Plaque imaging with CT—a comprehensive review on coronary CT angiography based risk assessment

Abstract

CT based technologies have evolved considerably in recent years. Coronary CT angiography (CTA) provides robust assessment of coronary artery disease (CAD). Early coronary CTA imaging—as a gate-keeper of invasive angiography—has focused on the presence of obstructive stenosis. Coronary CTA is currently the only non-invasive imaging modality for the evaluation of non-obstructive CAD, which has been shown to contribute to adverse cardiac events. Importantly, improved spatial resolution of CT scanners and novel image reconstruction algorithms enable the quantification and characterization of atherosclerotic plaques. State-of-the-art CT imaging can therefore reliably assess the extent of CAD and differentiate between various plaque features. Recent studies have demonstrated the incremental prognostic value of adverse plaque features over luminal stenosis. Comprehensive coronary plaque assessment holds potential to significantly improve individual risk assessment incorporating adverse plaque characteristics, the extent and severity of atherosclerotic plaque burden. As a result, several coronary CTA based composite risk scores have been proposed recently to determine patients at high risk for adverse events. Coronary CTA became a promising modality for the evaluation of functional significance of coronary lesions using CT derived fractional flow reserve (FFR-CT) and/or rest/dynamic myocardial CT perfusion. This could lead to substantial reduction in unnecessary invasive catheterization procedures and provide information on ischemic burden of CAD. Discordance between the degree of stenosis and ischemia has been recognized in clinical landmark trials using invasive FFR. Both lesion stenosis and composition are possibly related to myocardial ischemia. The evaluation of lesion-specific ischemia using combined functional and morphological plaque information could ultimately improve the diagnostic performance of CTA and thus patient care. In this review we aimed to summarize current evidence on comprehensive coronary artery plaque assessment using coronary CTA.

Introduction

After the first description of CT angiography (CTA) in 1992 (1,2), further technological advances, such as: more powerful X-ray tubes, faster gantry rotation times, multiple parallel detector rings and decreased slice thickness (3,4) were introduced, that allowed the visualization of the coronary arteries (5). Coronary CTA has emerged as a non-invasive alternative to invasive coronary angiography (ICA) for the diagnosis of obstructive coronary artery disease (CAD). With its excellent sensitivity and negative predictive value (6,7), coronary CTA is a robust diagnostic test to rule out severe coronary stenosis and is widely used as a “gate-keeper” for ICA (8,9). Nevertheless, modern CT scanners allow not only the visualization of the coronary lumen as ICA, but also the vessel wall, granting non-invasive analysis of atherosclerosis itself. This unique property of coronary CTA holds many advantages for patient risk stratification, that other non-invasive tests do not. Coronary CTA is currently also the only non-invasive imaging modality for the evaluation of non-obstructive CAD.

Cadaver studies of acute coronary syndrome (ACS) patients have shown that plaque morphology plays a crucial role in developing ACS (10). With around half of plaque ruptures occurring at lesion sites with smaller than 50% diameter stenosis (11-13), plaque morphology assessment seems equally as important as stenosis assessment. Furthermore, recent studies have indicated that coronary plaque burden also has a substantial effect on all-cause mortality (14,15). Due to its volumetric imaging capabilities coronary CTA depicts around twice as many atherosclerotic lesions as ICA (16,17), thus risk assessment based on atherosclerotic plaque burden may be a unique ability of coronary CTA.

Our objective was to review the current literature on stenosis, qualitative plaque morphology, plaque burden and composite plaque burden score assessment using coronary CTA, and examine their effect on patient morbidity and mortality.

Plaque morphology assessment with coronary CTA

Atherosclerosis is initiated by deposition of low density lipoproteins (LDL) in the intima. With oxidation of the lipids, an inflammatory response is triggered, which is characterized by macrophages engulfing oxidized LDL particles, thus becoming foam cells (18). Poorly understood genetic and environmental factors propagate inflammation resulting in further deposition of lipids, deterioration of the extracellular matrix and cell death (19). These processes lead to distinct plaque morphologies, which have been identified on histological samples, and have brought forward the concept of the vulnerable plaque (20). With submillimeter isotropic resolution coronary CTA is capable of imaging not only the lumen, but also the coronary wall, thus allowing non-invasive morphological assessment of coronary atherosclerosis.

