Investigation of factors determining haemodynamic relevance of leaflet thrombosis after transcatheter aortic valve implantation

Abstract

Aims

To determine the conditions under which early hypoattenuated leaflet thickening (HALT) after transcatheter aortic valve implantation (TAVI) becomes haemodynamically relevant.

Methods and results

The study included 100 patients (age: 81.5 ± 5.5 years; female 63%), thereof 50 patients with HALT. After anonymization and randomization, blinded readers measured maximum thrombus thickness per prosthesis (MT_pr) and movement restriction (MR_pr) on electrocardiogram (ECG)-gated whole heart cycle computed tomography angiography. These measurements were compared with echocardiographic mean pressure gradient (mPG), its increase from baseline (ΔmPG), and Doppler velocity index (DVI). Haemodynamic valve deterioration (HVD) was defined as mPG > 20 mmHg. Age, body mass index, valve type, valve size, left ventricular ejection fraction, and atrial fibrillation were considered as influencing factors. Multiple regression analysis revealed that only valve size (P = 0.001) and MT_pr (P = 0.02) had a significant influence on mPG. In an interaction model, valve size moderated the effect of MT_pr on mPG significantly (P = 0.004). Sub-group analysis stratified by valve sizes showed a strong correlation between MT_pr and echocardiographic parameters for 23 mm valves (mPG: r = 0.57, ΔmPG: r = 0.68, DVI: r = 0.55, each with P < 0.001), but neither for 26 nor 29 mm valves (r < 0.2, P > 0.2 for all correlations). Six of seven prostheses with HVD had a 23 mm valve diameter, while one had 29 mm (P = 0.02).

Conclusion

Early HALT rarely causes significant mPG increase. Our study shows that valve size is a key factor influencing the haemodynamic impact of HALT. In small valve sizes, mPG is more likely to increase. Our study is the first to offer in vivo evidence supporting previous in vitro findings on this topic.

Graphical Abstract

To determine the impact of early HALT on haemodynamics, 100 patients (50 with HALT) received baseline echocardiography directly after transcatheter aortic valve implantation (TAVI), followed by computed tomography angiography (CTA) and echocardiography after 5 days. Maximum thrombus thickness (MT) and movement restriction (MR) were measured in CTA images. The mean pressure gradient (mPG), Doppler velocity index (DVI), and mPG increase from baseline (ΔmPG) were assessed by echocardiography. Valve type, valve size, ventricular ejection fraction (EF), body mass index (BMI), age, and atrial fibrillation (AF) were considered as confounding variables, which could influence the impact of HALT on haemodynamics. In addition to HALT quantity (MT, MR), valve size was the only relevant factor with influence on mPG. Only for small valve size (23 mm), HALT was associated with a relevant haemodynamic change from baseline on echocardiography (boxplots), and MR and MT then correlated with mPG, DVI, and ΔmPG. Small prostheses were more often associated with haemodynamic deterioration.

Open in new tabDownload slide

Introduction

Early leaflet thrombosis has been described in ∼10–13% of patients after transcatheter aortic valve implantation (TAVI), based on large cohorts with a systematic computed tomography angiography (CTA) follow-up.^1,2 Its morphologic correlate on CTA is a hypo-attenuating defect at the aortic side of the leaflets, termed hypoattenuated leaflet thickening (HALT).^3,4 As HALT is typically an incidental finding not causing any clinically significant valvular dysfunction, it is often called ‘sub-clinical’ leaflet thrombosis.^1–3 HALT can be quantified by maximal thickness (MT) in diastole.^2,4 However, such quantification alone has no clinical implication yet.³ HALT may lead to movement restriction (MR) of leaflets which may be mild (reduced leaflet excursion <50%) or marked (≥50%), which is then specified as hypoattenuation affecting motion (HAM).⁴ HAM of one or typically more leaflets can lead to increasing transvalvular pressure gradients, with mean transvalvular pressure gradient (mPG) > 20 mmHg defined as haemodynamic valve deterioration (HVD) which may be treated with oral anticoagulation.^3,5,6 The relation of MT and MR with echocardiographic parameters is still unclear as studies have shown differing results in this context.⁷ Whether HALT leads to HVD seems to be unpredictable and can vary individually. In an in vitro pulse duplicator model, Makkar et al. (Supplement⁸) found that a relevant increase in mPG required immobilization of at least two leaflets per prosthesis and that this increase depended on valve size. However, to our knowledge, this dependency has not been confirmed in vivo yet.

