Introduction

The need for PROMISE

The development of coronary artery disease (CAD) is a major, final common pathway in heart disease worldwide [1]. In the US, 50 million stress tests are performed each year and management of patients with suspected CAD approaches $ 14 billion per year, making it one of the most common and costly clinical scenarios. Functional testing including exercise and pharmacological testing with or without imaging has been the diagnostic cornerstone in the management of these patients for 40 years. The detection of myocardial ischemia in the appropriate clinical context typically triggers a referral to invasive coronary angiography and – if a significant stenosis is detected – to coronary revascularization. Several studies, among them the COURAGE trial [2], suggest that the success of an up-front revascularization strategy is rivaled by optimal medical therapy, with provisional revascularization for severe symptoms, heart failure, or electrical instability.

In this context, coronary computed tomography angiography (CCTA) has emerged as a noninvasive alternative to functional testing as a first-line test for CAD detection but is complimentary in its nature. The benefits of CCTA include the unique ability to visualize the presence and extent of CAD as well as morphology and composition similar to intravascular ultrasound (IVUS), noninvasively in a quick, robust, well-tolerated examination. A wealth of single and multicenter studies have demonstrated the impressive diagnostic performance of CCTA for the detection of significant luminal stenosis in selected patient populations ranging from low to high clinical risk [3–5]. A consistent hallmark of these studies has been the high negative predictive value (nearly 100 %) and more modest positive predictive value (closer to 80 %) compared with invasive coronary angiography. In addition, the presence and extent of CAD by CCTA are strongly related to prognosis, with normal CCTA associated with near-zero event rates over short-term follow-up [6]. Finally, observational data suggest that over 30–40 % of patients undergoing CCTA demonstrate nonobstructive CAD, with an accompanying increased risk for cardiovascular events [7].

Differences in clinical opinion favoring functional imaging of myocardial ischemia versus CCTA testing have culminated in divergent professional society guideline recommendations: While the American Heart Association guidelines on stable chest pain largely recommend functional evaluation of myocardial ischemia (Class IA recommendation) with provisional CCTA only for inconclusive stress tests or those unable to exercise [1], the National Institute for Health and Care Excellence (NICE) guidelines recommend a coronary calcium score for risk stratification in symptomatic patients with a pretest probability between 10–29 % [8]. Moreover, European Society of Cardiology guidelines recommend CCTA as an alternative to stress testing in patients with intermediate pretest risk, those with inconclusive tests, or patients with contraindications to stress testing [9].

Of note, while randomized studies in the acute coronary syndrome setting have demonstrated a potential role for CCTA to identify low-to-intermediate risk patients safe for discharge, they report increased rates of invasive coronary angiography that may be secondary to limited positive predictive value and potential increased prevalence of obstructive CAD (as compared with positive functional studies) [10]. The divergent professional society recommendations suggest equipoise in functional and anatomic cardiac risk assessment. Does the use of CCTA earlier in the evaluation of patients with CAD improve outcomes over functional testing? Is the potential for increased downstream angiography and/or revascularization a cost-effective and safe approach for initial risk stratification? How does radiation and contrast exposure figure into the risk calculation? The PROMISE trial was designed to address these specific questions in a practical clinical study.

The PROMISE trial

Design and methods

PROMISE was a randomized comparative effectiveness study of 10,003 patients across 193 sites in the United States and Canada (both academic and private practice) targeted at patients with symptoms suggestive of CAD referred for further noninvasive clinical evaluation [11]. The primary hypothesis of PROMISE was that individuals with suspected CAD undergoing CCTA as a primary mode of risk stratification would have superior health outcomes relative to those undergoing functional testing (including stress electrocardiography, echocardiography or nuclear testing). Inclusion criteria (aimed at enriching the study population for CAD) comprised men over 45 and women over 50 years of age with symptoms potentially related to CAD, without prior CAD. Individuals with symptoms or signs of unstable acute coronary syndromes, a history of (or recent evaluation for) CAD, or contraindications to CCTA or functional testing were excluded. In addition, PROMISE was funded by the National Institutes of Health, without any contributions from vendors involved in CCTA. Enrolled participants were randomized to CCTA or functional testing (with type of functional testing prescribed by the attending physician). In keeping with the aim of the study to test “strategies” of care, CCTA and functional testing was reviewed on site, with subsequent management at the behest of the attending physician (not the study investigators). Protocol-driven follow-up at 6‑month intervals after randomization for a period of at least 1 year was performed. The primary endpoint of PROMISE was a composite of major adverse cardiovascular events, including all-cause mortality, myocardial infarction, unstable angina hospitalization, and complications of diagnostic testing within 72 h of testing (e. g., stroke, bleeding, renal failure, anaphylaxis; all events adjudicated independent of study investigators). The investigators examined a variety of secondary endpoints as well, focused on safety (radiation) and effectiveness (downstream health-care expenditure and nonobstructive CAD patterns by angiography). The trial defined a “positive” (abnormal) CCTA if there was at least ≥ 70 % stenosis in any of three epicardial arteries or ≥ 50 % stenosis in the left main coronary artery. A functional (stress) test was defined as “positive” (abnormal) if there was a reversible perfusion defect with or without accompanying infarction in at least one myocardial region.

