Article Text

Download PDFPDF

Variable phenotypes of bicuspid aortic valve disease: classification by cardiovascular magnetic resonance
  1. Stefan Buchner1,
  2. Marion Hülsmann1,
  3. Florian Poschenrieder2,
  4. Okka W Hamer2,
  5. Claudia Fellner2,
  6. Reinhard Kobuch3,
  7. Stefan Feuerbach2,
  8. Günter AJ Riegger1,
  9. Behrus Djavidani2,
  10. Andreas Luchner1,
  11. Kurt Debl1
  1. 1Klinik und Poliklinik für Innere Medizin II, Universitätsklinikum Regensburg, Germany
  2. 2Institut für Röntgendiagnostik, Universitätsklinikum Regensburg, Germany
  3. 3Klink und Poliklinik für Herz-, Thorax- und herznahe Gefäßchirurgie, Universitätsklinikum Regensburg, Germany
  1. Correspondence to Dr Stefan Buchner, Klinik und Poliklinik für Innere Medizin II, Universitätsklinikum Regensburg, Franz-Josef-Strauss-Allee 11, 93042 Regensburg, Germany; stefan.buchner{at}


Background Recently, cardiovascular magnetic resonance (CMR) has been shown to allow accurate visualisation and quantification of aortic valve disease. Although bicuspid aortic valve (BAV) disease is relatively rare in the general population, the frequency is high in patients requiring valve surgery. The aim of the current study was to characterise the different phenotypes of BAV disease by CMR.

Methods CMR studies were performed on a 1.5 T scanner in 105 patients with BAV.

Results The pattern of BAV phenotypes was as follows: a raphe was identified in 90 patients (86%). Among patients with raphe, 76 patients had fusion between the right and left cusps (RL) and 14 patients had fusion between the right and the non-coronary cusps (RN). There were no significant differences in the aortic dimensions in the different BAV phenotypes.

Conclusion CMR allows excellent characterisation of valve phenotype in patients with BAV. The present data demonstrate that a raphe is present in the vast majority of cases and RL fusion is the predominant phenotype of BAV. No significant differences in the aortic dimensions were observed.

  • Bicuspid aortic valve
  • magnetic resonance imaging
  • aortic stenosis
  • aortic regurgitation
  • surgery-valve
  • echocardiography (transoesophageal)
  • MRI
  • aortic valve disease

Statistics from


The bicuspid aortic valve (BAV) is a common congenital cardiac malformation with an estimated prevalence of 0.5–2% in the general population.1–3 The BAV is increasingly recognised as a disease of the entire aortic root and ascending aorta with heterogeneous phenotypes including valvular and vascular complications.4–6 The BAV has further gained interest from a pathogenetic and surgical point of view.7 8 Valve dysfunction in BAV is characterised by an earlier onset and a higher rate of progression in comparison to patients with tricuspid aortic valve disease.9 Dilatation of the proximal aorta seems to be independent of coexistent valve dysfunction. An exact classification of valve morphology and lesion characterisation is essential in BAV with aortic regurgitation, because reconstructive surgical techniques are increasingly performed.10 Recently, a systematic classification of BAV with respect to the numbers and spatial position of the cusps and raphes, as well as the functional status, was found reliable and appropriate for a classification system from a surgical point of view.11 However, diagnosis and classification of BAV by imaging techniques such as echocardiography remains challenging, mostly due to impaired image quality. Cardiovascular magnetic resonance imaging (CMR) is increasingly applied in aortic valve disease for the assessment of the severity of valvular stenosis and regurgitation. New CMR sequences, such as steady-state free-precession techniques allow an exact visualisation of valve morphology6 12 13 and aortic root dimensions.14 Additionally, superior image quality of CMR in comparison to transoesophageal echocardiography has been shown.13 15 To the authors' knowledge, CMR has not been used previously to study the phenotype in patients with BAV.

Therefore, it was hypothesised that CMR could characterise BAV phenotypes in detail. The aim of the present study was to apply CMR to determine BAV phenotypes and aortic dilatation in a cohort of 105 patients who underwent evaluation for aortic valve surgery.



