Article Text

Download PDFPDF

Original article
Device-dependent association between paravalvar aortic regurgitation and outcome after TAVI
  1. Rafal Dworakowski1,
  2. Olaf Wendler1,
  3. Brian Halliday1,
  4. Peter Ludman2,
  5. Mark DeBelder3,
  6. Simon Ray4,
  7. Neil Moat5,
  8. Jan Kovac6,
  9. Tomasz Spyt6,
  10. Uday Trivedi7,
  11. David Hildick-Smith7,
  12. Dan Blackman8,
  13. Damian Marlee9,
  14. David Cunningham9,
  15. Philip A MacCarthy1
  1. 1Kings College Hospital, London, UK
  2. 2University Hospitals Birmingham NHS Foundation Trust, Birmingham, UK
  3. 3James Cook University Hospital, Middlesbrough, UK
  4. 4University Hospital of South Manchester, UK
  5. 5Royal Brompton and Harefield Hospital, London, UK
  6. 6University Hospitals Leicester NHS Trust, Leicester, UK
  7. 7Sussex Cardiac Centre, Brighton and Sussex University Hospitals, Brighton, UK
  8. 8Leeds General Infirmary, Leeds, UK
  9. 9National Institute for Cardiovascular Outcome Research, UCL, London, UK
  1. Correspondence to Dr Philip MacCarthy, King's College Hospital, Denmark Hill, London SE5 9RS, UK; philip.maccarthy{at}


Objective The aim of the current study was to identify predictors of paraprosthetic aortic regurgitation (AR) after transcatheter aortic valve implantation (TAVI) and examine its influence on short/medium-term mortality using the UK TAVI Registry. TAVI is an effective treatment for high-risk patients with severe symptomatic aortic stenosis (AS), but paraprosthetic AR has been associated with increased in-hospital and mid-term mortality.

Methods Between January 2007 and December 2011, 2584 TAVI procedures were performed in the UK. Patients undergoing ‘valve-in-valve’ procedures, patients with aortic regurgitation as the primary pathology and with no recorded severity of AR were excluded from this analysis (n=144). In total, therefore, 2440 patients were included. Balloon-expandable and self-expanding devices were implanted in 52.7 and 47.2%, respectively, using either transfemoral (67.7%) or non-transfemoral, surgical access (32.3%).

Results Postprocedural AR was observed in 68%, mild AR in 57% and moderate-severe in 10%. A large aortic annulus, high preprocedural transaortic gradient, and use of self-expanding valve were independent predictors of moderate-severe AR. Moderate-severe (but not mild) AR was associated with increased mortality, and this relationship appeared significant for the balloon-expandable but not the self-expanding device.

Conclusions Our data suggest that a large aortic annulus, high preprocedural transaortic gradient, and use of the self-expanding valve predict moderate-severe AR after TAVI. Such a degree of AR is associated with a significantly worse outcome with the balloon-expandable, but not with the self-expanding valve. Further studies are needed to verify this and explore potential mechanisms.

View Full Text

Statistics from


Transcatheter aortic valve implantation (TAVI) has been shown to be an effective alternative treatment for patients with severe symptomatic aortic stenosis (AS) who are considered high risk for surgical aortic valve replacement (sAVR).1–3 Multiple national registries have demonstrated the successful performance of TAVI procedures in nearly 100 000 patients worldwide.

Paraprosthetic aortic regurgitation (AR) is significantly more common following TAVI than with sAVR, probably because of the complex anatomy of the calcified native valve and the relationship of the calcified/fibrosed leaflets with the implanted prosthesis.3–8 AR of any severity occurs in 60–70% of patients undergoing TAVI, being categorised as ‘mild’ in 45–60% of cases, and ‘moderate’ or ‘severe’ in 10–17%. However, it is now widely accepted that paraprosthetic AR is associated with increased in-hospital and mid-term mortality following TAVI.2 ,3 ,5–7 ,9–11 European registries have demonstrated this relationship, while the 2-year follow-up of the PARTNER trial, using the Edwards Sapien prosthesis, also showed increased mortality with all degrees of regurgitation (including mild).3 ,5–7 ,12 ,13 The precise nature of this association and whether AR is predominantly a confounding marker of other features that lead to poor outcome remains unknown.

