The Perivascular Fat Attenuation Index Improves the Diagnostic Performance for Functional Coronary Stenosis

Background: Coronary computed tomography angiography (CCTA) is an established first-line test in the investigation of patients with suspected coronary artery disease (CAD), while the perivascular fat attenuation index (FAI) derived from CT seems to be a feasible and efficient tool for the identification of ischemia. The association between the FAI and lesion-specific ischemia as assessed by fractional flow reserve (FFR) remains unclear. Methods: In a total of 261 patients, 294 vessels were assessed for CCTA stenosis, vessel-specific FAI, lesion-specific FAI, and plaque characteristics. The diagnostic accuracies of each parameter and the combined approach were analyzed via the receiver operating characteristic curve (ROC) with FFR as the reference standard. The determinants of FAI were statistically analyzed. Results: The cutoff values of vessel-specific FAI and lesion-specific FAI scores calculated according to the Youden index were −70.97 and −73.95 HU, respectively. No significant differences were noted between them; however, they exhibited a strong correlation. No significant differences were noted between the area under the curve (AUC) scores of vessel-specific FAI (0.677), lesion-specific FAI (0.665), and CCTA (0.607) (p > 0.05 for all) results. The addition of two FAI measures to the CCTA showed improvements in the discrimination (AUC) and reclassification ability (relative integrated discrimination improvement (IDI) and category-free net reclassification index (NRI)), vessel-specific FAI (AUC, 0.696; NRI, 49.6%; IDI, 5.9%), and lesion-specific FAI scores (AUC, 0.676; NRI, 43.3%; IDI, 5.4%); (p < 0.01 for all). Multivariate analysis revealed that low-attenuation plaque (LAP) volume was an independent predictor of two FAI measures. Conclusion: The combined approach of adding vessel-specific FAI or lesion-specific FAI scores could improve the identification of ischemia compared with CCTA alone. The LAP volume was the independent risk factor for both tools.


Introduction
Coronary artery disease (CAD) remains the leading cause of death in upper-middle and high-income economies [1]. Coronary computed tomography angiography (CCTA) is an established first-line test in the investigation of patients with suspected CAD, because it has a high negative predictive value and high accuracy in the diagnosis and exclusion of CAD [2][3][4]. However, only about half of obstructive coronary stenosis cases lead to ischemia [5,6] and are associated with worsened survival [7]. Therefore, it is important to find other factors beyond coronary stenosis to improve the recognition of lesion-specific ischemia.
The perivascular adipose tissue (PVAT) interacts with the arterial wall in a bidirectional manner and has been associated with the process of atherogenesis characterized by calcification and adverse clinical prognosis [8], which implies that there might be a

Study Population
In this study, we retrospectively collected data from patients with suspected or known stable CAD from January 2012 to March 2020 at our institution. The inclusion criteria were as follows: (1) patients with angina or angina-equivalent symptoms; (2) the presence of at least one lesion with stenosis diameters ranging between 30% and 90% of the major epicardial vessels (diameter ≥ 2 mm) based on CCTA; (3) patients who underwent CCTA, invasive coronary angiography (ICA), and FFR measurements within 2 weeks. The exclusion criteria were as follows: (1) age < 18 years; (2) previous history of myocardial infarction; (3) previous history of coronary revascularization; (4) insufficient quality of CCTA images; (5) patients with anatomic variations in the heart or coronary arteries. This study was approved by the Institutional Review Board of our institution, and informed consent from all patients in this study was waived.

