Computed tomography coronary angiography for patients with heart failure (CTA-HF): a randomized controlled trial (IMAGE-HF 1C)

Abstract

Introduction

The rising global prevalence and burden of heart failure (HF) has resulted in the increase in testing and costs.^1,2 Since patients with ischaemic cardiomyopathy potentially benefit from coronary revascularization, invasive coronary angiography (ICA) is often used.³ Unfortunately, ICA is costly, a limited resource and has inherent risks; therefore, alternative, safe, cost-effective, non-invasive strategies are desirable.

Coronary computed tomography angiography (CCTA) is a non-invasive diagnostic test^4–6 with prognostic value,^7,8 and in HF patients, has a reported sensitivity and specificity of 73–98% and 99–100%, respectively.^9,10 The 2016 European Society of Cardiology guidelines recommend (Class IIb) the use of CCTA in HF patients with low–intermediate pre-test probability for coronary artery disease (CAD). The use of CCTA in HF patients with systolic dysfunction is deemed as ‘appropriate’.¹¹ Whether or not CCTA can reduce costs in HF patients for diagnosis and management is unknown.

The objective of this randomized controlled trial was to determine the financial impact, at 12 months, of an initial diagnostic strategy of CCTA in patients with HF of unknown aetiology.

Methods

The design of this multicentre, international randomized controlled trial has been previously described and was part of the IMAGE-HF (Imaging Modalities to Assist with Guiding therapy and the Evaluation of patients with Heart Failure) study.¹² Patients with HF of unknown aetiology were screened. Eligible patients had a documented history of left ventricular dysfunction (left ventricular ejection fraction < 50%), or New York Heart Association (NYHA) Class II–IV symptoms, or an HF admission to a hospital or emergency department within the past 12 months. Patients with a history of myocardial infarction (MI) could be enrolled if the treating physician believed that the aetiology of HF was uncertain. Exclusions were: age <18 years, lacked informed consent, glomerular filtration rate <45 mL/min, allergy to intravenous contrast agents, contraindications to radiation exposure, inability to perform 20 s breath-hold, history of coronary revascularization, or CCTA or ICA within the preceding 12 months. Enrolled patients were randomized to either the CCTA or ICA strategy and were stratified according to recruitment site. Enrolling centres were from academic institutions in Canada (three sites) and Finland (three sites) and are listed in the Supplementary data online, Appendix.

Computed tomography and invasive coronary angiography

Prior to CCTA, non-contrast prospective electrocardiogram-triggered computed tomography (CT) was acquired to measure coronary artery calcification.^13,14 CCTA was performed using ≥64-slice CT scanners and radiation reduction techniques were encouraged. Patients with uncontrolled heart rates, atrial fibrillation, or ectopy were excluded depending on the capability of each institution’s scanner and clinical practice. ICA was performed according to standard clinical protocols, with selective coronary injection and imaging from multiple views.¹⁵ CCTA and ICA were evaluated and reported as per local clinical practice. For data capture, a 17-segment model and a 4-point grading score [normal, mild (<50%), moderate (50–69%), and severe (≥70%)] was used.^5,7

Patients with obstructive CAD (defined as coronary diameter stenosis ≥ 50%) were further categorized as having high-risk or non-high-risk CAD. High-risk CAD was defined as: left main stenosis ≥50%, 3-vessel disease (≥70%), or 2-vessel disease involving the proximal left anterior descending artery (≥70%).⁷

The aetiology of HF (ischaemic vs. non-ischaemic) was determined using a modified Felker classification based on history of MI or coronary anatomy (≥70% stenosis of the left main, proximal left anterior descending artery, or ≥2 epicardial arteries).¹⁶

Outcome measures

Patient follow-up was performed at 3, 12 months, and annually thereafter until the end of the study, death, or study exit. The primary outcome was the direct medical costs (average per-patient healthcare cost) of the CCTA and ICA strategies up to 12 months post-randomization. Secondary outcome measures included: all-cause death, cardiac death, MI, acute coronary syndrome (ACS), resuscitated cardiac arrest, cardiac re-hospitalization for worsening HF or arrhythmia, procedural complications (death, MI, stroke, vascular complications, severe allergic reactions; contrast nephropathy), and rates of normal ICA. Composite outcomes were also collected [major adverse cardiovascular event (MACE): cardiac death and MI; and composite clinical endpoint (CCE): MACE, resuscitated cardiac arrest, ACS, and cardiac hospitalization]. An events committee (blinded to patient allocation) reviewed and adjudicated each clinical event.

