Performance of Computed Tomographic Angiography–Based Aortic Valve Area for Assessment of Aortic Stenosis

Abstract

Background

A total of 40% of patients with severe aortic stenosis (AS) have low‐gradient AS, raising uncertainty about AS severity. Aortic valve calcification, measured by computed tomography (CT), is guideline‐endorsed to aid in such cases. The performance of different CT‐derived aortic valve areas (AVAs) is less well studied.

Methods and Results

Consecutive adult patients with presumed moderate and severe AS based on echocardiography (AVA measured by continuity equation on echocardiography <1.5 cm²) who underwent cardiac CT were identified retrospectively. AVAs, measured by direct planimetry on CT (AVA_CT) and by a hybrid approach (AVA measured in a hybrid manner with echocardiography and CT [AVA_Hybrid]), were measured. Sex‐specific aortic valve calcification thresholds (≥1200 Agatston units in women and ≥2000 Agatston units in men) were applied to adjudicate severe or nonsevere AS. A total of 215 patients (38.0% women; mean±SD age, 78±8 years) were included: normal flow, 59.5%; and low flow, 40.5%. Among the different thresholds for AVA_CT and AVA_Hybrid, diagnostic performance was the best for AVA_CT <1.2 cm² (sensitivity, 85%; specificity, 26%; and accuracy, 72%), with no significant difference by flow status. The percentage of patients with correctly classified AS severity (correctly classified severe AS+correctly classified moderate AS) was as follows; AVA measured by continuity equation on echocardiography <1.0 cm², 77%; AVA_CT <1.2 cm², 73%; AVA_CT <1.0 cm², 58%; AVA_Hybrid <1.2 cm², 59%; and AVA_Hybrid <1.0 cm², 45%. AVA_CT cut points of 1.52 cm² for normal flow and 1.56 cm² for low flow, provided 95% specificity for excluding severe AS.

Conclusions

CT‐derived AVAs have poor discrimination for AS severity. Using an AVA_CT <1.2‐cm² threshold to define severe AS can produce significant error. Larger AVA_CT thresholds improve specificity.

Clinical Perspective.

What Is New?

• Among aortic valve areas (AVAs) measured by direct planimetry on computed tomography (CT) (AVA_CT) <1.0‐cm², AVA_CT <1.2‐cm², AVA measured in a hybrid manner with echocardiography and CT (AVA_Hybrid) <1.0‐cm², and AVA_Hybrid <1.2‐cm² thresholds, AVA_CT <1.2 cm² provided the best accuracy and correct classification for aortic stenosis (AS) severity.

• There was no significant impact of different flow states on the diagnostic performance of AVA_CT or AVA_Hybrid.

• AVA_CT and AVA_Hybrid have poor discrimination for AS severity. The AVA_CT <1.2‐cm² threshold has good sensitivity (85%) but poor specificity (26%). The AVA_CT cut points of 1.52 cm² for normal flow and 1.56 cm² for low flow provided 95% specificity for excluding severe AS.

What Are the Clinical Implications?

• CT‐derived AVAs have poor discrimination for AS severity. Using an AVA_CT <1.2‐cm² threshold to define severe AS can produce significant error.

• Just like AVA measured by continuity equation on echocardiography <1.0 cm², AVA_CT <1.2 cm² is a sensitive but a nonspecific marker for severe AS. Larger AVA_CT thresholds improve specificity.

Nonstandard Abbreviations and Acronyms

AS

aortic stenosis

AVA

aortic valve area

AVA_CT

aortic valve area measured by direct planimetry on computed tomography

AVA_Echo

aortic valve area measured by continuity equation on echocardiography

AVA_Hybrid

aortic valve area measured in a hybrid manner with echocardiography and computed tomography

AVC

aortic valve calcification

MG

mean transvalvular pressure gradient

SVI

stroke volume index

TAVR

transcatheter aortic valve replacement

Aortic stenosis (AS) is the most common valvular heart disease in the high‐income countries, ¹ with >200 000 aortic valve replacements performed worldwide annually. Severe AS is defined on the basis of echocardiography parameters, with a peak aortic jet velocity ≥4 m/s or mean transvalvular pressure gradient (MG) ≥40 mm Hg and an aortic valve area (AVA) <1.0 cm². ² , ³ However, up to 40% of patients with severe AS have discordant or low‐gradient AS, with AVA measured by the continuity equation on echocardiography (AVA_Echo) <1.0 cm² and MG <40 mm Hg. ⁴ , ⁵ The management of this subset of patients is challenging because the AVA‐gradient discrepancy raises uncertainty about the actual stenosis severity.

