Author + information
- Received June 21, 2016
- Revision received August 22, 2016
- Accepted August 25, 2016
- Published online December 12, 2016.
- Mohamad Alkhouli, MD,
- Mohammad Sarraf, MD,
- Elad Maor, MD,
- Saurabh Sanon, MBBS,
- Allison Cabalka, MD,
- Mackram F. Eleid, MD,
- Donald J. Hagler, MD,
- Peter Pollak, MD,
- Guy Reeder, MD and
- Charanjit S. Rihal, MD∗ ()
- Divisions of Cardiovascular Diseases and Internal Medicine, Mayo Clinic College of Medicine, Rochester, Minnesota
- ↵∗Reprint requests and correspondence:
Dr. Charanjit S. Rihal, Mayo Clinic College of Medicine, Division of Cardiovascular Diseases, 200 First Street SW, Rochester, Minnesota 55905.
Objectives The aim of this study is to provide a summary of the currently applied aortic paravalvular leak (PVL) closure techniques and describe the procedural and long-term outcomes in a large consecutive cohort of patients.
Background Percutaneous repair has emerged as an effective therapy for patients with PVL. To date, clinical outcome data on percutaneous closure of aortic PVL are limited.
Methods All patients who underwent catheter-based treatment of aortic PVL between 2006 and 2015 were identified. Procedural and short-term results were assessed. Patients were contacted for clinical events and symptoms.
Results Eighty-six procedures were performed in 80 patients. The mean age was 68 ± 15 years, and 70% were men. The primary indications for PVL closure were symptoms of heart failure, hemolysis, and both in 83%, 5%, and 12%, respectively. Successful device deployment was accomplished in 94 defects (90%). Reduction in PVL to mild or less was achieved in 62% of patients. In-hospital major adverse events occurred in 8% of procedures. Symptomatic improvement at 30 days was achieved in 64% of patients. Patients who had reduction in the PVL grade to mild or less experienced more improvement in New York Heart Association functional class (from 2.93 ± 0.62 to 1.72 ± 0.73) compared with those with mild or greater residual leak (from 3.03 ± 0.57 to 2.52 ± 0.74) (p < 0.001). In patients with severe hemolysis (n = 8), transfusion requirements were eliminated in 7 (88%) after PVL closure. Kaplan-Meier survival analysis showed that the cumulative probability of freedom from repeat surgery at 2 years was 98 ± 2% in patients who had mild or less residual leak compared with 68 ± 10% in patients with higher grades of residual PVL (log-rank p = 0.004).
Conclusions Percutaneous reduction of aortic PVL is associated with durable symptom relief and lower rates of repeat cardiac surgery. The magnitude of benefit is greatest with PVL reduction to a grade of mild or less. Therefore, attempts should be made to reduce PVL as much as possible.
Paravalvular leak (PVL) occurs in 5% to 17% of patients after valve replacement surgery (1–4). For symptomatic patients, repeat surgery has been the traditional treatment of choice, but it is associated with high operative mortality and variable results even in the modern era (1,4–7). Percutaneous repair has emerged as an effective therapy for patients with PVL, with feasibility and efficacy demonstrated in multiple studies (8–12). Although the principles of transcatheter closure of mitral and aortic PVL are similar, the techniques and procedural complexity differ significantly (13). Mitral PVL closure is more intricate than aortic PVL because of some procedural and leak-specific characteristics. To date, the largest series examining the outcomes of percutaneous PVL closure included only a small number of patients with aortic PVL (10,12,13).
We hypothesized that aortic PVL closure can be achieved with high success and low complications rates and that the degree of PVL reduction correlates with symptomatic improvement and clinical outcomes. In this report, we review the commonly applied techniques in aortic PVL closure and provide comprehensive data on the procedural and long-term outcomes of a large consecutive cohort of patients referred for aortic PVL closure.
The Mayo Clinic Institutional Review Board approved this investigation. We retrospectively identified patients who underwent percutaneous repair of aortic PVL at the Mayo Clinic (Rochester, Minnesota) before January 10, 2016. The indications for percutaneous repair were moderate or severe PVL with severe or life-style-limiting dyspnea (New York Heart Association [NYHA] functional class III or IV or class II with significant life-style or occupational impairment) or clinically significant hemolytic anemia. Patients who had active endocarditis and those who had large leaks involving more than one-half of the circumference of the sewing ring, or rocking motion of the valve, were referred for surgical repair. Clinically significant hemolytic anemia was defined as symptomatic anemia (hemoglobin <11 g/dl in women and <12.5 g/dl in men), with laboratory evidence of intravascular hemolysis.
