Introduction

Mitral valve prolapse (MVP) is a valvular anomaly characterized by superior displacement of one or both mitral valve leaflets into the left atrium [1] which affects approximately 2–3% of the general population [2].

While, in unselected cohorts, the prognosis is mainly dictated by the severity of mitral regurgitation (MR), a subgroup of patients is at risk of malignant ventricular arrhythmias (VA) and sudden cardiac death (SCD). The incidence of SCD in the community MVP population is low and between 0.1% and 0.4% per year [3, 4]

However, a subgroup of patients in retrospect defined as affected by “Arrhythmic mitral valve prolapse” (AMVP) [5] may be at substantially higher SCD risk. The characteristics of AMVP are poorly characterized, accordingly, the diagnostic strategies to identify these patients represent an unmet need. A comprehensive non-invasive arrhythmic risk stratification of patients with MVP might allow for tailor monitoring (i.e., implantable loop recorder) and preventive strategies (beta-blockers and implantable cardioverter defibrillator) in high-risk patients [5].

Clinical, Electrocardiogram (ECG), and imaging characteristics including syncope, T-wave inversion, longer QTc interval, ventricular arrhythmia burden and complexity, bi-leaflet prolapse, longer anterior mitral valve leaflet, mitral annular disjunction (MAD), and late gadolinium enhancement (LGE) have previously been associated with complex ventricular arrhythmias (co-VAs) in MVP patients [6].

According to the 2021 ESC/EACTS Guidelines for the management of valvular heart disease (VHD) [7], significant gaps in evidence exist in the following elements of VHD concerning the relationship between MR, SCD, and VA. In addition, it has been emphasized that non-invasive assessment with three-dimensional echocardiography, cardiac computed tomography, cardiac magnetic resonance (CMR), and biomarkers are becoming increasingly important in VHD. According to the European Heart Rhythm Association (EHRA) expert consensus statement [5] CMR plays a central role in risk stratification for MVP patients and should include measurements of left ventricle (LV) size and function, MR severity, leaflet length/thickness, MAD characterization, and curling, and LGE assessment.

The impact of these CMR features has been analyzed in only a few publications [8,9,10,11,12,13,14,15,16,17,18] and to the best of our knowledge a systematic revision is lacking. Therefore, the aim of our study is to perform a systematic review and meta-analysis of studies of MVP patients undergoing CMR to investigate the features that could discriminate between patients with co-VAs and without (N-co-VAs), thus providing comprehensive evidence on how to best leverage CMR as a diagnostic tool for arrhythmic risk stratification in MVP.

Protocol and registration

This systematic review and meta-analysis was conducted according to the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analysis statement [19]. PROSPERO ID number: CRD42023374185.

Eligibility criteria

Studies were considered eligible if they met all the following inclusion criteria:

  1. 1.

    Patients with MVP who underwent CMR with the assessment of at least one of the following: (1) LV size and function, (2) MR severity, (3) leaflet length/thickness/prolapse distance, (4) MAD, (5) curling, (6) LGE (papillary and/or myocardial), and (7) T1 mapping.

  2. 2.

    Arrhythmic profile reported.

  3. 3.

    The association of at least one CMR parameter with the arrhythmic profile was reported.

The primary endpoint was co-VAs. Co-VAs comprised non-sustained ventricular tachycardia (NSVT), sustained ventricular tachycardia (SVT), ventricular fibrillation (VF), and aborted SCD (aSCD). A sub-analysis was performed to further divide co-VAs into two groups based on the presence of NSVT vs SVT, VF, and aSCD, and a sensitivity analysis was performed to investigate the heterogeneity of the studies.

Information sources and search strategy

The databases utilized for this study included Excerpta Medica dataBASE, Medical Literature Analysis and Retrieval System Online (PubMed/MEDLINE), and Cochrane Central Register of Controlled Trials (CENTRAL). The search was conducted until March 1, 2023. The string utilized is reported in the Supplementary Material. In addition to the electronic search, a manual search was conducted on the reference lists of selected papers to find any additional research that met the eligibility criteria.

Selection and data collection process and data extraction

The inclusion criteria were used by two researchers, F.P. and A.S., in a two-stage approach to search for studies. Initially, they examined the title and abstract of the publications, followed by a review of the complete text. The rationales for the exclusion of studies during this subsequent phase were recorded. A comparison was made between the results of the two searches, and any inconsistencies that arose were then examined. In case of disagreement, the resolution was reached through the involvement of a third researcher (referred to as M.G.) in the process of consultation. The chosen articles were downloaded, imported, and de-duplicated in Microsoft Excel (Microsoft).