Computed tomography (CT) is now endorsed by the American and European guidelines to aid in the diagnosis of low‐gradient AS. ² , ³ Aortic valve calcification (AVC) assessed on CT is a robust arbiter of AS severity and prognosis, with proposed sex‐specific cutoffs. ⁶ , ⁷ In addition, AVA measured by direct planimetry on CT (AVA_CT), offering the theoretical advantage of direct orifice measurement. ⁸ Although AVA_CT is larger than AVA_Echo, ⁹ , ¹⁰ the specific threshold for severe AS has not been defined. A hybrid method where the left ventricular outflow tract (LVOT) area is measured by CT and velocities are measured by echocardiography has also been proposed to calculate the AVA (aortic valve area measured in a hybrid manner with echocardiography and computed tomography [AVA_Hybrid]), with a higher threshold of <1.2 cm² as the discriminant for severe AS. ¹¹ In distinction to AVC, these CT‐derived AVAs are less well studied, and moreover it is not known whether these are affected by different flow states.

In a retrospective cohort of patients with presumed moderate and severe AS based on echocardiography, we sought to explore the diagnostic performance of AVA_CT and AVA_Hybrid, stratified by different flow states.

Methods

Study Cohort

Consecutive adult patients with presumed severe AS (AVA_Echo <1.0 cm²) and moderate AS (AVA_Echo1.0–1.5 cm²) who underwent a cardiac CT between 2017 and 2020 were identified retrospectively from the TAVR (Transcatheter Aortic Valve Replacement) CT Registry at the University of Minnesota. Patients aged <18 years and those with nondiagnostic CT images, >6 months’ duration between echocardiogram and CT, rheumatic or radiation‐induced aortic stenosis, and moderate or greater other valve disease (aortic regurgitation, mitral stenosis, and mitral regurgitation) were excluded. Figure S1 shows the study flow diagram. The Institutional Review Board at the University of Minnesota approved the study with a waiver of informed consent. The data underlying this study cannot be shared publicly because of the privacy of individuals who participated in the study.

Echocardiography Protocol

All echocardiograms were performed in accordance with current American Society of Echocardiography guidelines. ¹² Machines used were Philips IE33 or Epiq 7C (Philips Healthcare, Andover, MA) or GE Vivid E95 (GE Healthcare, Chicago, IL). Aortic stenosis assessment was in accordance with the American Society of Echocardiography guideline for evaluation of valvular heart disease. ²

Echocardiography Valve Analysis

All echocardiograms were reviewed by a single investigator JA, with board certification in echocardiography, blinded to all other imaging and clinical findings. AVA_Echo was calculated using the continuity equation. The LVOT velocity time integral was multiplied by the aortic annular area (derived from the LVOT diameter) and indexed to body surface area, providing the stroke volume index (SVI). LVOT diameter was measured ≈3 to 10 mm from the aortic valve cusp insertion plane. In patients with atrial fibrillation, an average of 5 spectral Doppler profiles were used to calculate velocity and gradient. AS severity was assessed according to the American Society of Echocardiography guidelines. ²

CT Protocol

All CT scans were performed on a dual‐source 128‐slice scanner (Siemens SOMATOM Definition FLASH scanner) in accordance with the Society of Cardiovascular Computed Tomography consensus document. ¹³ Images were acquired with retrospective ECG gating with dose modulation following the administration of 100 to 120 mL (weight‐based) intravenous contrast (Isovue 370; Bracco Diagnostics, Monroe Township, NJ). Acquisition and reconstruction parameters included the following: tube voltage of 100 or 120 kVp (the latter used in patients with body mass index >30 kg/m²), 0.6‐mm slice thickness, 0.3‐mm slice increment, 250‐mm field of view, and 512×512 matrix. A noncontrast chest CT scan, gated in diastole with a tube voltage of 120 kVp and tube current adjusted to patient size, was also performed.

CT Valve Analysis

All CT scans were reviewed by a single investigator PSN, with board certification and 8 years' experience in cardiac CT, blinded to all other imaging and clinical findings. AVC was assessed on the noncontrast chest CT scan using the Agatston method on TeraRecon software (TeraRecon, Foster City, CA). ⁷ Briefly, AVC was quantified on contiguous 3‐mm axial slices starting at the aortic annulus, excluding calcium originating from extravalvular structures, such as the mitral valve annulus, the ascending aorta, and coronary arteries. The total AVC in Agatston units was calculated, with sex‐specific thresholds to determine severity (≥1200 Agatston units in women and ≥ 2000 Agatston units in men). ³ AVA_CT was calculated from the multiphase contrast CT scan, as previously described. ¹⁴ Briefly, a midsystolic phase with the maximum aortic valve opening was selected. Images were reformatted by using 3 orthogonal planes from multiplanar reconstruction and aligning the crosshairs parallel to the valve plane to produce a true short‐axis view through the aortic valve. The smallest orifice at the leaflet tips was obtained, and direct planimetry was performed along the inner edges of the leaflets to obtain the AVA_CT. AVA_Hybrid was derived by substituting the average aortic annulus diameter measured on CT for the LVOT diameter in the continuity equation. The aortic annulus was traced at the lowest implantation base of aortic cusps during midsystole.

