Our prospective study presents important findings with respect to the added value of CMR for diagnostic grading and risk stratification in symptomatic patients with AVR or MVR and PVL. First, SQ-TEE showed a moderate association with CMR PVL-vol and RF and underestimated the CMR severity of PVL in half of these cases. Those observations were not dependent on the type (mechanical or biological) or position (AVR or MVR). Both SQ-TEE and CMR PC quantification parameters of PVL were associated with NT-proBNP. However, CMR PVL-vol and PVL-RF, but not LV parameters, showed the strongest correlation and the best prediction for the severely increased NT-proBNP. As expected, the CMR PVL-vol revealed the strongest association with LV enlargement. Finally, none of the TEE or CMR parameters predicted PVL-related hemolysis.
To the best of our knowledge, we present the first study evaluating the added value of CMR to SQ-TEE in patient quantification and stratification with a PVL related to MVR or AVR.
Echocardiography and CMR imaging in PVL
Pflaumer et al. published a first case report (2005) of a severe AVR PVL confirmed on CMR, which was underestimated on echocardiography [
21]. There are only a few studies, which compared the utility of CMR and echocardiography in patients with PVL. However, the study groups were limited only to patients with PVL related to TAVI and used mostly TTE as a comparison modality for CMR. Although the study by Crouch et al. enrolled patients with either surgical aortic valve replacement (SAVR) or TAVI, the patients were all analyzed as one group [
22]. Orwat et al. showed that TTE has only a moderate agreement with CMR and strongly underestimated the degree of PVL in the TAVI group [
23]. Hartlage et al. also showed that CMR led to a reclassification of the severity of TAVI-related PVL in most cases compared to TTE. Their results suggest that SQ-TTE overestimated the degree of PVL in a considerable number of cases. However, the quality of TTE itself seems not sufficient for PVL grading, especially when it is a retrospective design and used only in symptomatic, non-consecutive patients [
20]. The meta-analysis by Papanastasiou et al. reviewed seven studies on the utility of TTE and CMR in patients with post-TAVI PVL [
13]. They found a significant disconcordance between TTE and CMR in grading of PVL. In most studies, TTE only had sufficient power to distinguish none or mild and moderate or severe PVL. Moreover, most of the TAVI studies showed that TTE underestimated the severity of TAVI-PVL, which is in line with our results in the group of patients with MVR and AVR [
5,
20,
24].
Underestimation of PVL in echocardiography may also underestimate our interpretation of patient symptoms, result in suboptimal pharmacotherapy, and limit the number of patients scheduled for percutaneous transcatheter closure or repeat surgery. This misclassification of AV-PVL severity may explain worse outcomes in patients with even mild PVL in TTE [
6,
25]. In our study group, a majority of patients with mild PVL in SQ-TEE were found to have a moderate PVL on CMR. It was demonstrated that patients with at least moderate PVL in TAVI prosthesis confirmed with CMR revealed worse outcomes and clinical prognosis [
20,
26].
We do not have a clinical follow-up in our study group yet, so we cannot relate our imaging parameters to patients’ prognosis. Instead, we used NT-proBNP, which is an important marker of clinical prognosis, cardiac overload and wall stress, especially in patients with PVL. While parameters acquired in both modalities were associated with NT-proBNP, CMR-derived quantification parameters showed stronger correlation coefficients with this natriuretic peptide. Schewel et al. found that an increased in NT-proBNP that was greater than 1640 ng/l in patients with post-procedural PVL was associated with significantly increased rate of death in a follow-up [
27]. In our study, CMR PVL-vol > 30 ml and PVL-RF > 27.5% showed a very high sensitivity in predicting NT-proBNP levels with a similar cut-off to the study by Schewel et al. and a worse prognosis [
27]. The lack of association between LV volume or function in CMR and NT-proBNP was unexpected and suggests that increased NT-proBNP values add to patient’s symptoms and the severity of PVL and not to the LV dysfunction. Given the time since the cardiac surgery and chronic type of PVL in our patients, CMR PVL-vol showed the strongest association with the LV enlargement. However, CMR PVL-RF might be a better estimate of PVL in subjects with an acute PVL or those in their early post-surgical periods.