Study population

Of the initial 10,003 patients entered into randomization, all but 404 patients underwent specified initial testing, with 1‑year follow-up available in nearly 94 % of patients. Study characteristics were balanced across functional versus CCTA groups, and had significant CAD risk factors (2–3 CAD risk factors on average). Over half of the enrolled individuals had hypertension, dyslipidemia, or prior or current smoking, and nearly half were receiving aspirin, statin therapy, or antihypertensive medication. The primary presenting symptom was chest pain (in over 70 %), primarily of an atypical nature (78 %). As an aggregate measure of risk, the average pretest probability of obstructive CAD by the combined Diamond–Forrester and Coronary Artery Surgery Study model was over 50 %. Individuals randomized to functional testing most commonly underwent stress testing with some form of cardiac imaging (nuclear in 67.5 % vs. echocardiography in 22 %), and 29.4 % tests required pharmacologic stress agents. Collectively, these characteristics were overall in keeping with a low-to-moderate-risk population, an ideal clinical circumstance for evaluation of comparative effectiveness of functional versus anatomic strategies.

Primary outcome

Over a median 25-month follow-up, the trial did not meet the primary endpoint with 164 (3.3 %) participants in the CCTA arm and 151 (3.0 %) in the functional test arm experiencing an adverse clinical event (HR = 1.04 %, 95 % CI = 0.93–1.29, p = 0.75). This result was similar when follow-up was truncated at 12 months, or for any combination of CAD-related endpoints examined. There was no specific subgroup in which CCTA or functional testing was associated with improved hazard of any outcome.

Safety, test results, downstream testing, and coronary revascularization

PROMISE demonstrated that both CCTA and functional testing are clinically safe: No severe adverse event related to testing occurred, and the rate of mild adverse events was very low (n = 58; 0.6 %). Overall radiation exposure was higher in the CCTA group, probably because one third of patients in the functional arm did not undergo any test associated with radiation exposure. Nevertheless, a direct comparison between nuclear perfusion imaging and CCTA demonstrated significantly lower radiation exposures by CCTA (median exposure in the functional testing group, 12.6 mSv; CCTA, 10.1 mSv; p < 0.001; Table 1).

Table 1 Radiation exposure in CCTA and SPECT in PROMISE
Full size table

Overall, a relatively small number of patients had either obstructive CAD (CCTA arm; n = 517, 10.7 %) or myocardial ischemia (functional arm; n = 556, 11.7 %). While patients randomized to CCTA underwent downstream angiography at 90 days after randomization more frequently as compared with those randomized to functional testing (609 in CCTA, 12.2 % vs. 406 in functional group, 8.1 %), the diagnostic yield was higher for CCTA as obstructive CAD was observed in a greater frequency in patients randomized to CCTA (72.1 %) compared with those randomized to functional testing (48.5 %). Not surprisingly, this resulted in a greater frequency of coronary revascularization (6.2 % in CCTA vs. 3.2 % in functional testing). The corresponding results are shown in Fig. 1. Finally, approximate total costs (including downstream testing) were similar between CCTA and functional testing.

Fig. 1
figure 1

Results of testing and downstream invasive coronary angiography (ICA), as well as results at ICA, stratified by CCTA and functional testing. CAD coronary artery disease, CCTA coronary computed tomography angiography

Full size image

Lessons from PROMISE

The strengths of the PROMISE study included: a large sample size (n > 10,000) over a geographically diverse area and involving both academic and private centers; the inclusion of women and minorities; recruitment of patients with a 50 % probability of obstructive CAD (by Diamond and Forrester criteria), most of whom had atypical chest pain and two to three major cardiovascular risk factors, which represents a guideline-conforming clinical referral population of patients with suspected CAD for advanced diagnostic testing. In addition, the PROMISE trial provided a snapshot of real-world clinical care across North America and its effect on clinical and economic outcomes. While the results of the trial did not support the investigators’ pre-specified hypothesis (specifically did not demonstrate the superiority of CCTA), the trial provides significant insights and lessons.