AS Definitions ⁵

Severe AS was dichotomized on the basis of flow, with normal flow having SVI ≥35 mL/m² and low flow having SVI <35 mL/m². Severe AS was categorized as concordant when findings from both the velocity/gradient and AVA provided the same assessment of disease severity (AVA <1.0 cm² and peak aortic jet velocity ≥4.0 m/s or MG ≥40 mm Hg), and discordant when the velocity/gradient and AVA provided different assessment of disease severity (AVA <1.0 cm² and peak aortic jet velocity >4.0 m/s or MG <40 mm Hg). Patients with normal flow were further subdivided on the basis of AVA‐gradient concordance into those with concordance (concordant severe AS) and those with discordance (discordant normal‐flow severe AS). Patients with discordant low flow were further subdivided into those with a low (<50%; classic low‐flow severe AS) or preserved (≥50%; paradoxical low‐flow severe AS) left ventricular ejection fraction.

Statistical Analysis

Descriptive statistics included the mean with SD or median with interquartile range for continuous variables, and frequency with percentage for categorical variables. To compare patient characteristics stratified by flow or AS severity, t‐tests or Kruskal‐Wallis tests were used for continuous variables, and χ² tests were used for categorical variables. For this study, we used sex‐specific AVC thresholds as the reference standard for AS severity, and to arbitrate whether AS was severe or nonsevere. Measures of diagnostic performance, including sensitivity, specificity, negative predictive value, positive predictive value, and area under the receiver operating characteristic curve were calculated for AVA_CT and AVA_Hybrid. Thresholds of <1.0 and <1.2 cm² were tested. Interobserver agreement for AVA_CT and AVA_Hybrid was calculated in a randomly selected sample of 20 patients. All analyses were performed in Stata, version 17 (StataCorp, 2021; Stata Statistical Software: Release 17; StataCorp LLC, College Station, TX).

Results

Participant Characteristics

A total of 267 patients with presumed severe and moderate AS based on echocardiography who had cardiac CT performed were identified. A total of 52 patients were excluded: 35 because of moderate or greater other valve disease, 7 because of poor CT image quality, 7 because of >6 months’ duration between CT and echocardiogram, 2 because of discordant findings (AVA_Echo >1.0 cm² and MG >40 mm Hg), and 1 because of rheumatic aortic stenosis. A final cohort of 215 patients (189 with presumed severe AS and 26 with moderate AS based on echocardiography) were included in this study (Figure S1). Bicuspid aortic valve was present in 27 (13%) patients. Symptoms were present in 165 (76.7%) patients, with angina in 34 (15.8%), dyspnea in 155 (72.1%), and syncope in 9 (4.2%). Table 1 provides a detailed description of baseline characteristics. Further evaluation of AS severity was performed with a dobutamine stress echocardiogram in 17 (7.9%) and catheterization in 15 (7%). The results of this additional testing are provided in Table S1.

Table 1.