BAVs diagnosed by CMR were retrospectively analysed between March 2003 and December 2007. CMR was performed in a total of 303 patients with aortic valve disease who were referred to the Universitätsklinikum Regensburg, Germany for evaluation of suspected or previously diagnosed aortic valve disease and for disease severity. All patients gave informed consent. In 10 patients, classification of the aortic valve was not possible due to unacceptable image quality (three pulsation artefacts, three movement artefacts, two coil artefacts, two respiratory artefacts). Of these 10 patients, three patients had a transoesophageal echocardiography and the valve morphology was tricuspid. The presence of BAV was verified by image review by two investigators (KD, BD), then again by another two investigators (SB, MH). Twelve patients were classified as unicuspid aortic valve (Sievers BAV type II) as previously published.16 One-hundred and five aortic valves were classified as bicuspid by the consensus of the two review processes. The aortic measurements were performed by a single observer (MH). In a subset of randomly 20 patients, the measurements were repeated by a second observer (SB) and by the same observer at a different time to assess interobserver and intraobserver variabilities. Follow-up information was obtained by review of the medical records of the BAV patients or direct contact. Follow-up was complete for 100 patients (95%).


From the 105 patients, 62 patients were examined by transoesophageal echocardiography. These studies were then reviewed by two experienced observers blinded to the CMR data. Aortic valve morphology was examined in systole and diastole in the short axis view around 60 degrees. The aortic valves were graded with respect to calcification according to a semiquantitative scale (0=none; 1=slight; 2=moderate; 3=severe). The aortic valves were classified by echocardiography according to the same system as described for CMR.

CMR studies

CMR studies were performed on a 1.5 T scanner (Sonata and Avanto, Siemens Medical Solutions). Cine images were acquired in multiple short axis and long axis views with fast imaging with steady-state free-precession (trueFISP, slice thickness 8 mm, echo time 1.53 ms, readout bandwidth 1.085 Hz/pixel, repetition time 3.14 ms, matrix 256*202). The number of Fourier lines per heartbeat was adjusted to allow the acquisition of 20 cardiac phases covering systole and diastole within a cardiac cycle. The field of view was 340 mm on average and adapted to the size of the patient. Calculation of LV volumes, mass and ejection fraction was performed in the serial short axis slices.

The imaging plane of the aortic valve was defined by acquiring a systolic three chamber view and an oblique coronal view of the aortic valve and proximal aorta, as described previously.13 The subsequent slices (slice thickness 5 mm) were defined parallel to the valvular plane and additionally, in case of orifices with an eccentric outlet, perpendicularly to the direction of the jet. At least four slices (range 4–7) of the aortic valve were acquired shift wise in 2.5 mm steps. Additionally, prospectively triggered through plane breath-hold phase velocity mapping (PVM) was performed in the same slice positions (flash 2D; slice thickness 5 mm, echo time 3.2 ms, readout bandwidth 391 Hz/pixel, repetition time 55 ms, matrix 256*125). Further, aortic regurgitant volume and antegrade stroke volume were quantified by through plane PVM in a retrospective gating technique (flash 2D; slice thickness 5 mm, echo time 3.2 ms, pixel bandwidth 391 Hz, repetition time 41 ms, matrix 256*192) during normal respiration to cover the whole cardiac cycle. Slice position was perpendicular to the ascending aorta as close as possible above the aortic valve.17 18

The aortic valve disease severity was determined by CMR. In aortic stenosis, aortic valve area was assessed by planimetry, as described previously.13 In cases with moderate and severe AR, the severity was determined quantitatively by planimetry of the aortic regurgitant orifice and/or aortic regurgitant volumes and regurgitant fraction, as described previously6 or semiquantitatively19 in less severe cases. Valve dysfunction was classified using standard criteria (severe aortic stenosis: aortic valve area <1.0 cm2, moderate aortic stenosis: aortic valve area 1.0–1.5 cm2, mild aortic stenosis: aortic valve area >1.5 cm2). Aortic regurgitation was graded in mild, moderate (regurgitant fraction 30–40%), moderately severe (regurgitant fraction 40–50%) and severe (regurgitant fraction >50%). The studies were graduated regarding image quality on a semiquantitative scale (1=good image quality; 2=moderate image quality; 3=poor image quality; 4= unacceptable).

Categorisation of BAV

BAV stands for an aortic valve disease with two functional cusps. A common pathology in BAV is the malformation of a commissure and, in most cases, the presence of a raphe. The term raphe defines the conjoint or fused area of two underdeveloped cusps representing a malformed commissure between these cusps.

