Article Text

Download PDFPDF

Original research
Multimodality imaging assessment of mitral annular disjunction in mitral valve prolapse
Free
  1. Valentina Mantegazza,
  2. Valentina Volpato,
  3. Paola Gripari,
  4. Sarah Ghulam Ali,
  5. Laura Fusini,
  6. Gianpiero Italiano,
  7. Manuela Muratori,
  8. Gianluca Pontone,
  9. Gloria Tamborini,
  10. Mauro Pepi
  1. Department of Cardiovascular Imaging, Centro Cardiologico Monzino IRCCS, Milan, Italy
  1. Correspondence to Dr Valentina Mantegazza, Department of Cardiovascular Imaging, Centro Cardiologico Monzino IRCCS, Milano 20138, Italy; valentina.mantegazza{at}cardiologicomonzino.it

Abstract

Objective Mitral annular disjunction (MAD) is an abnormality linked to mitral valve prolapse (MVP), possibly associated with malignant ventricular arrhythmias. We assessed the agreement among different imaging techniques for MAD identification and measurement.

Methods 131 patients with MVP and significant mitral regurgitation undergoing transthoracic echocardiography (TTE) and cardiac magnetic resonance (CMR) were retrospectively enrolled. Transoesophageal echocardiography (TOE) was available in 106 patients. MAD was evaluated in standard long-axis views (four-chamber, two-chamber, three-chamber) by each technique.

Results Considering any-length MAD, MAD prevalence was 17.3%, 25.5%, 42.0% by TTE, TOE and CMR, respectively (p<0.05). The agreement on MAD identification was moderate between TTE and CMR (κ=0.54, 95% CI 0.49 to 0.59) and good between TOE and CMR (κ=0.79, 95% CI 0.74 to 0.84). Assuming CMR as reference and according to different cut-off values for MAD (≥2 mm, ≥4 mm, ≥6 mm), specificity (95% CI) of TTE and TOE was 99.6 (99.0 to 100.0)% and 98.7 (97.4 to 100.0)%; 99.3 (98.4 to 100.0)% and 97.6 (95.8 to 99.4)%; 97.8 (96.2 to 99.3)% and 93.2 (90.3 to 96.1)%, respectively; sensitivity (95% CI) was 43.1 (37.8 to 48.4)% and 74.5 (69.4 to 79.5)%; 54.0 (48.7 to 59.3)% and 88.9 (85.2 to 92.5)%; 88.0 (84.5 to 91.5)% and 100.0 (100.0 to 100.0)%, respectively. MAD length was 8.0 (7.0-10.0), 7.0 (5.0-8.0], 5.0 (4.0-7.0) mm, respectively by TTE, TOE and CMR. Agreement on MAD measurement was moderate between TTE and CMR (ρ=0.73) and strong between TOE and CMR (ρ=0.86).

Conclusions An integrated imaging approach could be necessary for a comprehensive assessment of patients with MVP and symptoms suggestive for arrhythmias. If echocardiography is fundamental for the anatomic and haemodynamic characterisation of the MV disease, CMR may better identify small length MAD as well as myocardial fibrosis.

  • cardiac magnetic resonance (CMR) imaging
  • echocardiography
  • mitral regurgitation

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

Mitral annular disjunction (MAD) is an anatomic feature closely linked to mitral valve prolapse (MVP).1–6 It is defined as a detachment of the left atrial (LA) wall-mitral valve annulus (MVA) junction from the left ventricular (LV) myocardium1 7 and its significance has been investigated in different clinical and histological series.

MAD seems to contribute to MVA distortion5 8 and systolic dysfunction2 3 5 as well as to accelerated progression of the underlying mitral valve (MV) disease, leading to more complex MV lesions.2 5 8 9 Additionally, an association between MAD and malignant ventricular arrhythmias (VAs) has been suggested.3 4 9–11 MAD has been previously evaluated by different imaging techniques, either transthoracic echocardiography (TTE),3 6 8 two-dimensional (2D) or three-dimensional transoesophageal echocardiography (TOE)2 5 or cardiac magnetic resonance (CMR).4 10 Although some papers have raised concerns about the wide range of MAD length values obtained by different methodologies12 or questioned TTE diagnostic accuracy for the detection of MAD,10 11 a systematic comparison between different imaging modalities in detecting and measuring MAD has never been accomplished.