Plaque composition

Coronary plaques can be classified as being: non-calcified, partially calcified or calcified based on the amount of calcium in the lesion (Figure 1) (21). Large multicenter cohorts such as the COronary CT Angiography EvaluatioN For Clinical Outcomes: An InteRnational Multicenter (CONFIRM) registry (22), investigated the prognostic value of plaque composition on all-cause mortality. Based on 17,793 suspected CAD patients’ 2-year survival data, the number of segments with partially calcified or calcified plaque had a significant effect on mortality (hazard ratios: non-calcified: 1.00, P=0.90; partially calcified: 1.06, P≤0.0001; calcified: 1.08, P≤0.0001). After adjusting for clinical factors, none of the plaque components improved the diagnostic accuracy of the model (non-calcified: P=0.99; partially calcified: P=0.60; calcified: P=0.10) (15). Hadamitzky et al. found similar results when investigating the prognostic effect of plaque composition on 5-year mortality rate based on suspected CAD patients. After adjusting of clinical risk based on the Morise score (23), only the number of segments with calcified plaques improved significantly the diagnostic accuracy of the model (non-calcified: P=0.083; partially calcified: P=0.053, calcified: P=0.041) (24). Dedic et al. found similar results in a different patient population. When investigating the effects of different plaque components of non-culprit lesions on future major adverse cardiac events (MACE) in ACS patients, they found none of the plaque types to have a significant impact on future MACE rates (hazard ratios: non-calcified: 1.09, P=0.11; partially calcified: 1.11, P=0.35; calcified: 1.11, P=0.15) (25). Interestingly Nance et al. found very different results when analyzing 458 patients’ data, who presented to the emergency room with acute chest pain but based on ECG and serum creatinine had inconclusive results and thus underwent coronary CTA. All patients had low to intermediate risk for CAD. After a follow-up of 13 months, they split the patients into three groups: only non-calcified plaques; exclusively calcified plaques; any partially calcified plaque or both non-calcified and calcified plaques. After adjustment for clinical characteristics and Ca-score they found the following hazard ratios: 57.64 for non-calcified, 55.76 for partially calcified and 26.45 for patients with solely calcified plaques (26). The difference compared to other studies might be due to the different methodological approach used. While previously mentioned papers examined the effect of plaque composition on a segment based level incrementally, giving the hazard ratio of an increase in the number of segments with a given plaque type, Nance et al. reported the results on a patient based level dichotomized, giving the hazard ratio of having a specific type of plaque as compared to patients without any plaques.

Figure 1.

Representative images of plaque characteristics identifiable using coronary CT angiography (CTA).

Overall, the effect of plaque composition on mortality still remains controversial. It seems simply classifying plaques based on the amount of calcium present holds little information regarding clinical outcome. These findings suggest, that identification of more complex morphologies are needed for better prediction of adverse outcomes.

Plaque attenuation and pattern

Histopathologic examinations demonstrated that thin-cap fibroatheromas (TCFA) exhibit similar plaque morphologies as ruptured plaques (20,27). TCFAs are composed of a lipid-rich necrotic core surrounded by a thin fibrotic cap. Coronary CTA is capable of distinguishing between lipid-rich and fibrotic tissue based on different CT attenuation values, however the reliable classification of non-calcified plaques into these two categories remains challenging.

Several studies have investigated the use of region of interest (ROI) to define the plaque components using coronary CTA as compared to intravascular ultrasound (IVUS) as the gold standard of in vivo plaque characterization (28-31). These validation studies were able to find significant differences in mean Hounsfield Unit (HU) values for the different plaque components, however there is a considerable overlap between these categories (121±34 vs. 58±43 HU, P<0.001) (28). Several studies were inspired by these results and investigated the mean and minimal attenuation values of plaques in ACS patients as compared to plaques of stable angina patients and found lower attenuation values for ACS patients (32-34). However, there was still a significant overlap in attenuation values between the two groups. Nevertheless, Motoyama et al. showed that with the use of a strict cut-off value (<30 HU), ACS patients have significantly more low attenuation plaques as compared to stable angina patients (79% vs. 9%, P<0.001), suggesting low attenuation to be a useful marker for identifying vulnerable patients (35). Marwan et al. proposed a more quantifiable approach using quantitative histogram analysis. For all cross-sections for each plaque, a histogram was created from the CT attenuation numbers, and the percentage of pixels with a density ≤30 HU was calculated. They found similarly overlapping HU values for lipid-rich versus fibrous plaques. However, using a cut-off value of 5.5% for pixels with ≤30 HU, they were able to differentiate between predominantly lipid-rich plaques versus predominantly fibrous plaques (sensitivity: 95%; specificity: 80%; area under the curve: 0.9) using IVUS as reference standard (36). Despite these encouraging results, there is still a major concern because of the overlapping HU values of different plaque components. Furthermore, several studies have shown slice thickness (37), imaging protocols (38), tube voltage settings (39), intracoronary contrast attenuation values (40), reconstruction algorithms, filters and noise (37,41) all to influence CT attenuation values. Overall, it seems discrimination of non-calcified plaques based on HU value thresholds into lipid-rich and fibrous categories has additional prognostic value, but the different modifying effects of image acquisition and reconstruction limit the robust use of attenuation values for patient risk prediction.