Each aortic valve was analysed and characterised in the acquired subsequent orthogonal views of the aortic valve area (figure 1). Special care was taken to identify the cusps separation and the site of leaflet fusion, because BAV might show a false tricuspid appearance in diastole based on the presence of the raphe. Particular attention was paid to the opening motion of the aortic cusps during systole. The two commissures can be visualised only during systole with a clear separation of the leaflets. Additionally, PVM images in the aortic valve area were analysed to better identify the two commissures by detecting antegrade flow through the aortic valve. The remaining ridge and non-separating border during systole represents the raphe (figure 2). Of note, systolic and diastolic images were critically analysed for the detection of the raphe.

Figure 1

To assess the exact phenotype of the aortic valve (especially to detect a ‘raphe’), the imaging plane was moved up shift wise in 2.5 mm steps for several slices. In this patient, a fusion of the right and left coronary cusps was detected (BAV-RL). A. Oblique coronal view of the aortic valve and the ascending aorta. Slice positions for image planes are indicated by white lines. B. Diastolic orthograde images of the aortic valve. C. Systolic orthograde images of the aortic valve.

Figure 2

Bicuspid aortic valve stenosis in BAV-RL. A. Diastolic steady-state free-precession (SSFP) image demonstrating fusion of left and right coronary cusps (asterisk). B. Velocity-encoded image during systole. The velocity pattern reveals a predominately slotted appearance of the aortic orifice of a bicuspid valve with fusion of the right and left cusps. C. Systolic SSFP image demonstrating fusion of left and right coronary cusps (asterisk). Image shows severely stenotic bicuspid aortic valve (planimetry of 0.9 cm2) with thickened valve leaflets (arrow).

Taken together, BAV phenotypes were defined by the presence and orientation of cusps and raphes as following (figure 3):

  1. Two completely developed cusps and commissures without a raphe. The orientation of the free edge of the cusps defined either anterior-posterior (BAV-AP) or lateral (BAV-LA).

  2. A malformed commissure, more or less obliterated, defining a raphe, extending from the commissure to the free edge of the two underdeveloped conjoint cusps. With respect to the orientation of the raphe, the following phenotypes are defined: BAV-RL (fusion of the right and left-coronary cusps), BAV-RN (fusion of the right and non-coronary cusps) and BAV-LN (fusion of the left and non-coronary cups).

Figure 3

Definition and frequency in the current cohort of bicuspid aortic valve (BAV) phenotypes. The following phenotypes are defined: In cases with raphe, BAV-RL (fusion of the right and left-coronary cusps), BAV-RN (fusion of the right and non-coronary cusps) and BAV-LN (fusion of the left and non-coronary cups). In cases without raphe, the orientation of the free edge of the cusps defined either anterior-posterior (BAV-AP) or lateral (BAV-LA).

Cusp size

The area of each cusp (the conjoined cusps and non fusion cusp) of BAV was measured in the angulated view at end diastole.

Aortic root dimensions

Aortic dimensions were measured at the aortic annulus, the sinus of Valsalva, the sinotubular junction and the proximal ascending aorta at the level of the pulmonary artery from blood-wall boundary to another at end diastole (figure 4).

Figure 4

Cardiovascular magnetic resonance oblique view showing the four aortic root diameter measurements at end diastole: A, annulus of the aortic valve; B, sinus of Valsalva; C, sinotubular junction; D, proximal ascending aorta.

Statistical analysis

Unless noted otherwise, results are shown as mean (±SD) or median (range) and as percentages for categorical variables. Differences in mean values between two groups were compared by Mann–Whitney U test and between all groups by Kruskal–Wallis H test. A χ2 test was performed to compare frequencies between groups. Interobserver agreement for CMR interpretation was determined using Kappa statistics. Intraclass correlation coefficients (ICC) were calculated for the intraindividual and interobserver variation. A level of significance of below 0.05 was defined as statistically significant. SPSS version 15.0 was used for statistical analysis.


Patient characteristics

Patient characteristics are depicted in table 1. Patients were predominantly men. The mean age of 82 male and 23 female patients was 54±16 (range 15–79). Most patients were in NYHA stage II. The body surface area was 1.99±0.21 m2. Sinus rhythm was present in 98 patients with a mean heart rate of 74±15 min−1.