Coronary Computed Tomography Angiography Acquisition
All CCTA scans were performed using a dual-source CT scanner (Definition Flash, Siemens Healthineers, Forchheim, Germany). All patients were scanned by prospective electrocardiogram (ECG) gating technology, and images were acquired at 35-75% of the R-R interval. Beta-blockers were administered when the heart rate > 75 beats/min. Sublingual glyceryl trinitrate was administered before scanning in all patients. The scanning parameters were shown as follows: tube voltage, 100 kV or 120 kV (according to the body mass index of patients); tube current, automatic tube current modulation; rotation time, 0.28 s per rotation; slice thickness, 0.75 mm; gap, 0.70 mm. The optimal cardiac phase was selected by radiology technicians. During scanning, 60-70 mL of contrast medium (Iohexol, Shuangbei 350; Beilu Pharmaceutical Co., Ltd., Beijing, China) was injected into the antecubital vein through a dual-cylinder high-pressure syringe (Stellant; Medrad, Indianola, PA, USA) at a speed of 4.5-5.0 mL/s and then flushed with 30-40 mL of saline at the same speed. The calcification score (CS) of the coronary artery was measured as previously described by Agatston [16]. Stenosis severity was categorized as 30-49%, 50-69%, and 70-90% in coronary segments ≥ 2 mm by two experienced local radiologists who were blinded to the patient's condition. When there were different opinions, a consensus was drawn after discussion. Coronary stenosis ≥ 50% was considered as obstructive stenosis.

Coronary Plaque Analysis
Plaque areas > 1 mm 2 in coronary segments ≥ 2 mm were measured with a semiautomated dedicated plaque analysis software (Coronary Plaque Analysis, version 2.0, Siemens Healthineers, Forchheim, Germany). The quantitative plaque components were automatically generated according to scan-specific thresholds within the manually designated area. Adjustments were performed if necessary. The remodeling index, plaque length, total plaque volume (TPV), low-attenuation plaque (LAP), intermediate-attenuation plaque (IAP), and calcified plaque (CP) values were previously measured. The remodeling index is the ratio of the largest vessel diameter at the lesion site to the vessel diameter at the proximal reference point, while a remodeling index > 1.1 indicates positive remodeling (PR) [17]. Two experienced observers performed the analyses and the average values were used for further analysis.

Perivascular Fat Attenuation Index Acquisition
Coronary PVAT around the epicardial coronary arteries with an attenuation window of −190 to −30 Hounsfield units (HU), is defined as the adipose tissue located within a distance from the outer vessel wall equal to the diameter of the adjacent coronary vessel [10]. Perivascular FAI was defined as the mean CT attenuation of PVAT, and it was also performed by a semi-automated post-processing software (Coronary Plaque Analysis, version 2.0, Siemens Healthineers, Forchheim, Germany). Vessel-specific FAI values were acquired as described previously. To avoid the effects of the aortic wall, the most proximal 10 mm of the right coronary artery (RCA) was excluded and the proximal 10-50 mm of the vessel was analyzed [9]. In the left anterior descending artery (LAD) and left circumflex artery (LCX), the proximal 40 mm of each vessel was analyzed. Lesion-specific FAI scores for lesion plaques interrogated with FFR evaluation were measured from the proximal to the distal shoulder of the lesion. Additional manual optimization was performed, if necessary, to avoid the effects of non-adipose tissues, such as the small side branches or coronary veins. The time periods required for vessel-specific FAI and lesion-specific FAI analyses are between 6 and 8 min and between 3 and 5 min, respectively. Vessel-specific FAI and lesion-specific FAI scores of 30 consecutive vessels were measured by two experienced radiologists who were blinded to clinical, CCTA, and FFR data to evaluate the reproducibility between observers.

Invasive Coronary Angiography and Fractional Flow Reserve Measurements
Invasive FFR measurements were assessed during the ICA inspection, and all operations were performed by a cardiovascular physician with extensive work experience. ICA and FFR were performed according to standard practices, as previously described [18]. FFR was the ratio of the pressure of the distal coronary artery to the aortic pressure during the maximum hyperemia, and it was measured by using a 0.014 inch pressure guidewire (St Jude Medical Systems, Minneapolis, MN, USA). The FFR was measured at approximately 2 cm distal to the lesion stenosis. A hyperemic state was induced by continuous administration of intravenous adenosine at a rate of 160 mg/kg/min. Invasive FFR ≤ 0.80 was considered that the stenosis was physiologically significant and causal of lesion-specific ischemia [19,20].

