients with a low risk Duke treadmill score, 45% of individuals were referred to angiography during follow-up. The vast majority of low risk treadmill patients who underwent coronary angiography had a prior history of coronary disease; with catheterisation more often performed >90 days after testing. Consequently, coronary revascularisation occurred in 35% of patients with established coronary disease who did not have inducible ischaemia during ExECG.

There was a proportional relationship between rates of revascularisation and the extent of ischaemia for both ExEcho and ExECG (p<0.0001). However, revascularisation with increasing levels ofrisk was applied differently in both study groups; procedures being more frequently performed in coronary disease patients without inducible ischaemia during ExECG and more frequently performed in high risk patients undergoing ExEcho (p<0.0001). This observation applied equally to percutaneous coronary intervention (p<0.0001) and coronary artery bypass surgery (p=0.001).

The frequency of cardiac death or myocardial infarction infarction similarly increased with both higher risk ExECG (p=0.001) and ExEcho (p<0.0001). Fig. 1a and b illustrate the risk-adjusted cumulative event-free survival in patients undergoing ExEcho and ExECG. Cumulative survival was 94%, 88%, 83%, and 76% for patients with no, 1-vessel, 2-vessel, and 3-vessels with inducible ischaemia on ExEcho (χ²=74, p<0.0001, Fig. 1a). Cumulative survival was 93%, 91%, and 89% for patients with low risk, intermediate risk, and high risk Duke treadmill score during ExECG (χ²=15, p=0.002, Fig. 1b).

3.4 Cost effectiveness analysis

The results of a decision analytic cost effectiveness model are reported in Table 5. Average life expectancy was 23.54 years in the group undergoing ExEcho and 23.52 years in those having ExECG—an average difference of 7.3 additional days of life in patients undergoing ExEcho. Conversion of life expectancy to total life years for each group yielded 89459 life years for ExEcho and 88464 life years for ExECG (p<0.0001).

Total short-term (3-year) costs were ∈17419657 and ∈16842611 for ExEcho and ExECG (p<0.0001), a greater lifetime cost of ∈2603141 for echocardiography. The total higher costs for ExEcho were noted for patients with intermediate-high post-test risk, but despite the higher cost, the clinical effectiveness measure (i.e. life years saved) was greater for echocardiography, yielding an enhanced cost effectiveness ratio of ∈2615 dollars per life year saved for ExEcho as compared with ExECG.

Table 6summarises risk-adjusted predicted total costs of care from a multivariable linear regression model that accounts for the influence of history of coronary disease (Table 7). In patients without a prior history of coronary disease, the overall costs of care were substantially higher for ExECG (test for interaction p<0.0001). In particular, predicted costs were 64% higher for ExECG patients without inducible ischaemia and substantially higher for those with a history of coronary disease. Despite the higher cost, average gain in life expectancy was similar between patients with a history of established coronary disease (only 2.9 days); yielding a more cost efficient strategy with echocardiography especially in the absence of provoked ischaemia. Additionally, predicted costs were also highest for those high-risk ExEcho patients. The 40% higher costs for ExEcho were offset by gains in life years averaging 1.02 years. These results yielded total cost effectiveness ratios of ∈3903 per life year saved for ExEcho in patients with establishedcoronary disease.

4 Discussion

In this study of clinical decision-making in contemporaneous groups of patients, matched for clinical markers of risk, ExEcho substratified levels of risk somewhat more effectively than the Duke treadmill score, leading to an anticipated improvement of life expectancy. A major difference between the techniques pertained to the likelihood of a patient undergoing angiography in the follow-up period, which was greater in patients identified by ECG criteria as being at lower risk (even after exclusion of those with pre-existing coronary disease), and greater in patients identified by ExEcho criteria as being at higher risk. Consequently, despite the lower cost of the ExECG study, this strategy was only slightly less expensive than ExEcho, because of more frequent performance of interventions, reflecting the higher rate of angiography. When the outcome implications were combined with the cost data, the lesser cost of the ExECG approach was outweighed by the survival benefit conferred by more reliable sub-stratification of risk with ExEcho, which was the more cost-effective technique.

4.1 Selection of patients for angiography

The standard approach to stress testing, used in previous decision models,^14,19–21 is based on the concept that the tests are used to identify coronary disease, so that all positive tests require referral to coronary angiography. In a patient having ongoing angina despite maximal therapy, this approach (or direct angiography-depending on the symptoms), is appropriate because the angiogram is a prelude to procedural intervention. However, most patients with known or suspected coronary disease do not complain of intractable angina, and the reason for referral to angiography is to confirm the diagnosis. While this activity may identify a minority of individuals at high risk, the likelihood of revascularisation (especially percutaneous intervention) is much greater once the coronary anatomy isdefined,¹² without necessarily changing outcome.