Another more qualitative approach is not to look at the absolute HU values, but rather to classify non-calcified plaques as homogeneous or heterogeneous in attenuation patterns (Figure 1). Heterogeneous plaques are characterized by at least two regions with different attenuations values, while homogeneous plaques do not have regions visually differentiable. Based on cross sectional images, heterogeneous plaques can be divided into ones with napkin-ring sign (NRS) and ones without (42). NRS plaques are characterized by a low attenuation central area, which is apparently in touch with the lumen, encompassed by a higher attenuation ring-like peripheral area (43). Maurovich-Horvat et al. showed based on ex vivo examinations that NRS plaques have excellent specificity and low sensitivity (98.9%; 24.4%, respectively) to identify plaques with a large necrotic core, which is a key feature of rupture prone TCFA’s (44). Histological evaluation of NRS plaques showed that NRS plaques had greater area of lipid-rich necrotic core (median 1.1 vs. 0.5 mm², P=0.05), larger non-core plaque area (median 10.2 vs. 6.4 mm², P<0.01) and larger vessel area (median 17.1 vs. 13.0 mm², P<0.01) as compared to non-NRS plaques (45). Interestingly, these results are in line with Virmani et al. who investigated the morphology of ruptured plaques (20). Furthermore, results of the Rule Out Myocardial Infarction/Ischemia Using Computer-Assisted Tomography-II (ROMICAT-II) trial strengthen the concept of NRS plaques being precursors of ruptured plaques. Based on the results of 472 patients suspected of ACS they found NRS plaques to be an independent predictor of ACS (odds ratio: 8.9; 95% CI: 1.8–43.3; P=0.006) independent of stenosis severity (46). Kashiwagi et al. found similar results, when analyzing the results of ACS patients and stable angina patients. They found NRS plaques to be more frequent at culprit and also at non-culprit sites in ACS patients as compared to stable angina patients (culprit: 49.0% vs. 11.2%, P<0.01, respectively; non-culprit: 12.7% vs. 2.8%, P<0.01, respectively) (47). Otsuka et al. conducted the first prospective clinical trial to assess the predictive values of NRS plaques for future ACS events (48). They showed that NRS plaques were significant independent predictors of later ACS events [hazard ratio: 5.55 (95% CI: 2.10–14.70)]; P<0.001. Similarly, a recently published study by Feuchtner et al. showed NRS to have the highest hazard ratio (7.0; 95% CI: 2.0–13.6) over other high risk features when investigating 1,469 patients with a mean follow-up of 7.8 years (49).

Overall it seems both plaque attenuation and pattern have additive information beyond simple plaque composition information. However, with many factors effecting assessment of attenuation values and patterns, we only have limited censored information regarding the prognostic effect of these entities.

Spotty calcification

Histological examinations identified calcified nodules in patients with coronary thrombosis (50). Several histological studies have shown the frequency of such findings to be around 2–7% in sudden death cases (51-53). Intra-plaque microcalcifications are thought to destabilize plaques and promote plaque rupture (54,55). Unfortunately, spatial resolution of current CT scanners is under the threshold needed for identifying microcalcification. Nevertheless, coronary CTA has excellent sensitivity to identify calcium, thus spotty calcification defined as a <3 mm calcified plaque component with a >130 HU density surrounded by non-calcified plaque tissue has been proposed as a CTA marker of histological microcalcifications (35,56) (Figure 1). van Velzen et al. suggested to further classify such lesions as small (<1 mm), intermediate (1–3 mm), and large (>3 mm) (57). They found small spotty calcifications to be more frequently present with TCFA’s identified by IVUS as compared to large spotty calcifications (31% vs. 9%; P<0.05). These results support the hypothesis, that small calcified nodules are indicators of high-risk plaques, and that CTA is at the limits of identifying real calcified nodules, which have been identified using histological studies. Even so, several studies have shown culprit lesions of ACS patients to have spotty calcifications as compared to stabile angina patients or non-culprit lesions (32,34,35,58). However, there are only few prospective studies evaluating the prognostic effect of spotty calcifications, thus the relationship between intra-plaque calcification and future cardiac events remains uncertain (59).

A promising technique for the identification of microcalcifications beyond the resolution limits of CTA is PET imaging (60). Dweck et al. used 18F-sodium fluoride to mark microcalcifications not visible on CTA. 18F-sodium fluoride has been used previously for decades to image new bone formation, primarily cancer metastases, and recently has been used to image active calcification in coronary plaques. In their study based on 119 volunteers, they showed higher uptake values in patients with prior cardiovascular events, angina and higher Framingham risk scores, as compared to control subjects (P=0.016; P=0.023; P=0.011, respectively).

Altogether, it seems spotty calcifications have additional additive values for identifying vulnerable plaques. However current resolution of CT scanners prohibits the imaging of microcalcifications that are seen as one of the common features of ruptured plaques. Nevertheless, spotty calcification detectable using CTA seems to correlate well with adverse cardiac events, and 18F-sodium fluoride PET is also a promising new technique to visualize microcalcifications. However prospective studies are needed to evaluate the predictive value of these markers.

