Use of Computed Tomography in the Assessment of Severity of Aortic Valve Stenosis

Abstract

The workhorse in the diagnosis of aortic stenosis (AS) has been transthoracic echocardiography (TTE) with clear-cut validated threshold values for grading it mild, moderate, or severe. However, up to one-third of patients may present with discordant findings on echo sonogram and may need further evaluation with other imaging modalities such as computed tomography (CT). CT is useful in determining aortic valve area (AVA) by planimetry and outperforms TTE in identifying severe AS in bicuspid aortic valve (BAV), but it is not routinely ordered for those purposes. It has been widely used in helping, determining, and grading the severity of AS by calculating aortic valve calcium (AVC) load with a scoring system. AVC scores of 2000 AU or more for men and 1300 AU for women are highly indicative of severe AS and have been associated with the poor outcomes. AVC score will underestimate AS in a minority of circumstances where the process is driven more by fibrosis than calcification. CT use is limited by its recent adoption into medical practice and, therefore, is still not universally available in every center. It requires additional training for providers and low-dose radiation exposure may be a concern for some patients.

1. Introduction

Aortic stenosis (AS) is one of the most common valvular diseases in the developed world and its prevalence increases with age. It is estimated that up to 10% of octogenarians suffer from it [1]. It is expected that with aging populations, the prevalence will only increase, being an important condition for most healthcare systems given its progressive nature and associated morbidity and mortality. However, the spike in symptomatology and mortality occurs when the stenosis becomes severe. There is no effective medical treatment to reverse or slow progression of the disease so most therapeutic solutions have been focused on replacing the stenotic valve through surgery or, more recently, via catheter [2]. Given increased safety and recent therapeutic advances in transcatheter aortic valve replacement (TAVR), more centers are performing an increasing number of them. However, this is still an expensive procedure and complications do occur. Therefore, increased importance has been given to determining severity of the disease to better assess which patient, at which time, would benefit the most from a therapeutic intervention.

Severity of AS is guided by transthoracic echo sonogram (TTE) findings. According to the most recent guidelines by the American and European Societies of Cardiology (American College of Cardiology [ACC], American Heart Association (AHA), European Society of Cardiology [ESC], and European Association for Cardio-Thoracic Surgery [EACTS]), AS can be assumed to be severe when aortic valve area (AVA) is equal or less than 1.0 cm² (or 0.6 cm²/m² of body surface area [BSA]) and its mean pressure gradient (MG) is equal or higher than 40 mmHg (alternatively, peak aortic jet velocity of at least 4 m/s is also accepted) [3, 4].

Frequently, patients present with TTE measurements of AVA and MG that would place the severity of their AS in different grading categories. Most commonly, this scenario implies an AVA of 1.0 cm² or less (putting the patient in the category of severe AS), but an MG less than 40 mmHg (which would establish the patient’s AS as moderate). This grading inconsistency can be present in up to one-third of patients [5] and is usually referred to as low-flow-low-gradient AS (LFLG) if stroke volume index (Svi) is less than equal or less than 35 ml/m². Elevated gradient is the most robust parameter when assessing a stenotic lesion of the aortic valve, and a high gradient AS is indicative of its severity. However, AVA less than 1.0 cm² has been the best predictor for severe outcomes in AS [6] so its presence should still prompt thorough evaluation beyond TTE regardless of low gradient. LFLG AS can be further classified based on associated left ventricular ejection fraction (LVEF).

LFLG AS with reduced LVEF needs to be teased out from pseudo-severe AS and is usually assessed with dobutamine stress echo. An increase in AVA with increased flow through the valve is indicative of pseudo-severe AS. However, no change in AVA or gradient with no increase in flow through the valve is indicative of no reserve in the left ventricle. AS severity in those cases is also hard to tease. LFLG AS with preserved LVEF, also called paroxysmal, is challenging to assess for true severe AS versus other clinical conditions that would explain low flow independently from the aortic valve, usually atrial fibrillation, mitral stenosis, mitral regurgitation, or right ventricular (RV) failure [7]. Errors in TTE measurements of AVA are also common, given the anatomical characteristics of left ventricular outflow tract (LVOT), especially given its diameter is squared for calculation of AVA. This is when multidetector computed tomography (MDCT) plays a crucial role in the assessment of the severity of AS, especially when dobutamine stress echo is inconclusive or cannot be performed [8].