The first objective of this study was to assess the relation of the CTA parameters MT and MR. The second objective was to analyse the effect of MT and MR on mPG, on mPG increase from baseline (ΔmPG) and on the Doppler velocity index (DVI). Prosthetic valve type, valve size, age, body mass index (BMI), left ventricular ejection fraction (EF), and presence of atrial fibrillation (AF) were considered to determine conditions in which HALT had a considerable haemodynamic impact.

Methods

The Ethics Commission of the University of Freiburg approved this retrospective study which complies with the Declaration of Helsinki (EK-Nr.:472/12, 6 December 2012).

Patient selection

Patients who underwent TAVI between 2012 and 2017 at our institution routinely received post-TAVI CTAs before discharge. Reasons for not performing CTAs were described previously.⁹ The following balloon- or self-expandable transcatheter heart valve types were used: SAPIEN XT, SAPIEN 3 (both Edwards Lifesciences, Irvine, CA, USA), CoreValve, Evolut R (both Medtronic, Minneapolis, USA) with transcatheter valve sizes 23, 26, and 29 mm.

Fifty anonymized patients with HALT were randomly selected from our clinical database. Fifty anonymized control patients without HALT were chosen from the overall database by using SPSS random selection to form a matched group regarding valve type and valve size. Diagnosis of HALT was confirmed by consensus reading of two experienced readers of retrospectively ECG-gated multi-phase whole heart cycle CTA (authors P.R., M.H.). As previously described, HALT was defined as hypoattenuated thickening with or without rigidity of one or more leaflet segments in at least two different MPR projections and two different reconstruction time intervals.^1,9

The same study population has been part of a previously published CTA study by our research group.¹⁰

CTA scan protocol

CTA of the aortic root was performed on a second-generation dual-source CT scanner (Somatom Definition Flash, Siemens Healthcare, Forchheim, Germany) using retrospective ECG gating in a cranio-caudal direction with a temporal resolution of 75 ms. Fifty millilitres of iodinated contrast agent (Imeron 400, Bracco, Konstanz, Germany) were injected intravenously at a flow rate of 4 mL/s. CTA scan was initiated by a bolus tracking technique with a region of interest placed in the left atrium (LA). All CTA images were reconstructed at 50 ms steps throughout the cardiac cycle with a slice thickness of 1 mm, an increment of 0.8 mm, and a stent-specific reconstruction kernel (B46f).

CTA image reading

Image analysis was performed with dedicated post-processing software (Syngo Multimodality Workplace, Siemens Healthcare, Erlangen, Germany) using multiplanar reformations.

Two readers (authors F.C., M.T.H.) measured the maximum thrombus thickness of prosthetic leaflets (MT) in all 100 patients in a consensus reading using whole heart cycle CTA. Whenever HALT was detectable, MT was determined at mid-diastole (70–75% of the heart cycle) as previously described.⁴ When HALT was not clearly discernible at mid-diastole, MT was measured in the cardiac phase in which it could be best delineated. The crosshairs in a cross-sectional MPR projection were aligned through the leaflet, generating a corresponding longitudinal projection (Figure 1A and B). MT was determined for each leaflet separately, and if HALT was detected at multiple leaflets of the same prosthesis, the maximum MT of all three prosthetic leaflets was defined as the maximum prosthetic HALT thickness (MT_pr).