Positive remodeling

Atherosclerotic plaques initially tend to grow outwards leaving luminal integrity unchanged (61). Thus while many coronary plaques accumulate lipids and become TCFAs, they might not cause any clinical symptoms. This phenomenon is referred to as positive remodeling (Figure 1). Varnava et al. examined 88 sudden cardiac death cases and showed that plaques with positive remodeling have larger lipid cores and more macrophages, both which are considered vulnerability markers (62). Using coronary CTA, the remodeling index is calculated as the vessel cross-sectional area at the level of the maximal stenosis divided by the average of the proximal and distal reference sites’ cross-sectional areas (63). Coronary CTA has a tendency to overestimate remodeling index, thus Gauss et al. proposed a cut-off value of ≥1.1, meaning a 10% increase in the vessel cross sectional area at the site of the maximal stenosis compared to the average of the reference cross sectional areas (64). This resulted in an increased sensitivity and a moderate drop in specificity as compared to a lower cut-off value of ≥1.05 (sensitivity: 78% vs. 45%; specificity: 78% vs. 100%) using IVUS as reference standard. Motoyama et al. showed positively remodeled plaques to be more frequent in ACS patients as compared to stable angina patients (87% vs. 12%, P<0.0001, respectively) (35). Positive remodeling had the best sensitivity and specificity (87%; 88%, respectively) as compared to low-attenuation and spotty calcification to identify ACS patients (65).

Overall it seems that positive remodeling is an important plaque feature for the identification of vulnerable plaques. Being less conditional to image noise as plaque attenuation, and having a more quantitative definition as the NRS, positive remodeling might become a more robust marker for vulnerable plaques. However, more prospective studies are needed to assess the effect of positive remodeling on later outcomes.

Altogether we can say that distinct plaque morphologies can be identified using coronary CTA which are associated with adverse cardiac events. However, many question the concept of the vulnerable plaque (66,67). Kubo et al. showed that 75% of TCFA’s lost their vulnerability characteristics after 1 year, while only 25% remained to be classified as TCFA using IVUS (68). Furthermore, the Providing Regional Observations to Study Predictors of Events in the Coronary Tree (PROSPECT) study demonstrated that the maximum annualized risk of myocardial infarction or cardiovascular death is only 0.06% per year for a single vulnerable plaque, which is much smaller risk than what is conventionally considered high risk (69,70). These results indicate that even though coronary CTA is capable of identifying vulnerable markers which correlate with gold-standard histological findings, these characteristics currently are not able to explain all aspects of later outcomes and thus we must also consider stenosis severity and overall CAD burden.

Stenosis assessment with coronary CTA

Based on guidelines published by the Society of Cardiovascular Computed Tomography (SCCT), luminal stenosis can be graded as: minimal (<25% stenosis), mild (25% to 49% stenosis), moderate (50% to 69% stenosis), severe (70% to 99% stenosis) and occluded (71) (Figure 2). As coronary plaques grow, blood flow is eventually disrupted causing ischemia distal to the lesion. Originally coronary CTA was considered as a non-invasive alternative to ICA, thus many studies have looked into the predictive value of luminal stenosis seen on CTA. All studies focus on obstructive lesions, since most patients referred to coronary CTA have minimal and mild plaques, thus the predictive value of such stenosis is poor. Two cut-off values: ≥50% and ≥70% diameter stenosis are frequently used to determine obstructive lesions. Min et al. investigated the prognostic effect of obstructive lesion on a patient-based level and found both ≥50% and ≥70% lesions to be significant predictors of later outcomes (hazard ratio: 2.89; 4.31, respectively) (72). Several other studies have also found obstructive lesions to be significant predictors of later outcome, but considerably different hazard ratios have been reported (24,73-77). Interestingly, there is a discrepancy in the results when correcting the models for clinical risk factors. The significant hazard ratios reported by Min et al. became non-significant when including cardiac risk factors in the model, but when looking at a segment-based level not a patient based level, the presence of obstructive stenosis remained significant, but with a smaller hazard ratio (1.05; 95% CI: 1.02–1.09) (72). However, Nakazato et al. and Andreini et al. did not find similar tendencies when correcting for clinical factors, as obstructive CAD remained a significant predictor (74,76). Increasing further the discrepancy in the results, a 5-year follow-up study published by Hadamitzky et al. reported that while non-obstructive CAD was a significant predictor of hard endpoints (hazard ratio: 3.33; 95% CI: 1.40–7.91), one and two vessel obstructive disease was not (hazard ratio: 1.46; 95% CI: 0.50–2.43; 3.85; 95% CI: 0.96–15.04) (24).

Figure 2.

Representative images of stenosis categories using coronary CT angiography (CTA).

