We prospectively enrolled 73 patients (58 ± 15 years; 57 men), 61 with a wide spectrum of AR of the native valve referred to our center for evaluation of the pathology (AR group), and 12 patients with non or trace AR on TTE with clinical indication for a CMR study other than heart valve disease (control group). Patients were eligible if they had sinus rhythm during the study. Those with an associated cardiac valve lesion more than moderate, with aortic coarctation, or with typical contraindications for CMR imaging were excluded.

The systolic and diastolic flow profiles of the LSA were evaluated by use of pulsed wave Doppler ultrasound (3.4–9 MHz linear probe). Patients were examined in a supine position with a subclavicular approach. Higher-frequencies (> 7 MHz) were used for assessment of the morphology, and lower-frequency (< 7 MHz) was preferred for Doppler examination. LSA was documented with gray-scale imaging and color Doppler to rule out relevant stenosis. The depiction of the LSA was modified to align the Doppler angle parallel to the vector of blood flow and to avoid the Doppler signal of the adjacent vein. The sample volume was placed just nearby the origin of the LSA. Patients with vascular shunts of the left upper arm were excluded. As previously described [2], the outer edge of the dense envelope of the spectral recording (i.e. modal velocity) was used to measure the VTI [12]. The RF was calculated as follows: RF(%) = LSA derived diastolic reversed flow VTI × 100/LSA systolic forward flow VTI. At least two measurements were performed. The LSA-Doppler examination was carried out blindly to the results of echocardiography and the CMR evaluation of AR.

All TTE were performed by experienced echocardiographers using commercially available ultrasound machines (Vivid E9, General Electric Healthcare, Wauwatosa, USA; or Acuson SC2000, Siemens Healthcare GmbH Erlangen, Germany) equipped with M5S or 4V1c 2D TTE probes. All recordings were stored digitally for offline analysis. Left-ventricular (LV) volumes and ejection fraction (EF) were calculated using the biplane Simpson disk method. Doppler measurements were evaluated as the average of at least three cycles. An effort was made to perform the flow convergence (PISA) method, from apical views or, in case of eccentric jets, from parasternal long-axis views. The AR severity was graded according to a recommended TTE “multiparametric approach” [10, 11], conceivable as comprehensive and integrative process, based on structural (i.e. aortic valve morphology and LV size), qualitative (i.e. Jet width and density, Jet deceleration rate -PHT-), semiquantitative (i.e. vena contracta width, end-diastolic flow velocity in the descending aorta), quantitative parameters (flow convergence method and quantitative pulsed Doppler), and all adjunctive parameters collected during the echocardiographic examination to consolidate the evaluation of the severity of AR. In intermediate cases the quantitative methods were conclusive. Evaluation of AR was performed by an experienced echocardiographer (RS, > 10 years of experience in echocardiography with ESC certification), who was blinded to the results of the CMR and LSA Doppler exams.

All CMR examinations were performed in our cardiology department on a 1.5-T MRI system (Ingenia, Philips Healthcare, Best, the Netherlands) equipped with a 28-element array coil with full in-coil signal digitalization combined with optical transmission. All scans were accomplished without sedation. Image data acquisition and subsequent analysis were carried out according to current guidelines. For cine imaging, a balanced steady-state free precession (b-SSFP) sequence with retrospective gating was used during short-periods of breath-holding. All standard cardiac geometries were acquired (multiple, gapless short-axis slices covering the entire left ventricle and 2-, 3- and 4-chamber views). Imaging parameters were chosen as follows: echo time (TE) and repetition time (TR) were set to shortest resulting in an average TR of around 4 ms and a TE of 2 ms, with a reconstructed in-plane resolution of 1.0 × 1.0 mm²; the slice thickness was 8.0 mm. The typical temporal resolution of the cine b-SSFP sequences was 30–25 ms depending on the heart rate. The imaging plane for the through-plane phase-contrast flow measurement was placed in the ascending aorta approximately 10 mm above the aortic valve and positioned perpendicular to the flow direction. On a coronal image of the aorta together with the 3-chamber view the CMR operator checked that the image plane was truly perpendicular to the aortic flow direction. To avoid aliasing, velocity encoding was individually adapted, starting at 200 cm/s, and if aliasing occurred, the maximum velocity was increased by 50 cm/s steps until aliasing disappeared. Image data acquisition was gated to the ECG signal with a temporal resolution of 35 phases per cardiac cycle and acquired during a 12–15 s breath-hold. Outlining the region of interest within the aortic lumen for each cardiac phase, the instantaneous flow volume (cm³/s) was calculated and graphically displayed over the entire cardiac cycle.

