Cardiovascular magnetic resonance evaluation of symptomatic severe aortic stenosis: association of circumferential myocardial strain and mortality

146 patients were prospectively recruited with severe trileaflet degenerative AS who were referred for either TAVI (n = 91) or SAVR (n = 55) at the University Hospitals of Leeds and Leicester, UK, between July 2008 and December 2013. Severe AS was classified by transthoracic echocardiography (TTE) as an aortic valve area of ≤1.0 cm² or peak velocity >4 m/s. The decision for TAVI or SAVR was taken by a multidisciplinary heart team in accordance with international guidance. Higher-risk (higher STS score) SAVR patients were recruited in preference to ensure baseline demographics were more comparable to the TAVI group. Exclusion criteria included any contraindication to CMR. Five age-comparable healthy controls were also scanned for comparison.

TAVI was performed under general anaesthesia. Either an 18 F CoreValve Revalving system (CVR, Medtronic, Minneapolis, Minnesota, USA) or an 18 or 20 F Lotus™ Aortic Valve system (Boston Scientific Corporation, Natick, MA, USA) were deployed as previously described [12, 13]. A transfemoral route was used preferentially when vascular access was suitable. In the presence of significant peripheral vascular disease, a subclavian artery approach was performed. The invasively measured LV end diastolic pressure was recorded from procedural details.

SAVR was performed by standard midline sternotomy with cardiopulmonary bypass and mild hypothermia. Biological or mechanical prostheses of varying sizes were used according to surgical preference; concomitant coronary artery bypass grafting was performed as indicated.

For each patient, identical preoperative and 6 m postoperative scans were performed on the same 1.5 T MRI vendor platform (Intera, Phillips Healthcare, Best, Netherlands or Avanto, Siemens Medical Systems, Erlangen, Germany). Both sites used the identical CMR protocol as previously described [14]; in brief this comprised standard steady-state free procession pulse sequences to image the entire left and right ventricle, through-plane velocity encoded phase contrast imaging to quantify aortic valve function and late gadolinium enhancement imaging 10 min after the administration of 0.2 mmol/kg of gadoteric acid to assess for fibrosis.

Complementary spatial modulation of magnetisation (CSPAMM) imaging was carried out during a single breath hold at end expiration in the short axis orientation, at the apex, mid-, and basal LV (multishot echo planar imaging, flip angle sweep applied to the radiofrequency excitation pulses of subsequent cardiac phases, two orthogonal line tags acquired per slice, field of view: 300 mm, matrix 128 × 128, slice thickness 10 mm, tag separation 8 mm, typically 18 phases, repetition time/echo time [TR/TE] 30 ms/6 ms, flip angle 25°). The “3 of 5 technique” was used to minimize variation in slice positioning between visits and has been demonstrated to be highly reproducible [5].

All analysis was performed blinded off-line, using commercially available software (QMass 7.5 and QFlow 7.2, Medis Medical Imaging Systems, Leiden, The Netherlands). Standard ventricular and valvular assessment was performed as previously described [14]. Late gadolinium enhancement images were reviewed by two experienced observers for focal myocardial fibrosis and scarring (secondary to infarction) and reported qualitatively, as either present or absent, and then quantified using the full-width half-maximum technique.

CSPAMM analysis was performed for each myocardial slice using a dedicated tagging analysis package (inTag© software, Creatis, Lyon, Fr) and an example is depicted in an additional figure (see Additional file 1: Figure S1). Endocardial and epicardial contours were drawn for each slice, and a mid-myocardial contour was automatically calculated; contours were propagated through all cardiac phases. Strain is defined as an index that refers to the amount of myocardial deformation in one direction normalised to its initial dimension. Strain rate is the rate of deformation within in a time unit [2, 4]. Circumferential Lagrangian strain and strain rates (between epicardial and endocardial contours) and rotation were calculated for the three short axis slices. Left ventricular twist was determined as the difference between rotation at the apex and rotation at the base. Torsion was calculated as twist corrected for by length and radius of the LV cavity.

Intra-observer (12 data sets 6 months apart) and inter-observer (12 data sets) agreement was assessed using the intra-class correlation coefficient.