One explanation for the disparity in the results might be that luminal stenosis on CTA is a poor indicator of hemodynamically significant lesions (78), which has been shown to be a good predictor of adverse events (79-81). Using computational fluid dynamics it is possible to simulate blood flow in the coronaries and estimate the hemodynamic characteristics of a lesion with good diagnostic accuracy (82,83). The Prospective LongitudinAl Trial of FFRct: Outcome and Resource IMpacts (PLATFORM) study further demonstrated, that the use of CT derived fractional flow reserve (FFR-CT) values significantly lowered the rate of ICA and used less resources at lower costs (84-86). However, current FFR-CT simulations are expensive and time consuming, as the simulations are performed off-site. Recently on-site FFR-CT techniques have been introduced, which are able to calculate FFR-CT in a couple of seconds to minutes (87-90). Further studies are warranted to explore the diagnostic accuracy and utility of these novel on-site algorithms. As an alternative to FFR-CT transluminal attenuation gradient (TAG) measurement was proposed, which tries to estimate hemodynamic significance by calculating the drop in HU values proximal and distal to the lesion (91,92). Unfortunately, validation studies comparing TAG to invasive FFR values showed poor accuracy results (sensitivity: 58%; specificity: 86%; positive predictive value: 64% negative predictive value 83%) (93), with area under the curve values equal to the flip of a dime (AUC =0.50) (94). A more qualitative method is the area of stenosis, lesion length, and APPROACH (ASLA) score (95), which uses plaque burden, minimal luminal area and diameter, stenosis diameter, area of stenosis, lesion length, remodeling index, plaque morphology, calcification severity, and the Alberta Provincial Project for Outcome Assessment in Coronary Heart Disease (APPROACH) score (96), which estimates the amount of myocardium at risk to identify hemodynamically significant lesions. Based on 85 patients, Ko et al. demonstrated that the ALSA score was superior to area stenosis, lesion length and the APPROACH score in predicting significant FFR values (95). Another approach to evaluate hemodynamic relevance of coronary lesions is CT myocardial perfusion (97). CTA therefore could provide comprehensive assessment of CAD incorporating both morphological and functional information to accurately detect significant coronary stenosis. Yang et al. evaluated 72 patients with rest CT perfusion imaging, on-site CT-FFR and invasive FFR (90). Combination of anatomical assessment with on-site CT-FFR or CT perfusion imaging significantly improved diagnostic accuracy (0.86 vs. 0.92, P=0.004; 0.86 vs. 0.91, P=0.004, respectively).

All in all, it seems the effect of significant stenosis on patient outcomes is controversial, which is mainly caused by the poor correlation between stenosis seen on CTA and hemodynamic information. Furthermore, several studies have shown that the more vessels are involved the higher the hazard ratio. This shows the need not only to look at single plaque features or stenosis, but to also look at the extent of CAD.

Plaque burden assessment with coronary CTA

Several studies, such as the Clinical Outcomes Utilizing Revascularization and Aggressive Drug Evaluation (COURAGE) trial showed that plaque burden assessment may be more important than ischemic myocardium burden for predicting later major adverse outcomes (98). Furthermore, Bittencourt et al. demonstrated that patients with extensive CAD (> four segments involved) have similar hazard ratios for developing major adverse outcomes as compared to patients with obstructive disease with less than five segments involved, thus also emphasizing the importance of quantifying plaque burden (14). Min et al. proposed a score system, the segment stenosis score (SSS) and the segment involvement score (SIS) to quantify plaque burden (72). SSS is calculated by grading all coronary segments as: 0, no plaque; 1, <50% stenosis; 2, 50–69% stenosis; 3, ≥70% stenosis. SIS is the number of affected segments. Based on 1,127 patients, SSS had a hazard ratio of 1.99 (95% CI: 1.48–2.67), while SIS had a hazard ratio of 1.23 (95% CI: 1.13–1.34). Similarly, results from the CONFIRM registry also showed SIS to be an independent predictor of later major adverse events (hazard ratio: 1.22; 95% CI: 1.03–1.44) (15). Several other studies have also demonstrated SSS and SIS to be significant independent predictors of later outcomes (24,74,99,100). While SSS and SIS are simple and elegant concepts for describing plaque burden, they are conceptually flawed. SSS and SIS scores assume that plaque burden is additive, meaning that adding one plaque to already two diseased segments or 12 diseased segments has the same effect. Furthermore, SSS and SIS lose all anatomical information, thus they assume a moderate stenosis on the left main has the same effect as a moderate stenosis on the second diagonal branch, which clearly is not true. Results of the CONFIRM trial also emphasize the importance of lesion characteristics and location. They found that excluding distal segments and only considering the number of proximal segments with obstructive plaques significantly improved their prediction model (15). Another approach for quantifying the magnitude of plaque burden is the 3-vessel score, which counts how many major epicardial vessels (left anterior descending, left circumflex, right coronary) have obstructive stenosis (72). Andreini et al. demonstrated that having only one major epicardial vessel effected with an obstructive lesion (≥50%) has a hazard ratio of 3.18 (95% CI: 2.16–4.69), if all three vessels are effected, the hazard ratio increases to 7.10 (95% CI: 4.61–10.93) (74). Similar tendencies have been reported by several studies (24,72,99-102). A more quantitative approach originally developed to characterize CAD severity using ICA (103), later adopted for coronary CTA is the Duke CAD Index (72,104). Patients are assigned a risk score between 0–100 based on former patient prognosis data (103). The score is an extension of the 3-vessel disease score. It also incorporates stenosis severity and calculates with left main stenosis and proximal left anterior descending stenosis (Table 1). Min et al. showed that there was a significant difference between patients’ cumulative survival for the different categories (72). Left main plaque with any additional moderate or severe stenosis had the worst outcome, while patients without any disease or only mild CAD had almost no events.