Eight (25%) patients had laboratory evidence of hemolytic anemia related to PVL. Although CMR led to an upgrade of the class of PVL in most of them, none of the imaging parameters of PVL severity or LV remodeling predicted the PVL-related hemolysis. No other reference reports describe this prediction. This finding suggests that the mechanism behind this phenomenon is complex and that there is no straight association between the severity of leak, shear stress, and the degree of red blood cell damage [
28].
The CMR PC technique is based on the assessment of velocities in the selected image plane. Different lengths of valve prostheses used in SAVR or TAVI require using different levels of assessment in the ascending aorta, which might impact the grading of PVL. Therefore, our results are not strictly comparable to the studies assessing PVL related to TAVI [
29].
Finally, the great majority of studies used only TTE in comparison to CMR, which has a modest agreement with TEE in the degree of PVL [
30]. Although patients with PVL have a multiparametric echocardiography assessment of PVL according to the guidelines [
12], a large number of qualitative or SQ parameters still give a wide range of final conclusion in most cases. Therefore, we focused on the SQ-TEE in a further analysis, which is currently the most prevalent classification used in clinical practice. The complex anatomy of leak channels preclude the use of the same echocardiography quantification as in native valve regurgitations. Besides, only a moderate agreement between echocardiography and CMR in those parameters using proximal isovelocity surface area (PISA) [
11] was found. Finally, the quality of images (prosthesis) obtained in particular patient affects the measures of severity of PVL, especially in echocardiography. It depends on the type and location of prosthesis and exact anatomy of leak channel. Our results showed no differences in the strength of associations between SQ TEE and CMR in patients with either mechanical prosthesis or bioprosthesis.
Study limitations
Our study group included mostly patients with AVRs-related PVL with a minority of MVR-PVL. The total number of study patients was too low to provide a separate analysis for both subgroups. The clinical characteristics of the study patients is complex and reflects the real clinical practice. However, although a prosthetic PVL was a main cause of HF symptoms and LV remodeling, it was not possible to separate the minor effects of other factors (such as long-standing native valve defect prior to surgery) or comorbidities. We only had a 6-month clinical follow-up with vital status of study patients (alive or dead), and this period was not enough to relate our results to clinical prognosis. We did not use three-dimensional echocardiography parameters in this study, which could have improved the compatibility in measurements with CMR. Finally, although CMR is a reference tool for volumetric assessment, no true gold standard for measuring the severity of PVL exists. Moreover, the PC method may also be susceptible to artifacts in close proximity to prosthesis. Finally, quantification of MVR function is always an indirect method that is more prone to errors.
Conclusions and clinical perspectives
SQ-TEE shows moderate agreement with CMR, and the Valve Academic Research Consortium II classification underestimated a considerable number of AVR or MVR-PVL cases. The LV cavity enlargement assessed by CMR reflects the PVL-vol, and a significant increase in NT-proBNP is related to PVL severity and not to LV remodeling. While there is no true gold standard for the severity of PVL, a non-contrast CMR did appear to show superior prediction for the upper tertile of NT-proBNP (CMR: PVL-vol > 30 ml; sensitivity 100% or RF > 27.5%; sensitivity 100%) compared to SQ-TEE (PVL circumference > 21%; sensitivity 62%). Neither TEE nor CMR parameters are helpful in predicting PVL-related hemolysis.
TEE is a necessary imaging modality for screening, grading, and guiding the percutaneous interventions. However, CMR is a complementary tool, which should be implemented in routine practice in patients with at least moderate PVL and/or difficult, incoherent cases. Finally, a need for evidence supporting new cut-off values for CMR quantification of PVL and optimal time for intervention exists.