Most observers accept that paraprosthetic AR is more common with the CoreValve device, yet despite this observation and the proven relationship between AR and outcome, mortality is no greater with the CoreValve than the Sapien in most published series. This paradox has never been explained.

In an attempt to shed light on this phenomenon, the aim of the current study was to identify predictors of AR after TAVI, and examine the relationship between paraprosthetic AR and short/medium-term mortality using data from the UK TAVI Registry, with particular attention to device type. This dataset comprised 2440 patients receiving balloon-expandable and self-expanding devices.


A total of 25 centres in England and Wales developed active TAVI programs between January 2007 and December 2011, with 2584 TAVI procedures being performed during this period. The development of UK TAVI activity has been described elsewhere.14 ,15 Patient eligibility for TAVI was decided in each centre by a multidisciplinary ‘heart team’ comprised of interventional cardiologists, imaging cardiologists, cardiothoracic surgeons and anaesthetists. The self-expanding valve (CoreValve System, Medtronic) and the balloon-expandable valve (Sapien THV, Edwards) were used at the operators’ discretion. Most centres used devices from only one manufacturer, and nearly all cases were performed in a centre where another device was not used. The decision on device type was therefore not influenced by individual patient characteristics, and device selection bias is unlikely. All centres adopted a ‘transfemoral default’ selection policy with criteria for a non-transfemoral approach based upon the heart team's consideration of the quality of the ileofemoral arteries. The balloon-expandable implants were either by the transfemoral, transapical and transaortic routes, and the self-expanding implants were by transfemoral, subclavian, or occasionally direct aortic access. Annulus sizing is an aspect of the TAVI procedure which has evolved significantly over recent years. In this historical cohort, the majority of cases were assessed by echocardiography rather than CT (70% transoesophageal echo).

The Central Cardiac Audit Database (CCAD) established a web-based system for data entry, encryption, and transfer. All 25 units submitted their data on all patients treated from 2007, retrospectively, for patients treated before establishing the database, and prospectively thereafter. In 2011, CCAD became part of the newly established National Institute for Cardiovascular Outcomes Research (NICOR) (, which, in addition to the TAVI registry, also hosts a number of other national cardiovascular registries (University College, London). Patients undergoing ‘valve-in-valve’ procedures, and patients with aortic regurgitation as the primary valve pathology were excluded from this analysis (n=129). A further 15 cases were excluded because the severity of postprocedural aortic regurgitation was not recorded.

Mortality tracking was undertaken by the National Health Service Central Register using unique patient identifiers.

Postimplantation AR was assessed using standardised echocardiographic parameters by individual centres prior to patient discharge, as per standard clinical practice. AR was graded as none, mild, moderate, or severe, and these gradings were entered in the database. The assessment of AR was done with transthoracic echo in all cases (after the TAVI procedure before discharge), and this is therefore the method of assessment we have used in this analysis. Some centres, additionally, used periprocedural transoesophageal echo and aortography, but this data is variable in our dataset, and was therefore not included.

Outcomes were reported according to VARC-2 definitions: device success (absence of procedural mortality, and correct positioning of a single prosthetic heart valve into the correct anatomical location, and intended performance of the prosthetic heart valve without moderate or severe prosthetic valve regurgitation), and early safety (all-cause mortality, stroke, life-threatening bleeding, acute kidney injury requiring renal replacement therapy, coronary artery obstruction requiring intervention, major vascular complication, valve-related dysfunction requiring repeat procedure).1 ,7


Categorical data are presented as percentages, and numerical data are presented as the mean with 95% CIs for completed datasets. Intergroup comparisons were made using Student t test, ANOVA test followed by Tukey posthoc test, or Pearson's χ2 test as appropriate. Relations between variables are assessed by logistic regression.

Cox proportional hazard analysis was used to assess the association between variables and mortality. χ2 Value, HR, 95% CI and significance level on the likelihood ratio test for risk factors are given. All significant univariate prognosticators were subsequently entered into multivariate models in a stepwise forward fashion. Kaplan–Meier cumulative survival plots were tested using log rank test.

Statistical significance was defined as p<0.05. The data were analysed with SPSS Statistics 20.0 (SPSS, Chicago, Illinois, USA).