The aim of our study was to evaluate MAD via different imaging methods and assess the agreement among them on MAD localisation and measurement in a cohort of patients with MVP and significant mitral regurgitation (MR) requiring surgery. Moreover, we aimed to determine TTE and TOE diagnostic accuracy as compared with CMR.

Methods

We reviewed our database of patients referred at our Institute from 2016 to 2019 for MVP secondary to Barlow’s disease (BD) or fibroelastic deficiency (FED) with moderate-to-severe or severe MR, considered eligible for surgery according to the European Guidelines.13 We retrospectively identified consecutive patients undergoing both TTE and CMR before surgery, within 6 months of each other. For a subgroup of patients, TOE was also available in our digital archive.

Imaging

TTE was performed using either a GE echocardiographic system (Vivid E9 or Vivid E95—GE Vingmed, Horten, Norway) or a Philips ultrasound machine (iE33 or Epiq—Philips Medical Systems, Andover, Massachusetts). Complete standard 2D TTE protocol was performed according to clinical laboratory practice. Left chambers volumes and LV ejection fraction were computed from four-chamber and two-chamber views using the biplane Simpson’s method. MVP was defined as a systolic displacement ≥2 mm of one or both MV leaflets above the MVA plane as observed in a long-axis view. MR grade was established integrating qualitative and quantitative criteria.14 Pulmonary artery systolic pressure was calculated from the peak tricuspid regurgitant velocity.15 TOE was performed using Philips machines, equipped with X7-2t or X8-2t probe, either to assess the MV anatomy in case of inadequate or inconclusive TTE images and/or to guide the surgical procedure. Since MAD can discontinuously involve any segment along the left atrioventricular junction, apart from the mitro-aortic curtain,5 7 TTE and TOE clips were reviewed frame-by-frame for off-line assessment of MAD in standard long-axis views (four-chamber, two-chamber and three-chamber) by an experienced observer. Specifically, we defined and measured MAD as any distance detected between the LA wall-posterior MVA junction and the LV wall at end-systole. Curling motion of the basolateral LV wall was also visually assessed by TTE,4 10 as an outward movement of the MVA during systole.

CMR studies were performed using a 1.5 T scanner (Discovery MR450—GE Healthcare, Waukesha, Wisconsin, USA) with a 32-channel phased array receiver coil. After acquiring localiser images of the heart, breath-hold steady-state free precession cine imaging was performed applying the following parameters: temporal resolution=30 frames per cardiac cycle; repetition time=3.8 ms, echo time=1.7 ms, flip angle=85°, slice thickness=8 mm, field of view=300×300–360 mm2; matrix=512×512. The same observer assessed MAD on CMR cine images in a randomised and blinded order from four-chamber, two-chamber and three-chamber LV long-axis views and measured its length from the LA wall-posterior MV leaflet junction to the top of the LV wall during end-systole, as previously described.4

For each view, feasibility of MAD assessment by TTE, TOE and CMR was noted and those views in which MVA could not be adequately evaluated were excluded from per-view analyses. Patients with inadequate images from all long-axis views obtained by a single technique were excluded in per-patient analyses.

To assess reproducibility of MAD evaluation, intraobserver variability was performed by the same reader repeating MAD assessment after ≥1 month; interobserver variability was performed by a second experienced reader, blinded to the previous results.

All patients provided written informed consent and the institutional review board approved the study protocol.

Statistical analysis

Continuous data are presented as mean±SD or median (IQR), after testing for normal distribution with the Shapiro-Wilk test and categorical variables as frequencies and percentages. Continuous variables were compared using the unpaired Student’s t test or Mann-Whitney U test, as appropriate. The χ2 test or Fisher’s exact test (if the expected cell count was <5) were used for categorical variables. General characteristics of the population were compared based on the presence or absence of MAD at CMR (MAD+ versus MAD–, respectively). Cohen’s Kappa statistic was run to determine the strength of agreement between pairs of imaging techniques for the identification of MAD in the same view per single patient. Considering those views in which an agreement was observed, the correlation between MAD measurements was evaluated with the Spearman correlation coefficient (ρ) and the agreement was tested applying the Bland-Altman analysis. Because of its higher spatial resolution and better tissue characterisation, CMR was then assumed as reference, and TTE and TOE diagnostic accuracy for the correct identification of MAD was determined as compared with CMR, using 2×2 contingency tables and according to three different cut-off values for MAD length detected by CMR (≥2 mm, ≥4 mm and ≥6 mm).

