Author + information
- Received November 27, 2013
- Revision received February 9, 2014
- Accepted February 11, 2014
- Published online May 27, 2014.
- Ganesh Athappan, MD∗,
- R. Dilip Gajulapalli, MD†,
- Prasanna Sengodan, MD∗,
- Anju Bhardwaj, MD∗,
- Stephen G. Ellis, MD†,
- Lars Svensson, MD, PhD†,
- Emin Murat Tuzcu, MD† and
- Samir R. Kapadia, MD†∗ ()
- ∗Department of Cardiovascular Medicine, Heart & Vascular Institute, Case Western Reserve University, Cleveland, Ohio
- †Department of Cardiovascular Medicine, Heart & Vascular Institute, Cleveland Clinic, Cleveland, Ohio
- ↵∗Reprint requests and correspondence:
Dr. Samir R. Kapadia, Cardiac Catheterization Laboratory, Department of Cardiovascular Medicine, J2-3, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, Ohio 44195.
Objectives The study undertook a systematic review to establish and compare the risk of stroke between the 2 widely used approaches (transfemoral [TF] vs. transapical [TA]) and valve designs (CoreValve, Medtronic, Minneapolis, Minnesota vs. Edwards Valve, Edwards Lifesciences, Irvine, California) for transcatheter aortic valve replacement (TAVR).
Background There has been a rapid adoption and expansion in the use of TAVR. The technique is however far from perfect and requires further refinement to alleviate safety concerns that include stroke.
Methods All studies reporting on the risk of stroke after TAVR were identified using an electronic search and pooled using established meta-analytical guidelines.
Results 25 multicenter registries and 33 single-center studies were included in the analysis. There was no difference in pooled 30-day stroke post-TAVR between the TF and TA approach in multicenter (2.8% [95% confidence interval (CI): 2.4 to 3.4] vs. 2.8% [95% CI: 2.0 to 3.9]) and single-center studies (3.8% [95% CI: 3.1 to 4.6] vs. 3.4% [95% CI: 2.5 to 4.5]). Similarly, there was no difference in pooled 30-day stroke post TAVR between the CoreValve and Edwards Valve in multicenter (2.4% [95% CI: 1.9 to 3.2] vs. 3.0% [95% CI: 2.4 to 3.7]) and single-center studies (3.8% [95% CI: 2.8 to 4.9] vs. 3.2% [95% CI: 2.4 to 4.3]). There was a decline in stroke risk with experience and technological advancement. There was no difference in the outcome of 30-day stroke between TAVR and surgical aortic valve replacement.
Conclusions Our findings suggest that the risk of 30-day stroke after TAVR is similar between the approaches and valve types. There has been a decline in stroke risk after TAVR with improvements in valve technology, patient selection, and operator experience.
Transcatheter aortic valve replacement (TAVR) has seen an exponential utilization in high surgical risk patients and an expansion to the intermediate risk population (1,2) due to impressive results in randomized PARTNER (Placement of AoRTic TraNscathetER Valve Trial) (Online Refs. e18,e19). Despite the growth, TAVR as a procedure is still evolving and requires further refinement to reduce complications. Stroke remains a major concern after TAVR and an important cause of increased morbidity and mortality. In the PARTNER trial cohort A and cohort B (Online Refs. e18,e19), the occurrence of stroke was doubled in the TAVR arm when compared to surgery (4.6% vs. 2.4%) and medical therapy (6.7% vs. 1.7%), respectively. Similarly, the risk of stroke in the TAVR arm of the PARTNER trials was also higher than that reported in the surgical literature for isolated aortic valve replacement (1.5% to 4%), raising safety concerns (3,4).
An understanding of the mechanisms underlying stroke after TAVR is therefore essential for the implementation of appropriate preventive measures prior to further expansion in its utilization. The manipulation of bulky endovascular devices along the aortic arch and aortic root during the transfemoral (TF) approach and manipulation of the apex during the transapical (TA) approach have been implicated for embolic strokes (5,6). Similarly, difference in valve design and deliverability between the self-expanding CoreValve, Medtronic, Minneapolis, Minnesota, and balloon expanding Edwards Valve, Edwards Lifesciences, Irvine, California have also been speculated to alter the stroke risk after TAVR (5) (Online Ref. e100). However, neither of these theories regarding valve delivery or valve type has been conclusively shown to alter risk of stroke after TAVR. Therefore, we undertook a comprehensive meta-analysis firstly, to establish and compare the risk of stroke between the 2 widely used valves (CoreValve vs. Edwards Valve) and approaches (TF vs. TA) for TAVR. Second, we looked at the temporal trend in stroke risk with experience and advancement in valve technology. Third, we compared the risk of stroke between TAVR and surgical aortic valve replacement (SAVR) in matched patient cohorts.