Grading of aortic PVL
The assessment of the severity of aortic PVL incorporated a multifaceted approach. This included echocardiographic, invasive hemodynamic, and angiographic measures. Semiquantitative echocardiographic parameters were used in all cases to grade the PVL as mild, mild to moderate, moderate, or severe: 1) the PVL jet width (vena contracta) measured in the short- and long-axis views at the level of the prosthesis sewing ring and in the left ventricular outflow tract; 2) assessment of diastolic flow reversal in the descending thoracic and abdominal aorta; and 3) measurement of deceleration rate by pressure half-time (14). In the presence of multiple defects, the sum of regurgitation from these defects was used. Quantitative Doppler parameters (e.g., regurgitant volume) were also used to confirm the severity of PVL in the majority of cases. In equivocal cases, invasive measurement of left ventricular and aortic pressures were used. In addition, aortic root angiography was performed with grading according to Sellers criteria in selected patients (15).
Procedural, in-hospital, and 30-day outcomes were assessed by retrospective chart review. Technical success was defined as successful deployment of a closure device across the leak with reduction of the PVL to mild or less. Clinical success was defined as symptomatic improvement in NYHA functional class by at least 1 class and/or elimination of need for transfusion in cases of severe hemolysis. Patients were contacted by telephone survey to determine occurrence of adverse events, symptoms, and clinical status. In addition, the dates of death of deceased patients were verified by querying the National Death Index, a centralized database of death record information on file in state vital statistics offices (16).
PVL closure techniques
Multimodality imaging guidance
Successful PVL closure relies heavily on good understanding of the PVL anatomy. Meticulous planning shortens the procedure time and increases the chances of successful PVL reduction. The number, location(s), and severity of PVL defects were assessed via detailed analysis of transthoracic echocardiograms and transesophageal echocardiograms, which were available in all patients (Figure 1). Additionally, later in our experience we used electrocardiographically gated cardiac computed tomography in the pre-procedural assessment in selected patients. Cardiac computed tomography can be used to: 1) measure the dimensions and identify the path of a PVL; 2) measure the distance from the defect to the coronary ostium; and 3) identify the optimal fluoroscopic angles to cross the defect (Figure 2). Transesophageal echocardiography was used to guide the procedure when percutaneous repair of a posterior leak(s) was planned or in cases of hemodynamic instability. Otherwise, the procedure was guided with transthoracic echocardiography or intracardiac echocardiography at the operator’s discretion.
Percutaneous PVL repair
Anticoagulation was held and bridging with low–molecular weight heparin was used in a minority of high-risk patients (e.g., patients with double mechanical valves). In the vast majority of patients, 6-F arterial access was obtained in the common femoral artery. Aortic PVL defects were crossed with a 0.035-inch stiff angled Glide wire (Terumo, Tokyo, Japan) through a telescoped 125-cm, 5-F multipurpose coronary catheter and a 6-F multipurpose guiding catheter (Cordis Corporation, Hialeah, Florida), respectively. If initial steering of the wire into the defect was difficult, a 6-F Amplatzer left 1 guiding catheter (Boston Scientific, Marlborough, Massachusetts) was used as the second catheter of choice. Once the leak is crossed, 3 common techniques can be used to deliver the closure device: 1) a catheter-only technique; 2) an “anchor wire” technique; and 3) an arterioarterial rail technique.
The catheter-only technique (Figure 3A) is an expeditious method that is useful when crossing of the defect is smooth and the leak is likely to seal with 1 device. After crossing the defect with the telescoped system described earlier, the wire and the 5-F catheter are removed, and the device is advanced via the 6-F guiding catheter. A 6-F guide is compatible with up to a 12-mm Amplatzer Vascular Plug II (AVPII, St. Jude Medical, St. Paul, Minnesota). If a larger AVPII device or other devices are needed, the guiding catheter is exchanged over an Amplatzer extrastiff guidewire with a 6- or 7-F shuttle sheath (Cook Medical, Bloomington, Indiana). A disadvantage of this technique is the loss of guidewire position across the leak at the time of closure device deployment. Recrossing of the leak is more challenging with an existing device across the leak.