Lesson 1

Patients with stable chest pain have much lower rates of positive test findings and cardiovascular events as compared with 25 years ago – emphasizing the critical need for improved selection instruments for noninvasive diagnostic testing.

First, the overall event rates in PROMISE were lower than expected for the level of clinical risk anticipated. Indeed, only 11 % of the CCTA arm of the study population had obstructive CAD, far lower than would have been expected from pretest clinical risk estimates (given Diamond–Forrester risk at 53 %). Moreover, the event rates were low at 3.1 %, relative to the projected 8 % based on United Health Care data. Some of these results are not surprising: In a registry study of 14,048 individuals with clinically indicated CCTA tests, typical angina was associated with the highest prevalence of any epicardial vessel stenosis over 70 % (27 % in men and 11 % in women). In this study, the pretest risk (based on a composite of Diamond and Forrester angina classification and ACC/AHA prescribed pretest probabilities) overestimated the observed prevalence of CAD significantly, regardless of age or sex, which appeared driven primarily by atypical angina (the primary form of chest pain symptom in PROMISE) [12]. Aside from poor calibration of clinical risk estimators, the overall prevalence of myocardial ischemia in the past 20 years has declined from 40.9 % in 1991 to 8.7 % in 2009 [13].

Because pretest clinical risk estimators for CAD prevalence are not well calibrated to contemporary practice, more efficient utilization of either CCTA or functional imaging demands improvement in patient selection to optimize cost and benefit [14, 15]. Several studies have suggested a systematic overestimation of significant anatomic CAD by traditional Diamond–Forrester classification of chest pain [16] or Duke treadmill score [15], with lack of discrimination especially in women. In addition, clinical risk estimates for asymptomatic individuals do not necessarily apply to symptomatic individuals referred for evaluation of suspected CAD [17]. Nevertheless, recent additions, and details from ancillary investigations (e. g., left ventricular function or coronary artery calcification [18]), may help clinicians direct noninvasive tests to the most appropriate cadre of patients. In recent work, Genders and colleagues extended the traditional Diamond–Forrester classification to include more comprehensive lipid-based and historical risk factors (e. g., diabetes) as well as coronary calcification in prediction models for CAD, based on invasive angiography and CCTA [15]. These investigators noted improvement in risk discrimination and 35 % risk reclassification based on extended models, suggesting that a comprehensive approach based on clinical, historical, and extended (e. g., calcium score) risk factors for CAD improves risk prediction in symptomatic individuals. These results require confirmation in large, transnational populations, and PROMISE affords a unique opportunity to validate composite risk scores (inclusive of coronary plaque morphology and calcification).

Moreover, some advocates of limiting diagnostic imaging may support a “no testing” approach, wherein symptomatic individuals with multiple risk factors (as in PROMISE, with 2–3 CAD risk factors) automatically receive CAD prevention. Given the low event rate and the low test positivity rate in PROMISE [12], this approach may appear attractive. Moreover, there is a complete paucity of data as far as the feasibility of such a strategy is concerned. Furthermore, patients and caregivers would face uncertainties around needs for revisits and follow-up, legal and cost considerations, and a general strategy. While it appears on the surface as an attractive approach, data are needed to warrant further consideration.

Ultimately, the development of more efficient approaches to identify candidates for imaging are warranted, including risk stratification algorithms that are based on contemporary populations and perhaps novel biomarkers that have discriminatory value. This would enable more efficient diagnostic imaging (whether it be nuclear perfusion or CCTA) and will avoid unnecessary exposure to radiation, need for follow-up of incidental findings, and normal results on diagnostic angiography for suspected CAD.

Lesson 2

Combining anatomic with functional information is the key to improve patient selection for invasive angiography and coronary revascularization.