Characteristics of Patients With AS Stratified by Flow

Characteristic	Entire cohort	Normal flow	Low flow	P value
No.	215	128	87
Age, y	78±8	77±9	79±8	0.32
Women	82 (38)	56 (44)	26 (30)	0.040
Race and ethnicity
White	205 (95)	120 (94)	85 (98)	0.27
Black	4 (1.9)	4 (3.1)	0
Other*	6 (2.8)	4 (3.1)	2 (2.3)
BMI, kg/m2	29.0±6.6	28.2±5.8	30.0±7.4	0.047
BSA, m2	1.94±0.26	1.90±0.25	2.00±0.26	0.004
Aortic valve cusps
Bicuspid	27 (13)	22 (17)	5 (5.8)	0.013
Tricuspid	188 (87)	106 (83)	82 (94)
CAD	152 (71)	88 (69)	64 (74)	0.45
CKD	91 (42)	51 (40)	40 (46)	0.37
Diabetes	83 (39)	46 (36)	37 (43)	0.33
Hypertension	182 (85)	106 (83)	76 (87)	0.36
Smoking
Ever	134 (62)	78 (61)	56 (64)	0.61
Never	81 (38)	50 (39)	31 (36)
AF	109 (51)	57 (45)	52 (60)	0.028
Echocardiographic findings
Stroke volume, mL	75±19	86±16	59±11	<0.001
Stroke volume index, mL/m2	39±10	46±8	29±4	<0.001
AV peak velocity, m/s	3.8±0.7	3.9±0.6	3.5±0.7	<0.001
AV mean gradient, mm Hg	35±13	38±12	31±13	<0.001
AVA continuity, cm2	0.85±0.18	0.90±0.18	0.79±0.18	<0.001
AV VTI	90.0±21.4	97.6±18.7	78.7±20.0	<0.001
DVI ratio	0.24±0.05	0.25±0.05	0.22±0.05	<0.001
LVOT VTI	21.3±5.9	24.1±5.3	17.2±4.2	<0.001
LVEF, %
>55	172 (80)	113 (88)	59 (68)	0.001
50–55	10 (4.7)	3 (2.3)	7 (8.1)
30–50	27 (13)	12 (9.4)	15 (17)
<30	6 (2.8)	0 (0)	6 (6.9)
LVEF, %	61 (58–63)	63 (58–63)	58 (50–63)	<0.001
RV dysfunction
No	190 (88)	121 (95)	69 (79)
Mild	17 (7.9)	7 (5.5)	10 (11)
Moderate	6 (2.8)	0	6 (6.9)
Severe	2 (0.93)	0	2 (2.3)
CT findings
AV annulus/LVOT, mm	25.0±2.4	24.9±2.3	25.3±2.4	0.29
AV annulus/LVOT area, cm2	4.97±0.95	4.91±0.93	5.05±0.97	0.29
AV calcium score, AUs	2414 (1597–3305)	2335 (1541–3520)	2453 (1902–3279)	0.52
AV calcium above threshold	169 (79)	100 (78)	69 (79)	0.84
AVACT, cm2	0.99±0.27	1.01±0.29	0.97±0.23	0.36
AVACT <1.0 cm2	125 (58)	73 (57)	52 (60)	0.69
AVACT <1.2 cm2	177 (82)	101 (79)	76 (87)	0.11
AVAHybrid	1.17±0.26	1.21±0.25	1.11±0.27	0.011
AVAHybrid <1.0 cm2	78 (36)	41 (32)	37 (43)	0.12
AVAHybrid <1.2 cm2	139 (65)	80 (63)	59 (68)	0.42

Values presented are frequency (percentage) for categorical variables or mean±SD for continuous variables. AF indicates atrial fibrillation; AS, aortic stenosis; AU, Agatston unit; AV, aortic valve; AVA, AV area; AVA_CT, AVA measured by direct planimetry on CT; AVA_Hybrid, AVA measured in a hybrid manner with echocardiography and CT; BMI, body mass index; BSA, body surface area; CAD, coronary artery disease; CKD, chronic kidney disease; CT, computed tomography; DVI, Doppler velocity index; LVEF, left ventricular ejection fraction; LVOT, left ventricular outflow tract; RV, right ventricular; and VTI, velocity time integral.

^*Other includes Hispanic, Asian or Native American.

Echocardiogram Findings in Entire Cohort

Mean±SD AVA_Echo was 0.85±0.18 cm², mean±SD MG was 35±13 mm Hg, and mean±SD peak velocity was 3.8±0.7 m/s. On the basis of SVI, 128 (59.5%) had normal flow and 87 (40.5%) had low flow (Table 1). The subtypes for the cohort with presumed severe AS (n=189) were as follows: concordant in 87 (46.1%), discordant normal flow in 44 (23.6%), classic low flow in 14 (7.3%), and paradoxical low flow in 44 (23.0%). Table S2 provides details on AS subtypes.

CT Findings in Entire Cohort

Median AVC was 2414 (interquartile range, 1597–3305) Agatston units, with 169 (79%) having AVC greater than sex‐specific thresholds. AVA_CT was feasible with acceptable image quality in 97.4% of the cohort. Figure 1 shows representative examples of AVA_CT. In a randomly selected sample of 20 patients for interobserver agreement, intraclass correlation coefficient for AVA_CT was 0.97 (interquartile range, 0.92–0.99); and for AVA_Hybrid, it was 0.98 (interquartile range, 0.96–0.99). Mean±SD AVA_CT was 0.99±0.27 cm², and mean±SD AVA_Hybrid was 1.17±0.26 cm². From the 169 patients with severe AS based on AVC, 67 (39.6%) had AVA_CT >1.0 cm², 26 (15.4%) had AVA_CT >1.2 cm², 105 (62.1%) had AVA_Hybrid >1.0 cm², and 59 (34.9%) had AVA_Hybrid >1.2 cm². From the 46 patients with nonsevere AS based on AVC, 23 (50.0%) had AVA_CT <1.0 cm², 34 (73.9%) had AVA_CT <1.2 cm², 14 (30.4%) had AVA_Hybrid <1.0 cm², and 29 (63.0%) had AVA_Hybrid <1.2 cm². Diagnostic performance measures of AVA_CT and AVA_Hybrid with <1.0‐ and <1.2‐cm² thresholds, using AVC sex‐specific thresholds as the reference standard, are listed in Table 2. AVA_CT cut points of 1.52 cm² for normal flow, and 1.56 cm² for low flow, provided 95% specificity for excluding severe AS.