We conducted this systematic review on published literature of stroke following TAVR using the QUOROM (Quality of Reporting of Meta Analysis) (7) and MOOSE (Meta Analysis of Observational Studies in Epidemiology) guidelines (8). A computerized search was performed to identify all relevant studies published until July 2013 in the PubMed database by 2 reviewers (G.A. and D.G.). The following search terms were used: TAVI, Percutaneous Valves, Transcutaneous Aortic Valve, and Transcatheter Aortic Valve. Citations were screened at the title and abstract level and retrieved as a full report if they reported on outcome of stroke after TAVR. Limiting the search parameters to the English language was applied subsequently. The full texts and bibliography of all potential articles were further reviewed in detail (G.A.) to seek additional relevant studies. Major conference proceedings were also searched to retrieve unpublished studies until November 2013.
Full text and references of all identified potential publications and conference proceedings were searched to select the reports for inclusion in the secondary analysis.
Studies were included if the following criteria applied: 1) enrollment for TAVR was based on existing and accepted guidelines; 2) enrolled consecutive patients; 3) reported data on stroke following TAVR using a particular approach or valve design; and 4) performed a minimum of 75 successful TAVR procedures and at least 50 by a particular approach or valve type when from a single center. When 2 similar studies were reported from the same institution or author, the most recent publication or the publication with most information on stroke post TAVR was included in the analysis.
Studies were excluded if any of the following criteria applied: 1) duplicate publication, overlap of patients, subgroup studies (nonconsecutive) of a main study; 2) lack of data on stroke by a particular approach or valve design; 3) if a valve other than the CoreValve or Edwards Valve was used; 4) if they were studies on valve in valve procedure; and 5) non-English reports.
Relevant information was collected by G.A./D.G. and included, but was not limited to, first author, year and journal of publication, study design, inclusion exclusion criteria, definition of stroke/transient ischemic attack, number of subjects included, subjects undergoing successful TAVR, type of device and approach used, study population demographics, follow up time period and primary and secondary outcomes.
The primary end points evaluated were: 1) 30-day risk of stroke after TF and TA approaches; and 2) 30-day risk of stroke after CoreValve and Edwards Valve implantation. The results were stratified into single-center or multicenter experience. Secondary end points of interest were: 30-day risk of stroke after: 1) TAVR feasibility studies; 2) early TAVR experience vs. overall experience of large volume centers; and 3) TAVR vs. SAVR.
For the purpose of the current analysis, we used the following: 1) study-reported 30-day stroke when available; 2) in-hospital/procedural stroke when 30-day stroke was not available; and 3) combined major and minor stroke if reported separately.
The definition of stroke was as reported by the primary study.
Represent the initial studies prior to European Conformite Europeenne approval for the particular valve design.
Centers that had performed 100 or more TAVR procedure for a particular valve/approach.
Early Experience of High-Volume Centers
The first 30% to 50% (or closest available) of patients enrolled by a center for a particular approach or valve design.
DerSimonian and Laird's (9) random effects model was utilized to pool the estimates of 30-day stroke from individual studies and subgroups. A random-effects model was also used to obtain a single pooled estimate of the odds ratios. The effect across subgroups was compared using a Q test based on analysis of variance. Statistical significance was set at a p value <0.05 (2-tailed). Heterogeneity, which was anticipated to be significant, was assessed by a Q statistic and I2 test. Significant heterogeneity was considered present for p values <0.10 and/or an I2 = >50%. Sensitivity analysis was performed by excluding reports that did not use the Valve Academic Research Consortium proposed endpoints/definitions. Sensitivity analysis was also performed by comparing 30-day risk of stroke by comparing the valve designs for the TF approach and comparing the type of approach for the Edwards Valve. Data analysis was performed using Comprehensive Meta-Analysis Software Version 2 (10).
