Putting It All Together: Which Test for Which Patient?

Goals of Testing for the Diagnosis of CAD

While test selection for the diagnosis of CAD has the immediate goal of trying to determine if obstructive CAD accounts for the patient’s symptoms, there are many other potential and salient goals that are both patient specific and system specific ( Fig. 15.1 ). Related short-term patient-centric goals include determining the presence, severity, and extent of CAD, lifestyle and medical treatment optimization, risk stratification, and referral for invasive angiography or possible revascularization, if necessary. The overall long-term goal is to improve clinical outcomes for both individual patients and the overall population. Additional longer-term patient-centric goals include safety, such as reducing radiation exposure to prevent adverse sequelae, while goals for the healthcare system include maximizing efficiency and minimizing cost.

However, selection of patients for testing and selecting initial tests for the diagnosis of CAD are not always straightforward. Major US and European guidelines differ fairly substantially in their basic approaches to determining both the pretest probability of CAD in symptomatic patients and how to proceed with test selection. This may be related to a variation in features at presentation and regional preferences for diagnostic strategies and may be influenced by differences in healthcare systems, access to testing technologies, and risk tolerance. Importantly, limited information on health-related outcomes exists in this stable as-yet-undiagnosed population, and there is little consensus about which test is preferable or even when one is required. The discrepancies between guidelines differ significantly from other areas in cardiology (i.e., therapy for acute coronary syndromes or chronic heart failure), in which general consensus exists largely based on the availability of randomized clinical trial data. To date, current guidelines for imaging stable chest pain of suspected cardiac etiology have not yet incorporated the results of recent, large, randomized trials comparing functional versus anatomic testing strategies.

Complicating the uncertainty are several potential adverse downstream consequences associated with noninvasive tests. These include patient-centered outcomes of false-positive testing such as discomfort during cardiac catheterization, procedural complications, and the effects of radiation, as well as the impact of changes in medical therapy after test result findings and cost. Recent reports of high rates of finding nonobstructive CAD on angiography may speak to the quality of clinical assessment, including the crucial step of patient selection for noninvasive testing. The majority of studies on outpatients with a clinical syndrome of possible ischemia show that, in contrast to past data, at present up to 90% of such tests are normal and approximately 99% of those patients will not experience an untoward clinical event. The risks of false-negative testing include those of a missed diagnosis and failure to treat CAD or risk factors properly. The high incidence of normal noninvasive testing and weak correlations between noninvasive testing results and the presence of obstructive CAD provide further impetus to improve patient selection for noninvasive testing. These issues have important implications for healthcare utilization as well as for the individual patient.

In addition, the fundamental concepts about how we define “significant” CAD have been recently evolving, contributing to the uncertainty of how to best evaluate specific patient populations with chest pain. Recent evidence has found that the association between angiographically defined coronary stenosis and ischemia is variable, as many patients have no ischemia despite the presence of significant stenosis and others may have ischemia with no severe stenosis. Furthermore, the extent to which routine revascularization to treat ischemia reduces death or myocardial infarction (MI), or improves quality of life in patients with stable ischemic heart disease, remains one of the most fundamental uncertainties in cardiology today.

Therefore, decisions regarding noninvasive test use and selection remain common but challenging for many clinicians and a controversial topic for practice guidelines. This chapter reviews important patient characteristics that influence noninvasive test selection for the diagnosis of CAD, including an emphasis on comparing major guideline recommendations and evaluating very recent data, including promising technologies. We review contemporary considerations in cardiovascular imaging, such as how best to evaluate special populations, and appropriate use, including the consequences of noninvasive testing that are sometimes overlooked, such as radiation and cost. Finally, we present a step-by-step practical proposal for a unified approach, incorporating the latest trial evidence, to optimize test selection for both functional and anatomic strategies.