Table 1

Patient characteristics

The underlying aortic valve condition was regurgitation (aortic regurgitation fraction >30%) in 29% (30/105) and stenosis (aortic valve area <1.5 cm2) in 61% (64/105). Thirty-eight patients had aortic stenosis as defined by an aortic valve area <1.5 cm2 and mild (24 patients) or moderate to severe (14 patients) aortic regurgitation.

Patients with moderate or severe aortic valve stenosis were significantly older (58±14 vs 46±16, p<0.001) and those with moderate or severe aortic regurgitation were significantly younger (45±16 vs 57±15, p=0.001) as compared to mild aortic valve stenosis or regurgitation.

BAV phenotypes

The results are summarised in detail in table 2. The pattern of cusp fusion was as follows: 76 patients with BAV-RL, 14 patients with BAV-RN, 11 patients with BAV-LA and four patients with BAV-AP (figure 5). There was no BAV-LN in the study population. Within this group the predominant finding was aortic stenosis.

Table 2

Pattern of bicuspid aortic valve phenotype

Figure 5

Visualisation of the bicuspid aortic valve at end diastole (top), maximum systole opening (middle) and the corresponding velocity encoded image (bottom). (A). BAV-RN, fusion of the right and non-coronary cusps. (B). BAV-RL, fusion of the right and left-coronary cusps. (C). BAV-LA, lateral orientation. (D). BAV-AP, anterior-posterior orientation. Arrow indicates raphe.

Patients with BAV-LA were more often female in comparison to the other four phenotypes. There were no significant differences between BAV types and age and ejection fraction.

Moderate or greater aortic valve stenosis was observed in 64%, 58% and 91% of patients with BAV-RN, BAV-RL and BAV-LA respectively. Aortic regurgitation was observed in 36% and 30% of the patients with BAV-RN and BAV-RL and was more frequent in BAV-AP (50%).

A raphe was identified along the mid-portion of the conjoined cusps in 90 patients (86%). The mean area of the fused cusps was significantly larger than the non-fused cusp in patients with BAV-RN and BAV-RL (8.42±2.42 cm2 vs 4.79±1.53 cm2, p<0.001). In patients without raphe, the two cusps were mostly unequal in size but the mean area of both cusps did not differ significantly (BAV-AP, 7.45±2.97 cm2 vs 5.18±0.78 cm2, p=0.131; BAV-LA, 6.60±1.78 cm2 vs 5.92±1.92 cm2, p=0.312).

Detailed surgical reports regarding the BAV phenotype were available for only 12 patients. In these patients, CMR and surgical inspection yielded identical results (two BAV-RN, nine BAV-RL and one BAV-LA).

The interobserver agreement for the BAV phenotype was 0.93. In one case, a bicuspid valve was classified as BAV-LA on first review and as BAV-RN on second review. In two other cases, the valve was classified as BAV-RL on first review and as tricuspid on the second review. Images were of overall good quality in 88 patients, moderate in 12 patients and poor in five patients.

Comparison between echocardiography and CMR

Results according to BAV phenotype in 62 patients with both image modalities are depicted in table 3. The distribution of calcification was as follows: 0, n=8 (13%); 1, n=13 (21%); 2, n=23 (37%); 3, n=18 (29%). All of the non-identifiable valve phenotypes by echocardiography had calcified aortic valves (six severely calcified and two moderately calcified). The classification of BAV phenotype did not differ between echocardiography and CMR.

Table 3

Bicuspid aortic valve comparison between CMR and TEE

Aortic root dimensions

The aortic root dimensions according to BAV phenotype are depicted in table 4. Overall, the mean aortic diameter at the annulus was 2.88±0.41 cm, the diameter of the sinus of Valsalva was 3.71±0.56 cm, the diameter of the sinotubular junction was 3.46±0.65 cm and the diameter of the proximal ascending aorta was 4.10±0.69 cm. No significant difference was present between the BAV groups.

Table 4

Aortic dimensions

Treatment characteristics of patients with BAV disease

The treatment characteristics from 100 out of 105 (95%) patients are depicted in table 5. The distribution of BAV phenotype was similar in all treatment groups. Three patients died prior to surgery (one sudden cardiac death, one malignancy, one of unknown origin). Seventy-six patients underwent aortic valve surgery and 24 patients were treated medically/conservative. Twenty-nine of the 76 patients (38%) had had biological aortic valve prosthesis, 29 (38%) a Ross procedure and 18 (24%) a mechanical aortic valve replacement. In addition, 13 of the 76 (17%) patients underwent surgery of the ascending aorta (nine prosthesis, four reconstructions). The ascending aorta dimension was significantly dilated in patients with combined aortic valve replacement and ascending aorta surgery.