The anchor wire technique (Figure 3B) preserves access across the defect and allows sequential deployment of multiple devices if necessary. In this method, once the telescoping system is advanced in to the left ventricle, the Glide wire is exchanged with an anchor wire (300-cm 0.032- or 0.035-inch Amplatzer extrastiff guidewire). The arterial sheath is then upsized to a 45-cm 8- to 10-F bright-tip sheath (Cordis Corporation) to minimize blood loss from the arterial site during catheter exchanges. The AVPII device does not fit alongside an anchor wire within a 6-F guiding catheter. Therefore, the 6- to 8-F shuttle sheath is advanced over the anchor wire and inside the 8-F sheath across the defect. The operator should pay attention to the stiff tip of the dilator of the shuttle sheath to prevent left ventricular puncture or damage. The closure device is advanced alongside the anchor wire into the left ventricle, is deployed across the defect, but remains attached to the delivery cable. The delivery catheter is then removed and placed back on the anchor wire only, leaving the device cable outside the delivery catheter. This remaining rail can be used to recross the defect if additional devices are needed, and if the operator fears losing access across the defect during attempts of closure device deployment.
In rare occasions, a more stable rail is needed for device deployment. In these cases, an arterioarterial rail technique (Figure 3C) can be used. The Glide wire used to cross the defect is advanced through the aortic valve into the descending aorta and is then snared and exteriorized to the contralateral femoral artery, creating an arterioarterial rail. The reminder of the procedure can be completed in a similar fashion to the anchor wire steps. This technique can primarily be used in patients with bioprosthetic valves. Although we have used this technique successfully in patients with mechanical prosthesis, we recommend against its routine use with mechanical valves. A wire across the mechanical valve can lead to stuck leaflets with severe aortic regurgitation and result in rapid hemodynamic compromise. If the patient has a mechanical valve and the device could not be delivered without a rail, the left ventricular wire can be snared through a transseptal or transapical puncture, creating an arteriovenous rail and an arterioapical rail, respectively. In our experience, an arterioapical rail was used in only 1 of 86 cases of aortic PVL closure.
Important technical considerations should be taken into account to maximize the efficiency and minimize the risks of the procedure.
First, device and sheath compatibility and fit are crucial to procedural execution. Knowledge of sheath-sizing requirements for each device with and without an anchor wire shortens the duration and reduces the cost of the procedure. Compatibility charts are provided in Online Figure 1.
Second, a feared complication of aortic PVL closure is the impingement of a mechanical or bioprosthetic valve leaflet during device deployment. Mechanical leaflet impingement is usually readily recognized with fluoroscopy. However, impingement of bioprosthetic leaflets, albeit rare, is more difficult to recognize but can be suspected when there is echocardiographic evidence of sudden valvular regurgitation or an increased transvalvular gradient. Detailed assessment of valve leaflet motion should be performed in every case before releasing the closure device. In rare occasions, device interference with the prosthetic valve leaflets can occur after plug release because of tilting of the device, which would require device removal with a snare or a long, flexible bioptome.
Third, para-aortic leak closure may lead to coronary artery obstruction. Aortography or selective coronary angiography may be needed to assess ostial clearance before device release. Meticulous cardiac computed tomographic measurement and analyses can identify low takeoff of the left or right coronary ostia before the procedure.
The primary efficacy endpoints of the study were the change in NYHA functional class at 30 days and the elimination of transfusion requirement in patients with severe hemolysis requiring blood transfusion. The primary safety endpoint of the study was the occurrence of acute and 30-day major adverse cardiovascular events, defined as stroke, major vascular complications, tamponade, acute coronary syndrome, or death. Secondary endpoints were: 1) long-term survival from death within the duration of the study; and 2) long-term event-free freedom from repeat cardiac surgery.
Continuous parameters of the study groups were compared using the Student t test. For comparison of categorical data, we used the chi-square or Fisher exact test. To examine the impact of residual regurgitation on outcomes, patients were grouped according to the residual PVL (mild or less, greater than mild). The probability of death or repeat surgery according to residual PVL groups was graphically displayed according to the method of Kaplan and Meier, with comparison of cumulative survival across strata by the log-rank test. Binary logistic regression analysis was used to identify variables independently associated with the failure to reduce the PVL to less than mild (technical failure). Variables included in the model were age, sex, diabetes, prosthesis type (bioprosthesis vs. mechanical, transcatheter heart valve), history of infective endocarditis (IE), location of the PVL, procedure status (elective vs. urgent), imaging guidance (transthoracic vs. transesophageal echocardiography), and device type (AVPII vs. others). All analyses were performed with SPSS version 22 (IBM, Armonk, New York). Statistical significance was inferred at p ≤0.05.