Two different observers (authors M.S., T.K.) blinded to the aforementioned mid-diastolic MT measurement assessed motion restriction (MR) of prosthetic leaflets on whole heart cycle CTA in another consensus reading. The cardiac phase with maximal leaflet excursion was selected. Here, the distance between the frame margin and the maximally open leaflet tip of the affected leaflet was taken as leaflet width (w) and the distance between the frame margin and the centre of the frame was taken as the frame radius (r) (Figure 1C and D). The percentage of reduction in leaflet motion was defined as % MR = (w/r)*100%.⁴ MR was scored as 0: no restriction; 1: MR < 50%; 2: MR = 51–75%; and 3: MR >75%. As a commonly accepted cut-off, MR <50% was defined as mild (score 0 and 1) and MR > 50% as notable (score 2 and 3), the latter also called hypoattenuation affecting motion (HAM).^3,4 To quantify total leaflet restriction per prosthetic valve, the summed MR of all three leaflets per prosthesis was defined as total MR per prosthesis (MR_pr).

Echocardiographic assessment

Transthoracic echocardiography was performed by experienced cardiologists using a Philips IE33-system (Philips, Leiden, The Netherlands) at two time points: as a baseline assessment directly (<24 h) after implantation as well as ∼5 days later at the time of CTA (<2 days apart from CTA). Echocardiographic assessment was blinded to CTA imaging. The mean transvalvular pressure gradient (mPG) was calculated using the Bernoulli formula. mPG >20 mmHg was defined as HVD in line with current guidelines and according to the VARC-2 criteria.^3,5,6,11 mPG increase was calculated as ΔmPG = mPG (at the time of CTA)—mPG (baseline). To account for potential low flow, low gradient prosthetic valve stenosis in patients with low left ventricular ejection fraction, the quotient of left ventricular outflow tract flow velocity (VLVOT) and maximum aortic flow velocity (VAO) was calculated as the DVI (DVI = VLVOT/VAO).¹²

Statistics

Descriptive statistics are given as mean and standard deviation for continuous variables, median and interquartile range (IQR) for ordinal variables and absolute, and relative frequencies for categorical variables. Comparison of two or more groups was performed with Student’s t-test or analysis of variance (Bonferroni corrected post hoc tests) for continuous variables, with Wilcoxon rank-sum test or Kruskal-Wallis test for ordinal and not normally distributed continuous variables, and χ² test or Fisher's exact test (expected frequency <5) for categorical variables. The association between variables was determined with Pearson’s (r) or Spearman’s rank (rho) correlation coefficient. To identify parameters with possible influence on mPG, we fitted a multiple regression model including MT_pr, valve type [balloon-expandable (SAPIEN XT/SAPIEN 3) vs. self-expandable (CoreValve/Evolut R)], valve size (23, 26, 29 mm), age, BMI, EF, and AF as covariates and mPG at the time of CTA as the outcome of interest. For variables with large main effects on the outcome mPG, we tested for a possible interaction between these variables and MT_pr. This was done by comparing the fit (increase in R²) of a linear regression model with and without the interaction, where the interaction variable was calculated as the product of the tested variables (MT_pr and the moderator) after mean centring. Statistical tests were one- or two-sided as appropriate. A P-value of ≤0.05 was considered significant. All analyses were performed with statistical software SPSS Version 28.0.1.0 (IBM, Armonk, NY, USA).

Results

Baseline characteristics

All baseline characteristics are shown in Table 1. CTA was performed at a median of 5 days after TAVI [IQR (4; 5)] and no statistically significant difference between both groups (P = 0.634).

Analysis of prosthetic leaflet thrombosis

CTA analysis per leaflet

Supplementary data online, Table S1 summarizes thrombus characteristics for all leaflets. As there were 100 prostheses and every prosthesis had 3 leaflets, 300 leaflets were analysed. Among those, 89 leaflets with HALT could be detected. MR per leaflet was significantly higher for thick than for thin leaflet thrombosis (for MT ≤1.5 mm: median MR = 1; for MT >1.5 mm: median MR = 2, P < 0.001). There was a strong correlation between MT and the degree of MR (Spearman’s correlation coefficient rho = 0.63, P < 0.001).