Table 5

Treatment characteristics of patients with BAV

Interobserver and intraobserver reliability

The intraobserver measurements of aortic annulus (ICC 0.92, range 0.82-0.97), aortic sinus (ICC 0.96, range 0.91–0.99), sinotubular junction (ICC 0.93, range 0.83–0.98) and ascending aorta (ICC 0.98, range 0.95–0.99) showed good reliability. Similarly, a high concordance in interobserver measurements of aortic annulus (ICC 0.89, range 0.72–0.96), aortic sinus (ICC 0.97, range 0.92–0.99), sinotubular junction (ICC 0.95, range 0.87–0.98) and ascending aorta (ICC 0.95, range 0.88–0.98) was present.


This is the first study to utilise CMR to further characterise BAV according to phenotype. BAV phenotypes could be assessed with good image quality in a large cohort of 105 patients. A raphe was present in 90 BAV patients and the frequencies of the BAV-RL phenotype and of the BAV-RN phenotype were 76 (72%) and 14 (13%) respectively. Valvular stenosis was the predominant haemodynamic lesion. There were no significant differences of the aortic dimensions. According to the present results, CMR is a promising imaging tool to assess phenotypes of BAV disease.

Assessment of BAV phenotypes

The recognition of the BAV in patients with aortic valve disease remains an important challenge to the clinician. Echocardiography usually allows an accurate quantification of aortic valve disease. However, sensitivity and specificity to assess valve morphology is limited in transthoracic echocardiography, but are substantially increased by transoesophageal echocardiography.20 21 Nevertheless, transoesophageal echocardiography has an invasive character, whereas CMR is a new non-invasive imaging technique allowing an exact quantification of the severity of valvular stenosis and regurgitation.6 13 19 22 The present findings that CMR usually allows classification of the aortic valve in the majority of patients (97%) are in line with and extend a recently published study by Gleeson et al,23 who demonstrated in 38 patients that CMR of the aortic valve is useful in distinguishing normal and bicuspid aortic valves compared to transthoracic echocardiography. Furthermore, the role of CMR for assessment of BAV morphology was compared to the established standard transoesophageal echocardiography in the present study. According to the present results, the classification of BAV phenotype did not differ between CMR and transoesophageal echocardiography. However, all of the non-identified valve phenotypes by transoesophageal echocardiography had moderate to severe calcified aortic valves, which makes the characterisation of BAV phenotype difficult. Especially in RL-phenotype, the raphe between the right and left cusp is usually posterior to the two commissures in the typical orthogonal imaging plane of 60 degrees. In advanced calcified valves, the imaging quality is impaired by acoustic shadowing from valvular calcification posterior to the two commissures. Therefore, due to the acoustic shadowing, visualisation and characterisation of the raphe in RL-phenotype may be challenging even by transoesophageal echocardiography. Otherwise, CMR had a lower rate of non-diagnostic studies due to the lesser susceptibility of the image signal to calcification artefacts.

Based on the present results, the authors strongly recommend a standardised approach in CMR to visualise and analyse the aortic valve in several subsequent orthograde views and to perform additional through plane PVM in the aortic valve area to better discriminate antegrade transvalvular flow. With this approach, BAV phenotypes could be accurately defined with an adequate image quality even in heavily calcified valves. A raphe was present in 86% of the present cohort and 14% had BAV without any raphe. The frequencies of the BAV-RL phenotype and of the BAV-RN phenotype were 72% and 13% in the present referral population, which is comparable to previous reports.24–26 Valvular stenosis was the predominant haemodynamic lesion. Taken together, the present data suggest that CMR allows an exact and reliable assessment of BAV phenotypes.

