The RETORIC (Rule out Transcatheter Aortic Valve Thrombosis with Post Implantation Computed Tomography) trial (NCT02826200) aimed to determine the predictors, incidence and clinical relevance of subclinical leaflet thrombosis in patients with severe, symptomatic AS following TAVI. We prospectively enrolled a total of 154 patients who underwent multimodality imaging (serial CTA, brain MRI and echocardiography) as part of the trial and clinical data were collected in a central database at a tertiary academic center. The current investigation is a sub-study of the RETORIC trial focusing on the clinical value of global strain parameters.

The imaging protocol included transthoracic echocardiography (TTE) including strain analysis to assess LA and LV anatomy and function immediately after (baseline) and 6 months (follow-up) following TAVI. Patients also had CT angiography (CTA) prior to and 6 months after the procedure. Follow-up CTA scans were performed on the same day as the follow-up TTE. Serial CTA images were used to determine LV mass (LVM) changes due to the excellent reproducibility and well-established correlation with gold standard magnetic resonance imaging (19).

Patients were excluded if the quality of echocardiographic images was inadequate for strain analysis, or if CTA was contraindicated per institutional standard of care (inability to cooperate with scan acquisition and/or breath-hold instructions, history of severe and/or anaphylactic contrast reaction, severe renal insufficiency defined as GFR ≤ 30 ml/min/1.73 m²) (Figure 1).

The study was approved by the local and national ethical committees (IRB: SE TUKEB 234/2017; ENKK 034489-004/2016/OTIG) and was registered on the ClinicalTrials.gov (NCT02826200) and was performed in accordance with the Helsinki declaration. Written informed consent was obtained from all patients.

All patients underwent echocardiography before hospital discharge after TAVI and at the 6-month follow-up according to current recommendations, using an EPIQ 7C system (Philips Medical System, Andover, MA), equipped with an X5-1 matrix transducer. 2D gray-scale images (frame rate: 60–80 Hz) were acquired over 3 heart cycles, and stored and analyzed off-line using QLab software (version 15.0 Philips Medical System, Andover, MA) by a single experienced echocardiographer, blinded to clinical and CTA imaging data. Volumetric analysis of the LA and LV were performed using standard methodology recommended by the current chamber quantification guidelines (20). The 2016 guideline document on echocardiographic assessment of diastolic function was utilized to determine filling pressures using the recommended algorithm for patients with preserved and reduced ejection fraction based on the following parameters: LAVi, E/A ratio, E/e′ ratio, e' velocity and peak velocity of the tricuspid regurgitant jet (21). LAVi was determined using biplane Simpson method. Elevated filling pressure was defined as diastolic dysfunction grade 2 and 3 according to the above referenced guideline algorithm. To calculate E/e', both the lateral and septal early diastolic peak longitudinal mitral annular velocities (e′) were obtained and averaged. Pulmonary artery systolic pressure (PASP) was derived from the simplified Bernoulli equation from the peak tricuspid regurgitant velocity and the addition of an estimated value of right atrial pressure.

For LASr and LV-GLS measurements, 2D speckle tracking software (Qlab15 AutoStrain LA for LASr and aCMQ for LV-GLS) were used. Both packages provide automatic endocardial border recognition, that was corrected manually, if needed. LASr was measured in the apical 4-chamber view, the reference point for zero strain was set at LV end-diastole (i.e. beginning of the QRS complex) as recommended by the EACVI/ASE/ Industry Task Force to standardize deformation imaging (22). Severely reduced LASr was defined as <20% and was used in the statistical methods as a categorical variable (23). LA stiffness is a further parameter of atrial function reflecting the change in pressure required to increase the volume of LA in a given measure. It was calculated as ratio of E/e′ to LASr (24). LV-GLS was derived from the apical 4- and 2-chamber views.