Similar to randomized trials in the acute chest pain setting, PROMISE observed increased downstream utilization of invasive coronary angiography and revascularization with CCTA. Within 90 days of initial diagnostic test, 609 patients (12.2 %) in the CCTA group underwent invasive angiography, compared with 406 (8.1 %) in the functional group. However, while 28 % of the individuals undergoing angiography after CCTA had normal coronary arteries at catheterization (3.4 % of the overall CCTA group), nearly 58 % of individuals in the functional group demonstrated normal coronary arteries (4.3 % of the overall functional group, p = 0.02 vs. CCTA). Further interpretation of these comparisons is limited by the PROMISE study design, which left the decisions for downstream angiography and revascularization to the treating physicians. In effect, rates of revascularization were higher in individuals who had previously undergone CCTA versus those who had functional tests (6.2 vs. 3.2 %, p < 0.001).

While thorough site certification and quality control were performed in PROMISE, image interpretation and downstream clinical management were left to the discretion of each local site, which in their entirety of 193 sites represented a large amount of heterogeneity in experience, equipment, and clinical management style, which is desirable given that that the results should be broadly generalizable. It is quite easy to imagine that the quality of the test is associated with its interpretation and suggested further management. The quality of CCTA depends on the equipment (e. g., 64- vs. 128-slice or higher-slice CT scanners), appropriate use of medication to dilate the coronary arteries and reduce cardiac motion (e. g., nitroglycerin and beta-blockade therapy), and composition of atherosclerosis (e. g., blooming artifacts with greater coronary calcification). It is further clear that the expertise of readers has a substantial influence on further management, and studies have shown that the diagnostic yield of CCTA imaging improves with reader expertise [19, 20]. Not directly test-related factors are differences in a cardiologist’s interpretation of the test results and the subsequent actions such a medical treatment or referral for intervention. The interpretation of test results, medical guidelines, and personal management styles (e. g., “confirmatory” diagnostic angiography for any lesion > 50 % vs. functional testing after CCTA) certainly affect downstream testing and safety of diagnostic imaging [21]. Overall, the results support the notion that real-world clinical management of angiographic findings after CCTA is largely based on anatomic but not functional information, despite a wealth of randomized evidence supporting a combined approach of anatomy and physiology. Tonino and colleagues reported the results of 1005 patients with multivessel CAD who underwent anatomic and functional assessment during angiography using fractional flow reserve (FFR) assessments. Individuals who underwent FFR-guided therapy had a lower rate of cardiovascular events at 1 year (13.2 %) relative to the anatomic-guided only approach (18.3 %; p = 0.02) [22]. The merits of a composite anatomic and functional approach to invasive CAD risk stratification have since been extended to longer follow-up [23].

In light of these results, CCTA has adopted this approach: Recent data involving CT-based assessments of fractional flow reserve (FFR-CT) have been put forth to address this important aspect of CAD risk stratification [24]. In an early study of 103 individuals with CCTA, FFR-CT methods were validated against FFR by diagnostic angiography, with sensitivity 84 % and specificity 82 % for per-vessel ischemia [25]. In this study, invasive FFR and FFR-CT were highly correlated (r = 0.72, p < 0.001). In a study by the same group, Min and coworkers demonstrated that FFR-CT methods improved discrimination for significant ischemic territories over obstructive CAD by CCTA (with catheterization as the gold standard) [24]. These observations have been replicated by other groups [26–28], including a recent large meta-analysis reaffirming the utility of CT-FFR [29]. The summary of most recent FFR-CT studies is shown in Table 2. While the FFR-CT calculations depend on several key assumptions about coronary flow (including a simulation of hyperemic flow), the assessment of FFR and anatomic lesion severity appears to improve discrimination of risk. Additional evaluation in the PROMISE study to assess whether FFR-CT assessments impact prognosis are underway, and should provide an added dimension to the data offered by CCTA.

Table 2 Selected large studies in CCTA evaluating performance of FFR-CT
Full size table

Lesson 3

Testing choice was not associated with better clinical outcomes, but it may not be the fault of the test.

A central result of PROMISE was that choice of the initial diagnostic modality (functional vs. anatomic) did not affect clinical outcomes. While at first blush this suggests equipoise between the different modalities, one has to consider that many things happen between the test and a future event, which may or may not be related to the test itself. We discussed test-related and nonrelated factors affecting management in the prior paragraph.