Clinical relevance of BAV phenotypes

BAV is the most common congenital cardiac malformation in adults with an incidence of 1–2%.27 Additionally, BAV is even more frequent in patients undergoing aortic valve replacement because valve dysfunction in BAV is characterised by an earlier onset and a higher rate of progression in comparison to patients with tricuspid aortic valve disease.9 26 Frequency of BAV was reported to be 49% in a study of 932 operatively excised stenotic aortic valves.28 However, BAV stands for a common congenital aortic valve malformation with heterogeneous morphologic phenotypes with respect to the number of raphes, spatial position of cusps or raphes and the functional status of the valve.29 Pathologically, a raphe is a ridge with many elastic fibres. There is a broad spectrum of morphological characteristics ranging from a completely missing commissure and raphe to a more or less underdeveloped commissure and the adjunct cusps resulting in a raphe, a common finding in most patients with BAV.3

Interestingly, the morphology of BAV may also be predictive of clinically important end points and prognostic relevance.30 4 24 26 In a longitudinal study by Thanassoulis et al, with a much younger population as compared to the present study cohort, RL fusion was suggested to be a predictor for rapid progression of aortic dilatation.26 However, RL fusion was also not a predictor for aortic dilatation as the present cross-sectional study demonstrates. Hence, the association between BAV phenotype and aortic dilatation is still controversial. Thus, a study by Schaefer et al demonstrated different types of aortic dilatation in different phenotypes of BAV.31 In contrast to the present study population, the study group was about 10 years younger, had less severe aortic valve disease and transthoracic echocardiography was used to determine the aortic dimensions, which may influence the association between phenotype and aortic dilatation. Otherwise, the present results are in concordance with a recent study by Fazel et al, who found that cusp fusion pattern was not associated with any particular aortic morphology performed by CT.32 Additionally, Novaro et al reported that fusion of the right and non-coronary cusps compared with left/right cusps was not associated with greater significantly mid-ascending aortic dimensions.33 Therefore, further studies are warranted to address this discrepancy in more detail. Furthermore, the RN fusion is associated with a more rapid progression of aortic stenosis and regurgitation and a shorter time to valve intervention as compared to RL fusion in childhood and young adults.4 30 In contrast, Sabet et al described more frequent stenosis in adult patients with RL fusion.24

With respect to the surgical point of view, recent developments for valve-preserving surgery were extended to patients with BAV and reconstructive techniques are increasingly performed to restore normal aortic valve function.7 8 34 35 To apply these operations, understanding of normal and pathological valve anatomy and physiology is of utmost importance.36–38 The American College of Cardiology/American Heart Association recommendations highlight the role of CMR as a complementary tool to echocardiography for the diagnosis and surveillance of the morphology of the aortic valve and ascending aorta in subjects with BAV.39 Therefore, an accurate preoperative assessment of morphology in bicuspid aortic valve disease is essential. Although echocardiography allows accurate measurement of the size of the proximal aorta, it has limitations in the assessment of the dimensions of the ascending aorta, usually the site of maximum dilatation. Therefore, the authors strongly recommend targeted imaging of the ascending aorta by CMR for exact evaluation of aortic dilatation in BAV. In addition, CMR allows assessment in one examination of the myocardial morphology and function, morphology of the aortic valve and aortic valve disease severity, and an exact measurement and characterisation of the aortic root and ascending aorta to identify the BAV phenotype.

Study limitations

Some limitations of the present study should be acknowledged. Patterns of referral for further evaluation of aortic valve disease may affect the overall demographics of the study population. A large proportion of patients in the present study with aortic valve disease were admitted with onset of symptoms. The study is, therefore, subject to referral bias and patients with only mild and moderate valve dysfunction may be underrepresented. Furthermore, the entity of unicuspid aortic valve (also called type II BAV with two raphes by Sievers classification) was excluded in the present study of BAV phenotypes. Despite the increasing utilisation of CMR, it seems unlikely CMR will replace echocardiography as the imaging modality of first choice in evaluating heart valve diseases. The relatively low costs, widespread availability and fast results favour echocardiography as the primary imaging modality for the initial evaluation of patients with aortic valve disease.


In conclusion, CMR is a promising imaging tool to assess valve morphology in congenital BAV with respect to the presence and orientation of raphes and cusps, as well as the aortic dimensions. In the vast majority of cases a raphe was present and the more frequent RL phenotype is present in approximately two-thirds of patients. No significant differences between BAV phenotypes regarding severity of valve disease and aortic size were found.


The authors thank Marion Merdian, Kerstin Kubernus, Daniela Spanja, Katja Ziczinski and Heike Koitsch for excellent technical assistance.



  • Competing interests None.

  • Patient consent Obtained.

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

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.