Figure 1. Representative examples of computed tomography (CT) images highlighting reclassification of aortic stenosis (AS) severity based on CT‐derived aortic valve area (AVA).

AU indicates Agatston unit; AVA_CT, AVA measured by direct planimetry on CT; AVA_Echo, AVA measured by continuity equation on echo; AVC, aortic valve calcification; echo, echocardiography; MG, mean transvalvular pressure gradient; and SVI, stroke volume index.

Table 2.

Diagnostic Performance of CT‐Derived AVA

AS type	No.	Sensitivity, %	Specificity, %	PPV, %	NPV, %	Accuracy, %	AUC
AVACT threshold 1.2 cm2
Whole cohort	215	85	26	81	32	72	0.55
Normal flow	128	81	29	80	30	70	0.55
Low flow	87	90	22	82	36	76	0.56
AVACT threshold 1.0 cm2
Whole cohort	215	60	50	82	26	58	0.55
Normal flow	128	59	50	81	26	57	0.55
Low flow	87	62	50	83	26	60	0.56
AVAHybrid threshold 1.2 cm2
Whole cohort	215	65	37	79	22	59	0.51
Normal flow	128	62	36	78	21	56	0.49
low flow	87	70	39	81	25	63	0.54
AVAHybrid threshold 1.0 cm2
Whole cohort	215	38	70	82	23	45	0.54
Normal flow	128	33	71	81	23	41	0.52
Low flow	87	45	67	84	24	49	0.56
AVAEcho threshold 1.0 cm2
Whole cohort	215	91	22	81	40	76	0.56
Normal flow	128	91	32	83	50	78	0.56
Low flow	87	91	6	79	14	74	0.55

AS indicates aortic stenosis; AUC, area under the curve; AVA, aortic valve area; AVA_CT, AVA measured by direct planimetry on CT; AVA_Echo, AVA measured by continuity equation on echocardiography; AVA_Hybrid, AVA measured in a hybrid manner with echocardiography and CT; CT, computed tomography; NPV, negative predictive value; and PPV, positive predictive value.

Comparison of Normal‐ and Low‐Flow AS

Compared with the normal‐flow cohort, the low‐flow cohort had fewer women (30% versus 44%; P=0.040) and higher body mass index (30.0±7.4 versus 28.2±5.8 kg/m²; P=0.047). There were no differences in comorbidities, except atrial fibrillation was more frequent in the low‐flow cohort (60% versus 45%; P=0.028). The mean±SD SVI (29±4 versus 46±8 mL/m²), MG (31±13 versus 38±12 mm Hg), and peak velocity (3.5±0.7 versus 3.9±0.6 m/s) were all expectedly lower in the low‐flow cohort (P<0.001). The mean±SD AVA_Echo (0.79±0.18 versus 0.90±0.18 cm²) and Doppler velocity index ratio (0.22±0.05 versus 0.25±0.05) were lower in the low‐flow cohort (P<0.001). Median (interquartile range) left ventricular ejection fraction was lower in the low‐flow cohort (58 [50–63] versus 63 [58–63]; P<0.001). There was no difference in the percentage of patients with AVC above sex‐specific thresholds (79% in the low‐flow cohort versus 78% in the normal‐flow cohort; P=0.84). There was no difference in the AVA_CT (mean±SD, 0.97±0.23 cm² in the low‐flow cohort versus 1.01±0.29 cm² in the normal‐flow cohort; P=0.36), whereas AVA_Hybrid was smaller in the low‐flow cohort (mean±SD, 1.11±0.27 versus 1.21±0.25 cm²; P=0.011). There was no difference in the percentage of patients with AVA_CT or AVA_Hybrid below the <1.0‐ or <1.2‐cm² thresholds.