CTA analysis per patient

Among patients with leaflet thrombosis, 34 of 50 patients (68%) had at least one thrombus with MT >1 mm and 28 of 50 patients (56%) had thrombosis at multiple leaflets. Supplementary data online, Table S2A and B summarizes HALT characteristics on a per-patient basis sorted by the number of affected leaflets (see Supplementary data online, Table S2A) and by maximum thrombus thickness per prosthesis (see Supplementary data online, Table S2B).

Comparison of CTA findings with echocardiography

Patients with HALT had slightly higher mPG at the time of CTA than patients without HALT (mPG 13.0 ± 5.8 vs. 11.3 ± 4.3 mmHg, P = 0.048) and little higher ΔmPG (mPG increase from baseline) (ΔmPG 0.4 ± 3.5 vs. 2.1 ± 5.6 mmHg, P = 0.037). Patients with HAM did not have significantly higher mPG than patients without HAM (13.1 ± 6.5 vs. 12.0 ± 4.8 mmHg, P = 0.2). Supplementary data online, Table S2A and B shows that there were no significant differences of mPG or ΔmPG for different numbers of affected leaflets (P = 0.43 and P = 0.34) or for different maximum prosthetic thrombus thickness (P = 0.96 and P = 0.78). There was no correlation of MT_pr with mPG among all patients with HALT (Pearson’s r = 0.035, P = 0.813; Spearman’s rho = −0.096, P = 0.517) or among the whole study collective (Pearson’s r = 0.141, P = 0.170; Spearman’s rho = 0.072, P = 0.485). The same applied for MR_pr, which did not correlate with mPG for all patients (Spearman’s rho = 0.071, P = 0.494). A significant but weak correlation could be observed for DVI with MT_pr (Pearson’s r = −0.29, P = 0.004; Spearman’s rho = −0.22, P = 0.03) and DVI with MR_pr (Spearman’s rho = −0.24, P = 0.02).

Identification of parameters that influence mPG

The multiple regression model with mPG as the dependent variable had a moderate to high goodness-of-fit (R² = 0.26; adjusted R² = 0.21) and was overall significant (P < 0.001). In the model, only MT_pr (P = 0.014) and valve size (P < 0.001) had a significant influence on mPG at the time of CTA. Age, body mass index, EF, presence of atrial fibrillation, and valve type had no significant influence on mPG (P > 0.2).

The interaction model with mPG as the dependent variable, MT_pr as the independent variable, and valve size as the moderator was significant (P = 0.004), predicting 27.4% of the variance. Valve size moderated the effect of MT_pr on mPG significantly (Delta R² = 0.064, P = 0.004). The interaction explained 6.4% of the variation in mPG.

Analysis stratified by prosthetic valve diameter

Prostheses sizes were similarly distributed (33 prostheses with 23 mm valve size, 42 prostheses with 26 mm and 25 prostheses with 29 mm, P = 0.114). There was a significant, high correlation of MT_pr with mPG among patients with 23 mm valve diameter (Pearson’s r = 0.566, P < 0.001; Spearman’s rho = 0.511, P = 0.002), but neither with 26 mm (Pearson’s r = −0.006, P = 0.968; Spearman’s rho = 0.031, P = 0.846) nor with 29 mm valve diameter (Pearson’s r = −0.021, P = 0.923; Spearman’s rho = 0. 045, P = 0. 838). Concordant correlations could be observed for ΔmPG and DVI in the 23 mm valve diameter group, but not in the 26 mm nor in the 29 mm group (Figure 2).

For 23 mm valve diameter, mPG significantly differed between patients with HALT and without HALT (17.9 ± 6.4 vs. 13.1 ± 4.43 mmHg, P = 0.009). This was not evident for 26 mm valve size (P = 0.326) or 29 mm valve size (P = 0.927). For 23 mm valve prostheses with HALT, mPG significantly increased from baseline measurement at the time of implantation to the time of valve thrombosis detection (from 11.8 ± 4.0 to 17.9 ± 6.4 mmHg, P = 0.001) and mean increase ΔmPG was significantly higher for prostheses with HALT than without HALT (ΔmPG 6.1 ± 6.4 vs. 0.8 ± 3.1 mmHg, P = 0.002) (Figure 3). ΔmPG for patients with HALT was significantly higher for 23 mm prostheses (ΔmPG 6.1 ± 6.4) than for 26 mm (ΔmPG 0.5 ± 4.2 mmHg, P = 0.002) or 29 mm (ΔmPG 0.0 ± 4.2 mmHg, P = 0.003) prostheses. Figure 4 shows an exemplary CTA measurement of HALT in a 23 mm prosthesis and Figure 5 in a 29 mm prosthesis, and corresponding mPGs are given for comparison.