Statistical Analysis
Continuous variables were presented as means ± standard deviation (SD) in the case of normal distribution and medians (interquartile range) in the case of non-normal distribution, while categorical variables were expressed as numbers and percentages. To evaluate interobserver reproducibility, intraclass correlation coefficients were used to evaluate the interobserver variability of vessel-specific FAI and lesion-specific FAI scores. The receiver operator characteristic curve (ROC) was created to predict the area under the curve (AUC), while p-value, diagnostic accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) scores were calculated with a 95% confidence interval (CI), using invasive FFR as the reference standard. The AUCs of different methods were compared using the method previously described by Delong et al. [21]. The accuracy, sensitivity, and specificity of different methods were compared using Cochran's Q test, then the post-Dunn test and Bonferroni correction were used for intergroup comparison [22], and PPV and NPV were compared using a chi-square test. Additive values of vessel-specific FAI and lesion-specific FAI measures were evaluated by relative integrated discrimination improvement (IDI) and category-free net reclassification index (NRI) scores [23,24]. The best cutoff values of vessel-specific FAI and lesion-specific FAI scores were selected according to the Youden index (defined as %sensitivity + %specificity − 1). Data were compared using Student's t-test, Mann-Whitney U test, Kruskal-Wallis test, or chi-square test as appropriate. Univariable and multivariable logistic regression analyses were performed to determine predictive factors of vessel-specific FAI and lesion-specific FAI, respectively. All statistical analyses were performed with SPSS 25.0 (IBM Corp., Armonk, NY, USA), Med-Calc 19.0.4 (MedCalc Software, Ostend, Belgium), and R 3.3.3 (R Foundation for Statistical Computing, Vienna, Austria) software. Statistical tests were two-tailed, and p-values < 0.05 indicated statistical significance.

Clinical Characteristics
The flowchart of patient selection is shown in Figure 1. Finally, a total of 261 patients were included in this study, in whom 294 vessels were interrogated by invasive FFR, including LAD (210; 71.4%), LCX (35; 11.9%), and RCA (49; 16.7%), respectively. The baseline characteristics of the patients are shown in Table 1. The mean age of the patients was 56.3 ± 8.7 years, and men accounted for 72.0% (188) of patients in this study. Lesionspecific ischemia was found in 45.2% (133/294) of vessels in 44.1% (115/261) of patients by obtaining FFR as a reference; the mean FFR value was 0.80 ± 0.11. There were 206 (70.8%) vessels with obstructive stenosis based on CCTA results.