Comparison of Moderate and Severe AS Arbitrated by AVC

Compared with patients with moderate AS, those with severe AS had fewer women (33% versus 57%; P=0.004) and higher body mass index (29.4±6.7 versus 27.3±5.9 kg/m²; P=0.055). There were no significant differences in comorbidities between the 2 groups (Table 3). AVA_Echo (mean±SD, 0.84±0.17 versus 0.92±0.21 cm²; P=0.008) and Doppler velocity index ratio (mean±SD, 0.23±0.05 versus 0.27±0.06; P<0.001) were significantly lower in severe AS. There was no difference in SVI (mean±SD, 39±10 versus 39±11 mL/m²; P=0.80), whereas MG (mean±SD, 38±13 versus 26±7 mm Hg) and peak velocity (mean±SD, 3.9±0.6 versus 3.3±0.5 m/s) were significantly higher in severe AS (P<0.001). There was no difference in the median (interquartile range) left ventricular ejection fraction (60 [58–63] versus 62 [58–63]; P=0.79). AVA_CT was lower in severe AS but not statistically significant (mean±SD, 0.97±0.26 versus 1.05±0.28 cm²; P=0.071), with no difference in the percentage of patients with AVA_CT <1.2 cm² (85% versus 74%; P=0.092). There was no difference in the AVA_Hybrid (mean±SD, 1.16±0.26 versus 1.20±0.27 cm²; P=0.37) or percentage of patients with AVA_Hybrid <1.2 cm² (65% versus 63%; P=0.80).

Table 3.

Characteristics of Moderate Versus Severe AS Based on AVC (n=215)

Characteristic	Moderate AS	Severe AS	P value
No.	46	169
Age, y	80±8	77±9	0.12
Women	26 (57)	56 (33)	0.004
Race and ethnicity
White	45 (98)	160 (95)	0.48
Black	1 (2.2)	3 (1.8)
Other*	0	6 (3.6)
BMI, kg/m2	27.3±5.9	29.4±6.7	0.055
BSA, m2	1.83±0.24	1.97±0.26	0.001
Aortic valve cusps
Bicuspid	6 (13)	21 (12)	0.91
Tricuspid	40 (87)	148 (88)
CAD	29 (63)	123 (73)	0.20
CKD	24 (52)	67 (40)	0.13
Diabetes	18 (39)	65 (38)	0.93
Hypertension	43 (93)	139 (82)	0.067
Smoking
Ever	29 (63)	105 (62)	0.91
Never	17 (37)	64 (38)
AF	28 (61)	81 (48)	0.12
Echocardiographic findings
AVA continuity <1.0 cm2	36 (78)	153 (91)	0.024
Stroke volume, mL	71±21	77±19	0.061
Stroke volume index, mL/m2	39±11	39±10	0.80
AV peak velocity, m/s	3.3±0.5	3.9±0.6	<0.001
AV mean gradient, mm Hg	26±7	38±13	<0.001
AVA continuity, cm2	0.92±0.21	0.84±0.17	0.008
AV VTI	77.4±15.8	93.4±21.4	<0.001
DVI ratio	0.27±0.06	0.23±0.05	<0.001
LVOT VTI	20.8±6.4	21.5±5.8	0.49
LVEF, %
>55	35 (76)	137 (81)	0.72
50–55	2 (4.4)	8 (4.7)
30–50	7 (15)	20 (12)
<30	2 (4.4)	4 (2.4)
LVEF, %	62 (58–63)	60 (58–63)	0.79
RV dysfunction
No	39 (85)	151 (89)	0.18
Mild	3 (6.5)	14 (8.3)
Moderate	3 (6.5)	3 (1.8)
Severe	1 (2.2)	1 (0.6)
CT findings
AV annulus, mm	24.0±2.3	25.3±2.3	<0.001
AV calcium score, AUs	1193 (889–1517)	2798 (2172–3911)	NA
AVACT, cm2	1.05±0.28	0.97±0.26	0.071
AVACT <1.0 cm2	23 (50)	102 (60)	0.21
AVACT <1.2 cm2	34 (74)	143 (85)	0.092
AVAHybrid, cm2	1.20±0.27	1.16±0.26	0.37
AVAHybrid <1.0 cm2	14 (30)	64 (38)	0.35
AVAHybrid <1.2 cm2	29 (63)	110 (65)	0.80

Values presented are frequency (percentage) for categorical variables or mean±SD for continuous variables. AF indicates atrial fibrillation; AS, aortic stenosis; AU, Agatston unit; AV, aortic valve; AVA, AV area; AVA_CT, AVA measured by direct planimetry on CT; AVA_Hybrid, AVA measured in a hybrid manner with echocardiography and CT; AVC, aortic valve calcification; BMI, body mass index; BSA, body surface area; CAD, coronary artery disease; CKD, chronic kidney disease; CT, computed tomography; DVI, Doppler velocity index; LVEF, left ventricular ejection fraction; LVOT, left ventricular outflow tract; RV, right ventricular; and VTI, velocity time integral.