The angiographic evaluation of risk is based upon the definition of prognostic groups on anatomic grounds in the initial studies of medical and surgical therapy.^22,23 The prognostic power of functional testing is at least equivalent to that of angiography.^24,25 In the CASS trial, follow-up of exercise testing results²⁶ showed that 12-year survival was enhanced by coronary artery bypass surgery compared to medical therapy in high-risk men, but that outcomes were similar in the intermediate- and low-risk subgroups. Indeed, prognostic power incremental to clinical evaluation and resting evaluation of LV function has been shown with radionuclide ventriculography,²⁷ radionuclide perfusion imaging^1,28,29 ExEcho.^2,25,30

Thus, the former reliance on risk stratification by angiography is changing, and the results of stress imaging tests appear to be increasingly incorporated into decision-making. Most importantly,studies have indicated that mildly abnormal results are prognostically benign,³¹ and recent evidence suggests that clinicians are not performing angiography on these individuals,^29,32 The angiography and revascularisation rates with ExEcho in this study conform to the experience with myocardial perfusion imaging,²⁹ showing referral rates significantly increase with worsening scan results. In contrast, the responses to exercise ECG shown in this study are discordant to these findings, with little difference in intervention with increasing risk. We believe that this reflects the preponderant use of the ST segment results, and therefore a binary response to ‘positive' or ‘negative' studies.

4.2 Cost-effectiveness

In addition to a significant evidence base of direct comparisons, several recent meta-analyses have confirmed the superiority of stress imaging to the stress ECG for the diagnosis of CAD.^33,34 However, this improvement in accuracy is attended with greater cost, and there is a need to scrutinise the cost implications of these alternative investigation strategies.

The simplest approaches to cost analyses of diagnostic testing in chronic coronary disease involved Bayesian analysis of different clinical algorithms to evaluate the costs for the diagnosis of disease. In these models, effectiveness was based upon the diagnosis of angiographic evidence of CAD, and various assumptions were made regarding the prognostic implications of correct and incorrect diagnosis of disease, so that utility (number of quality-adjusted life years) could be projected.¹⁹ These models showed that as the likelihood of CAD increased, cost increased in a linear pattern, but testing became more cost-effective because the likelihood of impacting outcome increased. A similar study, based on a meta-analysis of published accuracy suggested that in patients at intermediate pretest probability of disease, ExEcho or SPECT showed an incremental cost-effectiveness ratio of ∈41900 to ∈54800 per QALY saved compared with ExECG.¹⁴ The problems with these models are that they primarily assess the diagnostic strength of the tests, and better outcomes (rather than greater accuracy) would be considered to be a stronger justification for a more expensive test.

Several recent studies have compared the costs of various approaches for the assessment of coronary disease using an outcomes-based cost analysis, that differs from the cost per life year saved definition often used in therapeutic comparative trials. In the first of these studies, myocardial perfusion imaging provided incremental prognostic value in all patient subgroups, but was cost-effective only in patients with an intermediate to high post-exercise test likelihood of coronary artery disease.²⁸ The END study¹² found that the outcomes of 11372 consecutive stable angina patients, who underwent stress myocardial perfusion tomography or cardiac catheterisation, were the same. However, in all clinical risk subsets, composite costs of care were found to be higher for direct cardiac catheterisation than with stress myocardial perfusion imaging and selective catheterisation, a finding that reflected higher rates of revascularisation in the direct catheterisation group. Finally, the EMPIRE study³⁵ compared the cost-effectiveness of four diagnostic strategies (exercise ECG/coronary angiography, exercise ECG/myocardial perfusion imaging/coronary angiography, myocardial perfusion imaging/coronary angiography and coronary angiography alone) in 396 newly presenting patients with possible coronary artery disease. The 2-year patient outcome was the same, irrespective of strategy, but the use of imaging was associated with a lower total cost because of the performance of less angiography and revascularisation. These results are all concordant with the current study, suggesting that the performance of angiography is linked to intervention, but this is not necessarily translated into a different outcome.

4.3 Limitations

This study is based on the test results, physician responses and costs experienced at a large Cardiology department at a large cardiac referral centre in the USA. To the extent that these may vary from practice elsewhere, the generalisability of the results of this study may be questioned. While costings vary across the world, it is likely that similar ratios between tests, treatments and the cost of adverse events would hold in other environments. In this respect, while the monetary implications may differ, the conclusions are likely to be similar in different environments. Moreover, the results pertain to expert interpretation of ExEcho, and the difference between the tests in circumstances of less skilled application is likely to be less.

The outcomes aspect of the study focused on cardiac events. Different management strategies may be associated with differences in quality of life, which could perhaps justify differences in cost. However, we did not track quality of life and therefore cannot address these issues.

5 Conclusions

While these results will not alter the management of patients who require intervention because of significant ongoing angina, they suggest that patients with symptoms controlled by medical therapy are less likely to be referred for angiography (and therefore intervention) if an exercise echocardiogram is performed in preference to the exercise ECG. Thus, ExE has been shown in this study to offer a more cost-efficient approach to the evaluation of patients with known or suspected coronary disease.

Supported in part by the American Society of Echocardiography and the National Heart Foundation of Australia.

References