All patients underwent contrast enhanced, multi-detector CT scans using a 256-slice CT scanner (Brilliance iCT 256, Philips Healthcare, Best, The Netherlands, 270 msec rotation time, tube voltage 100–120 kV depending on body weight). For TAVI planning, retrospectively gated helical CTA of the aorta and the heart was acquired during a single breath-hold, in cranio-caudal direction with 1 mm slice thickness and 1 mm increment. We administered 75 ml 400 mg/ml iodinated contrast agent (Iomeron 400, Bracco Ltd; Milan, Italy) with 4.5 ml/s flow rate. Follow-up CT imaging was performed using retrospective gating with a reduced scan range covering the volume of the heart and the ascending aorta. All axial images were reconstructed with 1 mm slice thickness using iterative reconstruction (iDose4 Philips Healthcare, Cleveland, OH, USA). We measured LVM on pre-TAVI and follow-up CTA images in end-systolic or end-diastolic phases (25) using a semi-automated software (Philips Intellispace v6.0.4., Functional analysis tool, Best, The Netherlands). After automatic segmentation of the heart, a single reader manually corrected the LV boundaries excluding papillary muscles on short-axis stack if needed (on 9 slices with 3 mm slice thickness). Papillary muscles were excluded from LVM measurements. Then the software automatically calculated myocardial mass based on volumetric analysis of the LV. LVM values were indexed to the patients' body surface area.

Functional LA reverse remodeling was defined as any improvement in LASr, while anatomical LA reverse remodeling was defined as any reduction in LAVi during follow-up. We defined functional LV reverse remodeling as any improvement in LV-GLS on follow-up compared to baseline, whereas anatomical LV reverse remodeling as a reduction in LVMi assessed on follow-up CT images compared to pre-TAVI CT scans. Predictors of LA and LV changes (ΔLASr, ΔLAVi, ΔLV-GLS, ΔLVMi) were analyzed.

Continuous variables are presented as mean and standard deviation, whereas categorical parameters are presented as frequencies with percentages. Continuous clinical and imaging variables between baseline and follow-up were compared using paired t-test or were compared between groups by the two-sample t-test (patients with severely reduced LASr (<20%) vs. patients with LASr ≥20%) (Supplementary Table 1). Categorical variables were compared using the Chi-Square test. Based on former studies, we evaluated baseline LASr as a categorical variable in the statistical models as a predictor with a cut-off value defined as <20% (23). However, we have also calculated the models using LASr as continuous variable as seen in Supplementary Tables 2, 3. Relevant imaging parameters (including baseline LASr <20%) and comorbidities as predictors of LA and LV functional (LASr and LV-GLS) and anatomical (LAVi and LVMi) reverse remodeling were analyzed using linear regression analysis. Hypertension was defined as systolic blood pressure >140 mmHg and/or diastolic blood pressure >90 mmHg or antihypertensive medication use verified by medical records. Variables with p < 0.05 in univariate analysis were entered into the multivariate model. Spearman correlation was used to assess the relationship between LA and LV parameters. Intraclass correlation coefficient (ICC) was calculated to assess intrareader reproducibility for LASr measurements including 15 patients.

Mortality data were collected and verified via official death records of the National Health Insurance Fund that ensured that no lost-to follow-up occurred. All-cause mortality was used to generate Kaplan–Meier curves for patients with more and <20% LASr at baseline. All analyses were conducted using STATA v13.0. A two-sided p-value < 0.05 was considered statistically significant.

Any improvement in LASr was detected in 60.6% of all patients during follow-up. We found significant improvement comparing baseline and follow-up LASr values (17.7 ± 8.4 vs. 20.2 ± 10.2%, respectively; p = 0.024) (Figure 2). There was no significant difference in the proportion of severely reduced LASr (<20%) at baseline and follow-up, respectively [66.7 (60/90) vs. 44.4% (40/90); p = 0.134] (Table 2).

LASr correlated significantly with diastolic parameters both at baseline (E wave velocity Spearman's ρ = −0.46, LAVi ρ = −0.47, PASP ρ = −0.30; p < 0.01 for all) and follow-up (E wave velocity Spearman's ρ = −0.28, LAVi ρ = −0.48, PASP ρ = −0.37; p < 0.01 for all). Diastolic parameters differed significantly between both time points (all p < 0.05, Table 2). Mean LAVi was 53.8 ± 21.7 mL/m² vs. 50.6 ± 24.9 mL/m² at baseline and follow-up (p = 0.109). Baseline LA stiffness was 1.6 ± 1.1 for patients with LASr <20 and 0.5 ± 0.2 for patients with LASr ≥ 20% (p < 0.001).

