Guiding Therapy by Coronary CT Angiography Improves Outcomes in Patients With Stable Chest Pain

1Coronary computed tomography angiography (CTA) has high sensitivity and specificity for the detection of coronary artery disease,. This has prompted the evaluation of coronary CTA as a diagnostic test for coronary artery disease in patients presenting with stable chest pain. The short-term benefits of coronary CTA in this population have included better diagnostic certainty, lower rates of normal coronary arteries at the time of invasive coronary angiography, and improved targeting of symptomatic and preventative therapies,. Large-scale clinical trials have also suggested that short-term coronary heart disease events are reduced,. In the SCOT-HEART (Scottish COmputed Tomography of the HEART Trial), we recently reported the effects of coronary CTA on the pre-specified 5-year clinical outcomes including investigations, treatments, and clinical events. We demonstrated that an initial strategy of coronary CTA was associated with a 41% relative risk reduction in coronary heart disease death or nonfatal myocardial infarction at 5 years. This major reduction in events has prompted questions about the mechanisms of benefit, the potential for bias, and the plausibility of the effect size.

2We here present further analyses of the 5-year data from the SCOT-HEART trial to assess the robustness of the event reductions seen with coronary CTA with respect to the participant subgroups, the changes in diagnosis, and the alterations to procedural and pharmacological treatments. Drawing these disparate effects together, we wanted to determine the overall attribution of benefits in relation to the primary endpoint and the study intervention effect size.

3The study population and trial design were reported previously,,. In brief, adult patients age ≤75 years who attended the outpatient cardiology clinic with stable chest pain were invited to participate in the trial in 12 cardiology centers across Scotland between 2010 and 2014. They were randomized 1:1 to standard care alone or standard care plus coronary CTA. Attending clinician-directed changes in diagnosis, investigations, and treatments were documented at 6 weeks, and changes in investigations, treatments, and clinical outcomes over 5 years were obtained from nationwide routinely collected health care data through National Health Service (NHS) Scotland.

4Participants without a prior history of coronary heart disease comprised 91% of the study population and were categorized into those with nonanginal chest pain and a normal 12-lead electrocardiogram (nonanginal chest pain), and those with either typical or atypical chest pain, or nonanginal chest pain and an abnormal electrocardiogram (possible angina) as per the National Institute for Health and Care Excellence guidelines,.

5The 10-year cardiovascular risk of participants was determined by the ASSIGN score. This score is based on traditional cardiovascular risk factors but also incorporates other risk markers including social deprivation and family history of cardiovascular disease and has been calibrated for the Scottish population. To avoid potential for bias, changes in diagnosis, investigations, and treatments were documented following prompting with either the coronary CTA result (standard care and coronary CTA) or the ASSIGN score (standard care alone).

6Coronary revascularization procedures (percutaneous coronary intervention and coronary artery bypass graft surgery) were identified from inpatient and day-case episodes, and coronary disease burden was quantified using centralized review of individual coronary angiograms blinded to allocated intervention,,.

7The Scottish national community drug-prescribing database of the Information and Statistics Division in NHS Scotland maintains a detailed record of all NHS prescriptions dispensed in the community, which are linked to individual patient identifiers,,. All prescriptions are dispensed by community pharmacies, dispensing doctors, and a small number of specialist suppliers.

8Clinical data and deaths were obtained from the Information and Statistics Division and the electronic Data Research and Innovation Service of NHS Scotland up to March 31, 2018. To account for reporting delays in data completeness (6 to 8 weeks), outcomes were determined up to January 31, 2018. Previous analyses have demonstrated an excellent correlation (>95%) between clinical events identified through Scottish health record linkage and clinical events recorded and adjudicated as part of a regulated clinical trial. This has been successfully applied in other settings, and for longer-term clinical trial follow-up.

9This was a post hoc analysis of the SCOT-HEART trial. Clinical outcomes were analyzed using Cox regression, adjusted for center and minimization variables, with cumulative event curves constructed. We performed a landmark analysis at 12 months, reasoning that coronary CTA–driven alterations in invasive coronary angiography and coronary revascularization should have been completed by this time point. Data are presented as mean ± SD, median (interquartile range), and hazard ratio (HR) or odds ratio (95% confidence interval [CI]) as appropriate. No correction for multiplicity was undertaken when testing secondary or other outcomes.