A total of 2584 patients were enrolled between January 2007 and December 2011 at 25 centres. Patients undergoing planned ‘valve-in-valve’ procedures and patients with aortic regurgitation as the primary valve pathology were excluded from this analysis (n=129). A further 15 cases were excluded because the severity of postprocedural aortic regurgitation was not recorded, and 2440 patients were, therefore, included in the final analysis. An additional eight patients were lost to follow-up at 12 months, and these cases were excluded from survival analysis.

The mean (±SD) age was 81.6±7.2 years; 53% of the patients were male. All patients were highly symptomatic and were at high surgical risk for aortic valve replacement with logistic EuroSCORE 21.6±13.6%. Balloon-expandable and self-expanding devices were implanted in 52.7 and 47.2% of patients, respectively. Approaches were either transfemoral (67.7%) or surgical (transaortic, transapical, subclavian) (32.3%). The summary of baseline clinical and procedural characteristics is shown in tables 1 and 2. The device success rate as defined by VARC-2 was 86.9%. Rates of death at 30 days and 1 year were 6.3% and 19.3%, respectively.

Table 1

Clinical characteristics in the groups by severity of aortic regurgitation (percentages represent a ratio for the completed dataset)

Table 2

Procedural characteristics based on severity of aortic regurgitation

Postprocedural aortic regurgitation was observed in 68% of patients, with mild AR accounting for most of the cases (57%). Moderate-severe AR was observed after TAVI in 10% of all cases.

There were 51 patients requiring emergency valve-in-valve implantation during an index TAVI procedure (13 patients in the balloon-expandable valve group, and 38 patients in the self-expanding valve group). For our analysis, we used the degree of AR after the second valve deployment. The final breakdown for severity of AR after deployment of the second valve was 5, 28 and 18 for no AR, mild, and moderate-severe AR, respectively.

Predictors of aortic regurgitation

Predictors of severe AR using univariate analysis were: male gender, height, large aortic annulus, aortic valve gradient, absence of balloon aortic valvuloplasty, valve size, transfemoral delivery approach and using the self-expanding valve (tables 1, 2). Using multivariate modelling, only large aortic annulus, high transaortic gradient and use of the self expanding valve were independent predictors of moderate-severe AR (table 3).

Table 3

Predictors of moderate/severe aortic regurgitation versus no/mild AR (nominal logistic regression, effect likelihoods ratio test for multivariate analysis, χ2=43.3, p<0.0001)

Relation between postprocedural aortic regurgitation and mortality

Postprocedural AR was associated with complications, such as death and cardiogenic shock in the cath lab, blood transfusion postprocedure, and acute kidney injury (table 4). It was also associated with 30-day and longer-term mortality.

Table 4

Procedural outcome based on severity of aortic regurgitation

There was no significant difference in long-term mortality between patients with no and mild AR, but there was significantly higher mortality in the group with moderate-severe AR (figure 1).

Figure 1

Kaplan–Meier survival curves after TAVI according to severity of postprocedural AR. AR, aortic regurgitation.

The presence of postprocedural moderate-severe AR was significantly higher in patients receiving the self-expanding device compared to those receiving the balloon-expandable valve (13.6% vs 7.6%, p<0.0001), yet there was no difference in overall mortality between patient groups divided by device type (figure 2). Our data seemed to suggest that moderate-severe AR was associated with a significantly worse outcome with the balloon-expandable, but not with the self-expanding device (figures 3 and 4). This observation was confirmed in a Cox regression model, as moderate-severe AR was a strong predictor of mortality in the balloon-expandable valve group (HR 1.97, CI 1.47 to 2.61), but not in the self-expanding valve group (HR 1.13, CI 0.83 to 1.51).

Figure 2

Kaplan–Meier survival after transfemoral TAVI in groups according to device (balloon-expandable valve vs self-expanding valve). TF TAVI, transfemoral transcatheter aortic valve implantation.

Figure 3

Kaplan–Meier survival by grade of AR in TF self expanding valve cases. AR, aortic regurgitation.

Figure 4

Kaplan–Meier survival by grade of AR in TF balloon expandable valve cases. AR, Aortic regurgitation.

To further understand this difference in the mortality distribution between the two devices, we analysed the patients with moderate-severe postprocedural AR (performed transfemorally) (n=190) (table 5 online supplementary materials). There were no major differences in clinical characteristics between these two groups to account for the mortality difference.


