CaRi-Heart for predicting cardiac risk in suspected coronary artery disease: early value assessment

Considerations for HTE4

In the original assessment, the external assessment group (EAG) found 1 study that assessed the prognostic performance of CaRi‑Heart for predicting cardiac death for people with suspected stable CAD (Oikonomou et al. 2021). The study was a development and validation study in a German dataset, which included 3,912 people having CTCA to assess stable CAD.

The results of the study showed that it was better at predicting risk than a risk model based on traditional clinical risk factors. These include smoking, hypercholesterolaemia, hypertension, diabetes, Duke Activity Status Index, presence of high-risk plaque features, and epicardial adipose tissue volume. The committee agreed that, based on the results of Oikonomou et al. (2021), CaRi‑Heart was likely to improve risk prediction for cardiac death.

The comparator used in the CaRi‑Heart study did not reflect UK clinical practice, which limits the generalisability of the study. The committee said that the comparator for future studies should include assessing CTCA images alongside clinical risk factors, and that QRISK3 should be used in the 'no CAD' group. It recommended external validation of CaRi‑Heart in a UK setting and that data should be collected to demonstrate prognostic performance in people who do not have CAD identified on CTCA.

In the original assessment some data was presented that suggests the prognostic performance of CaRi‑Heart is consistent by groups including age, sex, CAD status (obstructive and non-obstructive) and ethnicity. The committee said that ongoing validation of CaRi‑Heart should include groups by sex, age, ethnicity, socioeconomic status and CAD status if possible.

New evidence

A prospective nested cohort of 3,393 people in the UK evaluated the prognostic value of FAI‑Score and the performance of the AI‑Risk algorithm (Chan et al. 2024a). CaRi‑Heart analyses CCTA scans then generates the FAI‑Score and uses an AI‑Risk algorithm to assess the risk for each patient. The FAI‑Score provides an estimate of the inflammation in the coronary arteries. The AI‑Risk algorithm is an AI-assisted prognostic algorithm that incorporates the FAI‑Score of 3 coronary arteries, the coronary atherosclerotic burden and traditional risk factors. The AI‑Risk algorithm classifies people into very high risk (≥10% 8‑year risk for fatal cardiac events), high risk (5% to <10%), and low or medium risk (<5%) categories.

In the study the AI-Risk algorithm output was analysed as a continuous variable and as a categorical variable. Both outputs predicted cardiac death and MACE in the overall population, as well as in people without obstructive CAD. The algorithm appeared to overestimate risk in people with obstructive CAD.

The company submitted an unpublished abstract with results on the performance of the AI‑Risk algorithm in people from different demographic and socioeconomic backgrounds (Chan et al. 2024b). The results are confidential and so cannot be reported here.

Considerations for HTE4

The EAG found no evidence on how CaRi‑Heart analysis changes risk assessment or clinical decision making for people with suspected CAD who have a CTCA. The Oikonomou study presented data on how risk groups (low, medium and high risk) changed when a CaRi‑Heart score was used, compared with a clinical risk score. But the clinical experts said that the clinical risk score used in the study was not used in UK clinical practice. So, it was still uncertain how using CaRi‑Heart would affect the outcome of a risk assessment compared with CTCA. The committee concluded that it was currently uncertain how CaRi‑Heart influences risk assessment but that the ongoing study would likely address this.

The committee discussed the treatments available for people identified with no CAD (low risk), non-obstructive CAD (medium risk) and obstructive CAD (high risk) after a CTCA and how these might change with the introduction of CaRi‑Heart. It noted that the company's suggestions included starting statins in the low-risk group, increasing the intensity of statins in the medium-risk group, and introducing other anti-inflammatory medicines such as colchicine in the high-risk group.

For the low-risk group, the clinical experts said there is good evidence on the effectiveness of starting and intensifying statins. So, it said that treatment strategies for people with no CAD or non-obstructive CAD may be clearer. For the high-risk group, colchicine is not licensed or recommended in the UK for this indication. The clinical experts discussed treatments that may be more widely used in the future, such as PCSK9 inhibitors and inclisiran. But they said that higher quality evidence is needed to show that these treatments could reduce cardiac events and mortality in this population, because they were expensive.

The committee concluded that further evidence is needed on how CaRi‑Heart affects clinical decision making and clinical management compared with UK standard clinical practice (CTCA alongside clinical risk assessment). It noted that QRISK3 should be included as a comparator for people who have no CAD identified on CTCA.

New evidence

A prospective real-world evaluation exercise was done to understand the impact of CaRi-Heart reports on clinical decision making in the UK (Chan et al. 2024a). This was done in 4 NHS trusts:

Considerations for HTE4

During development of HTE4, no evidence was found on how CaRi‑Heart affects patient outcomes such as cardiac mortality and morbidity. The EAG found no studies on targeting treatments using any measure of coronary inflammation. The committee noted that there is evidence that shows statins may benefit everyone, regardless of CAD status. Because changes to the guidelines mean that statins can be prescribed more widely, it is expected that this would impact the extent to which CaRi‑Heart influences treatment options.

The committee recommended further research to address the uncertainty around clinical outcomes for people with suspected CAD having CTCA for chest pain who have had CaRi‑Heart testing. The committee said that a clinical outcome study using CaRi‑Heart to determine a treatment strategy, with people followed up for long enough to observe a reduction in cardiac events or death would be ideal. But because this may be difficult it agreed that a linked-evidence approach would be acceptable.

New evidence

There is no direct evidence on how cardiac risk prediction and changes in clinical decision making following a CaRi‑Heart test impact on clinical outcomes. The company used a linked-evidence approach to develop a cost-effectiveness model.

Considerations for HTE4

No evidence was identified on the costs or cost effectiveness of CaRi‑Heart. The committee heard that there was no evidence on how CaRi‑Heart might affect resource use because of changes in treatments or the potential reduction in cardiac events. The committee recommended that data is collected on the costs associated with using CaRi‑Heart, including implementation costs, training costs, and impact on costs and resource use later in the treatment pathway.

New evidence

A new study reports a decision analytic model that compares AI-Risk assessment with CaRi‑Heart plus standard care versus standard care alone over a lifetime time horizon (Tsiachristas et al. 2025). The model used data from:

The meta-analysis results were used to inform the effect of statins in people in different risk categories. It is uncertain whether the effect of statins would be the same in people identified:

The base-case economic analysis uses costs for CaRi‑Heart of £300, £500 and £700 per scan. Implementing CaRi‑Heart in routine clinical practice and assuming full compliance with treatment reduced the number of cardiovascular events when modelled over a lifetime horizon. The model results suggested an 11% reduction in myocardial infarction, 4% reduction in stroke, 4% reduction in heart failure events and 12% reduction in cardiac mortality when using CaRi‑Heart. These reductions resulted in an increase in quality-adjusted life years (QALYs) of 0.21. The incremental cost-effectiveness ratios (ICERs) are presented in table 1.

Abbreviations: CI, confidence interval; ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life year.

To test the robustness of the results, 5 different scenarios analyses were done:

The sensitivity analyses gave ICERs ranging from £1,837 to £6,592 per QALY gained, assuming that the test cost £700.

In 2023 to 2024, during which NG238 was updated, almost 900,000 more people were given a NICE-recommended statin or ezetimibe by their GP to help reduce their cholesterol than in 2022 to 2023. This may result in increased ICERs for CaRi‑Heart because of a reduced clinical benefit compared with current practice.

The budget impact of implementing the AI‑Risk model (at £700 per scan) was estimated over 5 years, assuming annual changes in uptake of AI‑Risk in the NHS at 2%, 5%, 10%, 15% and 20% (see table 2).