MEDLINE and Cochrane Library were searched for relevant articles using the following search algorithm: ((“mitral regurgitation”) OR (“mitral insufficiency”)) AND ((“magnetic resonance imaging”) OR (“cardiovascular magnetic resonance”) OR (“CMR”) OR (“MRI”) OR (“cardiac magnetic resonance”)).

After duplicate removal, titles and abstracts were independently screened by two reviewers (IB and AE) in order to identify studies fulfilling the following inclusion criteria: (1) studies comparing MR severity between 2D-TTE and CMR, irrespective of parameters assessed and cut-off values used, (2) studies reporting Cohen’s kappa coefficient or provided sufficient data to calculate it, (3) published in any language up to March 2020. Two authors (CAP and DGK) independently assessed the eligibility of the potentially included studies.

A pre-specified form was used to extract the following demographics and baseline characteristics of the included studies: author’s name, year of publication, country, study design, time between TTE and CMR performance in days, number of patients, basic characteristics of participants (age, sex, MR etiology), TTE and CMR parameters used to assess MR severity and their cut-offs, kappa coefficient and its 95% confidence interval (CI), and the number and severity of MR cases identified by each imaging modality.

The primary objective was to assess the correlation of MR severity grading between CMR and 2D-TTE. The secondary objective was to evaluate the degree of concordance between 2D-TTE and CMR in detecting firstly severe MR and secondly moderate to severe or severe MR.

Wherever the kappa coefficient was not reported, we calculated it using relevant formulas along with its standard error (SE) and 95% CI. A value of kappa lower than 0.20 indicates poor agreement, while k = 0.21–0.40 a fair, k = 0.41–0.60 a moderate, k = 0.61–0.80 a substantial, and k = 0.81–1.00 an almost perfect agreement [26].

Given the considerable heterogeneity in the parameters used for MR assessment across the included studies, the hierarchical summary receiver operator characteristic (HSROC) curve and the corresponding area under the curve was used to determine the level of concordance between 2D-TTE and CMR in detecting severe or moderate to severe MR [27].

Revman version 5.3 (Copenhagen: The Nordic Cochrane Center, The Cochrane Collaboration, 2014) and Stata 13.0 (StataCorp, College Station, Texas, USA) was used for all analyses.

In total 858 studies were identified through the literature search. After screening of titles and abstracts 65 studies were eligible for full-text assessment. Finally, eight studies, including 718 patients, fulfilled the inclusion criteria and were incorporated into the final analysis. The detailed search strategy is illustrated in Fig. 1.

The number of enrolled patients ranged from 33 to 258 among the included studies. The mean age of the total population was 58.1 $\pm$ 5.7 and 59.1% were males. Four of the studies included patients with both primary and secondary MR [22, 23, 28, 29], whereas three studies (344 patients, 48%) only included patients with primary MR [24, 30, 31]. Gelfand et al. [32] (31 patients) did not specify the MR etiology. In six out of eight studies, the time period between CMR and TTE was less than three days. Baseline characteristics of the included studies are presented in Table 1 (Ref. [22, 23, 24, 28, 29, 30, 31, 32]). There was considerable heterogeneity among the included studies regarding the echocardiographic parameters used to diagnose and grade MR severity (Table 2, Ref. [22, 23, 24, 28, 29, 30, 31, 32]). Importantly, a multiparametric approach was used in the majority of the studies for 2D-TTE evaluation of MR severity, whereas the standard method for quantitative assessment was used in all studies but one for CMR evaluation. The regurgitant volume on CMR was defined as the difference between LV stroke volume calculated using LV planimetry of sequential cine images and the aortic forward volume obtained by phase-contrast images. By this method, MR was considered severe when the regurgitant volume was $>$60 mL, which is the same cut-off applied for TTE proximal isovelocity surface area (PISA) method based on current ACC/AHA and ASE guidelines [33, 34] (Table 2).

Cohen’s kappa coefficient varied significantly among the included studies (ranging from 0.10 to 0.48). Heitner et al. [29] and Penicka et al. [24] found moderate agreement between TTE and CMR; [k = 0.47 (95% CI 0.29 to 0.65) and k = 0.48 (95% CI 0.38 to 0.58), respectively]. On the other hand, Jang et al. [30] and Uretsky et al. [22] reported poor agreement between the two imaging modalities [k = 0.10 (95% CI –0.04 to 0.24) and k = 0.14 (95% CI 0.04 to 0.24), respectively].

Seven out of eight studies provided the necessary data to plot HSROC curves and calculate the AUC. The AUC for detecting severe MR was 0.83 (95% CI 0.80 to 0.86) (Fig. 2) and 0.83 (95% CI 0.79 to 0.86) for detecting moderate to severe MR (Fig. 3).

Fig. 2.

HSROC curve and the AUC for detecting severe MR. The HSROC curve demonstrating the accuracy of TTE in diagnosing severe MR with CMR as reference standard. CMR, Cardiac Magnetic Resonance; HSROC, Hierarchical Summary Receiver Operator Characteristic; MR, Mitral Regurgitation; TTE, Transthoracic Echocardiography.

Fig. 3.

HSROC curve and the AUC for detecting moderate to severe and/or severe MR. The HSROC curve demonstrating the accuracy of TTE in diagnosing moderate to severe and/or severe MR with CMR as reference standard. CMR, Cardiac Magnetic Resonance; HSROC, Hierarchical Summary Receiver Operator Characteristic; MR, Mitral Regurgitation; TTE, Transthoracic Echocardiography.

