Proximal aortic stiffening in Turner patients may be present before dilation can be detected: a segmental functional MRI study

Fifty-five consecutive female Turner patients, aged between 13 and 59 years, referred for routine morphologic CMR of the aorta, were prospectively included. Pregnant or lactating women, patients with pacemakers or implantable cardioverter defibrillators (ICD’s), aneurysm clips, cochlear implants, neural stimulators, epileptic seizures, large tattoos, significant claustrophobia or morbid obesity that would not enable the subject to fit in the scanner, were excluded from the study. Patients who had had aortic valve replacement or arch stenting/surgery were excluded. The control group consisted of 38 apparently normal female subjects, aged between 12 and 58 year, recruited for a previous study on aortic PWV [26]. MR studies in all patients and controls were performed at the Ghent University Hospital, Belgium and reviewed by a single radiologist with 10 years of experience in cardiovascular CMR (DGHD). Body height and weight were measured prior to the MR examination. BMI and BSA, using the formula of weight divided by body height squared (kg/m²) and Mosteller [27] respectively, were calculated. We consulted the treating physicians and reviewed the patient medical records for documentation of estrogen replacement therapy (ERT). The study protocol was reviewed and approved by the Ghent University ethics committee. All patients and controls gave a written informed consent.

All patients and controls were scanned on a 1.5 T magnet (Siemens, Erlangen). The scan protocol consisted of HASTE images of the thorax and abdomen (TR: 800 ms TE: 28 ms, FA: 160° ST: 6 mm). On four locations along the aorta, a retrospectively ECG-gated gradient-echo sequence with velocity encoding perpendicular to the aorta was applied to measure through-plane flow velocity (TR: 73 ms, TE: 4 ms, FA: 30°, ST: 6 mm, velocity sensitivity VENC: 150 cm/s, reconstructed phases: 40): ascending and descending aorta at the level of the pulmonary trunk were imaged in one series except in case of an extremely tortuous aorta, where two separate image planes were used; another series was placed at the level of the diaphragm, and a most distal series just above the aortic bifurcation [16]. A balanced fast field echo cine series with equal pixel matrix was acquired in exactly the same orientation (TR: 26–30 ms, TE: 1.2–1.5 ms, FA: 65–80, ST: 6 mm, number of images per cardiac interval: 40). During this breath hold, blood pressure was measured with a non invasive blood pressure monitor, using an adapted to size blood pressure cuff around the left upper arm (Tesla NIBP, Mammendorf, Germany). Baseline blood pressure was the average of all measurements that were performed during transverse scans.

In addition a standard keyhole high temporal, high spatial resolution MRA sequence of the entire aorta (Isovolumetric (1.3 mm) matrix 240x320 pixels, FOV: 420x320mm, TR: 2.6 ms, TE: 0.9 ms, symmetrically shared 33% peripheral K-space data for 20% central data (TWIST, time-resolved angiography with stochastic trajectories) was performed in patients. Aortic arch morphology was classified as normal arch (Romanesque, no stenosis and no arteria subclavia lusoria) or elongated transverse arch, tortuous arch, crenel shaped arch or gothic arch [4].

Postprocessing was performed off-line by a single radiologist (DGHD). It consisted first of manually drawing a centerline in the aortic lumen on a parasagittal angiographic planar reformatted image (so called candy cane view of the aorta). Along this centerline, the distance (Δx) was measured between the positions where the Phase Contrast images of the ascending aorta, the aorta at diaphragmatic level, and just above the aortic bifurcation were made (Additional file 1: Figure S1). On these phase contrast images the aortic lumen was segmented in an automated fashion [28]. Using in-house developed MATLAB code, the last 5 data points of each flow-time curve were cut and pasted in front of the time series, in order to translate the curve in time, ensuring that the onset of each curve was positioned after the start of the revised time series (Additional file 2: Figure S2). The graphs were subsequently processed in a custom-made PWV analysis tool as described by Grotenhuis et al. [29]: the onset of the systolic wave front was automatically determined from the resulting flow graph by the intersection point of the constant horizontal diastolic flow and the upslope of the systolic wave front, the latter of which was modeled by linear regression along the upslope from the flow values between 20% and 80% of the total range (Additional file 3: Figure S3). PWV was calculated as the ratio of distance Δx per time Δt, where Δx is the length of an aortic segment (thoracic, abdominal or entire aorta, see Additional file 1: Figure S1) measured on the MR image data along the centerline, and Δt is the time duration needed for the pulse wave velocity to travel that length through the aorta.

Maximum and minimum transverse aortic luminal areas were manually located and drawn on the cine series. Distensibility was calculated as:

where Amax and Amin are maximum and minimum area, Pmax is peak systolic blood pressure and Pmin is minimal (diastolic) blood pressure; it is expressed as 1/1000*mmHg. These blood pressures were taken from left arm sphygmomanometric measurement during the actual distensibility scan.

Statistical exploration was performed in Wizard v1.8.9 (OSX El Capitan v10.11.4).