Intraobserver and interobserver agreement for the identification of MAD in the same view for each patient was expressed by kappa coefficient (κ) and disagreement proportion. Considering those views in which an agreement was found, intraobserver and interobserver variability in MAD measurements was expressed as intraclass correlation coefficient and coefficient of variation. Bland-Altman analysis was used to express bias and limits of agreement. All results were considered significant with p<0.05. Statistics were performed with SPSS 26 (SPSS, Chicago, Illinois, USA).

Patient and public involvement statement

The patients and public were not involved in the creation of the study design, recruitment or statistical analysis. Patients were not consulted to develop patient relevant outcomes or interpret the results. Patients were not invited to contribute to the writing or editing of this document for readability or accuracy.

Results

Patient population

One-hundred and thirty-one patients with available TTE and CMR were enrolled in the study, while TOE was available in 106 patients, corresponding to a total of 393 views from TTE and CMR, and 318 views from TOE, respectively. Four patients were not included in per-patient analyses when examining TTE images, as MAD assessment was unfeasible from all long-axis views.

General characteristics and echocardiographic parameters of our study population are shown in table 1. Higher prevalence of BD compared with FED was observed. Dividing our population into two groups based on the presence of any-length MAD at CMR, there was a significantly higher representation of myxomatous MVP in MAD+ versus MAD–.

Table 1

General characteristics and echocardiographic parameters of the study population classifying patients based on the presence of any-length MAD by CMR

MAD assessment

MAD evaluation was feasible in 350/393 views (89.1%) by TTE, 295/318 views (92.8%) by TOE and 380/393 views (96.7%) by CMR, respectively (figure 1).

Figure 1

Feasibility of MAD assessment by TTE and TOE and study flowchart for the evaluation of TTE and TOE diagnostic accuracy as compared with CMR. TTE was compared with CMR in a total of 337 views; TOE was compared with CMR in a total of 285 views. CMR, cardiac magnetic resonance; MAD, mitral annular disjunction; MVA, mitral valve annulus; TOE, transoesophageal echocardiography; TTE, transthoracic echocardiography.

MAD prevalence is reported in table 2. Minimum and maximum values of MAD distance were 2–12, 4–12 and 4–13 mm, respectively, by CMR, TOE and TTE. Median MAD measurements and IQRs were 5.0 (4.0-7.0) mm by CMR, 7.0 (5.0-8.0) by TOE and 8.0 (7.0-10.0) by TTE. MAD prevalence was significantly higher in two-chamber than four-chamber and three-chamber views, regardless of the technique employed, whereas median MAD measurement was not significantly different among views (table 3).

Table 2

Overall MAD prevalence in the study population detected by TTE, TOE and CMR

Table 3

MAD prevalence and measurement (median and IQR, minimum and maximum values) obtained by TTE, TOE and CMR per single long-axis view, considering for each technique those patients whose posterior mitral valve annulus was assessable in all long-axis views

MAD could also be visualised in diastole in 5/22 patients by TTE, 10/27 patients by TOE and 22/55 patients by CMR, corresponding to a systolic MAD length of 10.0 (10.0-11.0), 8.5 (7.0-9.0) and 8.0 (6.0-9.8) mm, respectively (online supplementary figure 1).