By using the search keywords 6,922 reports were identified and reviewed at title and abstract level. Initial evaluation identified 1050 publications that were further evaluated using the search term stroke. This narrowed the selection to 346 potential publications. Search of conference proceedings further identified 15 relevant publications. When the inclusion and exclusion criteria were applied 25 multicenter studies (Online Table 1) and 33 single-center publications (Online Table 2) remained for assessment of the primary outcomes (Fig. 1). This included 3 randomized comparisons (Online Refs. e18–e20). All the studies included in the analysis were published between 2006 and 2013. Analysis was performed on 29,065 patients from multicenter registries and on 7,149 patients from single-center studies. Selected Baseline characteristics of the included patients are summarized in Online Tables 3 and 4. The VARC criteria were used to report endpoints by 13 and 23 multicenter and single-center studies, respectively (Online Tables 5 and 6).
Searching the full text and references of the previous publications and conference proceedings identified 12 feasibility studies on 992 patients (Online Table 7), 14 high-volume centers reporting on early and late experience (Online Table 8) and 11 reports comparing TAVR and SAVR in a randomized/matched patient population (Online Table 9).
The TF approach was used in 18,712 patients and the TA approach in 5,650 patients. The pooled estimate for overall incidence of 30-day stroke (Fig. 2A) in multicenter registries using the TF approach was 2.8% (95% CI: 2.4 to 3.4, I2 = 70.14) and that using the TA approach was 2.8% (95% CI: 2.0 to 3.9, I2 = 68.35). The VARC proposed definitions/endpoints were used in 12 registries reporting on the TF approach and by 4 registries reporting on the TA approach. There was no difference in stroke in comparing the 2 approaches in sensitivity analysis (Table 1). Nine multicenter registries involving 14,296 patients compared both approaches (66.9% TF and 33.1% TA). The pooled odds ratio (OR) for in-hospital/30-day stroke did not reach statistical significance (OR: 1.04, 95% CI: 0.83 to 1.31) (Fig. 2B). Similar results were obtained on sensitivity analysis (Table 1).
The self-expanding MC was implanted in 8,684 patients and the balloon expandable ES in 16,082 patients. The pooled estimate for overall incidence of 30-day stroke (Fig. 3A) in multicenter registries using the MC valve was 2.4% (95% CI: 1.9 to 3.2, I2 = 64.28) and that using the Edwards Valve was 3.0% (95% CI: 2.4 to 3.7, I2 = 73.89). The VARC proposed definitions/endpoints were used in 8 registries reporting on the MC valve and by 8 registries reporting on the ES. There was no difference in stroke on comparing the 2 valves (Table 1). Seven multicenter registries involving 8,758 patients compared both valves (41% MC and 59% ES). The pooled OR for in-hospital/30-day stroke (MC vs. ES) did not reach statistical significance (OR: 1.03, 95% CI: 0.78 to 1.35) (Fig. 3B). Similar results were obtained on sensitivity analysis (Table 1).
The TF approach was used in 4,556 patients and the TA approach in 2,588 patients. The pooled estimate for overall incidence of 30-day stroke (Online Fig. 1A) in single-center studies using the TF approach was 3.8% (95% CI: 3.1 to 4.6, I2 = 34.24) and that using the TA approach was 3.4% (95% CI: 2.5 to 4.5, I2 = 35.36). The VARC proposed definitions/endpoints were used by 22 centers for reporting outcomes on the TF approach and by 9 centers for reporting outcomes on the TA approach. There was no difference in stroke on comparing the 2 approaches (Table 1). Nine reports involving 2,448 patients compared both approaches (61.3% TF and 38.7% TA). The pooled OR for in-hospital/30-day stroke did not reach statistical significance (OR: 1.16, 95% CI: 0.72 to 1.85) (Online Fig. 1B). Similar results were obtained on sensitivity analysis (Table 1).
The self-expanding MC was implanted in 2,617 patients and the balloon expandable ES in 3,477 patients. The pooled estimate for overall incidence of 30-day stroke (Online Fig. 2A) in single-center studies using the MC valve was 3.8% (95% CI: 2.8 to 4.9, I2 = 35.71) and that using the ES was 3.2% (95% CI: 2.4 to 4.3, I2 = 49.27). The VARC proposed definitions/endpoints were used by 12 single-centers reporting on the MC valve and by 12 single centers reporting on the ES. There was no difference in stroke on comparing the 2 valves (Table 1). Five reports involving 1,429 patients compared both valves (59.34% MC and 40.6% ES). The pooled OR for in-hospital/30-day stroke (Medtronic CoreValve vs. Edwards Valve) did not reach statistical significance (OR: 1.81, 95% CI: 0.69 to 4.74) (Online Fig. 2B). Similar results were obtained on sensitivity analysis (Table 1).