Among all patients with HALT, six patients had HVD. From these six prostheses with HVD there were five prostheses with valve size 23 mm, none with valve size 26 mm, and one with valve size 29 mm so that HVD was significantly influenced by valve size (P = 0.016). Table 2 shows patient characteristics, CTA, and echocardiographic outcome for patients with HALT sorted by valve size.

Discussion

Our study had three major results: (i) maximum HALT thickness strongly correlates with movement restriction on CTA. (ii) CTA quantification of HALT does not necessarily correlate with echocardiographic parameters. (iii) Among the sub-group with 23 mm prosthetic valve size, there was a moderate to strong correlation of thrombus thickness on CTA with mPG, so that HALT on CTA was associated with a significant increase in mPG (11.8 ± 4.0 to 17.9 ± 6.4 mmHg, P = 0.001). In our study, the impact of HALT on haemodynamics consequently depended on valve size.

HALT is assumed to be a continuous process, progressing from normal leaflet motion over restricted movement to more severe HAM.¹³ Therefore, it seems natural that higher HALT thickness is associated with more advanced movement restriction on CTA which has already been shown by others.¹ This could be confirmed in our study, since MT showed a moderate to strong correlation with MR on independent CTA measurements (rho = 0.63, P < 0.001).

The impact of HALT on pressure gradients ranges from haemodynamically irrelevant leaflet restriction to symptomatic HVD¹⁴ as defined by echocardiography.^2,3,11 In a meta-analysis, the impact of HALT on haemodynamics diverged in multiple studies.⁷ In the CTA sub-study of the partner 3 trial, the presence of HALT did not significantly affect mPG at 30 days,² whereas in other studies, small but significant differences in mPG could be observed in patients with HALT (11.6 + 3.4 vs. 14.9 + 5.3 mmHg, P = 0.026).⁹ Analogously, in our study patients with HALT had only slightly higher mPG at the time of CTA (13.0 ± 5.8 vs. 11.3 ± 4.3 mmHg, P = 0.048) and ΔmPG increase (0.4 ± 3.5 vs. 2.1 ± 5.6 mmHg, P = 0.037) than patients without HALT. Moreover, in the partner 3 trial even greater HALT extent was not significantly associated with higher mPG (HALT > 50%: 14.6 ± 1.57 vs. HALT < 50%: 11.7 ± 0.23 mmHg P = 0.08).² This was confirmed in our study, as no significant correlation of mPG with thrombus thickness or movement restriction score on CTA could be observed. There was neither an association of mPG and ΔmPG with prosthetic thrombus thickness nor the number of affected leaflets. To account for low flow, low gradient prosthetic valve stenosis in patients with low left ventricular ejection fraction, which might have an effect on mPG, we also calculated DVI, but did not find any significant association with MT or MR. Another CTA study found an association of HALT thickness with HVD (cut-off 2.4 mm, specificity 94.1%, sensitivity 75.0%); however, the median thrombus burden in this study was extreme (median 2.9 mm).¹⁵