Among patients with severely reduced LASr (<20%), although some degree of LASr improvement was detectable (mean improvement of 4.4 ± 8.4%), the mean LASr at 6 months was still significantly lower (LASr <20% at baseline: 17.3 ± 9.6% vs. LASr ≥ 20% at baseline: 26.1 ± 9.0%, p < 0.001), in turn the LA stiffness was higher as compared to those with LASr ≥ 20% at baseline (1.4 ± 1.2 vs. 0.5 ± 0.4, respectively p < 0.001) (see Supplementary Table 1). Representative images of LA strain assessment are shown on Figure 3.

The LV showed an improvement in both functional and morphological aspects following TAVI: mean LV-GLS was −15.3 ± 3.4% at baseline vs. −16.6 ± 4.0% at follow-up (p < 0.001), whereas mean LVMi reduced from 96.2 ± 23.8 g/m² to 83.3 ± 19.5 g/m² on follow-up CTA (p < 0.001) (Table 2). Any improvement in LV-GLS was found in 74.5% of all patients during follow-up.

Based on Spearman correlation, ΔLV-GLS was significantly associated with ΔLASr (ρ = 0.594, p < 0.001). On multivariate linear regression analysis, ΔLASr was predicted by severely reduced baseline LASr (< 20%) (β = 5.24; p = 0.025) and ΔLV-GLS during follow-up (β = 2.10; p < 0.001). In a separate model using LASr as a continuous variable, baseline atrial strain was independently associated with ΔLASr (ß = −0.26; p = 0.035) (Supplementary Table 2).

LA anatomical reverse remodeling based on LAVi was independently associated with baseline age (β = 0.83; p = 0.006) and E/e' ratio (β = −0.86; p = 0.001, Table 3), but not with severely impaired LASr at baseline (p = 0.396).

When assessing the predictors of functional LV reverse remodeling, we found that only ΔLASr was linked to ΔLV-GLS values (β = 0.13; p < 0.001) and both parameters improved over time (Table 4). Severely reduced LASr (<20%) (β = 5.78; p = 0.036) and LVMi (β = 0.12; p < 0.001) measured at baseline were independent predictors of anatomical LV reverse remodeling based on ΔLVMi (Table 4). Regarding the magnitude of change in LVMi, patients with severely reduced LASr (<20%) at baseline demonstrated larger anatomical LV reverse remodeling based on LVMi as compared to patients with higher LASr (≥20%): −15.7 ± 12.2 g/m² vs. −8.3 ± 13.4 g/m²; p = 0.022 (see Supplementary Table 1). Similar results for anatomical LV reverse remodeling were not detected using baseline LASr as a continuous predictor (see Supplementary Table 3).

Regarding survival, no significant difference was observed in patients with severely reduced LASr (<20%) as compared to those with higher LASr (≥20%) as assessed immediately after TAVI (log-rank p-value 0.054, Figure 4). LASr is a parameter of atrial function that showed interaction with mortality. Baseline comorbidities (hypertension, diabetes mellitus, prior stroke, smoking, age, BMI, atrial fibrillation), parameters of ventricular function, valve type, LVMi or LAVi showed no association with mortality (all p > 0.05).

Former studies reported controversial results regarding atrial reverse remodeling based on functional and morphological LA changes following TAVI. Similarly to our findings, several studies found improvement in LA parameters. Significant improvement was seen in both LA and LV function using TEE in a study comprising 213 patients (38, 39). During a 3 months' follow-up, D'Ascenzi et al. found significant improvement in both functional and anatomical LA parameters: LASr increased from 14.4 ± 3.9 to 19.1 ± 4.7% (p < 0.001), LAVi and LA stiffness also decreased significantly (40). In another study, 12 months following TAVI only LASr improved significantly from 21.5 ± 11.2 to 26.6 ± 13.8% (p = 0.039), while LAVi remained unchanged (p > 0.05) (41). Short-term data by Spethmann et al. also noted an increase in LASr (13.9 ± 5.5 vs. 20.8 ± 8.1%, p < 0.001) without the reduction of LAVi (42). This might be explained by recent observations that LA function can improve independently from LA volume. Also, LASr might be a more sensitive parameter for atrial remodeling.

