Results for aim B: cohort characteristics, exposure and outcomes, and resource use/cost-effectiveness

Parts of this text have been reproduced from Brierley et al.27 © Article author(s) (or their employer(s) unless otherwise stated in the text of the article) 2018. All rights reserved. No commercial use is permitted unless otherwise expressly granted. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Aim B: to estimate the proportion of the target population who have cardiovascular magnetic resonance at two centres with dedicated cardiovascular magnetic resonance facilities and describe resource use and associated costs of having cardiovascular magnetic resonance after primary percutaneous coronary intervention

• Objective B1: to estimate the proportion of the target population who undergo a CMR following PPCI pathway activation.
• Objective B2: to describe resource use and associated costs of having CMR after PPCI pathway activation, including cost-effectiveness models.

In Chapter 4, we reported on the feasibility aspects of the study with respect to identifying the eligible population, implementing consent, and data linkage and extraction across sites and with hospital episode data. In this chapter, we report on the quantitative characteristics of the cohort with respect to index admission (baseline) data, follow-up data and the health economic analysis we conducted for this study.

A series of economic decision models were developed to identify the key drivers of cost-effectiveness. As uncertainty surrounds the performance of CMR imaging in practice and its benefit to patients in terms of more optimal treatment, a flexible framework was required to facilitate the assessment of uncertainty around parameter estimates. Models are a simplification of the real world and will never capture every possible scenario associated with an intervention, but the key consequences need to be included when models are conceptualised and structured.64 It is helpful if there are clinical events driving the models to define branches in the decision trees and associated event probabilities. The models were developed in consultation with clinical experts on the study team.

Two subgroups of patients who activated the PPCI pathway were chosen for the economic decision models: (1) patients with multivessel disease and (2) patients with unobstructed coronary arteries. Both subgroups were identified by formal consensus (see Chapter 6, Patient subgroups) as potentially benefiting from CMR. Patients with multivessel disease are the largest subgroup of patients in the population that activates the PPCI pathway (40–60%), and there is clinical uncertainty about how to treat this subgroup. Patients with normal coronary arteries are a distinct group which, at the outset, the research team expected to benefit most from CMR.

Characteristics of consented versus non-consented patients

Table 14 shows the baseline characteristics of consented and non-consented patients. The IMD and WIMD scores were higher among those who consented than those who did not (SMDs 0.31 and 0.27, respectively). The percentage of patients who underwent PPCI was also higher among those who consented than those who did not (90% vs. 65%, respectively; SMD 0.70). This imbalance arose primarily from hospital A.

Characteristics of participants by hospital

Table 15 shows the characteristics of participants by hospital. Participants from the four hospitals were similar with respect to demographics, comorbidities and previous cardiac interventions, although rates of hypertension were lower in patients from hospital B than in patients from hospitals A, C and D (18% vs. 41–52%).

A CMR scan was performed (within 10 weeks of the index admission) for 20% of consented patients in one CMR centre and for 14% of consented patients in the second CMR centre. CMR data from hospital A were submitted as free-text reports; individual CMR parameters were difficult to extract electronically from these. Hospital B extracted CMR parameters of interest from the free-text reports manually. The baseline characteristics of patients who did and patients who did not receive CMR are shown in Table 16. Patients referred for CMR were slightly younger than those who did not have CMR. No other differences were noted.

Participant follow-up

Table 17 shows the cardiac-related events in the 12 months following the index admission in the entire cohort. Overall, 55 out of 1670 (3%) patients died during follow-up (median age 73 years, IQR 66–82 years). MACE rate in this cohort was 13% (219/1670).

The recorded causes of death for patients who died within 1 year of admission are shown in Appendix 6. Only 32 out of 54 (59%) patients died from causes closely associated with ischaemic heart disease; an additional 2 (4%) died from other heart conditions and 8 out of 54 (15%) patients died of cancer.