More importantly, a recent meta-analysis of the PROMISE, SCOT HEART Trial, and COMPASS trials by Bittencourt and colleagues reported a 31 % reduction in the odds of having a downstream myocardial infarction with the use of CCTA (vs. functional approaches) [30], suggesting that the direct visualization of CAD may indeed enable improved outcomes. These data are corroborated by several observational studies demonstrating that a normal CCTA was associated with a very low 0.15 % annualized risk of death, similar to general population risk [21]. Another factor that may favorably affect outcomes after CTA is the ability to visualize the presence and extent of nonobstructive CAD. Several studies have shown that this finding, while not resulting in immediate events, is associated with a threefold increase in cardiovascular events.

Another discussion has evolved around the appropriateness and benefits of how to proceed with diagnostic angiography in CCTA lesions with > 50 % obstruction and myocardial ischemia [31]. There is no definite proof and consensus on which patients (except those with left main or triple vessel disease) may benefit from coronary revascularization. This question is being investigated in the International Study of Comparative Health Effectiveness With Medical and Invasive Approaches (ISCHEMIA) study – a large, contemporary investigation of medical versus interventional approaches in patients with moderate-to-severe ischemia. Appropriately, the design includes CCTA prerandomization to exclude left main disease [32]. We await the results of studies assessing the relative merits of ischemic and anatomic testing versus medical therapy alone (a “no testing” strategy).

Lesson 4

Diagnostic testing is very safe.

An important question within PROMISE was assessment of the relative safety profile of a functional strategy versus CCTA. An ongoing concern in contemporary literature on advanced cardiovascular imaging is safety with respect to contrast and radiation exposure. In PROMISE, the number of patients with any safety event was similar between CCTA and stress testing arms. While contrast reactions (mild and extravasation) were by definition more common with CCTA, renal failure and anaphylaxis were nonexistent. Although PROMISE did not show any differences in procedure-based risks between CCTA and functional testing, it is worth noting that angiography was used more frequently in individuals randomized to CCTA and is associated with further use of contrast agents.

Given increasing concerns over long-term malignancy risk, radiation exposure has become a major issue in modern imaging applications. While mean cumulative radiation exposure on average was higher in the CCTA arm versus the functional testing arm (12.0 vs. 10.1 mSv), this was largely due to confined radiation exposure in the nuclear stress testing arm. Indeed, when compared with individuals in the nuclear perfusion imaging arm, CCTA had a lower overall radiation exposure (12.0 vs. 14.1 mSv; p < 0.001). These results are consistent with prior results from other studies: In the Rule Out Myocardial Infarction Using Computer-Assisted Tomography (ROMICAT) study, CCTA radiation dose was 11.3 ± 5.3 vs. 14.1 ± 4.8 mSv for SPECT, with lower doses in those individuals who had a 128-slice dual-source scan [33]. Differences were smaller but still in favor of CCTA in the Coronary Computed Tomographic Angiography for Systematic Triage of Acute Chest Pain Patients to Treatment (CT-STAT) study (11.5 vs. 12.8 mSv SPECT) [34]. In a large observational registry of nearly 2000 patients with CCTA over 50 sites, the median dose was 12 mSv with variability by scanner platform and location [35]. While these studies suggest stability in CCTA radiation dose, over the past several years, innovations in CCTA technology and improved recognition of methods to reduce radiation exposure have progressively cut radiation exposure by over 50 % (in some cases, nearly to 1–2 mSv with prospectively ECG-triggered modalities). However, it should be noted that while in PROMISE (and in the US) nuclear perfusion stress testing is the most frequently performed functional test (67 % of patients), in other countries, such as Germany, other functional tests such as exercise treadmill or stress echocardiography are more frequently performed, which would have resulted in a significant lower radiation dose in the functional testing treatment arm. On the other hand, such practice patterns often result in a more frequent referral of patients to invasive coronary angiography.

Conclusion

Despite the limitations of PROMISE, it is clear that in a real-world, low-to-intermediate-risk patient population referred to noninvasive testing for delineation of CAD, both anatomic (CCTA) and functional (stress testing) approaches are associated with similar clinical, economic, and safety-based outcomes. However, several key questions remain as potential foci of future investigation: Does discovery of nonobstructive CAD by CCTA lead to increased utilization of preventative therapies and by extension to improved outcomes? Do assessments of coronary physiology by FFR-CT improve selection of candidates for invasive coronary angiography minimizing costs but perhaps also improving outcomes? Do novel clinical biomarkers of myocardial injury (e. g., high-sensitivity troponin [36]) add to (or obviate the need for) assessments of coronary anatomy or physiology? It is clear that tools enabling a more nuanced diagnosis and management of these patients are necessary to manage this large population appropriately and efficiently.