The following data was extracted: study title, authors, publication date, study design, number of patients enrolled, relevant baseline characteristics (including the definition of Co-VA with the respective standard of references used to evaluate it), CMR characteristics, and associations between CMR characteristics and study outcomes.

Study risk of bias assessment

Two reviewers (F.P. and A.S.) separately assessed the quality of the selected studies, and any discrepancy was resolved by discussion and consensus. The risk of bias in each study—classified as low, moderate, or high—was scrutinized in terms of selection, performance, attrition, detection, reporting, and overall risk of bias, according to the guidelines of the Agency for Healthcare Research and Quality [20].

Data analysis and synthesis

The association of each CMR parameter with co-VAs was meta-analyzed if reported in at least three studies. The association between the presence of CMR features and co-VAs was estimated by calculating Hedge’s g or pooled LogORs and their corresponding 95% CIs, when appropriate. Cochran’s Q test and I2 tests were used to measure heterogeneity between studies. For the Q test, p < 0.10 indicated significant heterogeneity; otherwise, heterogeneity was not statistically significant. For I2 tests, I2 between 0% and 25% was considered as low heterogeneity; I2 between 25% and 50% was considered as moderate heterogeneity, and I2 more than 50% was considered as high heterogeneity. Therefore, if p < 0.10 and I2 ≥ 50%, heterogeneity was present, and a random effect model was applied. Galbraith plots and Funnel plots were generated to assess heterogeneity within the included studies.

The sensitivity, specificity, positive (+ LR) and negative (− LR) likelihood ratios, and diagnostic odds ratio (DOR) with 95% confidence interval (CI) of dichotomous CMR parameters for the detection of co-VA were calculated using two-by-two contingency tables collected from each study. A bivariate random effects model was used to analyze, pool, and plot diagnostic performance measures from multiple studies. A hierarchical summary ROC curve (HSROC) was generated using logit values for sensitivity, specificity, and their respective variances. The clinical accuracy of each CMR parameter at the patient level was assessed using likelihood ratios to generate post-test probabilities based on Bayes’ theorem, Fagan’s nomograms, likelihood ratio scattergrams, and probability modifying plots. Stata (version 17.1, Stata Corp LP) was used to conduct the analyses.

Results

A total of 388 papers were found in a search of electronic databases from their inception to March 1, 2023. Eleven studies with 1278 patients met the inclusion criteria. The consort diagram is shown in Fig. 1.

Fig. 1
figure 1

Consort diagram

Full size image

Study characteristics and risk of bias

Table 1 summarizes the included studies, while Table 2 provides detailed baseline characteristics. The study period ran from 2015 to 2022, with one prospective and nine retrospective studies.

Table 1 Panel A: summary of the studies included. Panel B: summary of the studies included
Full size table
Table 2 Baseline characteristics
Full size table

The studies included a total of 1278 patients (median per study 52 (interquartile range 38–69.5) patients). Four studies with 213 patients were included in the evaluation of the left ventricular end-diastolic volume index (LVEDVI) and left ventricular ejection fraction (LVEF). Prolapse distance was assessed in three studies involving 181 patients. MAD was assessed in four studies involving a total of 653 patients. The presence of LGE was investigated in eight studies with 1173 patients, while the percentage of LGE was evaluated in three studies with 181 patients. Of note, only three studies with 556 patients assessed the predictive value of LGE specifically for the occurrence of SVT, VF, and aSCD).

Table 3 shows the assessment of bias for each study. On overall risk assessment analysis, only one study [12] was found to be at high risk. Of note, that study reported only the association between arrhythmia and LGE of papillary muscles and was not included in any meta-analysis.

Table 3 Risk of bias analysis
Full size table

Risk stratification

The association of CMR parameters with Co-VAs was largely different in terms of effect size and CI for different CMR parameters. MR severity, leaflet length/thickness, curling, MAD distance, and mapping techniques were not included in the meta-analysis, as they were reported in fewer than three studies. The Hedge’s g values for LVEDVi, LVEF, and prolapse distance were 0.02 (95% CI: [−0.31, 0.34]), −0.17 (95% CI: [−0.49, 0.16]), and 0.14 (95% CI: [−0.30, 0.58]), respectively, indicating small effect sizes with no statistical significance (Fig. 2A–C)). For MAD, the effect size was significant with a log odds ratio (OR) of 0.95 (95% CI: [0.30, 1.60]), while for LGE, the log OR was 2.12 (95% CI: [1.00, 3.23]) (Fig. 3A, B). The percentage of LGE showed a Hedge’s g of 0.45 (95% CI: [0.08, 0.83]) (Fig. 2D)