10As the SCOT-HEART trial randomized the application of coronary CTA as a diagnostic intervention, it cannot definitively determine the mediation pathway for the treatment effects demonstrated. To explore the plausibility of the treatment effect, we performed a modeling study aiming to estimate the mean outcome for patients in the coronary CTA group under a counterfactual scenario wherein the management of patients in the coronary CTA group was the same as that observed in the standard care group. The full set of assumptions are given in the Online Appendix. However, we wish to highlight that we modeled a scenario wherein the highest risk (most appropriate) patients were those who had received additional preventative therapies as a result of the coronary CTA intervention given that this would be the consequence of improved diagnostic precision. Under the observed difference between the trial arms in use of preventative therapies, this represents a best-case scenario.

11Statistical significance was taken as a 2-sided p < 0.05. All analysis was undertaken using R version 3.5.0 (R Foundation, Vienna, Austria).

12We randomized 4,146 patients with stable chest pain at 12 cardiology centers across Scotland to either standard of care alone (n = 2,073) or standard of care plus coronary CTA (n = 2,073). Participants were middle aged, had a high prevalence of cardiovascular risk factors, and had a slight male preponderance (Table 1).

13Of the 4,146 study participants, there were 1,447 patients with nonanginal chest pain and a normal resting electrocardiogram, 2,323 with possible angina, and 372 with prior coronary heart disease. Four patients were excluded due to incomplete symptom descriptions. Nonobstructive or obstructive disease was identified on coronary CTA in 49.9% of those with nonanginal chest pain and 66.9% of those with possible angina (Table 2). All 3 groups derived similar relative benefit from coronary CTA (p value for interaction ≥0.50), although the absolute magnitude and temporal pattern of benefit appeared to vary (Figure 1). Among patients with nonanginal chest pain, the event rate appeared relatively constant over time with the benefits of coronary CTA only apparent over a more prolonged time period (>1 year). In contrast, patients with prior coronary heart disease were at higher risk of events, especially in the early follow-up period, and the benefits of coronary CTA were more immediately apparent. Compared with these groups, patients with possible angina were at intermediate risk and demonstrated an intermediate time course of benefit.

14We have previously reported that the diagnosis of angina due to coronary heart disease was changed in 1 in 4 patients following coronary CTA. Among patients where the diagnosis of angina due to coronary heart disease was not made, those in the coronary CTA group had a lower incidence rate of coronary heart disease death or nonfatal myocardial infarction compared with the standard care group (incidence rate: 0.23; 95% CI: 0.13 to 0.35 vs. 0.59; 95% CI: 0.42 to 0.80 per 100 patient-years, respectively; p < 0.001) (Figure 2). The findings on coronary CTA were strongly predictive of future risk of coronary heart disease death or nonfatal myocardial infarction (Online Figure 1).

15Among patients referred for invasive coronary angiography, those randomized to CTA had more extensive burden of coronary artery disease (p = 0.056) (Table 3). This finding was particularly notable among participants without inducible ischemia on exercise testing (Online Table 1). At the end of 1 year, more patients who had undergone coronary CTA received coronary revascularization (246 vs. 208; HR: 1.21; 95% CI: 1.01 to 1.46; p = 0.042) (Figure 3). Coronary revascularization was predominantly percutaneous coronary intervention (n = 353; 8.5%) although one-quarter of patients (n = 111; 2.7%) underwent coronary artery bypass graft surgery. Beyond 1 year, 33 coronary revascularizations occurred in the coronary CTA group compared with 59 in the standard care group (HR: 0.59; 95% CI: 0.38 to 0.90; p = 0.015). Of these, 8 coronary revascularizations were prompted by myocardial infarction in the coronary CTA group, and 18 in the standard care group (HR: 0.45; 95% CI: 0.20 to 1.04; p = 0.061).