In an analysis stratified by prosthetic valve size, our study showed that only patients with 23 mm valve size had a significant association between the degree of HALT and mPG (Pearson’s r = 0.566, P < 0.001). This association could be confirmed in a subanalysis comparing mPG at the time of CTA and the increase in mPG from baseline in patients with 23 mm valve size with and without HALT (mPG 17.9 ± 6.4 vs. 13.1 ± 4.43 mmHg, P = 0.009 and ΔmPG 6.1 ± 6.4 vs. 0.8 ± 3.1 mmHg, P = 0.002). Similarly, when regarding all patients with HALT, only in patients with 23 mm valve size, a significant increase in pressure gradient from baseline could be observed (from 11.8 ± 4.0 to 17.9 ± 6.4 mmHg, P = 0.001). It is known that unaffected small prosthetic valves usually exhibit higher gradients than larger valves, but the difference is only small (e.g. for SAPIEN XT 23 vs. 29 mm: 10.4 ± 3.7 vs. 9.5 ± 3.6).¹⁶ To exclude inherent baseline differences in mPG according to valve size (irrespective of HALT) as a confounder, we also considered DVI and the increase in mPG (ΔmPG) in our analysis. It could be shown that DVI and ΔmPG were concordantly associated with MT exclusively in 23 mm prostheses. ΔmPG in patients with HALT was remarkably higher for 23 mm prostheses than for 26 or 29 mm prostheses (ΔmPG = 6.1 vs. 0.5 and 0 mmHg). This was also reflected by the number of patients with HVD which was higher in the sub-group with 23 mm valve size.

To our best knowledge, this is the first in vivo study considering valve size as a relevant factor for the haemodynamic impact of HALT on mPG. Other studies did not observe an association of valve diameter with HVD in case of HALT,⁵ but patients in these studies were either symptomatic or had very high-pressure gradients (mean 34 ± 14 mmHg). The impact of HALT on haemodynamics at an early sub-clinical stage, by contrast, has not been systematically examined yet. In an in vitro study, Makkar et al. (Supplement⁸) could already show that the relevant increase of mPG, which required in vitro immobilization of at least two leaflets, depended on valve size. To our knowledge, this is the first study which could confirm these results in vivo.

These results should be interpreted with caution. In the Galileo trial, a rivaroxaban-based antithrombotic strategy was effective in preventing sub-clinical leaflet motion abnormalities but associated with a higher risk of death, thrombo-embolic complications, and a higher risk of bleeding than antiplatelet-based therapy.¹⁷ Data from the multicentre initiative from the OCEAN-TAVI registry showed that untreated early leaflet thrombosis did not affect the cumulative event rates of death, stroke, and rehospitalization for heart failure in a 2-year follow-up.^13,18 However, a long-term follow-up study could observe an association of HALT with symptomatic HVD.^16,19

The primary aim of CTA diagnostic should be the identification of patients who benefit most from therapeutic anticoagulation. On the one hand, our results could indicate that smaller valve sizes require earlier treatment in case of HALT than larger valve sizes. On the other hand, our study could reflect a lower sensitivity of echocardiographic parameters for HALT of larger valves.

Limitations

Our study has several limitations. Compared with the ‘real world’, our study collective was distorted with a very high percentage of patients with HALT (50%). However, this allowed us to explore a reliable spectrum of different HALT. Furthermore, this study only included early post-interventional imaging and did not take late complications other than thrombosis, such as degeneration, into account. Moreover, the study was not sufficiently powered to correlate imaging findings with clinical events such as the cumulative event rate of death, stroke, and rehospitalization for heart failure. Therefore, no clinical impact can be directly derived from our results. The small number of patients with haemodynamic deterioration—five 23 mm valves and one 29 mm valve—may not be sufficient to draw a definitive conclusion regarding the relationship between haemodynamic deterioration and valve size, which requires confirmatory research. Regarding the sub-group analysis for different valve sizes, sample sizes per group were limited.

Conclusions

Our results suggest that valve size plays a key role when interpreting haemodynamic impact of prosthetic leaflet thrombosis. An increase in pressure gradient and concomitant reduction in DVI with higher HALT thickness was most prominent and significant only in small (23 mm) diameter prostheses. To our knowledge, this is the first in vivo study which confirms previously published in vitro results in this regard.

Supplementary data

Supplementary data are available at European Heart Journal - Cardiovascular Imaging online.

Funding

None.

Data availability

The data underlying this article cannot be shared publicly due to privacy and ethical reasons. Data may be shared on reasonable request to the corresponding author after an individual case review by the ethics committee.

References

Author notes