Cost-effectiveness models

Multivessel disease

The expected costs and QALYs gained under each of the ischaemia testing options are shown in Table 18 together with the cost-effectiveness results for CMR and pressure wire, each compared with stress ECHO. Comparisons are made with stress ECHO because this was the ischaemia test in use at the time of the study (and is considered the ‘existing’ comparator) and also because CMR and pressure wire were assumed to be equally effective; these strategies differ only in terms of their costs. Based on the point estimates, CMR was slightly less costly and more effective than stress ECHO and, therefore, dominated stress ECHO. Pressure wire was more costly and more effective than stress ECHO: the ICER is considerably higher than the accepted cost-effectiveness threshold of £20,000–30,000 per QALY, and, therefore, would not be considered cost-effective relative to stress ECHO. Pressure wire is convenient for cardiologists if the test identifies the fact that a patient needs revascularisation, as the patient can be treated immediately while in the catheter laboratory, and the cost of the pressure wire in addition to the revascularisation is modest. In the base-case analysis, 35% of patients having ischaemia testing were assumed to truly have ischaemia; the majority of patients are, therefore, having an expensive test without needing further treatment. If a greater proportion of patients undergoing ischaemia testing required revascularisation, it is likely that the estimated cost of a pressure-wire strategy would reduce. It would be possible to use a pressure wire on other vessels at the time of the index angiography and PCI, but it was beyond the scope of this work to consider the likely costs and consequences of such a strategy. Based on the point estimates, pressure wire is dominated by CMR, because CMR is less costly and equally as effective. Cost and QALY differences are small relative to the mean values under each strategy, and none of the differences in costs or QALYs is close to being statistically significant; indeed, the differences in QALYs were tiny. There is considerable uncertainty around these base-case results; however, as is clear from the plots of 1000 simulated cost and QALY differences from the PSA for CMR and pressure wire (each compared with stress ECHO) in Figures 10 and 11. In each figure, there are a large number of points in two or three quadrants of the cost-effectiveness plane. The black dot is the point estimate of the cost and QALY difference, and, in each figure, it is close to the origin.

FIGURE 10 A plot of the 1000 simulated cost and QALY differences from the PSA for CMR vs. stress ECHO. Reproduced with permission from Stokes et al.36 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/. (Figure not reproduced here; see the original article.)

FIGURE 11 A plot of the 1000 simulated cost and QALY differences from the PSA for pressure wire vs. stress ECHO. Reproduced with permission from Stokes et al.36 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/. (Figure not reproduced here; see the original article.)

For this feasibility study, it is not the cost-effectiveness results per se that are of interest, but rather the way in which they change in response to sensitivity analyses that helps us to determine the key drivers of cost-effectiveness. Table 19 reports the results of the sensitivity analyses around this model. Sensitivity analysis one, around the probabilities of ischaemia testing or a decision to revascularise a second vessel (or not) directly from the index angiography, affects the mean costs and QALYs under each strategy but does not alter the cost-effectiveness results, because these probabilities are the same under each strategy. Sensitivity analyses two and three, around the diagnostic accuracy of CMR and pressure wire, had the most effect on results. When the sensitivity and specificity of CMR were reduced to 80%, CMR became more costly than stress ECHO, rather than dominant, and, at accepted cost-effectiveness thresholds, would not be considered cost-effective relative to stress ECHO. When the sensitivity and specificity of pressure wire were reduced to 80%, conclusions did not alter: pressure wire was still more costly and slightly more effective than stress ECHO, but the ICER quadrupled. The diagnostic accuracy of the ischaemia tests, in particular CMR, are key drivers of cost-effectiveness. Varying the costs of CMR, angiography and pressure wire, and varying QALYs for patients with MACE and those without MACE, did cause cost-effectiveness results to vary, but did not alter cost-effectiveness conclusions.

An alternative way of assessing the sensitivity of results to the cost of an ischaemia test is to consider how the cost of the test would need to alter for the costs of alternative strategies to be estimated to be equal. In the base-case analyses, the cost of CMR was £264, and a CMR strategy was less costly than a stress ECHO strategy. Threshold analyses around the cost of CMR identified that if the cost rose to £371, there was no difference in the costs between a CMR and a stress ECHO strategy. In the base-case analyses, the cost of angiography and pressure wire was £1340, and a pressure-wire strategy was more expensive than a CMR or stress ECHO strategy. Threshold analyses identified that the cost of angiography and pressure wire would need to reduce to just £250 for there to be no difference in costs between a CMR and a pressure-wire strategy, and to £415 for there to be no difference between the cost of a stress ECHO and a pressure-wire strategy.