Discussion
The main findings of this study are as follows. No significant differences were noted between vessel-specific FAI and lesion-specific FAI scores, and there was a strong, almost collinear association between them. Secondly, the diagnostic AUCs and accuracy levels of vessel-specific FAI and lesion-specific FAI were not higher than for CCTA. However, the discrimination and reclassification ability for ischemia were significantly improved when vessel-specific FAI and lesion-specific FAI assessments, respectively, were added to CCTA. Lastly, the LAP volume was an independent risk factor for vessel-specific FAI and lesion-specific FAI values after adjusting for confounding factors.
Similar to previous studies [5,6], only 51.9% (121/233) of vessels with obstructive stenosis were hemodynamically significant in this study. Previous studies have also reported that lesion-specific ischemia diagnosed by FFR is associated with future adverse prognosis and that revascularization guided by FFR can improve event-free survival [19,20]. Consequently, this indicates that in addition to stenosis by CCTA, other non-invasive methods are needed to assist in improving the ability to discriminate ischemia.
Atherosclerosis is an inflammatory process [25] and inflammation is a key factor, not only for atherosclerotic development, but also for the progression of atherosclerotic plaques [26]. Perivascular FAI is a metric derived from CT, which reflects the presence of pericoronal inflammation [10]. The paracrine inflammatory signals from the inflamed vessel walls would prevent lipid accumulation by affecting biological processes such as adipocyte differentiation, proliferation, and lipolysis [9], thereby resulting in a shift from a lower to a higher water/lipid ratio, while the attenuation on CT images increases. Therefore, FAI would be higher when vascular inflammation occurs. Vascular inflammation is a chief contributor to endothelial dysfunction, leading to local "functional stenosis" [27,28]. FAI alone is a weak predictor of lesion-specific ischemia, because the diagnostic performance of the vessel-specific FAI or the lesion-specific FAI was not significantly higher than that of CCTA in this study. Compared to CCTA, the vessel-specific FAI or lesion-specific FAI aid in the identification of ischemia, resulting in relatively high-sensitivity and low specificity, rather than higher diagnostic AUC or accuracy. However, the discrimination and reclassification ability of hemodynamic significance stenosis were significantly improved when vessel-specific FAI and lesion-specific FAI assessments were added to CCTA, respectively, as reported previously [12][13][14][15]. Nevertheless, there were still some different results from previous studies, which may have been due to the different populations and software programs employed [29].
The measurement methods used for FAI in previous studies [12][13][14][15]29] were not consistent, which adds confusion in assessing the impacts of measurement methods on the diagnostic performance of the FAI. This study observed for the first time that there were no significant differences between vessel-specific FAI and lesion-specific FAI scores, and that there was a strong, almost collinear association between them. No significant difference in diagnostic performance was noted between the two. Therefore, the results of this study suggest that both FAI measurement methods could be applied to clinical practice. The vessel-specific FAI seems to be a more convenient and appropriate method than lesion-specific FAI because the former was measured automatically using software and it could reduce human error, meaning it had higher reliability. However, a large sample study is still needed for verification.
Patients with vessel-specific FAI scores ≥ −70.97 HU and lesion-specific FAI scores ≥ −73.95 HU tend to have higher LAP volumes. After adjustment, the LAP volume was shown to be an independent risk factor for vessel-specific FAI and lesion-specific FAI scores, and we also found that the LAP volume was positively correlated with both of these methods. Similar to our findings, Goeller et al. [30] demonstrated an increase in the burden of LAP marked by a significant increase in PVAT attenuation (p = 0.04), and there was an association between LAP burden and increased PVAT attenuation (R = 0.24, p = 0.01). LAP was the alternative to the presence of the necrotic core. The necrotic core causes inflammation and oxidative stress by improving the levels of vasoconstrictors and by reducing the production and bioavailability of vasodilators [28,31]. Moreover, vascular inflammation is associated with the FAI. Thus, the presence of necrotic cores is associated with the FAI, and this study supports these findings. In addition, CS is considered to be a specific marker of atherosclerotic burden, which might have a common inflammatory background with PVAT. A recent study speculated that there might be a potential correlation between PVAT and CS [32]. However, the results of this study found that there was no relationship between the FAI and CS, which may be related to the patient population and drug treatment received in this study.
This study still had some limitations. First, this study was a retrospective post hoc analysis of existing data. Thus, there may be potential selection bias in this study. Second, other CT-based techniques of functional assessment such as fractional flow reserve derived from computed tomography (FFR CT ) have been widely introduced and used. Intraindividual comparisons need to be performed in the future to determine the best single or combination approach. Finally, this study lacks clinical outcome data, and further clinical outcome studies are still needed to analyze the effectiveness of these methods.

Conclusions
The FAI is an additional tool used to identify patients with relevant stenosis, and the combined use of a vessel-specific FAI or lesion-specific FAI and CCTA could improve the diagnostic performance of ischemia compared with CCTA alone. Thus, the need for further invasive treatment can be better assessed in patients. The LAP volume is the independent risk factor used for both tools.