16Within the coronary CTA group, 99% of participants with evidence of inducible ischemia on exercise testing were already receiving preventative medications at the baseline visit. In contrast, only 50% of participants were receiving preventative treatments in those who had a normal exercise test and evidence of nonobstructive or obstructive coronary artery disease on coronary CTA (Online Table 2). Following clinic consultation, 1 in 20 patients (5%) had their treatment altered at 6 weeks in the standard care arm compared with nearly 1 in 4 (23%) in the coronary CTA group (Table 4). In particular, there were differences in the prescription of both antiplatelet and statin therapies that were sustained over the 5 years of follow-up (Figure 4). Of note, antiplatelet therapy use fell from 48% (baseline) to 41% (at 1 year) in the standard of care arm (p < 0.001), whereas it increased from 49% (baseline) to 52% (at 1 year) in those assigned to coronary CTA (p = 0.017). In contrast, overall statin use increased in both groups (standard care: 43% to 50%; coronary CTA: 44% to 59%; p < 0.001 for both groups) but this was greater in those assigned to coronary CTA (p < 0.001). Compared with those without coronary heart disease, rates of antiplatelet and statin therapies were markedly higher in patients who had coronary heart disease documented on the coronary CTA despite comparable 10-year cardiovascular risk scores (Figure 5). This was most apparent in participants with lower cardiovascular risk scores.

17For the baseline model, there were 48.5 and 82.4 estimated events in the coronary CTA and standard care groups, respectively. This was similar to the observed number of events in each group (48 and 81 events, respectively). Under the counterfactual scenario where the proportion of participants receiving preventative therapy was the same in both arms, the estimated number of events in the coronary CTA group was 84.4 (i.e., similar to that observed in the control arm).

18We report a post hoc analysis of the long-term follow-up of the SCOT-HEART trial participants in response to questions raised about the consistency and plausibility of the reported effect size. We demonstrate a consistency of effect on event rates across the differing trial subpopulations in keeping with their presentation and risk profile. We further demonstrate that greater diagnostic certainty led to better categorization and risk stratification with increased use of procedural and pharmacological interventions in the coronary CTA group. As a result of this improved diagnostic performance and risk classification, we can plausibly account for much of the observed effect size by the targeted application of these preventative interventions.

19The SCOT-HEART trial included a broad range of patients who were at low, intermediate, and high risk of coronary events, as well as patients with prior coronary heart disease,. In the current analysis, we categorized patients according to the National Institute of Health and Care Excellence classification into those with nonanginal chest pain, possible angina, and known coronary heart disease (Central Illustration). Those participants with nonanginal chest pain had the lowest event rates, and these were linear over time. Moreover, the apparent benefits of coronary CTA were slow to accrue with a continuous rate of separation in events with time that appeared to take >1 year to be realized. These features are consistent with the event rates and time course of effects seen in primary prevention trials. In contrast, participants with possible angina had a higher and time-varying event rate with increased early hazards consistent with some individuals having higher risk disease, namely new onset angina,,. Patients with known coronary heart disease were already on established preventative therapies and the benefit appears to be rapid and early, consistent with an effect of coronary revascularization.

20One of the main advantages of coronary CTA is its negative predictive value. As such, it moves beyond traditional symptom assessment and ischemia testing, providing confidence to clinicians and patients alike regarding the absence of disease. This was reflected in the increased number of patients in whom a diagnosis of angina was excluded and preventive therapies discontinued in the coronary CTA group,. We observed that the event rate among this group was much lower than the equivalent diagnostic classification in the standard care group, which is consistent with the high negative predictive value and diagnostic accuracy of coronary CTA and implies greater diagnostic misclassification of patients in the standard care arm contributed to their higher event rate. False negative misclassification is common and previous reports have highlighted that up to one-third of myocardial infarctions occur in patients who had previously been diagnosed with noncardiac chest pain. This finding was reproduced in the standard care group of the SCOT-HEART trial. Moreover, our modeling suggests that the use of coronary CTA can, via increased identification and treatment of previously unrecognized predominantly nonobstructive coronary heart disease, plausibly reduce event rates among those with nonanginal chest pain.