Fig. 2
figure 2

Summary forest plots for the association between CMR characteristics (LVEDVi (A), LVEF% (B), PD (C), and LGE% (D), and the presence of complex ventricular arrhythmia

Full size image
Fig. 3
figure 3

CMR images and summary forest plots for the association between CMR characteristics and the presence of complex ventricular arrhythmia (A MAD and B LGE)

Full size image

Sub-analysis to further stratify the arrhythmic risk of patients into two groups based on the presence of SVT, VF, and aSCD was only possible for LGE, which showed a log odds-ratio of 1.69 (95% CI: [0.81, 2.56]) (Fig. 4).

Fig. 4
figure 4

Summary forest plots for the association between CMR characteristics (LGE) and the presence of SVT, VF, or aSCD

Full size image

Details of the sensitivity analysis conducted to assess study heterogeneity are provided in the Supplementary Materials.

Sensitivity, specificity, and diagnostic performance

For LGE, the HSROC AUC was 0.83 (CI: 0.69, 0.91), indicating good overall predictive accuracy (Fig. 5A). Sensitivity and specificity were 0.70 (CI: 0.41, 0.89) and 0.80 (CI: 0.67, 0.89), respectively. The positive likelihood ratio for LGE was 3.5 (CI: 2.1, 5.9) indicating a small increase in the likelihood of disease after test discrimination. Conversely, the negative likelihood ratio was 0.37 (CI: 0.17, 0.83) suggesting a small decrease in the likelihood of disease after test discrimination. The DOR was 9 (CI: 3, 29), demonstrating the discriminatory ability of LGE. Fagan’s nomogram, with a pre-test probability for Co-VAs of 23%, showed a post-test probability of 51% in the presence of LGE and 10% in the absence of LGE (Fig. 5B).

Fig. 5
figure 5

Hierarchical summary receiver operating characteristic curve plot (A, C) and Fagan Plot (B, D) for LGE and MAD

Full size image

For MAD, the HSROC AUC was 0.69 (CI: 0.54, 0.81), with a sensitivity of 0.77 (CI: 0.51, 0.91) and a specificity of 0.43 (CI: 0.13, 0.80) (Fig. 5C). The positive and negative likelihood ratios were 1.3 (CI: 0.8, 2.2) and 0.54 (CI: 0.36, 0.82), respectively. The DOR for MAD was 2 (CI: 1, 5), indicating lower overall diagnostic efficacy compared to LGE. According to Fagan’s nomogram for MAD, with a pre-test probability of 17%, the post-test probability was 22% if MAD was present and 10% if absent (Fig. 5D).

Discussion

Our systematic review and meta-analysis evaluated the role of CMR in risk stratification of arrhythmia in patients with MVP.

The findings of the study emphasize the significance of LGE in the process of arrhythmic risk classification for patients diagnosed with MVP. Specifically, the presence of LGE demonstrated good overall predictive accuracy with balanced performance in identifying MVP patients at risk for Co-VA, with a Log OR of 2.12, positive likelihood ratio of 3.5, negative likelihood ratio of 0.37, sensitivity of 0.70, and specificity of 0.80 with an AUC of 0.83. Furthermore, LGE was also associated with SVT-VF-aSCD (log OR of 1.69), highlighting its utility in identifying MVP patients at higher risk of major arrhythmic events.

This observation is consistent with the growing body of evidence highlighting the predictive value of LGE for Co-VAs and SCD in MVP patients [8,9,10, 14, 16, 17, 21]. Indeed, higher extension of low voltages has been found in electrophysiological studies in the case of LGE involving papillary muscle and lateral/infero-lateral LV wall (in patients with MVP and complex arrhythmia, supporting the role of fibrosis as electrophysiological substrate [22].

Myocardial fibrosis in MVP seems to be the result of increased traction of the papillary muscles and mechanical stretch of the surrounding myocardium. LGE thus mainly involves the inferior and lateral basal wall of the LV, with a non-ischemic appearance (mid-wall or patchy) and less frequently with subendocardial pattern and the posteromedial papillary muscle [23]. Moreover, systolic curling of the mid-basal lateral wall may constitute a potential electromechanical trigger further enhancing the risk of VA [23].