Eighty-six aortic PVL closure procedures were performed in 80 patients. Baseline characteristics of the study population are shown in Table 1. The mean age was 67.8 ± 14.6 years, 70% were men, and 63% had bioprosthetic valves. The mean number of prior sternotomies was 1.4 ± 0.9 (range 0 to 6). The indication for aortic valve replacement was calcific aortic stenosis in 64%, IE in 9%, aortic regurgitation in 6%, and rheumatic aortic valve disease in 5%. The primary indication for percutaneous PVL closure was symptoms of heart failure in 66 patients (83%), hemolysis in 4 patients (5%), and both in 10 patients (12%). Eight procedures (9.4%) were done urgently in patients not in stable condition who presented with decompensated heart failure soon after aortic valve replacement. The mean times from valve implantation to first detection of aortic PVL and to percutaneous repair were 22 ± 40 months and 35 ± 46 months, respectively. Using the Society of Thoracic Surgeons risk calculator, the estimated operative mortality for open repair in the cohort was 5.7 ± 4.6.
Acute procedural outcomes
Echocardiographic guidance was used in all cases. Baseline echocardiographic data are summarized in Table 2. Percutaneous repair was attempted in a total of 105 PVLs in 80 patients. Successful crossing and device deployment were achieved in 94 defects using 114 closure devices (90%). The AVPII device was used in the majority of cases (88%). The reasons for failure to close the 11 remaining leaks were inability to cross the defect with a wire or with a sheath in 8 defects, impingement of prosthetic leaflet in 1, proximity to a coronary artery ostium in 1, and coronary dissection in 1.
Technical success (defined as mild or less residual PVL) was achieved in 53 patients (62%). The residual PVLs were mild to moderate in 17 patients (20%) and moderate or more in 16 patients (18%) (Table 3). Among multiple baseline and procedural characteristics, only history of IE predicted more than mild residual PVL after the procedure (Table 4). In-hospital major adverse cardiovascular events occurred in 6 patients (7.6%). These included death in 2 (2.5%) (1 due to tamponade and hemothorax after an apical puncture and 1 due to persistent cardiogenic shock in a patient who underwent successful salvage PVL closure immediately after surgical aortic valve replacement), stroke in 1 (1.2%), retroperitoneal bleeding in 1 (1.2%), tamponade in 1 (1.2%), and coronary dissection in 1 (1.2%). The mean length of hospital stay was 2.3 ± 3.2 days for patients who underwent elective PVL closure compared with 10.8 ± 15.7 days for patients who underwent urgent procedures (p < 0.001).
Short-term (30-day) outcomes
Clinical follow-up (NYHA functional class assessment) at 30 days was available in 70 patients (93%). The primary endpoint of the study was achieved in 51 patients (64%). The mean change in NYHA functional class was 0.9 ± 0.5 (Figure 4). Patients who had successful reductions in PVL grade to mild or less experienced significantly more improvement in NYHA functional class (mean change 1.21, from 2.93 ± 0.62 to 1.72 ± 0.73) compared with those with more than mild residual leak (mean change 0.51, from 3.03 ± 0.57 to 2.52 ± 0.74) (p < 0.001) (Figure 4). In patients with severe hemolysis requiring blood transfusion (n = 8), transfusion requirement was eliminated in 7 cases (88%).
Among the 78 patients discharged from the hospital, 5 major adverse cardiovascular events occurred in 4 patients (5.1%): 3 deaths and 2 hemorrhagic strokes. Two deaths occurred after successful PVL closure; 1 was at day 6 because of a hemorrhagic stroke, and 1 was at day 27 because of an unknown cause in a 35 year-old patient with a 27-mm Starr-Edwards Silastic Ball valve. One death occurred at day 28 because of decompensated heart failure after an unsuccessful attempt for PVL closure in an 86-year-old patient with a 23-mm SAPIEN-XT valve. Both hemorrhagic strokes occurred between 2 and 5 days after the procedure in patients who were bridged to therapeutic warfarin with low–molecular weight heparin.
Median follow-up for the study patients was 18.5 months, with a range of 1 to 101 months (mean 27 ± 25 months). During follow-up, all-cause mortality was 27.5%, and redo valve replacement for residual or recurrent PVL was 13% (Table 5). Compared with patients who had more than mild post-procedure leak, those who had mild or less residual leak had better freedom from repeat cardiac surgery but long-term survival rates were similar (Figure 5). In patients who achieved event-free survival, the initial improvement in NYHA functional class persisted at maximum follow-up (mean NYHA functional class 1.95 ± 0.78). The magnitude of NYHA functional class improvement was more significant in patients with mild or less residual leak (mean NYHA functional class 1.76 ± 0.74 vs. 2.24 ± 0.76; p = 0.027).