Feasibility studies and early experience of high-volume centers
There were 14 feasibility studies that enrolled a total of 992 patients, 7 using the TF approach (521 patients) and 7 using the TA approach (471 patients). The pooled estimate for overall incidence of 30-day stroke in feasibility studies using the TF approach was 7.1% (95% CI: 5.0 to 10, I2 = 10.75) and that using the TA approach was 2.6% (95% CI: 1.4 to 4.5, I2 = 0.00). There was a significant increase in the risk of stroke (Fig. 4A) with use of the TF approach (p = 0.003). The risk of stroke using the TF approach was higher than that seen in contemporary studies (7.1% vs. 2.9%). There was a progressive decline with gaining experience in the incidence of stroke after TAVR (Fig. 4B).
We identified 14 high-volume TAVR centers reporting on 3,813 TAVR procedures (9 centers for the TF approach and 5 centers for the TA approach). In high-volume centers the overall odds of 30-day stroke using the TF approach (Fig. 5A) in the early experience was 4.9% (95% CI: 3.6 to 6.6) and that in the overall experience was 3.4% (95% CI: 2.6 to 4.6). The overall odds of 30-day stroke using the TA (Fig. 5B) approach in the early experience from high-volume centers was 1.8% (95% CI: 0.8 to 4.4) and that in the overall experience was 2.8% (95% CI: 1.9 to 4.3).
TAVR versus SAVR
Eleven reports that included 2 randomized control trials and 9 propensity-matched analyses reported on patients. There was no difference in the risk of stroke between the 2 approaches in both high-risk (OR: 1.24, 95% CI: 0.58 to 2.65, I2 = 13.83) and intermediate-risk patients (OR: 1.69, 95% CI: 0.33 to 8.48, I2 = 33.69). The pooled OR for in-hospital/30-day stroke (TAVR vs. SAVR) did not reach statistical significance for both comparisons (Online Figs. 3A and 3B).
Subgroup analysis of Edwards Valve
We performed a subgroup analysis to compare the valve designs exclusively for the femoral approach (Online Fig. 4) . There was no difference between the Edwards valve and CoreValve for 30-day stroke post-TAVR in both multicenter (3.0% vs. 2.3%, p = 0.17) and single-center studies (4.4% vs. 3.9%, p = 0.7). We also compared the 30-day stroke between the 2 approaches (TF and TA) exclusively with use of the Edwards valve (Online Fig. 5). There was again no difference in 30-day stroke post-TAVR with either approach in both multicenter (3.2% vs. 2.8%, p = 0.49) and single-center studies (4.2% vs. 3.4%, p = 0.44).
TAVR was introduced as a revolutionary treatment option for inoperable patients with severe aortic stenosis in 2002 (11). It has since shown comparable results to SAVR, rapidly evolved with more than 150,000 procedures performed worldwide and has seen an expansion of its indications. Expansion in utilization and indications has shifted the focus from efficacy to combined efficacy and safety. In the PARTNER trial cohort B (Online Ref. e19), comparing TAVR to medical therapy, neurological events occurred more frequently with TAVR at 30 days (6.7% to 1.7%, p = 0.03). A similar trend was noticed in cohort A (Online Ref. e18) patients of the PARTNER trial comparing TAVR to SAVR (4.6% vs. 2.4%). An analysis of the Society of Thoracic Surgeons database of patients undergoing isolated aortic valve replacement between 2002 and 2006 showed a 1.6% risk of stroke (3). Others have reported a stroke rate in the order of 0.8% to 5% for SAVR compared to 0% to 10% for TAVR (4). Stroke has thus surfaced as a major safety concern after TAVR. Of note though, there is a lack of consistent definitions and inconsistencies in surveillance and reporting standards across the reports precluding any firm conclusions. Nevertheless, there does appear to be a risk of stroke with TAVR that has a potential for further improvement as we advance.