The study was approved by each institutional research ethics board and informed consent was obtained from each patient. This study was conducted in accordance to the Declaration of Helsinki, Good Clinical Practice and the TriCouncil Policy. The data underlying this article may be shared on reasonable request to the corresponding author.

Statistical analysis

Statistical analysis was performed using SAS 9.4 (SAS Institute, Cary, NC, USA) and STATA 15.1 (StataCorp LLC, College Station, TX, USA). Categorical variables are presented as frequencies and continuous variables are presented as means with standard deviations or medians with interquartile ranges. Continuous variables were compared using the Student’s t-test and categorical variables using Fisher’s exact test. Statistical significance was defined as P < 0.05.

An ‘intention-to-treat’ analysis was performed using all patients who were randomized to CCTA (121 patients) vs. ICA (125 patients), irrespective of whether they received the assigned strategy. Clinical decision-making was based on CCTA or ICA results and no downstream testing was mandated. Secondary analyses were performed using an ‘as-tested’ analysis based on the modality received (116 CT and 130 ICA patients) (Supplementary data online).

The financial impact in terms of health care costs was assessed in 2017 Canadian dollars (CDN$). Costs were grouped into five main categories: initial imaging modalities, medications, hospitalizations, cardiac procedures, and further cardiac diagnostic testing. Outpatient drug utilization data was collected at the time of baseline and follow-up visits: patients were asked to have medication available for all visits. Medications were reviewed directly with the patient in clinic/via telephone interview/or review of clinical notes. Current medication and dose in addition to any medications that may have been started or stopped during the follow-up period were captured on the case report form. Inpatient drug utilization was calculated using global hospital cost.

The total costs for each patient up to 12 months were estimated through patient follow-up surveys and through valuation of resource use using Canadian unit costs. To assess the financial impact of the management algorithm using CCTA we compared this to ICA (Supplementary data online, Appendix A).¹⁷ The costs savings associated with the use of CCTA and the underlying uncertainty around this was derived through non-parametric bootstrapping where the clinical trial sample was simulated through sampling from the clinical trial data set with replication.¹⁸ Given the Bayesian nature of the approach adopted, the analysis provides an estimate of the probability that management through CCTA will be cost saving compared to the use of ICA. The primary analysis was conducted using an ‘intention to treat’ approach and given the low degree of missing data at 12 months was based on patients with complete data. Further analyses were conducted adopting a 3-month time horizon and adopting ‘as treated’ approach.

CCE and MACE were analysed using survival analysis (time-to-major adverse cardiac event) using Kaplan–Meier curves and the log-rank test.

The proportion of various procedural complications was compared using χ² tests. A per-patient analysis of CCTA accuracy was also performed to determine the operating characteristics of CCTA in the identification of obstructive CAD and high-risk coronary anatomy.^5,19

Results

Of the 253 patients randomized, six patients were excluded prior to imaging (five withdrew consent, one inappropriate enrolment) and one patient was excluded immediately after baseline testing due to lost to follow-up (Figure 1); the remaining 246 patients were randomized to CCTA (121) and ICA (125). The mean age was 57.8 ± 11.0 years, 175 (71.1%) were men, and the mean left ventricular ejection fraction was 30.1 ± 10.1% (Table 1). Patients randomized to CCTA had a lower prevalence of diabetes and dyslipidaemia, and a higher prevalence of beta-blocker use. Mean follow-up was 19.2 ± 9.9 months.