Supplemental material

Agreement among different imaging techniques on MAD identification and measurement

To assess the agreement between imaging modalities, we performed per-view analyses considering views where MVA could be adequately assessed by both the techniques being compared with each other: TTE and TOE were compared in 268 views, TTE and CMR in 337, TOE and CMR in 285. A significant intertechnique agreement was found for the identification of MAD in the same view for each subject (TTE vs TOE: κ=0.73, 95% CI 0.68 to 0.78, p<0.05; TTE vs CMR: κ=0.54, 95% CI 0.49 to 0.59, p<0.05; TOE vs CMR: κ=0.79, 95% CI 0.74 to 0.84, p<0.05) (figure 2). Considering MAD ≥2 mm detected by CMR as reference, TTE and TOE accuracy were 88.7% (95% CI 85.3 to 92.1)% and 94.7% (95% CI 92.1 to 97.3)%, respectively (online supplementary table 1; figures 1 and 3). TTE and TOE accuracy improved to 97.0% (95% CI 95.2, 98.8)% and 96.5% (95% CI 94.4, 98.6)%, assuming a cut-off value for MAD length by CMR ≥6 mm and ≥4 mm, respectively (online supplementary table 1).

Supplemental material

Figure 2

MAD evaluation by multimodality imaging. Shown are two example of Barlow’s disease, where the LA-posterior mitral valve annulus junction was assessed by transthoracic echocardiography (A), transoesophageal echocardiography (B) and cardiac magnetic resonance (C). The three techniques are concordant on presence (upper panels) and absence (lower panels) of MAD in two-chamber long-axis view, at P3 level. In the upper panels, MAD is identified (yellow line) and measured at end-systole. LA, left atrium; LV, left ventricle; MAD, mitral annular disjunction.

Figure 3

Disagreement on MAD presence between echocardiographic imaging and cardiac magnetic resonance. Shown is the example of a patient with fibroelastic deficiency and P2 flail with a few-millimetres detachment of the mitral valve annulus from the LV wall. MAD (yellow line) is identified and measured at end-systole in two-chamber long-axis view by cardiac magnetic resonance (C). Transthoracic echocardiography (A) and transoesophageal echocardiography (B) show no clear signs of mitral valve annulus detachment from the LV myocardium. LA, left atrium; LV, left ventricle; MAD, mitral annular disjunction.

As concerns MAD measurements, the agreement was assessed considering those views in which MAD was detected by both the techniques being compared with each other. A total of 18 views were examined when comparing TTE with TOE, 28 when TTE was compared with CMR and 35 when TOE was compared with CMR. The correlation and the level of agreement between measurements are graphically represented by Spearman’s and Bland-Altman plots, respectively (figure 4).

Figure 4

Correlations and agreement on MAD measurement among imaging techniques. Results of Spearman’s correlation (upper panels) and Bland-Altman analysis (lower panels) to evaluate the agreement among different imaging techniques on MAD measurement obtained in concordant views. The blue dashed lines represent bias and the red solid lines represent ±1.96 SD. CMR, cardiac magnetic resonance; LOA, limits of agreement; MAD, mitral annular disjunction; TOE, transoesophageal echocardiography; TTE, transthoracic echocardiography.

Intraobserver and interobserver variability for MAD presence and measurement was performed on views where MAD evaluation was feasible and is reported in table 4.

Table 4

Intraobserver and interobserver agreement on MAD presence and MAD measurement in the same view per single patient by TTE, TOE and CMR

Discussion

Growing evidence on the role of MAD in the clinical course and prognosis of patients with MVP has emerged in the last few decades. MAD is associated with more severe MV lesions,2 5 9 and paradoxical systolic expansion of the MVA.2 3 5 Moreover, clinical series have demonstrated that MAD is potentially responsible for sustained ventricular tachycardia and sudden cardiac death (SCD).3 4 10 11 Therefore, the recognition of MAD in patients with MVP appears of utmost importance, especially in those reporting symptoms suggestive of arrhythmias. Imaging methods that have been applied so far for the evaluation of MAD were TTE, TOE and CMR but a reference standard imaging technique has not been established yet.

