We included consecutive patients who underwent TAVI for low-gradient (mean gradient <40 mm Hg), severe AS as assessed by TTE at the Leiden University Medical Centre (LUMC) and the Thoraxcenter, Erasmus University Medical Centre, The Netherlands (EMC), between April 2006 and September 2016. Patients without TTE and/or MSCT at baseline were excluded. A joint database was constructed including baseline demographics, procedural and clinical outcome data and selected imaging variables derived from TTE and MSCT. Incomplete clinical data were addressed by consulting referring physicians and patients whenever possible. Survival status was obtained from the Dutch Civil Registry. Written informed consent for the TAVI procedure and subsequent data analysis for research purposes was provided by every patient at the EMC. Both institutional review boards waived the need for patient written informed consent for retrospective analysis of clinically acquired data (EMC MEC no. 2019-0301). The study was conducted in accordance with the principles of the Declaration of Helsinki and did not fall under the scope of the Medical Research Involving Human Subjects Act according to the EMC Institutional Review Board.

All patients underwent TTE (at baseline and 1 year after TAVI) in accordance with a standard protocol. Two-dimensional TTE and Doppler data were acquired with commercially available systems, Philips iE33 (Philips Medical System, Best, The Netherlands) or Vivid‑7 and E9 ultrasound systems (General Electric, Horten, Norway). Images were stored offline and all analyses were performed in accordance with current guidelines using the Image Arena workstation (TomTec Imaging System, Unterschleissheim, Germany) or EchoPac (112.0.1, GE Medical Systems, Horten, Norway) [7]. Mean aortic pressure gradient was obtained by tracing the continuous wave envelope [8]. The AVA was estimated by the continuity equation and divided by body surface area (BSA) to obtain the indexed AVA (AVAi) [1, 8]. The LVOT was defined 5 mm below the aortic annulus (parasternal long-axis view) and its area was calculated based on the measured LVOT diameter, assuming circularity. The left ventricular ejection fraction (LVEF) was either visually assessed or calculated with the modified Simpson method [9].

Pre-procedural MSCT was performed in all patients with a dual source (Definition, FLASH or Force, Siemens Healthcare, Forchheim, Germany) or 64- and 320-detector row computed tomography scanner (Aquilion 64; Toshiba Medical Systems, Otawara, Japan and Aquilion ONE; Toshiba Medical Systems, Tochigi-ken, Japan) with electrocardiographic triggering or gated acquisitions in systole. CT scan settings for image acquisition were reported previously [6, 10]. All the reconstructions were stored on dedicated workstations for offline analysis (Vitrea 2, Vital Images, Plymouth, MN, USA and Intellispace, Philips, Best, The Netherlands). Aortic annulus and LVOT dimensions were analysed with 3Mensio software (Bilthoven, The Netherlands) and calcification was expressed using the score proposed by Rosenhek et al. and the Agatston calcium score [11, 12]. The LVOT was defined at the smallest area between 2 and 6 mm below the annular plane and measured by planimetry [13].

Fusion implied the use of MSCT and TTE data in the continuity equation in order to reclassify AS severity. LVOT area measured by MSCT was used to replace LVOT area measured by TTE as described previously [6]. To calculate fusion AVAi the following formula was used [6]:

$$\text=(\frac \boldsymbol \boldsymbol \boldsymbol \boldsymbol \boldsymbol \boldsymbol \boldsymbol} \boldsymbol \boldsymbol\,\boldsymbol\,\boldsymbol})/\boldsymbol\,\boldsymbol\,\boldsymbol$$

where VTI = velocity time integral, PW = pulse wave Doppler and CW = continuous wave Doppler. Reclassification was based on fusion AVAi ≥0.6 cm2/m2 (moderate AS) and AVAi <0.6 cm2/m2 (severe AS).

Normal distribution of continuous data was assessed by the Kolmogorov-Smirnov test. Values are expressed as mean ± SD or median (interquartile range, IQR) depending on distribution. Categorical data were presented as numbers and frequencies. Comparison of baseline characteristics was done by means of the Student t-test, Mann Whitney U test or chi-squared test. LVEF at baseline was compared with 1‑year follow-up LVEF after TAVI using the paired t-test. The primary endpoint was a composite of all-cause mortality and rehospitalisations for heart failure (HF) at 1 year. Cardiac mortality, non-cardiac mortality and rehospitalisations due to HF were secondary endpoints. Kaplan-Meier curves were used to assess the primary and secondary endpoints (30 days and 1 year) after TAVI. To further evaluate the impact of systolic LV function a separate analysis looked at differences in clinical endpoints between reclassified moderate and severe AS in patients with LVEF <50% and ≥50%.

Statistical significance was assumed when the p-value was <0.05. Statistical analysis was done using SPSS 24.0 (IBM Corporation, New York, NY, USA).