Figure 1

IMAGE HF—1 C flowchart.

Open in new tabDownload slide

Of the 121 patients randomized to CCTA, six crossed over to ICA and one patient from the ICA arm crossed over to CCTA (Figure 1). Crossover from CCTA to ICA were due to deteriorating renal function (2), arrhythmia/elevated heart rate (HR) (3), and decompensated HF (1). One ICA patient refused ICA for personal reasons. Of the 116 patients who presented for CCTA, six patients had the CCTA cancelled due to severe coronary calcification and one CCTA was cancelled for a high HR. The prevalence of obstructive CAD, 1-, 2-, 3-vessel disease, and high-risk CAD were similar in both groups (Table 2). Using the Felker classification, the diagnosis of ischaemic cardiomyopathy was made in 45 (18.7%) patients. There were fewer normal ICAs performed in the CCTA than the ICA arm [11 (9.2%) vs. 97 (77.6%), P < 0.0001]. Radiation dose was lower with CCTA (6.0 ± 4.7 mSv) than the ICA strategy (7.9 ± 6.6 mSv, P = 0.008).

A total of 93 (76.9%) CCTA patients avoided the need for ICA. Rates of downstream ischaemia and viability testing were similar for both arms (Table 3).

When comparing the medical costs of the two interventions; the cost (CDN$in 2017) savings from the CCTA arm [CDN −$871, 95% confidence interval (CI) −$4116 to $3028] were not statistically significant (Table 4, Supplementary data online, Appendix C). The small overall cost reduction was attributed to the lower cost of the CCTA test. The follow-up costs were higher with CCTA and appeared to be driven by hospitalizations. There were a total 44 patients (22 CCTA and 22 ICA) requiring 46 cardiac hospitalizations (24 CCTA and 22 ICA) for a total of 515 days (283 and 232 days for CCTA and ICA, respectively) (Supplementary data online). The mean hospital stay for CCTA was 2.3 ± 8.1 (CI 0.86–3.74) and 1.8 ± 7.1 (CI 0.56–2.44) days for ICA. Two patients [cardiac transplantation (58 days) and HF/stroke (60 days)] accounted for 44% of CCTA hospital days.

Multiple imputation of missing values for costs did not significantly change the results. The total cost (CDN$in 2017) of CCTA was CDN$1267 (95% CI −$2045 to −$419) lower than ICA.

Using an ‘as-tested’ analysis, CCTA was associated with a non-significant reduction in healthcare costs (incremental cost saving of CDN −$2932, 95% CI −$6248 to $746).

Clinical outcomes

At follow-up, there was no difference in medication use between arms and CCE was similar between groups (Figure 2, Supplementary data online). Cardiac death was higher in the CCTA arm [6 (5.0%)] than ICA [1 (0.8%), P = 0.0499] but MI and ACS trended lower in the CCTA arm than the ICA arm [1 (0.8%) and 4 (4.2%) patients, respectively]. Of the six cardiac deaths in the CCTA arm, four patients had been diagnosed with non-ischaemic HF, and two of the deaths were attributed to non-ischaemic HF. The one death in the ICA arm was attributed to non-ischaemic HF.

Figure 2

Kaplan–Meier survival curves (days) of coronary CT angiography (blue) and invasive coronary angiography (red) patients for the clinical composite endpoint [cardiac death, MI, cardiac arrest, cardiac hospitalization (HF, ACS, arrhythmia), and procedural complication].

Open in new tabDownload slide

CCTA accuracy

In the CCTA patients referred to ICA, the sensitivity, specificity, positive predictive, and negative value for detecting obstructive and high-risk CAD were 92%, 13%, 63%, and 50%, respectively and 100%, 54%, 45%, and 100%, respectively. Adjusting for verification bias, the adjusted specificity was 86%.