Continuous variables are presented as mean ± SD. Normality was determined by the Shapiro–Wilk test. Frequencies are reported as number (%). The Student t test and Wilcoxon signed rank test were used to compare continuous variables as appropriate, and χ2 or Fisher’s exact test for categorical comparisons. Changes over time were assessed for differences between the treatment groups and clinical variables by two-way repeated measures analysis of variance (ANOVA). Cox proportional-hazards ratio regression analyses were performed to investigate univariate and multivariate correlates of all-cause mortality. Hazard ratios and 95% confidence intervals (CIs) were reported. Variables with univariate p < 0.05 were entered in the multivariate analysis in a stepwise forward approach. Receiver operating characteristic (ROC) curves were constructed to assess the sensitivity and specificity predictor variables. The cumulative event rates were calculated on the basis of the Kaplan-Meier method, and comparisons between groups were assessed by log-rank test. All statistical analyses were performed using the PASW software package (V.21.0 SPSS, IBM, Chicago, Illinois, USA) with the exception of ROC analysis that was performed with MedCalc version 9.3.1 (MedCalc Software, Mariakerke, Belgium). A two-sided significance level of p < 0.05 was considered statistically significant.

Calculation of intra-class correlation coefficients indicated good intra- and inter-observer reproducibility of CMR measurements respectively: circumferential strain (0.98, 0.96) and LV twist (0.97, 0.95).

Procedural data for TAVI and SAVR are summarised in an additional file (see Additional file 2: Table S1). For TAVI, 40(77%) patients received a Medtronic CoreValve and 12(23%) a Boston Scientific Lotus valve. A 29 mm valve was the most frequently used size (n = 26, 50%). In the surgical group, six patients received a mechanical prosthesis and the remaining 40(87%) a tissue bioprosthesis from various manufacturers. The modal valve size was 23 mm (n = 17, 37%). Twelve (26%) patients received concomitant coronary bypass grafting.

Results of the baseline and 6 m CMR scans are shown in Table 2. Significant reductions in peak aortic valve pressure gradient resulted in comparable reverse remodelling post-SAVR and TAVI; with reductions in both indexed LV EDV and mass.

At baseline, both groups undergoing SAVR and TAVI had comparable LV circumferential strain of the base (p = 0.081) mid (p = 0.128) and apex (0.318) with overall preserved LV ejection fraction. Similarly LV torsion (p = 0.845) and twist (p = 0.879) were comparable between groups. However, both systolic (p = 0.039) and diastolic (p = 0.037) strain rates were higher in the SAVR group.

At baseline for the TAVI group, there was moderate correlation between increasing LV end diastolic pressure (measured invasively during TAVI implantation) and both a deterioration in peak basal circumferential strain (r = 0.4, n = 33, p = 0.04, two-sided) and diastolic peak mid-ventricular strain rate (r = −0.5, n = 33, p = 0.003, two-sided).

No significant change in basal or mid-LV circumferential strain or of diastolic strain rate was seen following intervention, either post-SAVR or TAVI. A significant decline in peak apical circumferential strain following SAVR and an increase in circumferential systolic strain rate following TAVI were noted.

Both SAVR and TAVI were associated with a significant and comparable decline in LV twist and torsion at 6 months following intervention (Fig. 2) consistent with normalisation towards physiological values (Table 3).

Analysing the total severe AS patient population (n = 98), no change in LV strain at any level was seen following aortic valve intervention (Base: −0.186 ± 0.056 vs. −0.186 ± 0.054, p = 0.961; Mid: −0.201 ± 0.058 vs. −0.199 ± 0.006, p = 0.714; Apex: −0.194 ± 0.065 vs. −0.182 ± 0.064, p = 0.05).

Over a maximum 6.0y follow up (median 2.5 years); there were 23 deaths (all-cause, of which 14 had completed follow-up imaging). Stepwise logistic regression identified a number of demographics and measures of cardiac function that were associated with mortality. Notably, the presence of baseline myocardial fibrosis (detected by LGE imaging), indexed LV mass, mean aortic pressure gradient and history of myocardial infarction were not significantly associated with prognosis. In multivariate analysis, baseline mid LV circumferential strain remained independently associated with all-cause mortality (Table 4). ROC analysis indicated the optimal threshold for pre-procedure mid LV circumferential strain to be −18.7%, from which a Kaplan-Meier graph was derived (Fig. 3).

Our study population included patients with coronary artery disease and hypertension, rather than being restricted to pure aortic stenosis. This is however, generalisable to real world clinical practise and reduces the effect of selection bias. CMR assessment of cardiac rotational mechanics is sensitive to atrial fibrillation and regional wall motion abnormalities which can impair image quality. However, our quantification of strain using myocardial tagging CMR has demonstrated good reproducibility [5].

The control group comprised only 5 patients and ideally a larger sample would have facilitated a more robust comparison. There is paucity in the literature in regards to normal CMR strain ranges particularly in patients aged 80 years and above as it is challenging to recruit volunteers with structurally normal hearts from this age group.