The retrograde TF approach for TAVR is the default approach in most centers worldwide. It involves the advancement of a delivery catheter containing the valve from the common femoral artery to the ascending aorta, a course that traverses the aortic arch and origins of the carotids, with a potential for cerebral embolism of atherosclerotic aortic plaques. On the other hand, the TA approach allows direct access to the aortic valve via the left ventricular apex bypassing the aortic arch and the origin of the carotids. The retrograde approach further requires crossing of the stenotic aortic valve with a guide wire that may increase the risk of stroke. There are likewise suggestions that valve sizing, positioning, and implantation may be better with a TA approach with subsequent reduction in paravalvular leak and balloon post-dilation, need for repositioning, and shorter fluoroscopic time, factors that have been linked to neurological events after TAVR (12) (Online Refs. e99,e100,e104). An earlier meta-analysis on 7,541 patients provided support to the previous concerns (5). The TA approach was associated with a 2.7% risk of stroke, as opposed to 3.1% for the TF approach using the CoreValve and 4.2% using the Edwards Valve. This finding has however not borne out because with subsequent reports suggesting a similar risk of stroke unrelated to the chosen approach (Online Refs. e23,e100). In our analysis, the stroke risk was comparable between the 2 approaches, in large multicenter (2.8% vs. 2.8%) and single-center studies (3.8% vs. 3.4%). This is in parallel to etiological studies of stroke with TAVR that suggest a high-risk period during balloon pre-dilation, valve positioning, and implantation (crushing of the stenotic native valve) factors common to both approaches and not during catheter manipulation across the aortic arch (Online Refs. e104–e106,e112). Therefore, it is safe to conclude that the major source of atherosclerotic emboli during TAVR is from aortic root manipulation and much less or negligible from arch manipulation. A point to be remembered is that catheter manipulation, though decreased with the TA approach is not fully eliminated.
The CoreValve and the Edwards valve are the most extensively evaluated valves for TAVR thus far, with important differences in design, deliverability and implantation technique that have been speculated to influence the risk of stroke. Eggebrecht et al. (5) in a pooled analysis of 5,097 patients implied a higher stroke with the Edwards valve when compared to the CoreValve (4.2% vs. 3.1%). Others have suggested a trend toward a higher risk of stroke with the CoreValve (3.5% vs. 1.5%) (13). In our pooled analysis the risk of stroke was similar between the valve types (multicenter registries, OR: 1.03; 95% CI: 0.78 to 1.35; and single-center studies, OR: 1.81; 95% CI: 0.69 to 4.74). Despite the similar stroke risk there may be an important differences in timing of stroke inherent to each valve, an understanding of which may be crucial in developing future valve designs to reduce stroke. Keeping with the high-risk period of stroke during positioning and valve deployment, Kahlert et al. (Online Ref. e112) showed that the risk with the CoreValve is during the slow stepwise implantation, while that with the Edwards valve, it is during the slow positioning of the device prior to implantation. Valve dislocation and embolization are more commonly reported with the CoreValve (14). Strategies to manage these complications are well known to increase periprocedural stroke (15) (Online Ref. e100). Rapid pacing during implantation of the Edwards valve is crucial for accurate positioning unlike that for the CoreValve. Rapid pacing causes functional cardiac arrest with ensuing transient cerebral ischemia that may be linked to clinical stroke (Online Ref. e107). Balloon post-dilation to reduce paravalvular leak is a common practice with both valve designs. It has been linked to stroke (Online Refs. e100,e113) and should be avoided where possible, balancing the risk of mortality with even mild aortic insufficiency (16).
The risk of stroke has declined over the years with operator experience, advancements in valve technology, and improvement in patient selection. In the PARTNER trial (Online Refs. e18,e19) all patients received an Edwards SAPIEN device using a 22- to 24-F delivery catheter. The current generation CoreValve and SAPIEN XT valve use an 18-F delivery catheter and have a lower profile. In the PARTNER II trial (Online Ref. e20) the 30-day risk of stroke was 3.2% with use of the Sapien XT device compared to a 5.5% to 6.7% risk in the PARTNER I trial. Similarly, the risk of stroke in the CoreValve ADVANCE study (Online Ref. e3) that included over 1,000 patients was only 2.9%. Comparing the pooled, early versus late experience of high-volume centers, we found a comparable drop in stroke risk (4.9% vs. 3.4%). Along the same lines, the risk of stroke after TAVR as compared to SAVR has declined. The current risk is within the range reported by historical SAVR trials (3,4). It must however be pointed out that none of the historical surgical trials were scrutinized as closely as TAVR, and therefore the true stroke risk after SAVR may be under-represented.