Normal (unobstructed) coronary arteries

The expected costs, QALYs and cost-effectiveness results under standard ECHO and CMR compared with standard ECHO alone for patients with normal coronaries are shown in Table 20. The reduction in costs associated with treating fewer patients for MI with the introduction of CMR testing compensated only partially for the additional cost of the CMR test. Mean costs and QALYs gained were higher with standard ECHO and CMR than with standard ECHO alone, and the ICER would not be considered cost-effective at accepted thresholds. There is considerable uncertainty around these base-case results, however, as is clear from the plot of the simulated cost and QALY differences from the PSA in Figure 12. There are a large number of points in three quadrants of the cost-effectiveness plane.

FIGURE 12 A plot of the 1000 simulated cost and QALY differences from the PSA for standard ECHO and CMR vs. standard ECHO. Reproduced with permission from Stokes et al.36 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/. (Figure not reproduced here; see the original article.)

Table 21 reports the results of the sensitivity analyses around this model, designed to help identify the key drivers of cost-effectiveness. Sensitivity analysis one (reducing the diagnostic accuracy of CMR) and sensitivity analysis six (increasing the cost of CMR) had the greatest impact on the mean cost difference between standard ECHO and CMR and standard ECHO alone. Threshold analyses around the cost of CMR (£264 in the base-case analyses) identified that, if the cost reduced to £166, there was no difference in the costs between standard ECHO and CMR and a standard ECHO alone strategy. Sensitivity analyses 8–10, which all directly varied QALYs, had the greatest impact on the difference in QALYs between the arms. However, in all cases, standard ECHO and CMR remained both more costly and more effective than standard ECHO alone, and in none of the sensitivity analyses would standard ECHO and CMR be considered cost-effective at accepted thresholds. These sensitivity analyses suggest that the diagnostic accuracy of CMR is again a key driver of cost-effectiveness. Results are also influenced by the cost of CMR, and the difference between utility values assigned to MACE and no MACE, and the proportion of patients who have MACE, who are assumed to die.

Summary of main findings

Cohort characteristics and follow-up

Baseline characteristics were reasonably well balanced between participants who consented and those who did not consent in the study, but suggested that patients living in less deprived residential areas were, on average, less willing to take part. Characteristics were better balanced across participants recruited by the four hospitals, and participants who did and those who did not receive a CMR scan. Data completeness was good for most variables. Follow-up data were available for the entire follow-up period (12 months) for only 729 out of 1035 (70%) participants in England and 287 out of 635 (45%) participants in Wales. This is because data were requested soon after recruitment finished and the Welsh data were provided rapidly. The data for English hospitals took much longer to be released, resulting in a longer follow-up period. The limited follow-up was a result of this process being part of the testing feasibility for a full registry; in the context of a registry, the request for these data would be made at an appropriate stage of the study to achieve the desired duration of follow-up.

The proportion of the target population that gets cardiovascular magnetic resonance after activating the primary percutaneous coronary intervention pathway

Although > 10% of patients in hospitals with a dedicated CMR facility were referred for CMR, many of these referrals were for research (patient participating in a research study) rather than clinical purposes. It is not clear whether or not this reflects practice in other hospitals with dedicated CMR facilities. Referral for non-clinical reasons (unless the findings are used to change clinical management) would dilute any observed effect of CMR on outcomes in an extended registry. It was not possible to determine how many CMRs that were done for research purposes were subsequently reviewed by the treating clinician and resulted in changes in patient management.

Cost-effectiveness models

For each of the base-case models, the differences in QALYs between strategies were tiny; therefore, the results are driven largely by the differences in costs, although these were also modest. Sensitivity analyses around the two models identified the diagnostic accuracy of the ischaemia tests as the key driver of cost-effectiveness. In the model for patients with multivessel disease, the costs of ischaemia testing and the QALYs associated with MACE and no MACE exerted an influence on cost-effectiveness results, but not to the same extent as the diagnostic accuracy of CMR and pressure wire. Similarly, in the model for patients with normal coronaries, the cost of CMR, the difference between QALYs associated with MACE and no MACE, and the proportion of patients who have MACE who are assumed to die, exerted an influence on the results.