The principal findings of the present investigation are as follows. 1) Percutaneous repair of aortic PVL can be performed safely with a low incidence of major complications. 2) Cannulation and closure device deployment across aortic PVL was achieved in the vast majority of PVLs (90%). However, successful reduction of the PVL to mild or less was only achieved in 62% of patients. 3) History of IE was strongly correlated with significant residual leak after percutaneous PVL closure. 4) Successful PVL reduction to mild or less resulted in quick and durable symptomatic relief and reduced the need for redo cardiac surgery.
There is growing evidence from the global experience with transcatheter aortic valve replacement (TAVR) that PVL after TAVR is associated with poor long-term outcomes. However, the negative impact of PVL is seen predominantly in patients with moderate and severe PVL (17,18). We classified successful PVL closure on the basis of the residual leak to mild or less (technical success) and more than mild (technical failure). Similar to what has been observed in the post-TAVR PVL literature, our findings suggest that more than mild residual PVL after percutaneous closure is associated with less symptomatic improvement and a lower rate of event-free survival. Although there was no difference in all-cause mortality between the 2 groups, our study is underpowered to detect mortality differences given the small sample size.
Achieving complete or near complete obliteration of aortic PVL, however, can be challenging. As illustrated in this study, although successful crossing of aortic PVL and deployment of closure device(s) across the leaks were achieved in 90% of defects, reduction in PVL to mild or less was possible in only 62% of patients. This can be due to a number of reasons.
First, a significant predictor of more than mild residual PVL in this study was a history of IE. The process of inflammation and healing in IE can lead to scar formation and/or fibrosis and may result in tissue fragility, which can make complete abolition of the leak more challenging. Other predictors of technical failure are the presence of mechanical prostheses, transcatheter heart valve, and urgent indications for PVL closure. Bioprosthetic valves generally have lower risk for device-leaflet interaction. In contrast, the spatial relationship of mechanical prosthesis occluder discs relative to the defect(s) is a key determinant of the size and number of devices that can be implanted without leaflet impingement. Therefore, technical success rates are expected to be lower with mechanical prostheses (11). Patients who develop significant PVL after TAVR represent another challenging group. In these patients, if the transcatheter heart valve is properly positioned and adequately expanded, the PVL is due to the lack of surgical excision of the old calcified valve, resulting in multiple defects of small dimensions (19). Smaller profile devices such as the Amplatzer Vascular Plug IV have been used in these patients, but the technical success rate has been only modest (20–22).
Second, despite the growing prevalence of PVL, there are no devices that were designed, tested, or approved for PVL closure (23). The AVPII is the most commonly used off-label device because of its stability and low profile. However, many patients have crescentic or irregular shaped PVLs, and the AVPII may not be adequate for treatment of such defects (24). In recent years, the Amplatzer Vascular Plug III has been increasingly used for PVL closure given its oblong shape (25,26). Also, a new purpose-specific device, the Occlutech PLD occluder (Occlutech, Jena, Germany), was designed to tackle a large range of PVL sizes and morphologies (26). Early experience with the Occlutech PLD device showed very encouraging results (26,27). Unfortunately, neither the Amplatzer Vascular Plug III nor the Occlutech is available in the United States. Another consideration to increase the success rate of the procedure is the possible use of transapical access in selected cases. Transapical access has been used primarily in the treatment of mitral PVL for medial and anterior leaks or when concomitant aortic and mitral PVL closure is planned (12). Although transapical access is potentially associated with higher complication rates, these can be minimized by careful access planning with a pre-acquired computed tomographic angiogram and closure of the access site with vascular plugs or occluders (12).
Despite the low rate of adverse events in our series, further modification of the current technique may further lower the risk for these complications. First, we observed 2 hemorrhagic strokes in patients with atrial fibrillation who were bridged with low–molecular weight heparin. Recently, the BRIDGE (Bridging Anticoagulation in Patients Who Require Temporary Interruption of Warfarin Therapy for an Elective Invasive Procedure or Surgery) trial suggested no benefit and higher bleeding events in patients who underwent periprocedural bridging with low–molecular weight heparin for atrial fibrillation (28). Therefore, we believe that bridging anticoagulation should be avoided when possible. Second, a meticulous post-procedural protocol that includes frequent echocardiographic examinations should be implemented when an apical puncture is performed to aid in early detection and management of hemothorax or tamponade. Closing the apical puncture site with a vascular plug should also be considered when sheath sizes >4 F are used.