^*Other includes Hispanic, Asian or Native American.

AS Severity Classification

Figure 2 shows the AS severity classification by different AVAs using AVC threshold as the reference standard for AS severity. The percentage of patients with correctly classified AS severity (correctly classified severe AS+correctly classified moderate AS) was as follows; AVA_Echo <1.0 cm², 77%; AVA_CT <1.2 cm², 73%; AVA_CT <1.0 cm², 58%; AVA_Hybrid <1.2 cm², 59%; and AVA_Hybrid <1.0 cm², 45%.

Figure 2. Bar graph of aortic stenosis (AS) severity classification by echocardiography (echo)‐ and computed tomography (CT)–derived aortic valve area (AVA).

AVA_CT indicates AVA measured by direct planimetry on CT; AVA_Echo, AVA measured by continuity equation on echo; and AVA_Hybrid, AVA measured in a hybrid manner with echo and CT.

Figure 3 shows the reclassification of AS severity if AVA_CT <1.2‐cm² threshold is applied after grading by AVA_Echo <1.0 cm². From the 189 patients with severe AS based on AVA_Echo <1.0 cm², 24 (12.7%) are reclassified to moderate AS based on AVA_CT >1.2‐cm² threshold. Using AVC threshold as the reference standard for AS severity, only 7 of the 24 (29.2%) are correctly reclassified. From the 26 with moderate AS based on AVA_Echo >1.0 cm², 12 (46.2%) are reclassified to severe AS based on AVA_CT <1.2‐cm² threshold. Using AVC threshold as the reference standard for AS severity, 7 of the 12 (58.3%) are correctly reclassified.

Figure 3. Flowchart of reclassification of echocardiography (echo)‐graded aortic stenosis severity by computed tomography (CT).

AVA_CT indicates aortic valve area (AVA) measured by direct planimetry on CT; AVA_Echo, AVA measured by continuity equation on Echo; and AVC, aortic valve calcification.

Discussion

In a cohort of 215 patients with moderate and severe AS based on AVA_Echo, we made the following observations:

1. AVA_Hybrid, which provides an estimate of the effective AVA, is substantially larger than AVA_CT, which is the anatomic AVA. Because of the flow contraction phenomenon, the effective AVA should be smaller than the anatomic AVA. Hence, this suggests that AVA_Hybrid grossly overestimates the effective AVA.

2. Among AVA_CT <1.0‐cm², AVA_CT <1.2‐cm², AVA_Hybrid <1.0‐cm², and AVA_Hybrid <1.2‐cm² thresholds, AVA_CT <1.2 cm² provided the best accuracy and correct classification for AS severity. However, AVA_Echo had the highest correct classification for AS severity.

3. Although AVA_Echo and AVA_Hybrid are flow dependent, this effect is mitigated for AVA_CT. There was no significant impact of different flow states on the diagnostic performance of AVA_CT or AVA_Hybrid.

4. AVA_CT and AVA_Hybrid have poor discrimination for AS severity. AVA_CT <1.2‐cm² threshold (area under the curve, 0.55) has good sensitivity (85%) but poor specificity (26%). AVA_CT cut points of 1.52 cm² for normal flow, and 1.56 cm² for low flow, provided 95% specificity for excluding severe AS.

5. Comparing moderate AS and severe AS with AVC threshold as the reference standard, there were no significant differences between the groups in AVA_CT or AVA_Hybrid, or percentage of patients with AVA_CT <1.2 cm² or AVA_Hybrid <1.2 cm².

Measurement of AVA on echocardiography is a central criterion for AS assessment, and patients with AVA_Echo <1.0 cm² are presumed to have severe AS. ² , ³ However, the subset with discordant or low‐gradient AS is heterogeneous in AS severity. ⁴ AVA_Echo <1.0 cm² has prognostic value for mortality in concordant severe AS but lacks similar prognostic value in discordant or low‐gradient AS. ¹⁵ In other words, although AVA_Echo <1.0 cm² is a sensitive marker for severe AS, it lacks specificity. On the other hand, MG ≥40 mm Hg is less sensitive but more specific for severe AS. Although dobutamine stress echocardiography is recommended for the classic low‐flow subtype, it is of limited use for the paradoxical low‐flow and discordant normal‐flow subtypes. Given these challenges in the diagnosis of discordant or low‐gradient AS, we need additional tools to help in arbitrating AS severity. CT has emerged as 1 such potential tool, and we aimed to determine the diagnostic performance of CT‐derived AVAs.