Discussion

The IMAGE-HF (Imaging Modalities to Assist with Guiding therapy and the Evaluation of patients with Heart Failure) consortium begins to examine the utility of non-invasive techniques in the HF population.¹² To our knowledge, this is the first randomized controlled trial examining the clinical utility of CCTA in HF patients. The primary results of our study demonstrate that the financial impact of a CCTA strategy, in patients with HF of unknown aetiology, was not statistically different from an ICA strategy. The lower initial diagnostic costs of CCTA were partially offset by higher downstream costs.

It is important to acknowledge that our findings are specific to the HF population since these patients are more likely to require have higher resource utilization which dilute any potential initial cost-savings. Therefore, these results cannot be applied to other populations (e.g. suspected or stable CAD patients).

The financial burden of HF is partially driven by the ageing population, growing prevalence of HF, improved survival of HF patients, and increasing access to newer technologies.^1,2 There is a growing need for cost containment, thus, identifying strategies that are accurate, safe and cost-effective is important.

Operating characteristics and accuracy of CCTA

Two studies have examined the accuracy of CCTA in HF patients and reported sensitivities, specificities, positive, and negative predictive values of 73–98%, 99–100%, 92–99%, and 97–100%, respectively.^9,10 Our accuracy results were subject to referral and verification bias, however, support the notion that CCTA is a sensitive test reduced the need for ICA in 75% of CCTA patients.

CT strategies in HF patients have been studied. Abunassar et al.¹³ showed that an Agatston score = 0 effectively ruled out high-risk CAD and ischaemic aetiology for left ventricular dysfunction. Premaratne et al.²⁰ performed a systematic literature review and confirmed that an Agatston score = 0 was 98.4% specific for non-ischaemic cardiomyopathy. The ability of cardiac CT to function as a gatekeeper for HF patients was studied in 93 patients.²¹ A strategy using CCTA only for those with an Agatston score >0 had a sensitivity of 100%, specificity of 95%, positive predictive value of 67% and negative predictive value of 100% for detecting ischaemic HF. The cost implications of such strategies have not been studied.

In newly diagnosed HF patients, CAD should be considered²² and ICA should be considered for patients at intermediate to high pre-test probability for CAD, if ischaemia is a contributing factor or in the presence of refractory angina.^3,23 The American College of Cardiology considers the use of CCTA ‘appropriate’ for newly diagnosed systolic HF and may be appropriate in those with diagnosed diastolic HF.¹¹ The European Society of Cardiology gives a class IIb recommendation for the use of CCTA in HF patients with low-intermediate pre-test probability of CAD. The lower frequency of normal ICA in the CCTA group suggests that CCTA can function as a gatekeeper and potentially improves the diagnostic yield of ICA.

Clinical outcomes

The small sample size limits our ability to make conclusions about individual clinical outcomes. There were no differences in MACE and CCE between the two strategies.

Healthcare costs continues to be a focus of healthcare payers and is increasingly important in medical research. Our analysis focuses on the financial impact of CCTA compared to ICA—a full economic evaluation would also include a comparison of the health care outcomes of the two management strategies facilitating an assessment of the cost-effectiveness of adopting the CCTA algorithm. It is especially relevant when multiple diagnostic tests are available for the same indication. Although the ‘intention-to-treat’ CCTA strategy was not more costly than ICA, a secondary ‘as-treated’ analysis found greater cost savings with CCTA. Some would advocate that the ‘as-treated’ analysis better reflects clinical practice and that these results be considered. Others argue that the ‘as-treated’ analysis breaks the randomization within the clinical trial thereby introducing bias. In a sense, the ‘as-treated’ analysis takes a randomized controlled study and turns it into an uncontrolled observational study. In our study, it is likely that specific patient clinical characteristics led to greater crossovers from the CCTA to the ICA arm. Thus, reliance on the ‘as-treated’ analysis will lead to bias and false inference and cannot be considered an equally valid analysis to the randomized results.