Forward and reversed flow volumes were measured, and the regurgitation fraction was calculated as follows: aortic RF (%) = diastolic reversed flow volume × 100/systolic forward flow volume.

Data are presented as mean (SD), median (25th to 75th percentile), or frequency (percent) as appropriate. Statistical differences between groups were assessed using Student’s t-test for continuous variables or Fisher’s exact test for categorical variables. Multigroup comparisons of continuous variables were performed using an analysis of variance (ANOVA). Spearman correlation coefficient was used to assess correlations between methods. Linear regression and c-statistics were used to assess the cut-off values of CMR and LSA-Doppler derived RF that best reflect the three-scale grading of AR by multiparametric TTE approach. Using these cut-offs the rate of agreement for AR grading was evaluated by calculating a κ-statistics. Additionally, the mean difference with the 95% confidence interval was also determined. Two-tailed p-values < 0.05 were considered statistically significant. Analyses were performed using SPSS software (IBM-SPSS Statistics, Version 20, IBM Corp.). The study was conducted in accordance with the Declaration of Helsinki, and was approved by the local research ethics committee (067/17-ek). All patients received informed consent.

In the whole cohort the RF measurements assessed by PISA-method, LSA-Doppler, and CMR imaging were strongly correlated (Fig. 3). But the mean RF values were significantly greater on TTE compared with LSA-Doppler and CMR (39 ± 16% vs. 35 ± 18% vs. 32 ± 20%, respectively; p < 0.037). The mean differences of RF values were significant in the groups with mild and moderate AR, with overestimation by TTE (Table 3 and Fig. 4).

Using a linear regression and a receiver operating characteristic (ROC) curve analysis the RF values that best defined the multiparametric TTE derived AR severity using CMR were: mild, < 21%; moderate, 22 to 41%; and severe, > 42%; and using LSA-Doppler: mild, < 29%; moderate, 30 to 44%; and severe, > 45% (Fig. 5). Applying these RF cut-offs, the level of agreement between LSA-Doppler, CMR, and multiparametric TTE grading of AR was strong (Table 4). Importantly, in patients with severe AR, there were only three misclassified patients by LSA-Doppler and CMR, resulting in a good sensitivity, specificity, and diagnostic accuracy for severe AR (88.5%, 94.3%, and 91.8%; respectively). The ROC curve analysis showed similar cut-off values for severe AR when using the best sum of sensitivity and specificity giving an excellent area under the curve. Applying the same grading scale for all methods (mild, < 30%; moderate, 31 to 49%; and severe, > 50%) the level of agreement between modalities was numerically lower with kappa values of 0.628 and 0.616 for CMR and LSA Doppler, respectively (Online Resource 1).

Seven of ten patients, with PISA derived RF just below the recommended limit for severe AR (borderline values of 45 to 48%), were recategorized as severe AR when a multiparametric approach was used. Five of these 7 patients were symptomatic; all of them had dilated LV (LV end-diastolic volume 295 ± 55 ml, LV end-systolic diameter 48 ± 6 mm); and the decision to grade them as severe AR were made based on proposed echocardiographic parameters like a descending aorta end-diastolic flow reversal velocity > 20 cm/s, quantitative Doppler derived RF > 50%, and a vena contracta > 6 mm. Finally, there were comparable feasibility and kappa values for the agreement between all three methods and multiparametric TTE approach (Table 4).

On the other extreme of the grading scale there was a significant overlap between non/trace AR (control group) and mild AR. RF cut-offs of < 7% for CMR and < 14% for LSA-Doppler showed a sensitivity and a positive predictive value of 100% for presence of non/trace AR but a poor specificity (59% for CMR, 47% for LSA-Doppler) due to overlap between non/trace and mild AR (Fig. 5).