Performance of CT‐Derived AVA for AS Severity

In a cohort with mostly severe AS based on AVC threshold, we found that both AVA_CT and AVA_Hybrid have poor discrimination for AS severity. There was no significant impact of different flow states on the diagnostic performance of AVA_CT or AVA_Hybrid. Prior studies have reported on performance of AVA_CT and AVA_Hybrid. Mittal et al performed AVA_CT in 450 patients with severe AS referred for TAVR CT and stratified them by flow status. ¹⁴ AVA_CT >1.0 cm² was present in 35%, with no significant variation by flow status. A threshold of AVA_CT <1.2 cm² was not tested. Although AVC was significantly higher in those with AVA_CT <1.0 cm² compared with AVA_CT >1.0 cm², arbitration of AS severity by sex‐specific AVC thresholds was not performed. We found that, compared with a threshold of <1.0 cm², a <1.2‐cm² threshold for AVA_CT provided better accuracy for AS severity. A possible explanation for AVA_CT being larger than AVA_Echo is that the latter represents the area of the smaller vena contracta distal to the anatomic AVA. ¹⁶

Clavel et al calculated AVA_CT and AVA_Hybrid in a prospective study of 269 patients with AS. ¹¹ They found that AVA_CT correlated poorly with AVA_Echo and had considerable dispersion of values. In contrast, AVA_Hybrid had better correlation with AVA_Echo. Use of AVA_Hybrid values did not reduce the incidence of discordant AS, and a threshold of AVA_Hybrid <1.2 cm² was predictive of outcomes on spline curves. In our study, AVA_CT performed better than AVA_Hybrid. However, as most patients in this registry went on to have a TAVR, we did not assess relation to clinical outcomes.

We found that just like AVA_Echo <1.0 cm², AVA_CT <1.2 cm² is a sensitive but a nonspecific marker for severe AS. Using an AVA_CT <1.2‐cm² threshold to define severe AS can produce significant error. Approximately one‐sixth with severe AS based on AVC had AVA_CT >1.2 cm², and three‐quarters with nonsevere AS based on AVC had AVA_CT <1.2 cm². In further support of these findings, there were no significant differences between the groups with moderate and severe AS in AVA_CT or AVA_Hybrid, or percentage of patients with AVA_CT <1.2 cm² or AVA_Hybrid <1.2 cm². When comparing the different valve areas, we found that AVA_Echo had the highest correct classification for AS severity.

Arguably, in patients with presumed severe AS who are referred for TAVR CT, the goal of CT evaluation for AS severity is to rule out nonsevere disease. To this extent, using commonly accepted thresholds of AVA_CT <1.0 or <1.2 cm² will lead to a significant proportion of patients with true severe AS being misclassified as having nonsevere AS. AVA_CT cut points of 1.52 cm² for normal flow, and 1.56 cm² for low flow, provided 95% specificity for excluding severe AS.

Limitations

Although our study is 1 of the larger cohorts of patients with AS to be evaluated with CT‐derived AVA to date, it is single center and retrospective in design. It is prone to selection bias as we only included patients who underwent clinically indicated CT, and the results may not be generalizable to all patients with severe AS. Because there is no single imaging parameter that can identify severe AS in isolation, guidelines recommend a multiparametric approach. ² , ³ AVC sex‐specific thresholds, although guideline recommended especially for discordant AS, do have limitations. For example, AVC cannot quantify fibrosis, an important contributor to valve stenosis, and may therefore misclassify disease severity, particularly in young women and those with bicuspid valves. ¹⁷ The CT reader, although blinded from clinical data and AVC scores, could be biased from a visual assessment of AVC while calculating AVA_CT. LVOT stroke volume calculation on echocardiography has many known pitfalls; however, every study was read by a single investigator in a systematic way.

Conclusions

CT‐derived AVAs, AVA_CT and AVA_Hybrid, have poor discrimination for AS severity. The AVA_CT <1.2‐cm² threshold has better diagnostic performance for AS severity than AVA_CT <1.0 cm² or AVA_Hybrid <1.2 cm². CT‐derived AVAs do not provide better diagnostic performance than AVA_Echo. Just like AVA_Echo <1.0 cm², AVA_CT <1.2 cm² is a sensitive but a nonspecific marker for severe AS. Higher AVA_CT thresholds improve specificity. These results should be confirmed in an adequately powered prospective sample of patients with AS with moderate and severe disease.

Sources of Funding

None.

Disclosures

None.

Supporting information

Tables S1–S2

Figure S1