CT imaging for planning transcatheter aortic valve replacement (TAVR) is also crucial as it helps define the anatomy of aortic annulus and LVOT, reduces post-TAVR complications, and aides with the selection of vascular access [9, 10]. However, that role of CT is not in the purview of this chapter and we will only focus on its role in helping define its severity.

2. Aortic valve calcium scoring

Aortic valve calcium (AVC) score can determine whether true AS is present, regardless of flow [11]. For acquiring a validated AVC score, obtained images have to be non-contrast, electrocardiogram (ECG) gated in diastole (60–80% of RR interval), slice thickness of 3 mm, applied tube voltage of 120–140 kilovolts (KV), and a tube current of 30–80 milliampere seconds (mAs) based on patient body weight. Contrast-enhanced CT images have not been validated for accurately predicting calcium load or outcomes in AS [12].

The way to measure AVC is through a modified Agatston method. For every cluster of four pixels with an attenuation of 130 Hounsfield units (HU) or more, one arbitrary unit (AU) gets assigned. There is a density weighing factor (DWF) that derives from the highest Hounsfield unit in the lesion when it was originally designed for coronary artery calcium scoring. The area of the lesion gets multiplied by the DWF (130–199 HU = 1, 200–299 HU = 2, 300–399 HU = 3, and > 400 HU = 4) and then the areas with calcification are summed to give a total AVC score [11, 12, 13]. The software identifies those calcific regions but then the operator must manually select the ones that will be included in the score calculation. Areas that are considered for the AVC score are the AV leaflets as well as the annulus in axial slices. LVOT calcification is sometimes difficult to differentiate from AV and should not be included in total AVC score even though its presence is associated with post-TAVR peri-valvular leak [12]. Possible anatomical structures apart from the LVOT that may get erroneously included in the calculation of AVC score are calcium in the aortic root, right coronary ostium, and anterior mitral valve annulus. Use of different orientations and reconstructions of CT images, such as the “en face” (short axis) may help differentiate structures when there is a high calcium burden in surrounding structures. However, the AVC score should be calculated in axial views rather than in those reconstructions [12].

The presence of AVC has been independently associated with an increased risk of all-cause mortality [14], but a more specific and validated score can be helpful in grading the severity of AS and maybe determining who is a candidate for a life-saving intervention, such as TAVR. Initially, an AVC score of more than 1274 AU in women (sensitivity 86%, specificity 89%) and 2065 AU in men (sensitivity 89%, specificity 80%) were found to be highly indicative of severe AS with a sensitivity and specificity close to 90% [15]. A subsequent larger study found similar thresholds for severe AS; 1377 AU in women (sensitivity 87%, specificity 84%) and 2062 AU for men (sensitivity 80%, specificity 82%) [16]. Most of the patients in this study had a reduced EF (average 21 ± 4.6%) [16]. These are just absolute AU numbers. However, some patients with paroxysmal LFLG AS may have a smaller AV annulus but still have absolute AU values that do not reach the above-mentioned threshold but may still have severe AS. Indexing the calcium score to the valve area provides the AVC density, which was a more powerful predictor of survival than AVC load but threshold may need to be revised, especially for women [17]. AVC density for severe AS differs between gender; 420 AU/cm² (292 AU/cm² in the previous study [17]) or more for women and 527 AU/cm² for men [16]. In a prospective study, AVC density has been found to correlate well with severe AS and bicuspid aortic valve (BAV) in the age group of over 51 years of age but not in younger individuals with BAV [18]. AVC score may underestimate AS in young patients. An observational study found higher AVC score in patients with BAV compared to tricuspid AS (510 AU vs. 0 AU) in addition to earlier calcification of the AV (as early as 4th decade of life). The fusion raphe was the most common location for calcific deposits in BAV followed by the cusp in relation with the left coronary artery. For tricuspid AV, the noncoronary cusp was the most common location with evidence of calcification [19].