Clinical Classification of Chest Pain and the Pretest Probability of Obstructive CAD

Classically, chest pain symptoms are categorized as typical, atypical, or noncardiac chest, which in combination with age can be used to quantify the pretest probability of underlying coronary disease ( Table 15.1 ). This must be distinguished from other tools such as the Framingham Risk Score or newer atherosclerotic cardiovascular disease (ASCVD) score, which are helpful in evaluating overall risk burden and prognosis using baseline clinical characteristics, but are not designed to evaluate the likelihood of obstructive CAD in the presence of symptoms. While the Coronary CT Angiography Evaluation for Clinical Outcomes: An International Multicenter (CONFIRM) registry risk score provides incremental prognostic information from plaque burden and stenosis found on coronary computed tomographic angiography (CCTA), it was derived in a mixed population of patients (asymptomatic and symptomatic) and requires noninvasive testing by definition, and thus does not aid in the decision to test or in test selection itself.

The Diamond and Forrester algorithms represent the gold standard for prediction of obstructive CAD in symptomatic patients. However, calculation of pretest probability using this score differs depending on which guideline is followed, as each country or region has adopted a different modification. In the United States, current guidelines recommend use of a Diamond and Forrester likelihood score combined with data from the Coronary Artery Surgery Study risk score ( Table 15.2 ). The UK National Institute for Health and Care Excellence (NICE) guidelines advocate calculating CAD likelihood using another modified Diamond and Forrester clinical prediction rule by Pryor et al. This score incorporates the additional high-risk features of diabetes, smoking, hyperlipidemia, and resting ECG changes. More recently, a clinical prediction rule by Genders et al., which aimed to validate, update, and extend the Diamond and Forrester model to a more contemporary population and especially women, had the effect of reclassifying 16% of men and 64% of females. This revised risk score has been incorporated into the European Society of Cardiology (ESC) guidelines (see Table 15.2 ). For example, the pretest probability of obstructive CAD in a 55-year-old female with typical angina differs depending on whether the score used is based on the American College of Cardiology (ACC)/American Heart Association (AHA) guidelines (73%), ESC guidelines (47%), or NICE guidelines (38–92%). A discussion of the impact of diagnostic testing in specific subgroups is discussed subsequently (see “Noninvasive Diagnostic Testing Considerations for Special Populations” ).

Pretest Probabilities and the Degree of Obstructive CAD

While all of these adapted Diamond and Forrester scores are easily implemented at the bedside, mounting evidence demonstrates that they largely overestimate the degree of obstructive CAD. Estimates based on a contemporary CCTA registry, as well as recent clinical trials, have found that the prevalence of obstructive epicardial CAD in patients with typical or atypical angina is much lower than that predicted by Diamond and Forrester in 1979, or any subsequent modification. There are also increasing questions about the need to predict other coronary abnormalities besides severe stenoses. Several studies have found that high rates of nonobstructive CAD are routinely identified during elective coronary angiography, with a significant variation in the rate of nonobstructive CAD between centers. Plaque burden and location each carries incremental prognostic value above traditional obstructive stenosis. Therefore, while we continue to rely on likelihood scores to predict pretest probability of CAD based mainly on age and symptoms alone, improved strategies for likelihood of CAD, risk stratification, and subsequent test selection are warranted and have been proposed or are in development (see “Noninvasive Imaging Integrating Functional and Anatomical Strategies” ). Validation in other populations may encourage future adoption.

Quantifying “Intermediate” Pretest Probability of CAD

Determination of pretest probability also impacts the performance of the available diagnostic methods (the likelihood that this patient has obstructive disease if the test is positive, or does not have disease if the test is negative). Diagnostic testing is most valuable when the pretest probability of ischemic heart disease (IHD) is intermediate, since the application of a test result using Bayesian analysis drives the posttest probability sufficiently lower (negative test) or higher (positive test) to inform future decision-making—usually whether or not the patient should proceed to cardiac catheterization.

