Non-invasive testing in patients with suspected coronary artery disease: some may be more equal than others

This editorial refers to ‘The performance of non-invasive tests to rule-in and rule-out significant coronary artery stenosis in patients with stable angina: a meta-analysis focused on post-test disease probability’†, by J. Knuuti et al., on page 3322. Imaging is a growing and increasingly expensive field in cardiology.1 In patients with risk factors for the development of atherosclerosis or those with chest symptoms, it has become almost mandatory to do some kind of testing in order to rule-in or rule-out the presence of significant coronary artery stenosis.2 Often, as recommended in the guidelines,3–5 cardiac imaging is performed. The choice between tests is difficult. It depends on pre-test probability, patient characteristics, availability of techniques, local expertise, and other ill-defined factors. Is there a test which is better than all other tests? Which is the best test? Although the number of variables involved suggests that it is very difficult to provide a straight answer to such a question, it is asked over and over again. Obviously, there is a clinical need for objective advice. Several recent meta-analyses tried to provide answers.6–8 These meta-analyses often focused on coronary computed tomography angiography (CCTA). In this issue of the European Heart Journal, Knuuti et al.9 give us a broad look at the merits and the problems of imaging techniques employed to detect coronary artery disease (CAD) in patients presenting with stable angina pectoris. They also include the still widely used exercise ECG. Data were collected comparing non-invasive testing not only with invasive coronary angiography (ICA) but also with invasive fractional free flow (FFR) measurements. This also provided the opportunity to compare ICA with FFR measurements for the detection of relevant CAD. Instead of looking at sensitivities and specificities, they provide us with positive and negative likelihood ratios. These not so commonly used values can be determined in the following way: the positive likelihood ratio is the probability of a test being positive in a diseased patient divided by the probability that it is positive in a healthy person. In other words, it is the ratio between true-positive and false-positive tests. The negative likelihood ratio then is the probability of a test being negative in a person who has the disease divided by the probability that the test is negative in a healthy person. In other words, it is the ratio between the false negatives and true negatives. The likelihood of a person having the disease is determined by multiplying the pre-test probability by the positive likelihood ratio. On the other hand, the likelihood of a person not having the disease is determined by multiplying the pre-test probability by the negative likelihood ratio. Based on studies ranging from the late 1980s to now, the authors conclude that the stress ECG is not helpful either to rule-out or to rule-in CAD because both the positive and negative likelihood ratios are low. They confirm that CCTA is the ideal technique to exclude anatomic CAD accurately. For functional imaging, they recommend positron emission tomography (PET) and stress cardiac magnetic resonance (CMR) imaging because these two techniques demonstrate good performance with optimal application ranges for anatomical and functional CAD. The two imaging techniques most commonly used worldwide for functional ischaemia testing, stress echocardiography and nuclear myocardial perfusion imaging, do not perform as well in this meta-analysis. Potential reasons for this are that studies using these two techniques are older (usually from the 1990s) and that technical progress which has occurred over the past years thus is not reflected in the numbers. Surprisingly, ICA was the least effective technique for ruling out significant CAD when compared against the functional gold standard FFR. This confirms current European3 and US guideline4 recommendations to use ICA preferentially in patients with high pre-test probabilities. However, the data question the approach chosen in the guidelines not to give specific advice on which technique to use in which patient. Thus, after all, some techniques look a bit more equal than others. Although useful, this type of analysis faces three challenges. (i) In contrast to the updated NICE guidelines on the assessment and diagnosis of recent onset chest pain,5 the analysis by Knuuti et al. still pursues the quest of finding the optimal test for individual patients based on the pre-test probability of significant CAD. Many of these studies were performed and published at the end of the last century. Unavoidably, pre-test probabilities and prevalences of CAD were very different from the current situation. In the 1970s, Diamond and Forrester developed a table of pre-test probabilities depending on the type of symptom, the age, and the sex of the patient.10 These were the days when smoking, a sedentary lifestyle, and poor dietary habits were widespread. Since then the prevalence of CAD has dramatically declined,11 although the number of patients with chest sensations appears to be constant. The recent PROMISE study conducted in contemporary patients with symptoms suggestive of CAD found evidence of ischaemia in only 11.7% and of coronary narrowing in only 10.7% of patients despite a predicted prevalence of disease according to Diamond and Forrester of 53.3 ± 21.4%.12 It seems as if one could divide the Diamond Forrester values by 5 to reach realistic current pre-test probabilities. In such a population, reliable exclusion of disease is much more frequently needed and hence becomes more important than reliable detection of disease. As a consequence, the current version of the NICE guidelines on chest pain of recent onset has abandoned the concept of testing according to a certain pre-test probability.5 They now recommend CCTA in patients with anginal symptoms in whom stable angina cannot be excluded by clinical assessment alone. In patients with non-anginal chest pain, ST-T changes or Q-waves in the resting ECG are required as an indication for CCTA. Only patients with known CAD are recommended to have functional ischaemia testing if there is uncertainty as to whether chest pain is caused by myocardial ischaemia. (ii) Many of the studies available for meta-analyses probably suffered from referral bias.13 Such a bias is introduced by preferentially including such patients in studies comparing stress imaging with invasive angiography in whom CAD is suspected due to a pathological non-invasive test. In contrast, patients who have a normal stress test and appear not to have coronary disease would be less frequently included. This bias will result in overestimation of sensitivity and underestimation of specificity. One can avoid such bias by including consecutive patients in studies where ICA is mandatory before any form of testing has been performed.14 (iii) Is the focus on the presence of stenosing CAD still clinically relevant when it becomes increasingly apparent that revascularization offers little if any advantage over medical therapy?15 , 16 Do we spend our health moneys in the right area? Should we not rather focus on the detection of diffuse atherosclerotic disease which is not provided by stress imaging? CCTA will detect both coronary plaques and coronary stenoses. The presence of coronary plaques puts symptomatic patients at an increased risk for atherosclerotic events. More than 50% of patients with normal stress myocardial perfusion scans harbour coronary artery calcifications.17 Such patients are at higher risk than patients with completely normal coronary arteries,18 and yet specific preventive therapies are rarely instituted. By detecting subclinical atherosclerosis, CCTA may offer a better chance of initiating such therapies than stress imaging.18 , 19 Knuuti et al. remind us that not all tests are equal in patients with suspected CAD. They provide us with an easy to use tool to estimate post-test probabilities for a broad variety of tests. The declining prevalence of stenosing CAD, however, makes pathological test results difficult to interpret. Thus, the time may have come when we should not look back to studies of the past but design new ones to improve our current algorithms on how to approach patients with suspected CAD. Conflict of interest: U.S. reports that fees and honoraria are paid to his institution for his work as a consultant to Sanofi.