Assessment of cost-effectiveness and utility
In the past years, cost considerations have become increasingly relevant in clinical decision making[
38]. Cost-effectiveness analysis can help decision-makers to allocate limited resources and can guide the utilization of latest-generation and presumably high-cost imaging modalities[
39].
One of the key requirements of cost-effectiveness analysis is the identification of all relevant costs. If the analysis is carried out from the viewpoint of a health care payer, procedure costs usually represent charges or reimbursement fees[
40]. The German health care system is characterized by the presence of public and private health insurers. Within the public system, outpatient health care services such as SPECT are to be charged according to the EBM system[
30]. In the absence of a specific EBM procedure code for CMR, we chose to use the GOÄ fee schedule for privately insured patients. This is common practice in Germany, as the GOÄ fees much better reflect the actual costs incurred by CMR[
31,
34]. As a result, CMR was associated with nearly 40% higher costs than SPECT (Table
1). Importantly, this cost ratio is very similar to cost ratios described for the United States or the United Kingdom. Because our model depends on relative rather than on absolute costs, the results of the present study can be assumed to be generally valid for other health care systems as well[
18]. As the study takes a health care payer’s and not a societal perspective, other costs such as lost productivity due to missed days at work (indirect costs) were not included[
21,
40]. Rather than assessing the impact of diagnostic tests on the overall welfare of society, the goal of the study was to compare the cost-effectiveness and utility of CMR and SPECT to achieve the same objective, i.e. diagnosing CAD and thereby improving the clinical outcome[
20,
21]. The outcome variable was limited to a 10-year follow-up underlining the conservative nature of our analysis. If the effects of CAD therapy would have been simulated beyond the 10-year horizon, than the impact on outcome (ΔQALY) might have been even more favourable than indicated by our results.
The diagnostic accuracy of CMR to detect CAD was taken from the recently published CE-MARC study which prospectively evaluated the role of CMR in patients with suspected CAD[
17]. By comparison, CMR delivered a higher sensitivity and specificity than SPECT[
17]. Another large study, MR-IMPACT II (Magnetic Resonance Imaging for Myocardial Perfusion Assessment in Coronary artery disease Trial II), also detailed the accuracy of CMR compared with SPECT[
25,
41]. MR-IMPACT II confirmed CMR’s position as alternative to SPECT with higher sensitivity to detect CAD. Unlike CE-MARC, the specificity of CMR was inferior to SPECT in MR-IMPACT II[
25,
41]. However, CE-MARC had a more rigorous study design and included a larger patient population[
17,
42]. Moreover, CE-MARC used the full potential of CMR, including perfusion imaging, late gadolinium enhancement, left ventricular cine imaging, and non-invasive coronary angiography while MR-IMPACT II focused solely on perfusion abnormalities[
17,
25,
43]. Because coronary angiography by CMR cannot be regarded as standard part of routine examinations and because it is not feasible in all patients, it was not incorporated in our analysis (Table
1). This exclusion led to a slightly diminished diagnostic accuracy of CMR, again reflecting the conservative estimates used in our study[
17]. It is noteworthy, that future imaging protocols may even further increase the diagnostic accuracy of CMR[
44]. Dobutamine stress CMR, while clinically valuable, was not part of our analysis[
45]. For consistency and comparability, data on diagnostic accuracy of SPECT were also derived from the CE-MARC study[
17]. Although other studies have reported different and wide varying diagnostic accuracies of SPECT, the superior sensitivity of CMR in comparison to SPECT seems to be a common finding[
17,
25,
41,
46]. With respect to cost-effectiveness and utility, our results clearly indicate that a high sensitivity is more important than a high specificity (Figure
3A and B).
Although the cost per patient tested increased linearly along the prevalence of CAD (Figure
2A), both cost-effectiveness (cost per effect) and utility (cost per utility unit) showed a hyperbolic decrease in costs (Figure
2B and
2C). The observation that the effectiveness as well as the utility criterion yielded concordant results supports the validity of our findings. The decrease in costs is due to the fact that the underlying mathematical model defines a patient accurately as having CAD as the effect and an increase in QALYs (ΔQALY) as utility[
19‐
21]. Both effect and utility become more frequent with an increase in CAD prevalence[
19‐
21]. Specifically, the effect (i.e., the diagnosis of CAD) was based on functional testing (CMR and SPECT)[
17,
25]. In contrast, invasive coronary angiography relies on morphological criteria and direct visualization of the coronary arteries and may therefore be an imperfect standard of reference. Obviously, invasive fractional flow reserve is the most accurate parameter to assess the functional relevance of a stenosis, but those data were not available from CE-MARC or other large comparative studies of CMR and SPECT[
17,
47]. However, as coronary stenoses not causing ischemia may be judged as significant on the basis of angiographic severity alone, the actual diagnostic accuracy and therefore the cost-effectiveness of CMR is probably even better than simulated by our model.
At low CAD prevalences, both non-invasive tests were more cost-effective than invasive coronary angiography (Figure
2A and
2B). This is because the majority of negative CMR and SPECT examinations will correctly rule out significant CAD at low disease prevalences and will therefore reduce the number of invasive angiographies[
18]. At higher disease prevalences, though, both non-invasive tests (SPECT > CMR) start to miss patients who actually have CAD. The consequences of such false-negative test results decrease cost-effectiveness and utility because they may prevent patients having CAD from receiving adequate treatment. The lack of treatment may lead to increased mortality, decreased quality of life, and to additional costs related to complications of CAD (i.e., treatment of myocardial infarction). False-positive test results also lead to decreased cost-effectiveness and utility, mainly due to overtreatment and unnecessary invasive coronary angiographies.
In parallel, invasive coronary angiography as the initial test becomes more competitive in terms of cost per effect and cost per utility. Sensitivity analyses indicated that above a threshold value of CAD prevalence of 0.60, performing invasive angiography was the most cost-effective strategy. This threshold is in line with previous cost-effectiveness analyses examining non-invasive strategies to detect CAD[
19‐
21,
48,
49]. In addition, recent guidelines recommend invasive angiography as the most cost-effective first test if the pretest probability of CAD is >61%[
50].
Our study has some limitations. Firstly, our model necessarily simplifies some aspects of the underlying clinical reality and does not account for all complications associated with CMR (e.g. gadolinium-associated nephrogenic systemic fibrosis) or SPECT (e.g. radiation-induced malignancies)[
23,
49]. Secondly, diagnostic accuracy data were derived from studies with an intermediate prevalence of CAD. Extrapolation of these data to populations with high or low disease prevalences should be judged with caution. Thirdly, the analysis of further imaging modalities (e.g. computed tomography or stress echocardiography) was beyond the scope of the present study. Furthermore, coronary revascularization was not within the scope of the current analysis.