TTE is the guideline-directed first-line imaging test for the diagnosis and management of MR. A multiparametric approach that combines several TTE indices is sufficient for evaluating MR in most patients [10, 36]. In cases of poor image quality or discrepancy between clinical and TTE findings, ACC/AHA guidelines recommend either transesophageal echocardiography (TEE) or CMR for the accurate assessment of MR [33]. However, discordance in MR evaluation between CMR and 2D-TTE has been reported [18, 22, 34]. Herein, we confirm the presence of discordance between the two modalities and additionally report significant variation in MR grading. There are several reasons that could explain these findings. First, there was significant heterogeneity in the type of MR among the included studies. Secondary MR imposes a greater challenge for accurate echocardiographic evaluation [41]. Even the European and American guidelines disagree and have different cut-offs for defining severe secondary MR on TTE by the PISA method (regurgitant volume of 30 vs 60 mL/beat, respectively) while there is no specific consideration for the CMR cut-off [10, 33]. Second, the variability in the method used to calculate the regurgitant volume in either CMR or TTE could be a significant source of discrepancy [42]. Either the flow convergence method or the time-consuming and more prone to errors Doppler volumetric method [43, 44, 45], could affect the calculated regurgitant volume value significantly using echocardiography. The same applies to CMR, when determining the last basal slice at the mitral annulus in the planimetry-derived stroke volume calculation, especially when there is a marked descent of the base and the analysis software does not use long-axis planes for cross-referencing [46].

The PISA technique is limited in the following scenarios: (1) cannot be used in the presence of multiple jets, and (2) typically measured at a single-timepoint which could over or underrepresent the measurement across the length of systole, (3) becomes less accurate as the shape of the isovelocity shell deviates from hemispherical or is eccentric/wall-bound [47]. The MR jet is commonly elliptical and evaluation by vena contracta or proximal isovelocity surface area may underestimate MR severity [48]. Of note, quantitative methods are not routinely performed for mild MR. In addition, these 2D methods are dependent on significant geometrical assumptions in the measurement of the MR radius for the PISA equation and in evaluating the vena contracta, which are both not spherical in shape. Three-dimensional echocardiography allows for a more accurate assessment of MR severity with a high level of agreement with that obtained by standard CMR methods [47, 49]; however, such studies were not included in our analysis, considering that they are referring to transesophageal echocardiography. Another disadvantage of CMR is that image quality may be affected by irregular heart rhythms such as atrial fibrillation which is relatively common in this group of patients.

CMR allows for accurate LV volumetric assessment with excellent reproducibility, which can be used for quantitative MR evaluation [1, 32, 50, 51, 52, 53]. Additionally, there is a relatively small but growing body of literature regarding the association of valvular regurgitation evaluation by CMR with outcomes in different clinical settings [22, 24, 54, 55, 56]. Using CMR-derived regurgitant volume as a criterion, Myerson et al. [55] were able to discriminate accurately the patients that would develop an indication for MR surgery (RV $>$55 mL with AUC = 0.81) and Penicka et al. [24] predicted all-cause mortality and its combination with development of an indication for MR surgery (regurgitant volume $>$50 mL with AUC = 0.83).

The current meta-analysis suggests that patients with significant MR on CMR can be accurately identified by both TTE and CMR. However, moderate MR on CMR is usually overestimated by TTE. This finding, combined with the prior works by Myerson and Penicka, suggests that a lower regurgitant volume cutoff may be needed for CMR as compared to echocardiography. If the cut-off for CMR-based regurgitant volume was lower, the agreement between TEE and CMR would have been better. The cut-off of 60 mL was extrapolated from TTE to CMR. Nonetheless, when different techniques are used to define severity in other valvular diseases such as aortic stenosis, another cut-off for aortic valve area is applied to define severe stenosis [57, 58, 59]. Therefore, further studies correlating the MR severity through different modalities and patients’ outcomes should be conducted.

Despite the moderate, at most, agreement between 2D-TTE and CMR evaluation of MR severity across the entire spectrum of severity grading, our study shows that the two imaging modalities are highly correlated when focusing on severe MR. This is in agreement with the established role of 2D-TTE in the clinical management of MR. Indeed, current ESC and ACC/AHA guidelines suggest surgical or transcatheter mitral valve intervention for symptomatic patients with severe MR or asymptomatic patients with severe MR and signs of LV decompensation or pulmonary hypertension using established 2D-echocardiographic methods for evaluating all these parameters: MR type and grade, LV size and function and estimating the pulmonary artery pressures [10, 33]. Furthermore, TTE is the modality of choice for follow-up of asymptomatic MR, evaluation of LV function with the volumetric method of ejection fraction and assessment of LV myocardial intrinsic dysfunction by evaluating the global longitudinal strain by speckle tracking for both primary and secondary MR providing guidance regarding the timing of intervention [60, 61]. The use of 3D-TEE can further address the challenges imposed on 2D-TTE by complex jets and geometrical assumptions, allowing for accurate assessment and grading of MR, by providing direct visualization of MV leaflets and other cardiac structures, such as pulmonary veins [43]. On the other hand, time-resolved 3D (or else 4D) flow phase-contrast CMR with a demonstrated reliability in intracardiac flow visualization might prove to be the answer to some of the limitations of conventional CMR techniques [62].