Nevertheless, there remains no universal definition of intermediate pretest probability. The definition of 10% to 90%, first advocated in 1980, has been applied in several studies and is the current definition used in the ACC/AHA guidelines for stable IHD ( Table 15.3 ). Low and high pretest probabilities are thus less than 10% and greater than 90%, respectively. This risk stratification scheme is also used by the most recent ACC/AHA Appropriate Use Criteria Task Force. In contrast, the current ESC guidelines using the Genders et al .– modified Diamond and Forrester clinical prediction rule stratifies patients into four groups: less than 15%, 15–65%, 66–85%, and greater than 85%. In comparison to the US guidelines, the intermediate group is defined by a pretest probability of 15% to 85% by combining the two mid-risk groups. Based on these four groups, the ESC guidelines recommend specific test strategies (see later). Finally, the UK NICE guidelines differ compared to both the ACC/AHA and ESC guidelines, identifying an intermediate pretest probability as 30% to 60%.

Is There a Role for Watchful Waiting in Patients with Stable Chest Pain?

Due to low event rates in stable chest pain populations undergoing imaging or ECG testing, and the similar outcomes with medical treatment and coronary revascularization in trials such as Clinical Outcomes Utilizing Revascularization and Aggressive Drug Evaluation (COURAGE), some have recommended that a strategy of deferred testing may be preferable to performing any test at all. In this scenario, a patient with a sufficiently low pretest probability of obstructive CAD as the cause of their stable chest pain would not undergo any initial noninvasive cardiac testing, would be monitored clinically, and would be treated according to primary prevention strategies. This is on a background of the explosive growth in cardiac imaging in the United States, which has become central to discussions surrounding the high cost of healthcare, including a rapid escalation of costs for testing (twice that of other physician services). However, there is as yet no direct clinical trial evidence to support a deferred testing strategy. This is in contrast to the robust evidence now available that supports a testing strategy with either functional testing or CCTA as being safe and effective (see “A New Standard for Pragmatic, Imaging Trials: PROMISE and SCOT-HEART” ).

Another important aspect in the decision regarding whether to defer testing is first considering whether the patient would benefit from revascularization. If the patient has significant comorbidities or their quality of life is not expected to benefit from revascularization, then optimizing medical therapy may be a more reasonable approach than testing.

Approach to Choosing a Functional Test

For functional testing, the choice of stress must first be considered (exercise vs. pharmacologic) and, if exercise is employed, consideration must also be given to whether or not additional imaging should be performed. Several stress imaging modalities currently exist, each with their advantages and disadvantages ( Table 15.4 ). These include radionuclide stress myocardial perfusion imaging (MPI) using single photon emission computed tomography (SPECT) or positron emission tomography (PET), stress echocardiography, and stress cardiac magnetic resonance (CMR). SPECT, PET, and echocardiography may be coupled to either exercise or pharmacologic stress, whereas stress CMR is only performed with pharmacologic stress.

In the absence of contraindications, symptom-limited exercise with an exercise treadmill test (ETT) or bicycle ergometer is the preferred stress-testing modality (over pharmacologic stress) since it provides information concerning reproducibility of symptoms, cardiovascular function, exercise capacity, ECG changes, and the hemodynamic response during usual forms of activity. Exercise capacity alone is a powerful predictor of mortality. Furthermore, a score such as the Duke Treadmill Score when applied to data generated by the ETT can improve diagnostic certainty in addition to its prognostic implications. However, a patient may be unable to exercise due to one or more noncardiac reasons. These include obesity, orthopedic limitations, balance issues, pulmonary limitations, frailty, or limb dysfunction as a result of paraplegia from a prior cerebrovascular event. A detailed discussion on the various forms of exercise modalities (treadmill, or upright or supine bicycle) and protocols (Bruce, Modified Bruce, Naughton) is presented elsewhere (see Chapter 10 ). If absolute contraindications exist, then pharmacologic stress should be used; if relative contraindications exist, pharmacologic stress should be considered.

In addition to considering exercise capacity, several conditions interfere with the ability to make an ECG diagnosis of ischemia (e.g., left bundle branch block, right ventricular pacing, resting ST depression > 1 mm) and may result in an uninterpretable exercise ECG. When such conditions are present, imaging should be used regardless of the stress modality.