Table 1. Modified Duke Coronary Artery Disease Index for coronary CTA.

Extent of coronary artery disease	Points
Stenosis <50%	0
Stenosis ≥50%
1 vessel	23
2 vessel	37
3 vessel	56
Stenosis ≥50% and proximal LAD stenosis ≥50%
1 vessel	48
2 vessel	56
3 vessel	74
Left main stenosis
≥50%	80
≥70%	100

Score system is based on Miller et al. In the Modified Duke Coronary Artery Disease Index patients are assigned to the most severe category. CTA, CT angiography; LAD, left anterior descending.

Altogether, plaque burden assessment seems to be a very important concept to describe the severity of CAD and predict adverse outcome. Several methods have been proposed to properly quantify plaque burden, indicating the lack of a single best method. Furthermore, as we have seen, not only plaque burden, but plaque localization, stenosis severity, plaque composition and vulnerability features all play a role in later outcomes, thus necessitating a more complex holistic approach, which incorporates as many of these parameters as possible (43).

Composite plaque burden score assessment with coronary CTA

Based on research investigating the risk of plaque features and extension of CAD, several attempts have been made to create composite scores incorporating anthropometric vulnerability with extent of CAD, plaque localization and vulnerability features as assessed by CTA.

CONFIRM risk score

The CONFIRM registry is an international prospective observational cohort currently with seven participating countries (22). Structured interviews were used to collect information regarding patients’ anthropometrics and cardiovascular risk profile. Using this information the National Cholesterol Education Program Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (NCEP ATP III) (105), the Framingham (106) and the Morise clinical risk scores were calculated (23). The 16-segment coronary artery model was used to assess the CTA images (107). Each coronary segment was evaluated for the presence of plaque. Plaques were classified as calcified, partially calcified or non-calcified. The degree of stenosis was graded as: none (0% luminal stenosis); mild (1% to 49% luminal stenosis); moderate (50% to 69% luminal stenosis); or severe (≥70% luminal stenosis).

Overall, 17,793 patients’ data was used to create the CONFRIM risk score using multivariate Cox proportional hazard models (15). The resulting models were evaluated using a separate test set, which consisted of 2,506 patients’ data. After separate assessment of clinical risk scores and CTA imaging markers, a combined score was created. The COMFIRM risk score is a combination of the NCEP ATP III score, the number of proximal segments [proximal and mid right coronary artery (RCA), left main, proximal, and mid left anterior descendent, proximal circumflex, first obtuse marginal branch] with stenosis greater than 50%, and the number of proximal segments with partially calcified or calcified plaques. Adding these two additional parameters caused 32% of the patients to be reclassified, 22% to a lower risk category and 10% to a higher category. Overall, the combined risk score outperformed all clinical scores and significantly improved prediction of all-cause mortality. A online calculator is available for the CONFIRM risk score (108).

Leaman score

Originally the Leaman score was established based on ICA measurements. Since plaque features cannot be visualized using ICA, only the localization and the degree of stenosis is used to calculate the score. Obstructions are weighted based on typical amount of blood flow to the left ventricle going through that given segment. On average in case of right dominant coronary anatomy, the RCA receives 16%, while the left main trunk delivers 84% of the blood flow going the left ventricle (109). For left dominant coronary systems, all of the left ventricle is supplied by the left coronary artery. Weighting factors are equal to how many times more blood goes through a given segment as compared to the RCA. For left dominant systems, the RCA receives a weighting factor of zero, while the weighting factor of the LM and circumferential segments increases by one (110). The degree of stenosis was also accounted for. Occlusions receive a multiplication factor of five, 90–99% stenosis receive multiplication factor of three and obstructions between 70–89% receive a multiplication factor of one. Non-obstructive lesion (<70%) are not accounted for. A patients’ Leaman score is equal to the sum of all segment scores for all 16 segments (107).

CTA adapted Leaman score as proposed by de Araújo Gonçalves et al. (111) has minor modifications as compared to the original publication of Leaman et al. (110). To account of balanced coronary systems an intermediate value was used for segments where there was a difference in the weighting factors for left and right dominant systems. Plaque composition was also included. For non-calcified and partially calcified plaques weighting factor of 1.5 is added, while calcified plaques receive a weighting factor of one. Lesions with <50% stenosis receive a multiplication factor of 0.615 which is the relative proportion of the hazard ratios for mortality between obstructive and non-obstructive CAD, as reported by Chow et al. from the CONFIRM registry (112). A summary of the calculation can be found in Table 2.

