Article Text
PDF
Statistics from Altmetric.com
For nearly two decades, the National Institute for Health and Care Excellence (NICE) has been admired as a reliable and authoritative source for evidence-based practice pathways. Unfortunately, in the latest iteration of the stable chest pain guidelines, NICE has veered dangerously off course by advocating that we abandon Bayesian analysis and simply order CT coronary angiography (CTCA) in all patients with typical or atypical chest pain.1 The clinician should no longer consider a patient’s pretest probability, whether any imaging is necessary, and if so, which imaging test is best. These recommendations are not supported by high quality scientific evidence and put patients at risk for serious harm. At its foundation, the NICE guideline is antithetical to the current focus on a more personalised approach to medicine with conscientious use of diagnostic imaging, particularly when it involves ionising radiation. Moreover, given that cardiovascular events have dramatically decreased with improved risk factor modification,2 why should we advocate for more imaging and radiation exposure in patients at increasingly lower risk?
To support their radical changes to the diagnostic paradigm, the authors make several erroneous assertions. First, they contend that an assessment of the pretest probability of coronary artery disease (CAD) is useless and that exercise testing should not be performed given its limited diagnostic performance and the absence of prognostic value. In addition, they incorrectly assert that CTCA is more cost-effective and that trials of CTCA have consistently demonstrated its superiority. Given the implications, these assertions warrant critical appraisal (table 1).
Summary of arguments in support of and in opposition to the updated NICE stable chest pain guidelines
Should we disregard the pretest probability of CAD?
In their classic work,3 Diamond and Forrester stated that ‘the data presented… should not be considered as absolute standards but, rather, as preliminary estimates that will require modifications as more precise data become available.’ Indeed, several decades later, the systematic overestimation of CAD risk with the Diamond-Forrester approach is well recognised. With the decreasing yield of all imaging tests to evaluate for suspected CAD, a more selective approach to imaging is needed rather than the indiscriminate use of CTCA. Instead, the NICE guidelines take a defeatist perspective, claiming that we cannot effectively assess the probability of CAD. However, in the Coronary CT Angiography Evaluation for Clinical Outcomes (CONFIRM) registry, a model based on medical history was developed to serve as a ‘gatekeeper’ for CTCA.4 More recently, in a retrospective analysis from the Prospective Multicenter Imaging Study for Evaluation of Chest Pain (PROMISE) trial, a calculator was developed that could safely avoided testing in many of the patients.5
To further bolster their approach, Drs Timmis and Roobottom report that under-recognition of angina is common. Yet, this problem warrants more careful history taking and the broader use of validated chest pain instruments, not more imaging. They also report that non-anginal chest pain accounts for one third of cardiac events during follow-up.1 This finding highlights the need for improved primary prevention, as very few of these patients were taking statins and antihypertensives. Regardless, NICE is not yet advocating for CTCA in patients with non-anginal chest pain and normal ECGs, so their reference to these patients seems misplaced.
Should we abandon the exercise ECG test?
In disparaging the exercise ECG, Drs Timmis and Roobottom claim that the test has a poor diagnostic performance and no prognostic value. However, a decreased exercise capacity remains an important modifiable risk factor, and the exercise ECG provides a teachable moment, allowing clinicians to encourage anxious chest pain patients to resume exercise. With regard to diagnosis, the exercise ECG does not perform as well as imaging-based tests. However, in the meta-analysis discussed,1 variables that negatively impact test performance, such as upsloping ST depression, prior myocardial infarction, left bundle branch block, and pre-exercise hyperventilation, were included. Moreover, sensitivity of the exercise ECG for multivessel CAD does improve,6 though admittedly, the diagnostic performance is still inferior to imaging-based tests. Nevertheless, exposing all patients to imaging is misguided, as many of these patients have a low probability of CAD with extremely low-risk for morbidity or mortality.
More importantly, exercise capacity does effectively risk stratify patients. In their discussion of the lack of prognostic value of the exercise ECG,1 a minority of the patients actually had exercise data available, and the authors overemphasise the importance of the C statistic in risk prediction. For complex diseases, calibration is at least as important as the C statistic, and established cardiovascular risk factors may have little effect on the C statistic despite important effects on patient risk.7 In fact, in a study of over 100 000 patients that incorporated many cardiovascular comorbidities, exercise capacity remained the predominant risk factor.8
Is CTCA more cost-effective?
