Risk prediction by non-invasive coronary imaging: we are not there yet!

This editorial refers to ‘Prognostic value of coronary computed tomographic angiography findings in asymptomatic individuals: a 6-year follow-up from the prospective multicentre international CONFIRM study’†, by I. Cho et al., on page 934. In the current issue of this journal, Cho and colleagues report the findings of a large prospective multicentre international registry study investigating asymptomatic patients without known coronary artery disease (CAD) who underwent both coronary artery calcium scoring (CACS) and coronary computed tomographic angiography (CCTA).1 On top of contrast-free low-dose radiation CACS using computed tomography (CT), CCTA provides additional morphological data regarding plaque composition, severity of coronary artery stenosis, and coronary segment location, but also requires iodinated contrast and increased radiation exposure. In recapitulating the most salient findings of this study, CCTA just like CACS, improved prognostication of all-cause mortality on top of existing traditional risk factors (RFs), where the addition of dichotomous categorical (CACS or CCTA on top of RFs) as well as discrete and ordered categorical data sets (CACS intervals, plaque characterization by CCTA, stenosed segments by CCTA) significantly improved C statistics by augmenting diagnostic discriminative power in a total of 1226 asymptomatic subjects. Coronary plaque was defined as the presence of any tissue >1 mm2 in at least two planes located within or adjacent to the coronary artery lumen able to be distinguished from epicardial fat, pericardial tissue, or the artery lumen, and was separated into calcified, non-calcified, and mixed type of plaque. Additional categorization was performed to classify coronary plaque into strata of causing more or less than 50% luminal obstruction. Although CCTA by itself was proven to identify patients at risk for future cardiovascular events reliably in the current and previous studies,2 its added benefit to standard CACS remains highly questionable. Along these lines, the current study adds an important additional piece to already existing knowledge regarding the use of CT angiography in predicting total mortality and its suitability to be used as upstream screening methodology in patients with subclinical atherosclerosis.3 The authors must be congratulated not only for setting up a tremendous network of experts in CCTA, applying rigorous data collection up to 6 years of follow-up in a multicentre international registry of asymptomatic patients undergoing CCTA in addition to CACS, but also for reporting highly valuable findings with relevance for cardiologists, general practitioners, as well as a multitude of other medical disciplines with sound scientific evidence. Yet, despite these favourable attributes, the study also has important limitations, which deserve mentioning. First and foremost, the current study population is of moderate size (n = 1226) with only 78 deaths as major endpoints, and represents a highly selected subgroup from a total of 17 181 patients who underwent CCTA at 17 centres in nine countries. Owing to incomplete data regarding CACS assessment (n = 9626), CCTA stenosis information (n = 214), demographic data (n = 662), and n = 5453 patients for whom no exclusion criterion was given, the findings of the current study may not be representative for the general population with subclinical atherosclerosis. The presentation of the data may have greatly benefited from including a CONSORT diagram. Furthermore, considering an overall mortality rate of 6.4% at 6 years of follow-up, it remains questionable whether additional prognostic benefit over traditional RF assessment can be detected by any diagnostic means given such an overall low-risk population. Since both CACS and CCTA reflect specific changes in the coronary artery vessel wall, coronary heart disease (CHD) or cardiovascular disease (CVD) as endpoints may have provided stronger results. However, such data were not available. Also, the Framingham Risk Score (FRS) performed fairly poor in this study, with an area under the curve (AUC) of only 0.641, clearly facilitating a positive incremental value of CACS (ΔAUC = 0.07). Another important limitation refers to the nature of the study design, for which prospective registry data cannot account for potential interfering interventions during the follow-up period of 6 years, which may impact on overall mortality rates. Schematic illustration of an integrated approach for improved risk stratification and prediction. A large number of studies over the last 20 years have established a fairly strong positive association between CACS and CHD endpoints including an incremental value of CACS to traditional RFs in the majority of intermediate risk populations,4 , 5 but not in all.6 The latter may be due at least in part to methodological issues since it becomes increasingly evident that discrimination analyses using receiver operating characteristic (ROC) curves and establishing AUCs, and consequently assessing net reclassification improvements (NRIs) and integrated discrimination improvements (IDIs), are of limited value, and machine learning techniques may be more appropriate.7 Current prevention guidelines8 , 9 have assigned CACS a class IIb recommendation, level of evidence B. This is due to the fact that a number of issues regarding its clinical utility are still being discussed controversially, e.g. CACS showed no incremental value to the FRS in Caucasians and Hispanics,10 a CACS of zero does not exclude obstructive CAD,11 whether progression of CACS adds to risk prediction,12 , 13 the lack of a randomized trial to test the role of CACS for evidence of improved health outcomes,14 cost-effectiveness issues,15 and, finally, the observation that statin treatment promotes coronary calcification in the presence of plaque regression which would suggest that this represents a favourable effect on plaque stability, which may question not only the utility of serial CACS to monitor patients treated with lipid-lowering agents but also the value of CAC screening itself.16 Given the principal finding of the study that CCTA does not provide additional discriminative power on top of CACS with traditional risk stratification, the question arises of whether contemporary CCTA is inherently limited by its inability to derive sufficient pathobiological information needed to triage patients into low- and high-risk subgroups, and whether we have finally reached a critical limit of resolution with this highly developed technology, making further improvement in downstream diagnostic accuracy negligible. The unique strength of CCTA when compared with other non-invasive imaging technologies is its proposed capability to detect earlier stages of atherosclerosis owing to incremental morphological information derived from cross-sectional athero-phenotyping. While undoubtedly progress in morphological plaque characterization has been made in recent years with CCTA,17 its diagnostic accuracy in deciphering high-risk atherosclerotic plaque and identifying vulnerable patients may still be limited. CCTA has repeatedly been proposed to function as a gatekeeper for more invasive coronary angiography given its ability to up- and downgrade patients into risk categories justifying or avoiding the need to perform invasive procedures in selected patients.18 Yet, additional radiation in combination with contrast exposure when compared with stand-alone calcium scoring may represent a substantial hurdle in rendering this technology clinically adoptable. In keeping with this issue, not only has CCTA recently failed to improve risk stratification of subclinical atherosclerosis but rather the entire spectrum of cardiovascular risk prediction ranging from more sophisticated conservative risk scores to complex intravascular imaging providing 10–100 times greater resolution compared with CCTA. It is therefore to be questioned whether our understanding of atherosclerosis is sufficient to feed into existing risk prediction models using contemporary algorithms or whether we have finally reached a plateau in advancing imaging modalities owing to incomplete knowledge of basic concepts of atherosclerotic plaque progression. Along these lines, there has been continued debate regarding the priority of morphological vs. systemic features in predicting patient vulnerability and susceptibility to cardiovascular events. Seminal findings from post-mortem autopsy studies were paramount in determining plaque characteristics associated with atherothrombosis,19 and yet have failed to become the single discriminator in identifying patients at risk for future cardiovascular events in subsequent clinical imaging studies.19 Consequently, the concept of systemic inflammation triggering atherothrombotic events was proposed and became mainstream in cardiovascular basic research. Approximately 30 years later, a huge diversity of alternative hypotheses regarding atherosclerotic plaque progression have been suggested, yet none of them provide sufficient clarification in cardiovascular risk prediction to move the field forward. We must therefore not refrain from improving existing imaging technology by performing systematic clinical examination, and the investigators of the CONFIRM registry have certainly laid a foundation for future amelioration of this concept. Conflict of interests: none declared.