The diagnostic performance of CGM to detect patients with physiologically significant stable CAD is poor at rest. Although, the diagnostic performance of CGM improves with the administration of intravenous adenosine, it does not reach sufficient levels of accuracy to safely exclude patients without physiologically significant stable CAD.
One of the main objectives of this study was to see if CGM had adequate sensitivity to safely stratify patients who required FFR assessment, avoiding the additional cost of the pressure wires and increased procedural risks. However, even during maximal hyperaemia the sensitivity of CGM was not high enough to be relied upon. For every hundred patients with physiologically significant stable CAD tested, CGM would not detect the presence of stable CAD in twenty-nine of those patients.
The reported figures for the measures of diagnostic performance of CGM in this study are considerably worse than previously published work which used coronary angiography to define CAD.7,13,14,20 In previously published CGM studies, CGM demonstrated greater sensitivity (64–72%) and specificity (60–82%) when ≥ 50% DS was used to define significant CAD then seen in our results. Furthermore, previous CGM studies which used ≥ 70% DS to define significant CAD, also reported a higher sensitivity (75–84%) and specificity (74–81%) than results seen in our study.13,14 The difference in diagnostic performance of CGM in our study compared to previously published work may be because the pre-test probability of having angiographic CAD was greater in our study, as the patients were known to have angiographic evidence of disease prior to being enrolled.
One major issue of the automated algorithm used by the CGM device is that when it was originally validated, it was based on data from a study where patients were classified as having stable CAD based only on coronary angiography (defined as ≥ 50% diameter stenosis).13 This is a fundamental flaw of the device algorithm, as % diameter stenosis is not a reliable method of detecting significant lesions causing ischaemia.17 This may explain the low specificity seen in our study, as the cardiogoniometric variables it classes as being present in patients with stable CAD are not sufficiently specific for physiologically significant coronary disease. Nevertheless, it should be stated that CGM was first developed before the routine use of FFR. Therefore, it is not unreasonable that its initial development was based on the ability of CGM to detect angiographically significant coronary stenoses, and not FFR significant stenosis.
All of the previously published work did not limit study participants to having single vessel disease and this may also contribute to explaining the observed differences in results. The diagnostic performance of CGM may be greater in patients with multi-vessel disease as they may have a greater ischaemic burden. Furthermore, patients with multi-vessel disease tend to have worse long term outcome, increased procedural risk and significant comorbidities.4 Nevertheless, this is the only study investigating CGM using FFR, a robust method of identifying physiologically significant myocardial ischaemia.
The diagnostic performance of CGM in this study was considerably worse than other methods of non-invasive physiological assessment of coronary ischaemia, including stress echocardiography and myocardial perfusion imaging (MPI).6,12 However, when these methods of physiological assessment have been assessed using FFR as the gold standard, they have been shown to have poor agreement with FFR to detect physiologically significant stable CAD.8,9 Reported figures in the literature of the sensitivity and specificity of stress echocardiography at identifying FFR-defined CAD are 50 and 90% respectively.8 MPI with SPECT has also been shown to perform poorly, with a sensitivity and specificity of 76 and 38% at identifying FFR-defined CAD.9 Of note, MPI with SPECT has been shown to overestimate FFR-defined ischaemia in 22% of participants and underestimate it in 36% of participants. An argument could be made that it is inappropriate to critique SPECT against FFR, as FFR was first validated against SPECT.11 However, since its validation, FFR based decision making has been shown to improve clinical endpoints and as such, can be taken as a reliable and robust method for identifying significant coronary stenoses.17 It is therefore difficult to criticise the diagnostic performance of CGM on the basis of other non-invasive methods of physiological ischaemia assessment.
It has previously been proposed that the diagnostic performance of CGM could be driven by detection of myocardial scarring, as opposed to chronic reversible ischaemia.1,19 It would be expected that the specificity of CGM would be reduced by myocardial scarring. However the results of this study have demonstrated that there were no significant differences in the specificity or sensitivity of CGM when patients with previous MI are excluded (Appendix 1). This correlates with previously published work, and we can therefore be confident that myocardial scarring does not have a big influence on the diagnostic performance of CGM.19
It would have been interesting to see if other stressing agents like dobutamine, may have increased the sensitivity of CGM. Dobutamine, a β1 adrenergic receptor agonist, acts by directly raising the metabolic demands of the myocardium by increasing heart rate and the force of cardiac contractility. Whereas the mechanism by which adenosine may induce ischaemia is less defined. It is likely indirect, and may be due to its ability to produce a coronary steal effect.2 As patients were already undergoing pharmacological stress with intravenous adenosine as part of their clinical care, it was felt to be unethical to subject them to an additional CGM stress test and induce the unpleasant side effects associated with dobutamine (nausea, headaches and dyspnoea). One previous study reports similar figures seen in this study for the sensitivity and specificity of CGM during adenosine stress.19 However, the authors of this study used SPECT to confirm the presence of stable CAD, which as previously mentioned is not a robust method of quantifying physiological significance. Interestingly, this previous study also showed a reduction in the specificity of CGM, when patients underwent pharmacological stress with intravenous adenosine.
CARDIOFLOW was designed to be as clinically applicable as possible, therefore the interpretation of the CGM result was based solely on the automated ischaemia score alone, and not by review of the raw data by an experienced CGM reporter. The idea being that, if CGM were to be implemented into routine clinical practice, a recording could be performed by an operator without detailed knowledge of CGM. Our patients are typical of routine clinical practice, a reflection of our consecutive recruitment of participants and reduction of the risk of selection bias. Furthermore, there was a wide range in both the length and severity of the lesions in the participants recruited in our study. This again mirrors the picture seen in routine clinical practice and increases the external validity of the study
This was a single centre study and only enrolled a relatively small number of participants. In addition to this, patients with multi-vessel disease, atrial fibrillation or previous CABG were excluded. Myocardial scarring was assumed in patients with previous MI and not formally assessed by performing cardiac MRI with late gadolinium enhancement; therefore, patients may have been incorrectly excluded from the subgroup analysis. Additional testing of participants with stress echocardiography and myocardial perfusion imaging would have allowed direct comparison between CGM and other methods of assessment, however this was not performed as it is not part of their routine clinical care. Finally, the operator performing CGM was not blind to the results of coronary angiography which could have influenced their interpretation. However, the result given by CGM is automated and dichotomous (i.e. categorically positive or negative) and therefore the impact of seeing coronary angiography would have been low.