Clinical characteristics and cardiac parameters of both the control and Turner population are listed in Table 1. The Turner population was significantly shorter, and had a higher BMI compared to controls. Haemodynamically, TS patients had a similar mean systolic blood pressure compared to controls, but presented with a higher mean heart rate and diastolic blood pressure. Mean left ventricular volume including stroke volume was smaller in TS patients, even after normalization for Body Surface Area (BSA). There was no significant difference in mean left ventricular ejection fraction, nor in left ventricular mass normalized for body surface area. As expected, TS patients, being smaller, had significantly shorter descending thoracic aorta and abdominal aortic lengths. In contrast, their aortic arch length was similar to the controls’. While all controls had tricuspid aortic valves, 16 Turner patients had a bicuspid aortic valve (BAV).

The multivariate determinants of aortic wall properties were assessed in a Generalized Linear Model (GLM) including the presence of Turner syndrome and adjusting for possible confounders: age, height, weight, heart rate and systolic blood pressure. Age was included as a covariate to take into account the potential effect of the non-significant age difference between Turners and controls.

The effect of the presence or absence of Turner syndrome is shown in Fig. 1 where estimated marginal means were plotted for thoracic, abdominal and total aortic PWV, area and distensibility in Turner patients versus controls. As expected in all models, age was the strongest determinant. Turner patients had significantly higher thoracic aortic PWV (TA-PWV) compared to controls. Abdominal PWV was similar, the higher overall aortic PWV in Turner subjects being driven by the higher TA-PWV. No statistically significant differences in aortic area and distensibility were seen between Turner patients and controls.

We then analysed the impact of aortic valve morphology. Table 2 shows the baseline characteristics of the BAV Turner patients, Tricuspid Aortic Valve (TAV) Turner patients, and controls. There were no significant differences between baseline population characteristics of TAV and BAV Turner patients.

To assess the effect of aortic valve on the aortic properties, we performed a GLM analysis comparing controls, TAV Turner and BAV Turner, again adjusting for age, height, weight, heart rate and systolic blood pressure. The results are plotted in Fig. 2, showing the multivariate adjusted estimated marginal means (and 95% confidence interval) of the studied aortic parameters. Overall, differences in aortic properties were seen at the most proximal aortic level: BAV Turner patients had significantly higher PWV, compared to TAV Turner (p = 0.014), who in turn had significantly higher PWV compared to controls (p = 0.010). BAV Turner patients had significantly larger AA luminal area and lower AA distensibility compared to both controls (all p < 0.01) and TAV Turner patients. TAV Turner had similar ascending aortic luminal areas and ascending aortic distensibility compared to Controls.

No consistent differences were seen across groups in more distal aortic characteristics. Exceptions were marginal (but significant) differences in diaphragmatic area (larger in BAV Turner patients), and differences in overall aortic PWV (Turner higher than controls), again driven by the differences in thoracic PWV, as abdominal PWV values were similar.

Adding left ventricular stroke volume, which was significantly smaller in Turners compared to controls, to these models did not substantially change the results (data not shown).

We plotted F and p values of the used GLM model within the Turner group and the control group separately in Additional file 4: Table S1. Age is the main determinant of PWV, aortic luminal area and distensibility at all levels in both groups, only surpassed slightly by systolic blood pressure for descending and distal abdominal distensibility in Turner patients (and weight for distal abdominal aortic area in the control group).

Using the same model separately in the TAV Control group, in the TAV Turner patient group and in the BAV Turner patient group (F and p values in Additional file 5: Table S2), the independent association between aortic properties and age remain largely unchanged in both TAV groups. Strikingly, in BAV Turners almost all effects of age, blood pressure, height and weight were lost.

To further study whether the impact of Turner Syndrome on aortic properties is different in younger versus older subjects, we divided all three subgroups into two halves, above and below their respective median age. Age cut-offs were 27 years for TAV Turner subjects; 23.5 years for BAV Turner subjects; 32 years for Controls. A similar GLM model adjusting for age, height, weight, heart rate and systolic blood pressure was used and the results of proximal aortic properties are shown in Fig. 3. Overall, the functional changes (borderline increase in thoracic PWV and significant loss of proximal aortic distensibility) are present in both younger and older Turner subjects, whereas structural remodeling (enlargement of ascending aortic) clearly is only prominent in older BAV Turners.

Finally, we plotted the ascending aortic area in relation to the ascending aortic distensibility for young and old groups of controls, TAV and BAV Turner patients (Fig. 4). This graph illustrates the inverse correlation between ascending aortic area (BSA-normalised) and its distensibility (incremented and ln-transformed), or, larger aortas are stiffer in older controls and Turner patients (p < 0.001), regardless of aortic morphology. There is no such correlation in the younger subjects (either controls or Turner patients).

We found ERT in at least 43 of 55 Turner patients (78%). Four patients previously received ERT, but it was stopped or temporarily interrupted in the year of the MR scan. Retrospectively, we could not rule out that the remaining 8 patients ever had been on ERT (5 Turner patients probably not on ERT, and unknown in 3).