A major focus of the updated NICE guidelines, CTCA is touted as the most cost-effective approach. The important distinction, however, is that these analyses have looked primarily at alternative imaging-based strategies. Clearly, universal CTCA cannot be more cost-effective than a strategy that involves no testing in some patients. In addition, cost-effectiveness analyses generally do not account for non-coronary findings, which often occur with CTCA and can lead to further inappropriate testing. Furthermore, the discussion of cost-effectiveness fails to highlight that CTCA leads to unnecessary costly and risky procedures. Proponents of CTCA like to emphasise the lower rate of invasive coronary angiograms showing no obstructive CAD, an expected finding when comparing one anatomic test (CCTA) with another (invasive angiography). More alarming, CTCA leads to higher overall rates of invasive coronary angiograms and higher rates of revascularisation. The PROMISE trial clearly demonstrated significant harm associated with CTCA. Patients assigned to CTCA had an excess rate of coronary angiography (13.3% vs 5.1%), percutaneous coronary intervention (6.0% vs 1.8%) and coronary bypass surgery (2.9% vs 1.3%) without any improvement in clinical outcome.9
Do clinical trials demonstrate the superiority of CTCA?
The final argument presented to support the broader use of CTCA centres on the importance of clinical trials. Initially, though, the authors cite observational data claiming that stable CAD patients have improved outcomes with percutaneous coronary intervention.1 In the propensity-score analysis discussed, nearly a fifth of patients without revascularisation could not be matched, no adjustment was made for the burden of CAD, and the absolute risk difference was small (0.57%). Astoundingly, despite these limitations, the authors have chosen to emphasise this observational data over a landmark randomised controlled trial that showed no benefit for routine revascularisation in stable patients with CAD.10
With PROMISE, the authors have again focused on the purported benefit of better patient selection for invasive coronary angiography, neglecting to mention that the trial actually failed to reach pre-specified superiority and non-inferiority endpoints.9 A negative trial simply cannot be used to support the conclusion that CTCA is clinically equivalent to functional testing. In fact, the upper bound of the 95% CI for CTCA was 1.29, and most clinicians would not accept a possible 29% increased hazard of cardiovascular events with anatomic testing.
Drs Timmis and Roobottom also claim that the Scottish computed tomography of the heart (SCOT-HEART) trial demonstrated a reduction in cardiovascular death and non-fatal myocardial infarction with CTCA.11 However, the differences were not statistically significant (1.3% vs 2.0%, p=0.053), and represented only one of 13 endpoints, none of which achieved statistical significance, even without adjustment for multiplicity. Unperturbed by the methodological issues with continued post-hoc analyses, the SCOT-HEART investigators have more recently focused on the small number of patients with medication changes after CTCA, neglecting to mention how many of these patients should have been on these medications at baseline. In fact, with better background medical therapy, the Following Core 64 (FACTOR 64) trial showed no benefit for screening with CTCA in diabetics.12 Finally, the Prospective Longitudinal Trial of FFRct (PLATFORM) study has been erroneously referred to as a clinical trial, when in fact PLATFORM included two subsequent cohorts with 584 patients and a total of five cardiovascular events.13 Given this study design and few outcomes, widespread use of CT-Fractional Flow Reserve (CT-FFR) without a supporting clinical trial would be incredibly irresponsible.
Conclusions
In clinical research, practical conclusions should depend both on the strength of the available evidence and the societal implications. Given that the NICE update would require a dramatic increase in cardiac CT scanners and accredited practitioners to report the scans, as well as expose many patients to unnecessary radiation and downstream testing, is the available evidence strong enough to support CTCA for all stable patients with typical or atypical angina? The answer is a resounding ‘No’. On the contrary, this update to the NICE guidelines should have focused on the low event rate achievable with optimal medical therapy and the inability of imaging to modify outcomes for many of these low-risk patients.
Footnotes
Contributors Both authors contributed to this manuscript.
Competing interests None declared.
Provenance and peer review Commissioned; internally peer reviewed.