A few studies have investigated ethnic differences in AVC. In a large prospective cohort study in 6814 individuals without symptoms or known cardiovascular disease, after adjustment for risk factors, relative risk (RR) for AVC was similar between Caucasians and Hispanics (1.03 in Hispanics with 95% CI 0.82–1.28). Compared with Caucasians, RR was 0.72 in Blacks (95% CI 0.59–0.90) and 0.56 in Chinese (95% CI 0.40–0.80). These differences were not specific to patients with AS [20]. More recently another study suggested AVC score thresholds for severe AS are comparable in Asian (68% of study population) and Caucasian population but were less accurate for Asian women when compared to Caucasian women, suggesting fibrosis and not calcification as an important driver of stenosis in this population [21].

The 2019 consensus document from the Society of Cardiovascular Computed Tomography defines the cutoff for AVC score for severe AS as 3000 AU or more in men and 1600 AU or more in women [22], which has been added to the 2021 ESC/EACTS guidelines as highly likely for severe AS. Any score less than those previously mentioned but 2000 AU and above for men and 1200 AU for women is considered “likely” for severe AS. AVC scores of less than 1600 AU for men and 800 AU for women are considered “unlikely” to represent severe AS [23]. ACC and AHA in their 2020 guidelines have mentioned AVC scores of 2000 AU or more for men and 1300 AU for women as diagnostic for severe AS [4].

AVC scoring has also been shown to progress with time at an average of 152 AU/year and progression was faster with severe disease (342 AU/year) compared to moderate AS (289 AU/year) and mild AS (64 AU/year) [24]. This expands the possible use of CT also to track progression of disease in patients with LFLG AS with preserved EF or in patients with poor windows for TTE. Given its reproducibility and sensitivity, CT could also serve as a tool to track disease progression while researching medical therapies looking to prevent or slow progression of AS as it would mean a smaller sample size needed to detect a change in AS.

2.1 Limitations and advantages of aortic valve calcium score

MDCT use is limited by its recent adoption into medical practice and therefore is still not universally available in every center. It requires additional training for providers and low-dose radiation exposure may be a concern for some patients. As pointed out before, MDCT AVC score will underestimate AS in a minority of circumstances where the process is driven more by fibrosis than calcification (BAV and young females) [18, 19].

However, MDCT AVC score has many attributes that should make it easy to introduce in daily clinical practice. AVC score is reliable, reproducible, independent of flow, has low (average < 5% of score) interobserver and intraobserver variability [25, 26] (unlike TTE), and it can be performed with an array of scanners that already exist and established thresholds remain valid [16]. In addition, no contrast is needed. It is important to know that AVC score has not been validated in contrast-enhanced scans and it may greatly differ from non-contrast images [12].

3. Anatomic assessment

AV planimetry can be used to measure AVA and the LVOT. CT is superior to TTE in determining valve anatomy but has not shown its superiority in improving the correlation between AVA and MG or in predicting mortality [27]. To measure AVA, the CT has to be ECG-gated as well and the smallest AV opening is chosen during systole (15–35% of RR interval) when the valve is fully open [12] as you can see in Figure 1. Measured AVA by MDCT in severe AS is larger than AVA measured by TTE (1.2 cm² vs. 1.0 cm², respectively) [27], which has also been found in earlier studies [28]. A small meta-analysis of 9 studies with 262 men and 175 women found that AVA measurements by planimetry were very similar to AVA obtained by continuation on TTE, but consistently overestimated it [29], suggesting that the CT-measured AVA threshold for severe AS should be less than 1.2 cm², but this threshold difference has not been included in the guidelines. This discrepancy has been present in most studies, and it is thought that the difference stems from the fact that flow through the stenotic valve will not be equal in the middle and at the edges of the effective orifice area (EOA). This difference in measured AVA has been found comparable to other imaging modalities in a recent pairwise meta-analysis, with a mean AVA difference of 0.12 cm² over the one calculated by TTE (0.14 cm² specifically for the MDCT subgroup) [30].