If the patient is unable to exercise to sufficient workload, then pharmacologic stress is required. The decision regarding which imaging modality to use will depend on patient factors including suitability of the stress agents for that purpose as well as patient tolerance; ischemic endpoints may vary accordingly. If MPI is used, vasodilators are the preferred pharmacologic stress agents, and perfusion is assessed. If echocardiography is performed, inotropic agents are the most commonly used (although this can vary by country), and wall motion is assessed. For CMR, either inotropes or vasodilators can be used with corresponding endpoints. However, as for exercise testing, contraindications to vasodilator stress agents (adenosine, dipyridamole, and the selective A2A receptor agonists, including regadenoson, binodenoson, and apadenoson) or inotropic agents (typically dobutamine) should be taken into account in selecting the test modality.

If the patient is not a candidate for exercise or pharmacologic stress testing, an anatomic strategy with coronary artery calcium (CAC) scoring or CCTA should be considered. Moreover, based on recently published trial data, an anatomic-first strategy may be a reasonable alternative in selected patients (see “A New Standard for Pragmatic Imaging Trials: PROMISE and SCOT-HEART” ) as discussed later.

Diagnostic Accuracy of Functional Testing Strategies

There are distinct strengths and weaknesses associated with each imaging modality (see Table 15.4 ), and test selection ultimately depends on many factors, including local availability, local expertise, existence and relevance of prior imaging data, cost, the patient’s body habitus (e.g., morbid obesity), radiation exposure, and the need for concomitant assessment of hemodynamic function or valvular disease. Diagnostic performance should be considered when multiple options exist, ideally based on local laboratory performance rather than the literature. Since such detailed data are often not available, a cost-effectiveness meta-analysis by Garber and Solomon may be used. This analysis includes information on diagnostic accuracy of individual tests and is cited by the ACC/AHA guidelines as evidence for differing diagnostic accuracy between modalities (see Table 15.3 ). PET is the most sensitive noninvasive functional test, and exercise ECG is the least sensitive. SPECT is nearly as sensitive as and somewhat less specific than PET (specificity, 0.77 [range in individual studies: 0.53–0.96] for SPECT and 0.82 [0.73–0.88] for PET). Echocardiography is more specific than PET (0.88 [0.80–0.95] compared with 0.82 [0.73–0.88]) but less sensitive (0.76 [0.40–1.00] compared with 0.91 [0.69–1.00]). The Clinical Evaluation of Magnetic Resonance Imaging in Coronary Heart Disease (CE-MARC) study directly and prospectively compared CMR to SPECT. Compared to SPECT, CMR had greater sensitivity (0.87 [95% CI 0.82–0.90] compared with 0.67 [0.60–0.72]) and similar specificity (0.83 [0.80–0.87 for CMR]; 0.83 [0.79–0.86 for SPECT]). CE-MARC2 is a three-arm trial that is ongoing and will provide valuable insights by comparing outcomes following CMR-guided care, positron emission tomography–computed tomography (PET-CT)-guided care (according to ACC/AHA appropriate-use criteria), and NICE guideline-based management.

The ESC guidelines use multiple primary studies to summarize test performance. A major difference between the reference data used by each guideline is the lower sensitivity of the exercise ECG reported in the ESC guidelines—only 50% (despite an excellent specificity of 90%). The marked differences in these values are at least partially, if not fully, explained by the ESC’s use of data obtained from studies avoiding verification bias; ACC/AHA guidelines do not restrict data to studies avoiding verification bias. Because this lower sensitivity means that the number of incorrect test results will become higher than the number of correct test results in populations with a pretest probability of greater than 65%, the ESC does not recommend employing the exercise stress test without imaging in such higher-risk populations for diagnostic purposes. In general, it may be more appropriate to employ more specific testing for patients with a low-intermediate pretest probability of CAD and reserve more sensitive testing for those with high-intermediate pretest probabilities. Therefore the precision of pretest probability estimates to impact the choice the optimal noninvasive test is important even within the intermediate ranges.

Guideline Recommendations for Choosing a Functional Test