First, the retrospective nature of this study has known limitations, including potential for referral basis. Notably, serial echocardiography was not performed, and data on residual regurgitation at final follow-up were not available, because of the referral nature of our clinical practice.
Second, the practice of PVL closure might differ outside the United States because of the availability of purpose-specific devices. Therefore, the results of this study may not be generalizable to practices outside the United States.
Third, residual PVL grade in our study was correlated with symptomatic improvement and need for redo surgery. However, grading of aortic PVL can be quite challenging. Echocardiographic measures used to grade PVL are semiquantitative and suffer from limited validation. This could lead to interpretation variability, which may affect the results of this study.
Fourth, follow-up laboratory data on patients with hemolysis were limited because of the referral nature of our center. However, in those with severe hemolysis requiring blood transfusion, PVL closure eliminated blood transfusion in all but 1 patient.
Last, the small number of patients enrolled in this study may affect the reliability of the data, and the validity of certain statistical analyses (e.g., logistical regressions). However, to our knowledge this is the largest reported series of percutaneous aortic PVL closure. The descriptive data in our study are very valuable in closing the knowledge gap on the management of this complex entity.
Percutaneous aortic PVL closure can be performed with high technical success and low complication rates. Successful PVL reduction to mild or less results in quick and durable symptomatic relief and reduces the need for redo cardiac surgery. Attempts should be made to accomplish adequate reduction in PVL to optimize clinical outcomes.
WHAT IS KNOWN? Percutaneous closure has emerged as an alternative therapy to repeat valve surgery for patients with symptomatic aortic PVL. However, procedural and long-term outcomes of percutaneous closure techniques are not known.
WHAT IS NEW? Percutaneous reduction of aortic PVL is associated with durable symptom relief and lower rates of repeat cardiac surgery. The magnitude of benefit is greatest with PVL reduction to grade mild or less, which was achieved in two-thirds of the patients.
WHAT IS NEXT? Further studies are needed to assess the effectiveness of PVL closure with purpose-specific devices.
For a supplemental figure, please see the online version of this article.
The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- Amplatzer Vascular Plug II
- infective endocarditis
- New York Heart Association
- paravalvular leak
- transcatheter aortic valve replacement
- Received June 21, 2016.
- Revision received August 22, 2016.
- Accepted August 25, 2016.
- 2016 American College of Cardiology Foundation
- Hwang H.Y.,
- Choi J.W.,
- Kim H.K.,
- Kim K.H.,
- Kim K.B.,
- Ahn H.
- Davila-Roman V.G.,
- Waggoner A.D.,
- Kennard E.D.,
- et al.
- Genoni M.,
- Franzen D.,
- Vogt P.,
- et al.
- Smolka G.,
- Pysz P.,
- Jasinski M.,
- et al.
- Sorajja P.,
- Cabalka A.K.,
- Hagler D.J.,
- Rihal C.S.
- Sorajja P.,
- Cabalka A.K.,
- Hagler D.J.,
- Rihal C.S.
- Ruiz C.E.,
- Jelnin V.,
- Kronzon I.,
- et al.
- Zoghbi W.A.,
- Chambers J.B.,
- Dumesnil J.G.,
- et al.
- ↵National Center for Health Statistics. National Death Index User’s Guide. Available at: https://www.cdc.gov/nchs/data/ndi/ndi_users_guide.pdf. Accessed September 8, 2016.
- Jerez-Valero M.,
- Urena M.,
- Webb J.G.,
- et al.
- Athappan G.,
- Patvardhan E.,
- Tuzcu E.M.,
- et al.
- Eleid M.F.,
- Cabalka A.K.,
- Malouf J.F.,
- Sanon S.,
- Hagler D.J.,
- Rihal C.S.
- Kim M.S.,
- Casserly I.P.,
- Garcia J.A.,
- Klein A.J.,
- Salcedo E.E.,
- Carroll J.D.
- McElhinney D.B.
- Cruz-Gonzalez I.,
- Rama-Merchan J.C.,
- Arribas-Jimenez A.,
- et al.
- Bedair R.,
- Morgan G.J.,
- Bapat V.,
- et al.
- Yildirim A.,
- Goktekin O.,
- Gorgulu S.,
- et al.