Ever-increasing evidence supports the value of fibrosis assessment by LGE for arrhythmic risk stratification across several clinical settings. A recent meta-analysis conducted by Al-Sadawi et al [24] on more than ten thousand non-ischemic cardiomyopathy patients confirmed a significant 4.6 risk of VA and SCD among LGE-positive patients. In general, when evaluating a patient with MVP, it is essential to always consider other possible etiologies that may cause a similar pattern of LGE (e.g., myocarditis and coronary artery disease). Overall, our findings confirm and reinforce the value of LGE assessment for arrhythmic risk stratification also among patients with MVP.

Importantly, despite strong association with Co-VAs, a non-negligible proportion of events occurred also among patients without LGE, suggesting the multifactorial etiology of arrhythmias and supporting the need for further research to achieve a multiparametric risk stratification. Of note, interstitial fibrosis detected only by mapping techniques may partly explain this discrepancy. Indeed, increased native T1 and ECV values on the lateral wall in patients with MVP were associated with arrhythmic events even in the absence of LGE [15, 25]. However, we could not meta-analyze T1 mapping due to the limited study number available. Future studies will elucidate the addictive performance of these promising markers over LGE among MVP patients.

The analysis also supported the role of MAD in predicting arrhythmic risk. Although the presence of MAD showed a moderate association with co-VAs (log OR 0.95), with relatively high sensitivity (0.77) and negative likelihood ratio (1.3), the predictive value of MAD appears to be lower than that of LGE. Its low specificity (0.43) may indeed be related to the dichotomic evaluation of MAD rather than an evaluation based on MAD extension and site. In fact, a degree of MAD seems to be prevalent in the general population, possibly representing a normal variant of mitral annulus anatomy [26]. Conversely, MAD extension over a certain length, especially in the setting of MVP, resulted associated with Co-VAs and SCD, with pathological MAD reported for length higher than 5 mm or 8.5 mm [9, 27], albeit a precise cut-off that remains to be determined, and when localized in the infero-lateral wall and associated with MVP [9, 23, 26]. Associated with MVP, longer MAD seems to worsen mitral annulus dynamics with subsequent more severe mechanical wall stretch and fibrosis, concurring in eliciting arrhythmias.

A multimodality approach to MVP patients is increasingly recognized in importance, especially for arrhythmic risk stratifications. Indeed, the value of continuous rhythm monitoring either with Holter-ECG or implantable loop recorders demonstrated the importance of VA burden and complexity in predicting life-threatening events [28]. These tools could be used comprehensively together with CMR, although their combined performance remains to be evaluated. Moreover, new CMR parameters may complement LGE evaluation.

Contrary to previous studies [8, 10], our analysis showed that the prolapse distance, LVEDVi, and LVEF were not significantly associated with co-VAs. This discrepancy could be due to differences in patient selection criteria across studies. However, it also underscores the complexity of MVP and the multifactorial nature of its associated arrhythmic risks. Therefore, a single parameter might not be sufficient to capture the nuanced risk profile of an MVP patient.

Overall, our findings confirm the usefulness of a CMR approach in arrhythmic risk stratification in MVP patients, as recommended by the ESC/EACTS guidelines by the EHRA expert consensus statement [5, 7]. Given the multifaceted nature of MVP-related arrhythmias, combining several CMR features, particularly LGE and MAD features, might provide a more comprehensive and accurate risk assessment.

Some limitations of this analysis must be acknowledged. Due to the retrospective nature of the majority of included studies, inherent bias cannot be ruled out. In addition, heterogeneity in study design and patient populations may have influenced the results, however, only one study was assessed as having a high risk of bias. Moreover, we could not analyze the impact of MR severity, leaflet length/thickness, curling, MAD distance, and mapping techniques because a limited number of studies were available. This underscores the importance of further research to assess whether these parameters may improve current well-validated risk stratification markers. Finally, no attempt was made to find unpublished or grey data, as the focus of this study was solely on published literature.

In conclusion, this systematic review and meta-analysis emphasize the value of LGE as a key CMR characteristic for assessing arrhythmia risk classification in patients with MVP. Moreover, the presence of MAD may further stratify these patients. Thus, a multi-parametric CMR approach provides a more comprehensive risk assessment. Future prospective multicenter studies are needed to validate these findings and to establish standardized guidelines for the integration of CMR in the management of MVP. Finally, while CMR is a useful tool for risk stratification, its results must always be linked with clinical parameters and individual patient features for a thorough risk assessment.