On the other hand, Sabatino et al. did not find any significant improvement in LA morphology or function in 100 patients after TAVI (43). Weber and colleagues evaluated 150 patients and found significant reduction in LAVi by 2.1 ± 10 ml/m², without an increase of LASr, although LA stiffness decreased by 0.38 (p = 0.05) (44). In our study, we demonstrate functional, but nomorphological reverse remodeling of the LA 6 months after TAVI. Mean increase in LASr during follow-up was 2.4 ± 9.0% and mean decrease in LAVi was −3.2 ± 17.1 ml/m². Our results suggest that elevated LV filling pressure (based on E/e') hinders anatomical reverse remodeling of the LA based on LAVi change.

Currently there are no studies evaluating LASr improvement following TAVI in respect of LA stiffness. Although LASr improved to some extent (by 4.4 ± 8.4%) during follow-up in patients with severely impaired LA function at baseline, it mostly remained in the severely reduced range (LASr <20%) 6 months after a successful TAVI implantation and LA stiffness also remained significantly elevated. These findings underline the role of early intervention in patients with severe AS before irreversible damage occurs to the atrial tissue. LASr might be an excellent biomarker for early detection of atrial injury and thereby guide patient management and procedure timing.

In our study, improvement in LV functional remodeling was associated with greater improvement in LA function (ρ = 0.594, p < 0.001). D'Andrea et al. found that LVMi and LV-GLS measured before TAVI predicted LA functional improvement (β = 0.45 and 0.54, respectively, p < 0.001) (39). In other studies, LA volume, mean transvalvular gradient and its change were reported as predictors of LASr increase (40).

Data regarding the predictive value of LASr on cardiovascular outcomes in patients receiving TAVI are limited. The improvement of LASr after TAVI has been reported as an independent predictor of cardiovascular mortality and heart failure hospitalization (43). In previous studies LASr in TAVI patients was also a predictor of new-onset AF after the procedure (44). In our study, survival analysis revealed only a trend toward better survival of patients with baseline LASr ≥ 20%, although the difference was not statistically different.

Beneficial reverse remodeling of the LV after TAVI was predominantly demonstrated by TTE and CTA studies. Our research group previously found that LVM can regress in a magnitude of 43.4 ± 33.9 g during a mean follow-up time of 2.6 years (19). It has also been suggested that reverse remodeling based on LVMi occurs within the first weeks following implantation and is a reliable marker of improved prognosis (45). Interestingly, we found that despite severely reduced LASr values, morphological LV remodeling can occur to a substantial extent. This might be explained by the observation, that even despite long-standing AS, LV can recover following afterload normalization. In our analysis, we detected substantial difference in ΔLVMi when comparing patients with LASr <20 vs. ≥20% but not with LASr as a continuous parameter. Baseline LASr <20% was also linked to improvement in LVMi (ß = 5.78, p = 0.036) after adjustment for the baseline LVMi.

Also, in line with prior investigations we found that LV function described by LV-GLS improves after TAVI based on TTE or CT imaging. Echocardiographic data by Alenezi et al. found significant increase in the absolute value of LV-GLS (46). Also, Marwan et al. who measured LV-GLS using CT images reported the recovery of the LV (47). Shiino and colleagues demonstrated that LV-GLS improvement occurs already within 1 month after TAVI (48).

Predictors of LV-GLS improvement in other studies were pre-TAVI LVM, LV-GLS, glomerular filtration rate, diastolic blood pressure and post-TAVI peak creatine kinase-MB level after the procedure (39, 46). Interestingly, patients with normal LV-GLS before TAVI had lower improvement in LV-GLS during follow-up based on the results of Alenezi et al. (46). Based on our data, only improvement in LASr correlated with the change of LV-GLS.

We acknowledge the limitations of our study. After excluding non-diagnostic images, we could assess LVMi and strain parameters in only a subset of patients. This also highlights the limitations of current non-invasive imaging modalities for the assessment of TAVI patients. Data regarding the exact duration of AS before the procedure were not available. All-cause mortality was used for outcome analysis. Former studies included LA reservoir strain as a marker of diastolic assessment as a binary variable with the cut-off value used in current study (20%) that has been also widely used in the clinical setting for the evaluation of LA function. Using LASr as a continuous predictor was associated with ΔLASr but not with ΔLVMi (Supplementary Tables 2, 3).