Several randomized controlled trials have examined the utility of CCTA. Early trials examined CCTA in the emergency department and showed that it was associated with earlier discharges and had potential cost savings.²⁴ In patients with suspected stable angina, CCTA improved certainty of diagnosis and positively impacted upon downstream investigations and treatment.²⁵ PROMISE compared to functional testing to CCTA in stable in the patients with stable symptoms suspicious for CAD.²⁶ CT was not associated with improved patient outcomes but reduced the frequency of normal ICAs. CT-MAN showed that CCTA functioned as a gatekeeper to ICA, and reduced the length of hospital stay without increasing MACE.²⁷ Our study contributes to the literature by examining CCTA in the HF population.

Limitations

As with many studies, patient preference and physician willingness to enroll patients into trials is a potential source of population bias and impacts on the study population’s characteristics. The inherent willingness to enroll those who are less likely to have ischaemic cardiomyopathy may overestimate the utility and benefit of the CCTA arm. The lack of blinding may explain the high CCTA crossover rate, is likely a bias of treating physicians and might underestimate the true benefit of CCTA. However, these biases may reflect real-world clinical practice where ICA is reserved for those at higher CAD risk. Similarly, CCTA and ICA interpretation and subsequent patient management were determined by local clinical practice. Although this potentially increases heterogeneity in patient diagnosis and management, it would reflect real-world practice. Our study was not powered to detect differences in clinical outcomes.

Our study is conducted based on cost data from the Canadian setting applied to resource use data primarily from Canadian settings but inclusive of data from Finland. Thus, the generalizability of the findings of the financial impact analysis needs to be considered. Clinical practice in countries may differ but such differences can also exist within the same country. Unfortunately, this is a potential limitation of many multicentre trials including this study. In different clinical contexts, there may be greater willingness to change clinical practice based on financial implications and therefore the results should be interpreted specific to the context in which a decision might be made.

Non-invasive downstream testing was similar between groups; however, the indications for testing were not collected and thus could not be compared.

The objective of this study was to understand the potential costs directly and immediately associated with the initial diagnostic strategy. Although the authors believe that most cost implications would be captured within 12 months, it is possible that there are potential cost implications beyond this time. All expenditures were calculated as if there were performed in Canada in 2017. Acknowledging that institutions and payer systems may have different costs, our results should be calculated for each practice.

Conclusion

In patients with HF of unknown aetiology, costs and composite clinical outcomes were not statistically different between the CCTA and ICA strategies.

Supplementary data

Supplementary data are available at European Heart Journal - Cardiovascular Imaging online.

Acknowledgements

The authors extend their gratitude to the IMAGE HF organizing committee and principal investigators (B.J.W.C., L.M., E.O’M., I. Paterson, K. Chan, D. Birnie, J. White, and F. Prato) for their guidance.

Funding

B.J.W.C. is PI of the CTA-HF trial. He held the Saul and Edna Goldfarb Research Chair in Cardiac Imaging. R.S.B. is PI of the IMAGE HF team grant. He is a Career Investigator supported by the Heart and Stroke Foundation of Ontario (HSFO) and Tier 1 Research Chair supported by the University of Ottawa, and UOHI Vered Chair in Cardiology. L.M. is a phase 2 clinician-scientist supported by the HSFO. E.O’M. is sponsored by les Fonds de Rercherche en Santé du Québec (Junior 2 researcher grant 2011–2015) for research on biomarkers and imaging in HF and chronic kidney disease. K.A.C. is also supported by a New Investigator Award from CIHR and an Early Researcher award from the Ministry of Ontario. This study was supported in part by the IMAGE HF CIHR Team Grant (Team grant #CIF 99470), GE provided a small unrestricted grant (<10k allocated to this project) that was used to help support a research co-ordinator.

Conflict of interest: B.J.W.C. receives research support from CV Diagnostix, Ausculsciences and educational support from TeraRecon Inc and has equity interest in GE. E.O’M. receives research funding from Johnson & Johnson. J.K. is a consultant for AstraZeneca and GE Healthcare. R.S.B. is a consultant for: Lantheus Medical Imaging, Jubilant DraxImage, GE, and he receives research funding from: Lantheus Medical Imaging, Jubilant DraxImage, GE. All other authors have declared no conflict of interest.

References