Table 2. CTA adapted Leaman score weighing factors.

Segments	Dominance
	Right dominance	Left dominance	Balanced
Proximal RCA	1.0	0	0.5
Mid RCA	1.0	0	0.5
Distal RCA	1.0	0	0.5
R-PDA	1.0	–	0.5
R-PLB	0.5	–	–
Left main	5.0	6.0	5.5
Proximal LAD	3.5	3.5	3.5
Mid LAD	2.5	2.5	2.5
Distal LAD	1.0	1.0	1.0
1st diagonal	1.0	1.0	1.0
2nd diagonal	0.5	0.5	0.5
IM	1.0	1.0	1.0
Proximal LCX	1.5	2.5	2.0
1st OM	1.0	1.0	1.0
Mid-distal LCX	0.5	1.5	1.0
2nd OM	1.0	1.0	1.0
L-PDA	–	1.0	–
L-PLB	–	0.5	0.5
Stenosis severity
Obstructive	1.000	–	–
Non-obstructive	0.615	–	–
Plaque composition
Non-calcified or partially calcified	1.5	–	–
Calcified	1	–	–

CTA adapted Leaman-score is calculated by multiplying the weighing factors regarding plaque composition, stenosis severity and location for a given segment. Overall score is calculated by summing up scores for all segments. RCA, right coronary artery; R-PDA, posterior descending artery originating from right coronary; R-PLB, posterolateral branch originating from right coronary; LAD, left anterior descending; IM, intermediate branch; LCx, left circumflex; OM, obtuse marginal; L-PDA, posterior descending artery originating from left coronary; L-PLB, posterolateral branch originating from left coronary.

Mushtaq et al. evaluated the CTA adapted Leaman score using a single-center prospective registry including 1,304 consecutive patients (113). Hard cardiac events (cardiac death and nonfatal myocardial infarction) were considered primary end-points. Using multivariate Cox regression models which included clinical parameters and SSS or SIS or the CTA adapted Leaman score, were all independent predictors of adverse events. The Leaman score had the highest hazard ratio as compared to the other two scores (hazard ratio: Leaman score: 5.39, 95% CI: 3.49–8.33; SSS: 4.42, 95% CI: 2.97–6.57; SIS: 3.09, 95% CI: 2.00–4.75, respectively). The event free survival of patients with Leaman scores in the highest tercile (score >5) and obstructive CAD was similar to patients with similar Leaman scores but without obstructive CAD (78.6% vs. 76.5%; P=0.627).

SYNTAX score

Originally the SYNergy between percutaneous coronary intervention with TAXus and cardiac surgery (SYNTAX) score was developed to quantify the complexity of CAD, and to determine optimal revascularization strategies for multi-vessel CAD patients (114). SYNTAX score incorporates multiple score systems. As opposed to previously described CTA scores, the SYNTAX score is a lesion based scoring system, rather than a segment based system, thus multiple lesions can be present and also scored in the same segment.

The original 16-segment classification of the American Heart Association (107) is extended based on the arterial revascularization therapies study (115), to include additional side branches. Only vessels greater than 1.5 mm and lesions with a stenosis greater than >50% are analyzed. The SYNTAX score does not recognize balanced coronary dominance. Each lesion receives the Leaman score values for the segments in which it is present (Table 2). Each segment score is multiplied by two for non-occlusive lesions (50–99%) and by five for occlusive lesions (100%). Only one segment is allowed to be occlusive for each lesion. Additional lesion attributes are scored based on the ACC/AHA lesion classification system (116). Characteristics of occlusions (117), involvement of trifurcations, bifurcations (118,119) and aortal ostium, severe tortuosity, lesion length, heavy calcification, thrombus and diffuse coronary disease are all accounted for. Further adverse lesion characteristics are all additive. Details of the scoring system are described in Table 3.

Table 3. Scoring system of the SYNTAX score.

Characteristics	Points
Stenosis
Occlusion	×5
Significant lesion	×2
Aorto ostial stenosis	+1
Occlusion characteristics
Age >3 months or unknown	+1
Blunt stump	+1
Bridging	+1
First segment visible beyond occlusion	+1 per non-visible segment
Side branch	+1
Trifurcations
1 diseased segment	+3
2 diseased segments	+4
3 diseased segments	+5
4 diseased segments	+6
Bifurcations
Type A, B, C	+1
Type D, E, F, G	+2
Angulation <70°	+1
Severe tortuosity	+2
Lesion length >20 mm	+1
Heavy calcification	+2
Thrombus	+1
Diffuse disease of affected vessel	+1 per number of segments

SYNTAX score is calculated by multiplying the Leaman score (Table 2) of the segments which contain the given lesion by the stenosis factor. Further lesion characteristics are all additive. Overall, the SYNTAX score is the sum of all individual lesion scores. +, addition; ×, multiplication.