This large study, based on over 2400 patients who underwent TAVI with either the balloon-expandable valve or the self-expanding valve, demonstrates that moderate and severe degrees of AR are associated with increased mortality rates at 1 year, in line with previous studies.4–8 ,16 Our data also suggest this relationship may vary according to the device used.

The sudden change in haemodynamics associated with AR, and the maladapted ventricles seen in patients with longstanding severe AS could explain the association between AR and mortality.9 Severe AS subjects the LV to a period of prolonged pressure-overload leading to left ventricular hypertrophy, increased myocardial oxygen demand, increased myocardial stiffness, impaired relaxation and increased left ventricular end-diastolic pressure.12 In the setting of AR and volume overload, the hypertrophied, non-compliant ventricle is unable to increase the end-diastolic volume, and is therefore unable to cope with the sudden increase in volume.14 As a consequence, the stroke volume decreases and the left ventricular end-diastolic pressure increases. This results in reduced coronary perfusion and myocardial ischaemia with consequent pulmonary oedema. Therefore, the effect of longstanding aortic stenosis on the LV leaves it maladapted to cope with the sudden change from pressure-overload to volume-overload, and may explain the association between even moderate AR and increased mortality. Whether patients with significant pre-TAVI AR together with AS cope with post-TAVI AR remains to be studied.

Our data also indicates that a large aortic annulus dimension, high transaortic pressure gradient, and the use of the self-expanding valve are independent predictors of moderate or severe AR in a multivariate model. The significantly higher rate of moderate/severe AR seen with the self-expanding valve (compared with the balloon-expandable valve) has been reported previously, and is explained by several factors including the different composition and structure of the valve prosthesis.7 However, the literature is not clear on this point, and some have reported recently (in a multicentre European study) there is no significant difference in the rate of mild, moderate or severe AR between the two prostheses.16 This latter report was a significantly smaller (793 patients) study than ours, and there was a similar trend of a higher frequency of moderate/severe AR with self-expanding (5.2%) when compared to the balloon-expandable valve (2.8%). Most experts in the field accept there is a difference between the two devices.

Our registry is based on retrospective data acquired at a phase of TAVI technological evolution when there was no large valve size. We could therefore speculate that some implanted valves were undersized, resulting in moderate-severe aortic regurgitation, explaining why large annulus size predicted AR. In keeping with previous studies demonstrating more accurate sizing of the aortic annulus with CT as compared with TOE, there were fewer cases of moderate and severe AR when CT was used to measure the annulus preprocedure.8 ,17 ,18 This is possibly because 2D echo techniques were used at this time, and these are now known to be substandard in accurately sizing the aortic annulus, which is often a non-circular structure. Methods of annulus sizing have significantly evolved throughout the course of this patient cohort and since. We have also learnt that the assessment of valve morphology (eg, eccentric calcification) is vital to predict the severity and location of periprosthetic regurgitation or the risk of coronary artery obstruction. The decision about valve size is therefore more complex than TAVI operators initially appreciated, being multifactoral and depending on annulus size and also severity of valve calcification, size of the LV outflow tract and the aortic root/sinuses. We now know that the use of 3D CT or 3D TOE annulus measurement (many operators use both) is essential in selecting the correct device size in the preprocedure work-up process.8 ,19 Indeed, there is a recent movement towards using annulus perimeter or area-derived dimensions to further refine this choice.

Some of the predictors of AR we have identified can be explained intuitively. A high transvalvular gradient is directly linked with the severity of aortic stenosis, which is itself associated with a heavy degree of valve calcification and immobilisation of valve cusps. It is known that AR is more likely with heavily calcified valves and, in particular, extensive asymmetric calcification of the aortic valve has been linked with AR following TAVI.20 It is possible, therefore, that a high pre-TAVI gradient is a surrogate for advanced, calcified valve disease.

Despite the increased incidence of significant AR seen with the self-expanding valve, there was no demonstrable difference in mortality between the patients receiving the self-expanding valve and those receiving the balloon-expandable valve (figure 2). At first glance, this seems somewhat paradoxical. Interestingly, in our study, the increased rate of moderate and severe AR in the self-expanding valve group did not lead to a statistically significant increase in mortality rate. This contrasted with the balloon-expandable valve group, where there was a definite mortality effect in the group with moderate/severe AR (figure 4). This difference does not appear to be due to the device choice being a marker/confounder for other prognostically important clinical characteristics, as there are no such clinical differences between the groups with this degree of AR (table 5).