Figure 1.

Aortic valve area (AVA) calculation tracing the edge of the aortic leaflets that border the smallest aortic valve (AV) opening during systole. Measured AVA 0.8 cm² (78.53 mm²). Note the three distinct leaflets of this tricuspid AV with presence of calcium. “EnFace” view is used to better see the valve. Calcium present is not used to calculate aortic valve calcium (AVC) score.

LVOT measurement by MDCT is another tool that has been used to better study the valve. The reconstruction of the LVOT has been fundamental in planning for TAVR, specifically in selecting the correct valve size and preventing post-TAVR valve leaks [10]. However, measuring LVOT on MDCT has also been helpful in grading the severity of AS by calculating a hybrid AVA using Hybrid MDCT-Doppler imaging: the use of TTE and MDCT measurements in the continuity equation [31]. Inaccurate LVOT measurements by TTE are one of the most common ways error can be introduced in the continuity equation, leading to an underestimation of gradients across the AV. CT is able to obtain an accurate LVOT area that can be used in the continuity equation and eliminate some of the variability that standard TTE introduces. Several studies have shown that, when compared to MDCT, TTE underestimates AVA and LVOT areas and in some instances, the use of a hybrid AVA (or sometimes called fusion AVA) helped reclassify a big proportion of patients into a different severity grading [32, 33, 34]. One study performed on 359 consecutive patients with low gradient severe AS, who already had TAVR, recalculated AVA based on CT and TTE parameters combined and reclassified 35% of them as moderate based on the new AVA. Even though their reclassification did not affect clinical outcomes, it shows the extent that combined imaging can help correctly grade the severity of AS in this subset of patients [35]. A similar study in 422 patients found that about 30% of patients were reclassified after calculating hybrid AVA without any difference in clinical outcomes 2 years post TAVR [36]. It is important to note that clinical outcomes were similar in trials, regardless of reclassification or not; however, there was no control group as every patient received a TAVR. The biggest change when using a hybrid AVA is usually a higher number of concordant moderate AS (in the patients who had originally low gradient severe AS) and a higher number of discordant high gradient moderate AS (who previously had a high gradient severe AS and now the AVA is being recalculated to >1 cm²). However, revisiting the threshold for severe AS to 1.2 cm² when obtained by CT may help define better the latter group of patients. Due to lack of more trials and no apparent association with outcomes, the use of the hybrid AVA is not part of the guidelines yet [37] but it could be helpful in cases where the LVOT cannot be optimally visualized in TTE.

Cardiac CT can also be useful in detecting BAV when it is difficult by echo sonogram as it has better sensitivity and specificity (94.1% vs. 76.5% and 100% vs. 60.6%) when compared to TTE [28]. Presence of BAV has not been a limitation in measuring AVA [29]. BAVs have been found to be heavier than tricuspid AV in a study of excised severe AS when undergoing surgical AV replacement [38], suggesting a higher amount of calcium and fibrosis likely due to increased endothelial damage and increased repetitive mechanical stress on the valve.

4. Conclusion

Computed tomography is a useful tool in helping determine the severity of AS in patients, but it has not replaced TTE as the main tool in diagnosing it. CT-derived AVC score is most useful in establishing true AS severity in LFLG AS with preserved EF or in cases where dobutamine stress echo cannot be performed or results are inconclusive. AVC score is reliable, reproducible, independent of flow, and has low interobserver variability, but it can underestimate severity of AS in BAV and young women. Anatomic measurement of AVA is also possible but usually overestimates it compared to TTE. LVOT measurement is more reliable than its TTE counterpart and it could be used in the continuity equation to classify patients with AS more accurately.

Acknowledgments

We would like to acknowledge Dr. Pragya Ranjan for her help in guiding us in writing this chapter as well as helping with the analysis of the images.

No external funding was used for the development of this chapter.

Conflict of interest

The authors declare no conflict of interest.