The SYNTAX score includes many vulnerability parameters, thus utilization of the scoring system for long-term prognosis seems rational. Suh et al. evaluated the performance of the SYNTAX score based on 339 patients who underwent both CTA and ICA (120). Only characteristics assessable by both CTA and ICA were included in the SYNTAX score. Based on univariate Cox regression analysis age, 3-vessel or LM disease on CCTA, 2-vessel disease or 3-vessel or LM disease on ICA, and SYNTAX scores higher than 23 based on ICA were predictors of MACE. On the contrary, multivariate analysis showed that models incorporating the SYNTAX score or simply the number of involved vessels had similar predictive power, both in case of CTA (area under the curve: 0.701 vs. 0.659, respectively) and ICA (area under the curve: 0.706 vs. 0.676, respectively). Recently, the SYNTAX score II has been developed that combines the SYNTAX score with clinical variables (121). Long-term follow-up data are promising based on ICA, but we currently lack CTA-based results. Based on these results, it seems incorporating an exceedingly complex score system, such as the SYNTAX score, is not justifiable, since it has no proven additive value in risk prediction, as compared to simple CTA based CAD burden scores.

Overall, composite plaque burden scores seem to be a valid concept to determine the severity of CAD. One major flaw of simple plaque burden scores is that very different disease severities can have very similar scores (Figure 3). Composite scores seem to account for this, but one must not forget, that these scores are only useful if they are calculated. The calculation of composite scores can become very complex, adding an additional burden to the clinicians. In the future, with the use of structured reporting platforms, these values can be calculated and evaluated automatically by the program. Therefore, these scores could transition from simple research interests to clinically useful risk stratification systems.

Figure 3.

Representative examples of plaque burdens and composite plaque scores. For the CONFIRM score both patients were assumed to be 65-year-old smoking male patients with 230 mg/dL total cholesterol, 47 mg/dL HDL, 142 mmHg systolic blood pressure using hypertension medication. For the SYNTAX score calculations, the LAD-LCX bifurcation was assumed to be ≥70˚ and all plaques were shorter than 20 mm. The example shows, that patients with very different degree of disease can have very similar plaque burden scores. Composite plaque burden scores on the other hand seem to better differentiate between the severity of coronary artery disease. D, diagonal; IM, intermediate branch; LAD, left anterior descending; LCX, left circumflex; PDA, posterior descending artery; PLB, posterolateral branch; OM, obtuse marginal; RCA, right coronary artery; prefixes: d, distal; m, mid; p, proximal.

Disadvantages of coronary CTA

Even though coronary CTA is a non-invasive technique, there are still some risks one has to consider. CT uses radiation for image acquisition, thus there is an increased risk for radiation induced illnesses, such as cancer (122). Some estimates show that up to 1.5% to 2.0% of all cancers can be associated to CT use (123). Furthermore, to achieve proper image quality high intraluminal contrast media concentrations need to be reached. However, high contrast volumes and high injection rate increases the risk of contrast induced nephropathy and contrast media extravasation (124,125). In addition, coronary CTA provides detailed information regarding coronary plaques, but only provides limited information regarding the functional significance of CAD. New techniques such as CT myocardial perfusion can add valuable information regarding myocardial ischemic burden, but improvement of diagnostic performance is achieved through additional radiation exposure (90,97).

Future directions

Quantitative plaque assessment

Radiological images can not only be visually inspected, but can also be looked at as vast three-dimensional datasets. Using coronary segmentation software, one can delineate the lumen and the vessel wall to receive volumetric information regarding the composition of the coronary wall. One major concern is the fact that HU values are effected by several different factors (37-41), thus precise classification of plaque components is difficult. Nevertheless, there are encouraging results indicating that quantitative plaque assessment might hold additional information for predicting later outcomes (126-129).

Radiomics

Radiomics is the process of extracting quantitative features for radiological images and converting them into minable datasets, which can be further analyzed using data mining and machine learning algorithms (130,131). Radiomics also integrates clinical and genetic information to make a comprehensive multi-dimensional database, which can be used not only for population based statistical inference, but also for individual patient based risk assessment. This holistic approach has shown promising results in oncology imaging (132,133), but utilization to cardiovascular imaging is scarce.

Conclusions

Coronary CTA has evolved like no other non-invasive imaging modality over the past decade. A couple of years ago only considered as a “gate-keeper” for ICA, is now thought of as versatile imaging modality capable of depicting atherosclerosis in vivo. With the advances of image quality, so did our understanding regarding the different aspects of CAD. All the results indicate that coronary CTA is a valuable tool for predicting patient outcomes. However, a more holistic approach is needed, accounting for many different features of atherosclerosis. For this, new characteristics need to be defined and new statistical methods need to be employed. However strenuous this may be, the reward is not less than decreasing the burden of cardiovascular morbidity and mortality on a global scale.