21Coronary CTA is the only widely available and well validated noninvasive diagnostic technique able to identify nonobstructive coronary plaque disease. Standard approaches focusing on noninvasive ischemia testing have relatively poor sensitivity for obstructive coronary disease and, by definition, are unable to differentiate patients with normal coronary arteries from those with nonobstructive plaque disease. Furthermore, we have shown for the first time that the identification of nonobstructive coronary disease was a key factor associated with de novo provision of evidence-based preventative therapies among those who underwent coronary CTA.

22Coronary revascularization was more common in the coronary CTA group during the first year after randomization. We here show this reflects increased early detection of obstructive coronary artery disease that includes left main stem and triple vessel disease. Coronary revascularization may be particularly impactful in those with prognostic disease and those with new onset, rapidly progressing, or recurrent angina, in whom underlying atherosclerotic disease process may be more active. This is potentially important as coronary revascularization is known to reduce ischemic coronary events and improve clinical outcomes in these more unstable clinical contexts,. In contrast, beyond 1 year, rates of coronary revascularization were higher in those who had received standard care alone and many of these revascularization episodes were triggered by myocardial infarction. This suggests that standard care may be associated with missed diagnoses of coronary artery disease that later declare themselves, whereas coronary CTA correctly identifies those who require treatment and thereby reduces future events.

23There were important differences in the changes to pharmacological therapies seen within the 2 trial groups. First, antiplatelet therapy use fell in the standard care group but rose in the coronary CTA group, whereas statin use increased in both groups albeit there were greater increases in the coronary CTA group. Antiplatelet therapy at baseline in part reflects diagnostic uncertainty on the part of the primary care physician who may have initiated preventative therapy while awaiting specialist review. The high diagnostic sensitivity of coronary CTA resulted in an increased diagnosis of coronary artery disease (especially nonobstructive disease) and consequently increased prescribing of antiplatelet therapy in this group. In contrast, the reduced ability to detect nonobstructive coronary disease in the standard care group resulted in treatment discontinuation in many participants. Although the overall differences in prescribing may appear modest, it should not be forgotten, it is the distribution of such therapies that is important and many patients in the standard care group may be receiving futile treatments because many will not have underlying coronary disease. For the first time, we here report that those undergoing coronary CTA will have more directed and appropriate therapies. Indeed, when we look at those who underwent coronary CTA, the rates of antiplatelet and statin therapy use were almost 3-fold higher in those with coronary artery disease compared with those without disease despite identical 10-year cardiovascular risk scores. Thus, simply looking at the overall frequency of prescribing of preventative therapies ignores the importance of how such medications are distributed within a population. Again, these findings suggest that coronary CTA is consistent with a precision medicine approach by identifying the most appropriate treatment for each patient.

24Using treatment effect estimates from published randomized controlled trials, systematic reviews, and meta-analyses,,,,,,,, we explored for the first time whether the effect of additional treatment with antiplatelet, statin, and revascularization interventions could account for the observed effect size. Whereas this modeling required a number of assumptions, they were nonetheless reasonable, and more reasonable than assuming that patients receiving additional treatment in the coronary CTA group shared the same cardiovascular event rate as the entire cohort. This modeling allowed us to demonstrate that at least some of the effect size we observed was plausible and could be accounted for by current evidence. In addition, we would also highlight that the effect size reported in the SCOT-HEART trial is consistent with rates of myocardial infarction reported in other trials of coronary CTA in patients with stable chest pain, and meta-analyses,, as well as large-scale observation studies. Moreover, our model neither accounts for positive lifestyle changes, such as smoking cessation, nor treatment compliance, which may also have a beneficial impact given the greater uptake of antiplatelet and statin therapy in those diagnosed with coronary artery disease by coronary CTA. We accept that many of our modeling choices are necessarily subjective, and we are pleased to provide access to the data to allow other researchers to test the effect of different assumptions. Finally, we would highlight that the magnitude of the effect size we described in the SCOT-HEART trial is consistent with other diagnostic studies in patients with suspected coronary heart disease, as well as primary prevention trials in patients with interventions targeted to those at higher risk, such as the JUPITER (Justification for the Use of Statins in Primary Prevention: An Intervention Trial Evaluating Rosuvastatin) trial.