Aortic arch morphology was normal (Romanesque, no stenosis, no lusoric subclavian artery) in all 38 control subjects, compared to only 20 out of 55 Turner patients. Thirty-five Turner patients had aortic arch abnormalities: 21 elongated transverse arches, 5 tortuous arches, 4 crenel shaped arches, 2 gothic arches, 6 subjects had an aortic arch stenosis (5/6 were moderate stenosis), and 8 subjects had an arteria subclavia lusoria. Adding aortic arch abnormalities to the GLM models did not impact the models for PWV, but for AA distensibility and for AA area, the aortic arch abnormalities competed with the diagnosis of Turner Syndrome and/or presence of a bicuspid aortic valve (i.e. partial co-linearity, especially with BAV presence).

A limitation of our study is the small number of patients, especially when groups were divided in subgroups: the total number of bicuspid Turner patients was only 16. In addition the effect of estrogen replacement therapy (ERT) and previous growth hormone therapy was not studied. We did review the patient files for history of ERT, but with the vast majority of Turner patients on ERT, we were not able to divide into ERT/No ERT groups of sufficient size. However, the effect of ERT on aortic diameter was not statistically significant (p = 0.08) in a prospective 5 year follow up study (n = 78) [15]. In a smaller study the same group has investigated the effect of 6 months of ERT on ambulatory arterial stiffness index (AASI) derived from 24-hour ambulatory blood pressure: although Turner patients were found to have higher AASI when compared to controls, AASI was unchanged by ERT [45].

The age difference between control population and the 5 years younger Turner population was not significant, but once subdivided in valve morphology groups, the small BAV Turner patients were significantly younger than the TAV Turner patients. The older/younger subdivision of our control and Turner patients at the respective median line was chosen to preserve equal number of subjects in both young and old groups. Subdivision at one fixed age for all groups would hinder proper statistical analysis, as this would lead to comparing a large group of young Turner patients, with a small group of young controls.

Blood pressure measurement was performed during the actual distensibility scan, but we were limited to a left arm sphygmomanometric measurement. This may not be representative for the blood pressure in the ascending aorta proximal to a dysmorphic arch.

The postprocessing of PWV is another limitation. Detection of the foot of the curve is difficult in such a fast physiologic process, when performed with the relatively poor temporal resolution of MR. However this limitation applies equally to all studied groups in this study. Our data did not allow analyzing smaller aortic segments. Other studies have focused on increasing temporal resolution by acquiring in-plane velocity encoded images [46, 47]. At the time our data acquisition started, transverse images and detection of the foot-of-the-curve time shift was the major proposed method, a transit-time method focussing at the early systolic part of the curve in an attempt to avoid the effect of wave reflection, which, especially in a stiffer aorta, is important. It was used as early as 1993 by Mohiaddin et al. [48] and more recently, by means of computer phantoms, Dorniak et al. calculated the minimal required temporal resolution of 30 ms (35 image frames at a heartrate of 60 bpm) for accurate and precise quantitative flow data for CMR-PWV over the range 2–20 m/s in the thoracic aorta [49]. Newer methods involve complex wavelet cross spectrum analysis [50], which equally have the advantage of being robust at lower temporal resolution.

Our transverse aortic area measurements were not corrected for longitudinal strain, which in the ascending aorta reduces systolic dilation and thus leads to underestimation of distensibility [51]. However, ascending and descending aortic distensibilities in our study are comparable. To minimize through-plane motion error, the ascending aortic cine image was positioned carefully out of range of the motion of the sinotubular junction. The reported perpendicular aortic areas in our work are end-diastolic. As for comparison with other PWV values in literature, it is important to note that the most critical aspect of PWV postprocessing is the lack of standardization and all-in-one software solutions.

Our data is not longitudinal. We can therefore not make assumptions regarding the evolution of a stiffer proximal aorta at young age towards dilatation at older age. We do however think this should be studied further.

In Turner patients, structural (dilatation) and functional aortic wall changes (stiffening and loss of distensibility) are restricted to the proximal aorta, even more so –but not exclusively- in the presence of a bicuspid aortic valve.

Increased thoracic aortic wall stiffness is already present at a young age whereas the aortic dilatation was only seen in our older Turner patients, specifically in those with the most pronounced stiffening.

If proximal aortic stiffness measurements could be used to predict dilatation -and by extension, risk of dissection- at older age, the follow-up criteria and timing of both bicuspid and tricuspid aortic valve Turner patients could be redefined.

Turner patients exhibit a predominantly proximal aortic stiffening and dilatation, especially -but not exclusively- if their aortic valve is bicuspid. Aortic stiffening in Turner syndrome is already present at an early age, whereas ascending aortic dilatation is not. We put forward the hypothesis that proximal aortic stiffness at early age may predict aortic dilatation in both bicuspid and tricuspid aortic valve Turner patients. This hypothesis should be studied longitudinally.