It is difficult to explain this phenomenon from our data alone. One hypothesis could be that the self-expanding prosthesis continues to expand over the months following the procedure, resulting in a reduction in the degree of AR in the medium term (ie, after hospital discharge) without long-term haemodynamic implications. Our measures of AR were all early postprocedure, but were taken at the same stage, post-TAVI, irrespective of device type. Other explanations of this differential effect of AR on mortality include artefact in the echo measurement of AR with the different devices, which may cause different levels of acoustic shadowing due to their different structure, or different effects of AR on coronary flow between the two prostheses. It may also be a statistical error due to relatively low numbers of patients with moderate-severe AR, particularly in the balloon-expandable valve group. Clearly, further studies are needed to clarify these issues.

The increased mortality rate associated with moderate and severe AR highlights the importance of developing devices and techniques that minimise this complication in the future. There has also been a recent focus on reinterventions that reduce the degree of AR. An Italian study demonstrated a reduction in AR severity with postdilatation of the prosthesis in 81% of cases.21 There have been reports of successful valve-in-valve reinterventions, and repositioning of the CoreValve by snare catheter in cases of distal malposition of the prosthesis.7 ,21 It is important to further explore these techniques and determine the optimal timing and precise indications for reintervention to optimise clinical practice.


Our data are retrospective and observational, and are therefore subject to all the limitations of this kind of study. The assessment of AR with transthoracic echo is clearly not ideal, but it is what happens in practice in most centres in the world. More accurate, standardised, and thorough techniques of assessing AR would give more insight but would be difficult to do on this scale. Additionally, we do not have detailed information about valve morphology, in particular, the distribution of calcification, which is relevant to paravalvular AR.

This study is therefore a description of associations and does not offer any mechanistic explanations. However, this is a large patient cohort, and we describe what happens in real-world practice. Such observations should be seen as hypothesis-generating, and clearly, extensive further analysis is needed to explain the phenomena we describe.


This study demonstrates that moderate/severe AR predicts increased mortality following TAVI. Multivariate predictors of AR post-TAVI include a large annulus size, a high transvalvular gradient pre-TAVI, and use of the self-expanding valve. Although self-expanding prostheses are associated with increased rates of significant AR, this does not translate into increased patient mortality rates with this device. The relationship between post-TAVI AR and mortality appears to differ according to device used, with a more marked relationship in patients receiving the Sapien device, and the reasons for this are not clear from the current data. These data emphasise the clinical need for on-going endeavour to reduce post-TAVI AR, a phenomenon which will be improved by more accurate 3D annulus sizing, novel device design and clear indications for reintervention/postdilatation. Further studies are needed to identify why post-TAVI AR predicts a poor outcome, and why this might be device-specific.

Key messages

What is already known about this subject?

  • Paraprosthetic AR is associated with increased in-hospital and mid-term mortality following TAVI. Most observers accept that paraprosthetic AR is more common with the self-expanding device, yet despite this observation and the proven relationship between AR and outcome, mortality is no greater with the self-expanding than the balloon-expandable valve in most published series. This paradox has never been explained.

What might this study add?

  • Our data indicate a significantly higher rate of moderate/severe AR in the self-expanding, compared to the balloon-expandable valve. Our study confirms the findings of other smaller studies/registries that moderate-severe AR, but not mild AR, correlates with clinical outcome. We also show that moderate-severe AR is associated with increased mortality with the balloon-expandable but not the self-expanding valve.

How might this impact on clinical practice?

  • Our data emphasise that every effort should be made to avoid postprocedural AR, particularly AR, which is moderate or severe. This is especially important for balloon-expandable valves, while self-expanding devices may be more forgiving.


View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors RD, OW, BH, PL, MDB, SR, NM, JK, TS, UT, DHS, DB, DM, DC, PAM took part in designing the study and interpretation of all data. RD, BH and PAM prepared the first draft of the manuscript. RD analysed the data and performed statistical analysis. PAM is responsible for overall content as guarantor.

  • Competing interests None.

  • 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.

Linked Articles