25The identification of disease is inextricably linked to downstream changes in lifestyle, initiation and intensification of preventative therapies, and the judicious use of coronary revascularization. The combined effect of all of these interventions on coronary atherosclerosis reduces the future risk of myocardial infarction. In this regard, it was intriguing to observe that the benefits of coronary CTA were similar regardless of presentation symptoms and indeed the greatest proportionate benefit was seen in patients without a diagnosis of angina due to coronary heart disease. This begs the question of whether coronary CTA has a role in primary prevention of coronary heart disease, and this will be the subject of the forthcoming SCOT-HEART 2 trial.

26Some have highlighted that there may have been bias in the reporting of outcomes in the SCOT-HEART trial: something that should always be considered in an open-label trial. However, we think this is unlikely. First, the delayed separation of the event curves suggests that there was no early bias in event reporting because the CT result would be available after 2 weeks, whereas treatment changes took a further 4 to 6 weeks to implement. Second, the finding that coronary CTA was associated with less normal invasive coronary angiography and higher early rates of revascularization suggests that the coronary CTA more accurately identified the disease process. Third, coronary CTA increased the rate of diagnosis of coronary artery disease and intuitively would be more inclined to increase the diagnosis of myocardial infarction, which would act against any potential benefits of the trial intervention. Fourth, the early increase followed by later reductions in coronary revascularization in the coronary CTA group does not support the postulation of ascertainment bias. Why would these rates suddenly change direction? Fifth, the patterns of event reduction are consistent across the trial subpopulations with both lower risk nonanginal chest pain and higher risk new onset angina pectoris responding to changes in treatment as previously reported in primary and secondary prevention trials. Sixth, although we did not undertake clinical endpoint adjudication, we do not believe that this is likely to have affected our findings. A large analysis involving 10 cardiovascular trials has demonstrated the process of blinded endpoint adjudication changes neither the number of endpoint events identified nor the overall treatment effect estimate for the randomized intervention. A recent Cochrane Systematic Review has also reported similar trial outcomes irrespective of central versus local endpoint determination. Ultimately, for a pragmatic trial determining the effect of introducing a new diagnostic test into a health care system, the main goal is to establish the impact on health care outcomes as reported by that health care system. We believe clinical outcomes reported by routine health records data remain the most appropriate, independent, and accurate method of clinical endpoint ascertainment for the SCOT-HEART trial. Finally, we acknowledge our modeling approach does assume that the highest risk patients were those who had received additional preventative therapies as a result of the coronary CTA intervention, and this may have overestimated some of the benefits. This is particularly applicable to the assumption of treatment benefit derived from coronary revascularization. Although we chose treatment effects applicable to an unstable angina population, we believe this is justifiable as the shape of the instantaneous hazard curves supports a high short-term risk that plateaus after 6 to 12 months in a pattern consistent with acute coronary syndrome populations.

27We have presented a multifaceted analysis that consistently and robustly demonstrates the plausibility of a reduction in long-term coronary events consequent on investigating patients with stable chest pain using coronary CTA. If we are to improve the prevention of future myocardial infarction, coronary CTA would appear to be the most effective and indeed the only proven investigative approach in patients with stable chest pain.COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: Adding coronary CTA to standard care in patients with stable chest pain syndromes can reduce myocardial infarction and death over 5 years by facilitating evidenced therapy for those with coronary artery disease independent of symptom characteristics.TRANSLATIONAL OUTLOOK: Future studies should evaluate the value of coronary CTA to guide treatment of asymptomatic patients at risk of coronary events.

28COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: Adding coronary CTA to standard care in patients with stable chest pain syndromes can reduce myocardial infarction and death over 5 years by facilitating evidenced therapy for those with coronary artery disease independent of symptom characteristics.

29TRANSLATIONAL OUTLOOK: Future studies should evaluate the value of coronary CTA to guide treatment of asymptomatic patients at risk of coronary events.