These three imaging modalities were compared with each other in our study population of MVP patients with severe MR. A significant difference in MAD prevalence emerged between echocardiographic imaging and CMR. When considering any-length separation between the LV myocardium and the LA wall-posterior MVA junction, we detected MAD in 42.0% patients by CMR, which was slightly higher than reported in a recent paper describing MAD in 35% patients with MVP (including both BD and FED) evaluated by CMR.11 Such difference may be due to the higher prevalence of BD in our population (66.4% vs 44.9%). Although several studies indicated that MAD is not associated exclusively with BD,6–8 10 11 higher prevalence of myxomatous MVP than FED was shown in patients with MAD+, which is in agreement with recent papers.8 11 The lower detection rate of MAD by echocardiography could be ascribed to different mechanisms, such as inadequate acoustic window, shadowing or reverberations caused by posterior MVA calcification and a lower spatial resolution compared with CMR. Assuming CMR as reference in per-view analysis, diagnostic accuracy of echocardiographic imaging was mainly affected by underdiagnoses, especially by TTE. As a unique definition of MAD is not available in the literature, per-patient and per-view analyses were performed first considering as reference any MVA detachment detected by CMR (≥2 mm) and then considering as reference a MAD length ≥4 mm and ≥6 mm. The highest diagnostic accuracy of TTE and TOE was shown selecting a MAD cut-off value by CMR ≥6 mm and ≥4 mm, respectively.

With respect to MAD length, no significant difference was determined among the three methods. The correlation between measurements was moderate for TTE versus CMR and strong for TOE versus CMR. Although the presence of positive bias at Bland-Altman analysis highlights a tendency to slightly overestimate this distance by TTE, and to a lesser degree by TOE, compared with CMR, it is reasonable to believe that such difference may not have clinically significant impact.

The intrinsic limitations in MAD measurement are reflected also by intraobserver and interobserver variability, showing quite high coefficients of variation, still in the absence of clinically significant bias. However, the highest correlation between measurements was documented by CMR, for both intraobserver and interobserver variability. As regards concordance on MAD presence in the same view, the strongest intraobserver and interobserver agreement was shown by CMR; however, the disagreement proportion, which is less affected by MAD prevalence, was similar among all techniques. As a whole, these findings may suggest that a standardisation of MAD definition and measurement is strongly needed in clinical practice.

Although maximal MAD distance has been reported to be most frequently located at P2 scallop5 and although MAD has been frequently assessed exclusively from three-chamber3 4 6 11 and sometimes from four-chamber view,2 4 we found that MAD distance was not significantly different among views, although it was most frequently observed in two-chamber view at P3 level. Our data are in line with Dejgaard and colleagues, who showed higher prevalence of MAD along the LV inferior wall compared with the posterolateral wall in a heterogeneous population of patients with MAD+.10 Although two-chamber view has rarely been analysed for MAD evaluation, it seems plausible that MAD can be clinically identified by cardiac imaging along the LV inferior wall similarly as the lateral wall, since pathological studies described MAD at any point around the MVA.7 These observations further support the importance of a routine evaluation of MAD spanning the entire MVA circumference.5 10

The results of the present study confirm that different diagnostic methods are not interchangeable for all MAD values.12 Although several studies have suggested that the greater the extent of MAD, the higher is the arrhythmic risk, there is no consensus on whether there is a specific MAD length associated with malignant VAs.3 4 10 11 Further studies are desirable to deeply investigate the clinical significance of MAD and possibly to establish the cut-off value discriminating patients at high arrhythmic risk, thus guiding the clinicians in the choice of the most appropriate imaging test for MAD diagnosis.

The ability to visualise the posterior MVA varies among different imaging techniques.16 CMR has a unique capacity to distinguish adjacent structures and to characterise the myocardial tissue. If on one side, it allows the detection of a 2 mm-separation of the posterior MVA from the LV wall, on the other hand, it accurately identifies myocardial fibrosis in the posterior papillary muscle and in the LV inferobasal segment by late gadolinium enhancement.17 18

However, echocardiography remains an essential diagnostic tool for the evaluation of MVP because of its widespread availability and low costs and for its excellent capacity to evaluate the haemodynamic consequences of MR and to assess MV morphology.19 20 In clinical practice, TOE is mainly advocated in patients with inconclusive or technically difficult TTE to better define the MV anatomy before MV surgery, and this study also shows a strong agreement between TOE and CMR for MAD measurement and MAD identification for values ≥4 mm. Still, it is not routinely performed because of its semi-invasive nature.21