There is growing interest in embolic protection filters to further reduce the risk of stroke after TAVR. In a cohort of 40 patients who underwent TAVR with use of an embolic protection device, Van Mieghem et al. (17) showed that embolic debris was captured by the filter in 75% of the procedures. The captured emboli consisted of both tissue fragments and thrombotic debris. The potential of embolic protection devices to clinically impact outcomes is currently under investigation, as stroke after TAVR is multifactorial with the risk prevalent at all steps of the procedure and also beyond (Online Table 10). Approximately 50% to 60% of strokes occur within 24 h and the majority of the rest within a week of the procedure (18,19). Furthermore, progressive device iterations, improved patient selection, and advancing operator experience are continually decreasing the risk of procedural stroke. Attention to antithrombotic therapies and risk factors beyond the procedure may also be rewarding.
Moving forward, uniform prospective surveillance and reporting of stroke per the VARC2 criteria (20) will define the true stroke burden after TAVR and enable valid interpretations and comparison across trials. Of concern though is the reliance on neuroimaging to detect stroke. Clinically silent new ischemic brain lesions are relatively common after TAVR on diffusion-weighted cerebral magnetic resonance imaging studies (21). Supplanting a comprehensive neurological examination with neuroimaging will therefore likely overinflate the risk of stroke after TAVR. Imperative to accurate stroke identification is anatomical localization by sound neurological examination in order to identify clinically relevant lesions. Pre-imaging localization by clinical exam should be stressed and incorporated in the VARC2 recommendations to precisely identify strokes in the TAVR population. Clinical exam should not be replaced by imaging.
First, publication bias is an important limitation of our analysis. Data from included studies and registries may be outdated as the true utilization of TAVR is not matched by the pace of reporting. Also, single-center studies that were excluded because of low volumes are most likely to have progressed to high volumes. Second, there were no randomized comparisons between the approaches (TF or TA) or valve designs (MC vs. ES). This introduced significant heterogeneity. Third, the definition of stroke was variable within the included reports, as was surveillance and reporting. Minor stroke is likely under-represented in our analysis. Nevertheless, despite these limitations our analysis on >25,000 patients is the largest sampling to date and thus provides valuable insights into the risk of stroke post-TAVR.
In this large meta-analysis there was no difference in stroke between the approach (TF or TA) or type of valve (MC or ES) used for TAVR. Importantly, there was a decline in stroke over time, reflecting on continually improving outcomes after TAVR. Reduction in device profiles over time and advancements in technique are likely to make TAVR safe, to potentially expand its indications to a low-risk population.
For supplemental references, tables, and figures, please see the online version of this article.
All authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- confidence interval
- Medtronic CoreValve
- odds ratio
- surgical aortic valve replacement
- transcatheter aortic valve replacement
- Valve Academic Research Consortium
- Received November 27, 2013.
- Revision received February 9, 2014.
- Accepted February 11, 2014.
- American College of Cardiology Foundation
- Piazza N.,
- Kalesan B.,
- van Mieghem N.,
- et al.
- Wenaweser P.,
- Stortecky S.,
- Schwander S.,
- et al.
- ↵Borenstein MHL, Higgin J, Rothstein H. Comprehensive meta-analysis version 2. 2005. Available at: http://www.meta-analysis.com/index.php. Accessed April 2014.
- Cribier A.,
- Eltchaninoff H.,
- Bash A.,
- et al.
- Chieffo A.,
- Buchanan G.L.,
- Van Mieghem N.M.,
- et al.
- Geisbusch S.,
- Bleiziffer S.,
- Mazzitelli D.,
- Ruge H.,
- Bauernschmitt R.,
- Lange R.
- Athappan G.,
- Patvardhan E.,
- Tuzcu E.M.,
- et al.
- Van Mieghem N.M.,
- Schipper M.E.,
- Ladich E.,
- et al.
- Kappetein A.P.,
- Head S.J.,
- Genereux P.,
- et al.