Implications

MVP is a complex disease, which primarily involves the MV apparatus morphology and secondly affects cardiac haemodynamics, heart chambers dimensions and ventricular function. VAs have also been reported with an estimated incidence of SCD ranging from 0.2%/year to 0.4%/year18. The risk factors for VAs include extravalvular features, LV tissue characteristics, MVP-related functional and morphological factors, including MAD12 possibly via myocardial stretch of the LV inferobasal wall and papillary muscles, with relative hypertrophy and mechanically-induced fibrosis.4 10 Therefore, MAD should be routinely evaluated in patients with MVP10 22 and those with an incidental finding of MAD should be first investigated for symptoms suggestive of arrhythmias and undergo a non-invasive clinical evaluation with baseline ECG and Holter monitoring. A recent review reported that the majority of patients with MVP with SCD were symptomatic before the index event or had abnormal electrocardiographic or Holter findings,9 so it seems reasonable that patients without MAD by TTE showing these characteristics should be evaluated by TOE or CMR. Specifically, if MAD<4 mm (which is most frequently underdiagnosed by echocardiographic imaging) will be demonstrated to increase the arrhythmic risk, CMR may constitute a complementary technique for the evaluation of the MVA and a valuable method to better identify MAD in symptomatic patients. In this context, a multimodality imaging approach in MVP would add value to individual diagnostic tests. However, since suggested mechanisms responsible for VAs are still speculative, the relationship between MVP and VAs remains to be clarified and requires a great effort yet. Large multicentre prospective trials reporting clinical variables, genetic testing results, longitudinal data on cardiac imaging, recorded data on VAs and long-term outcomes are needed.

Study limitations

Some limitations have to be acknowledged in this work. It was a single-centre study conducted in a selected population of relatively small size. We focused on patients with MVP with surgical indication, so our results may not hold true for patients with mild or moderate MR or in other clinical settings. Furthermore, limitation of CMR in terms of temporal resolution may have influenced the identification of the precise end-systolic time-point. Finally, the present study had a retrospective nature and all the examinations were executed for clinical reasons; therefore, the evaluation of MAD was performed in standard long-axis views. Echocardiographic off-axis views or dedicated CMR long-axis cine images could allow a complete visualisation of the posterior MVA and thus increase MAD detection rate.

Conclusion

A widely accepted definition of MAD and its clinical implications still remain to be clarified in the clinical scenario. An integrated imaging approach could be necessary for a comprehensive assessment of MVP. Echocardiographic imaging is fundamental for the anatomic and haemodynamic characterisation of the MV disease, though CMR may better define MAD of minor degree and myocardial fibrosis, helping the physician in identifying patients with an arrhythmogenic substrate.

Key messages

What is already known on this subject?

  • Mitral annular disjunction (MAD) is suggested to be a structural abnormality in mitral valve prolapse (MVP) associated with malignant arrhythmias. Despite different imaging techniques are employed for the assessment of MAD, there is a paucity of data evaluating their agreement on MAD presence and measurement.

What might this study add?

  • The agreement between transthoracic (TTE) or transoesophageal echocardiography (TOE) and cardiac magnetic resonance (CMR) for the detection of MAD varies considering different cut-off values for MAD length. Assuming CMR as reference, TTE and TOE demonstrate the highest diagnostic accuracy for values of MAD distance by CMR ≥6 mm and ≥4 mm, respectively.

  • The agreement on MAD measurement is moderate between TTE and CMR and strong between TOE and CMR.

How might this impact on clinical practice?

  • Patients without MAD by TTE reporting symptoms suggestive for arrhythmias or showing abnormal ECG or Holter findings should be evaluated by TEE or CMR. Specifically, if large-scale studies will demonstrate MAD of small length (<4 mm) being responsible for malignant ventricular events, CMR may be desirable to better identify patients with MVP with high arrhythmogenic risk.

References

Footnotes

  • Contributors VM and MP contributed to the conception and design of the study. VV, PG, SGA, LF, GI, MM and GT contributed to data collection, analysis and interpretation. VM, GP and MP drafted and revised the manuscript. All authors read and approved the manuscript before its submission.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests GP reports research grants and/or honorarium as speaker from GE Healthcare, Bracco, Bayer, Medtronic and Heartflow.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Patient consent for publication Not required.

  • Ethics approval The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Ethics Committee (R1168/20-CCM 1230). Informed consent was obtained from all subjects.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement Data are available on reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.

Linked Articles