Early Transcatheter Aortic Valve Function With and Without Therapeutic Anticoagulation

Abstract: Objectives. Prosthetic leaflet thrombosis is a growing concern in transcatheter aortic valve replacement (TAVR). Given the uncertainty of best practices for antiplatelet and anticoagulation therapies in the post-TAVR period, additional evidence regarding the impact of anticoagulation on prosthetic valve function after TAVR is needed. Methods. Patients undergoing native-valve TAVR at a single academic institution between 2012 and 2015 were analyzed based on any anticoagulant use at hospital discharge post TAVR. Changes in prosthetic valve peak velocity and mean gradient were assessed based on transthoracic echocardiograms performed immediately following valve implant and at 4-week follow-up. Multivariate regression analyses were performed to explore the impact of anticoagulation status on early TAVR valve performance. Results. For 403 patients, there were no available data to analyze. Of those, 29.6% were discharged on anticoagulation. Following TAVR, the average mean prosthetic valve gradient was 11.8 ± 5.6 mm Hg and peak velocity was 2.33 ± 0.52 m/s.  There were no significant differences between anticoagulated and non-anticoagulated groups in the mean or peak gradients or velocity immediately following implant or at 4 weeks, which remained true following multivariate adjustment (P=.80 for delta mean gradient; P=.91 for delta peak velocity). Conclusion. Our data suggest that the absence of anticoagulation is not associated with short-term degradation in TAVR performance and do not support the routine use of anticoagulation following native-valve TAVR.  

J INVASIVE CARDIOL 2017;29(11):391-396.

Key words: anticoagulation, stenotic aortic valve disease

Transcatheter aortic valve replacement (TAVR) has revolutionized treatment for stenotic aortic valve disease.¹ Although procedural risk has declined and accessibility has improved, subclinical valve thrombosis remains a concern. Prosthetic valve thrombosis is a known early complication of surgical aortic valve replacement (SAVR) and is associated with an increased risk of valve degeneration, transient ischemic attack, stroke, and death.^2-8 Computed tomography (CT) studies observing subclinical valve changes show thrombosis to be more common after TAVR compared with SAVR.³ Subclinical early valve thrombosis is now recognized as a prevalent phenomenon post TAVR, and is observed on cardiac multidetector CT in approximately 13% of TAVR registry patients.^2,3 This prevalence is heavily dependent on anticoagulation status; for instance, one study found that subclinical valve thrombosis was not observed in those on warfarin with a therapeutic international normalized ratio (INR) >2.0, whereas it was seen in 51% of patients without anticoagulation or with subtherapeutic INR.² Thrombosis is associated with abnormal leaflet motion on multidetector CT, which normalized on anticoagulation studies,^2,3 suggesting a direct therapeutic benefit of anticoagulation. 

The clinical implication of subclinical prosthetic valve thrombosis after TAVR remains unclear. Subclinical thrombosis is highly associated with restricted leaflet motion, and up to 18% of patients with valve thrombosis develop overt valve obstruction.⁴ A recent larger study including 752 post-TAVR subjects found that prosthetic valve subjects with leaflet thrombosis were more likely to have a gradient increase >10 mm Hg and a gradient >20 mm Hg than those without thrombosis.³

The 2017 American Heart Association/American College of Cardiology focused update to the guideline for management of patients with valvular heart disease now gives a IIB indication for anticoagulation with a vitamin K antagonist (VKA) in patients at low risk of bleeding.¹ This shift reflects recent concerns of valve thrombosis and escalates therapy from the general practice stemming from the precedent of the initial TAVR trials in using 1 month of dual-antiplatelet therapy (DAPT) in patients not already on anticoagulation. Anticoagulation therapy is not benign in the TAVR patient population, however, with frequent comorbidities that place them at substantial risk of bleeding. While findings regarding valve thrombosis in the early post-TAVR period are concerning, the clinical benefit of anticoagulation remains uncertain. We therefore sought to evaluate the association between anticoagulation status (used for other clinical indications) and changes in early prosthetic valve function measured on transthoracic echocardiogram (TTE) following TAVR. 

Methods

Study sample. Patient information was abstracted from a registry of consecutive patients undergoing native-valve TAVR at University of Washington Medical Center between 2012 and 2015. Patients who underwent valve-in-valve TAVR or were missing post-procedure or 1-month follow-up TTE parameters were excluded. Patient variables captured included: (1) demographics and medical comorbidities; (2) procedural characteristics; (3) echocardiographic characteristics immediately after TAVR, including mean aortic valve gradient and peak velocity; (4) echocardiographic characteristics at initial follow-up (approximately 4 weeks after TAVR); and (5) anticoagulant and antiplatelet therapies administered after TAVR. 

Transthoracic echocardiography. TTE was performed in the immediate post-TAVR period, typically within 48 to 72 hours, and repeated at 4-week follow-up as part of routine care. Echocardiographic variables such as left ventricular ejection fraction (LVEF) and valvular hemodynamics were evaluated according to American Society of Echocardiography guidelines. Variables collected included LVEF, mean aortic gradient, peak aortic gradient, and maximum aortic velocity. When endocardial definition was inadequate to directly measure LVEF, a visual assessment of LVEF was performed (13 subjects) and all but 1 subject was deemed to have normal LVEF. Images were correlated by visual assessment to prior studies utilizing echocardiographic contrast or higher imaging quality as a reference in these special cases. All echocardiograms were performed by experienced sonographers at a TAVR center. 

Implanted valve types were Sapien, Sapien XT, Sapien S3, (Edwards Lifesciences), CoreValve, Evolut R (Medtronic), and Lotus valve (Boston Scientific). Anticoagulant agents used included warfarin, rivaroxaban, apixaban, and enoxaparin. Antiplatelet agents included low-dose aspirin, high-dose aspirin, clopidogrel, and ticagrelor.

Statistical analyses. Baseline patient variable comparisons were performed using student’s t-test with equal variances for continuous variables, and using Chi-squared testing for categorical variables. The change in individual subjects’ aortic valve (AV) peak velocity and mean AV gradient between the immediate post-TAVR period and 1 month follow-up were determined as the delta (D) AV velocity and D mean gradient, respectively. Outliers were assessed by assessing the distribution of mean AV gradient change from implant to 1 month. The number of patients with gradient changes above 5 mm Hg, 10 mm Hg, and 20 mm Hg were quantified. Linear regression analyses were performed to compare groups defined by anticoagulation status with respect to change in mean gradient and peak AV velocity at 1 month. Multivariate linear regression analyses adjusted for age, sex, LVEF, valve type, and valve size. Two-sided P-values of <.05 were considered statistically significant. All data analysis was performed using Stata software v. 11 (Statacorp). The University of Washington Institutional Review Board approved this study. 

Results

Four hundred and three patients with available data were included in this analysis. Baseline clinical characteristics are presented in Table 1. A total of 122 subjects (29.6%) were anticoagulated at hospital discharge. There were no significant differences between the anticoagulated and non-anticoagulated subgroups when comparing age, sex, race, body mass index, valve type, valve size, or days between TAVR placement and follow-up echocardiogram. A history of atrial fibrillation was present in 50.1% of all patients, including 83% of anticoagulated patients and 37.8% of non-anticoagulated patients (P<.01). Prior deep vein thrombosis was present in 3.9% of anticoagulated patients and 0.6% of non-anticoagulated patients (P=.02) and pulmonary embolism in 5.2% of anticoagulated patients compared with 1.7% of non-anticoagulated patients (P=.04). 

The median interval to initial post-TAVR echocardiogram was 2.35 ± 1.69 days, and median interval to 4-week follow-up echocardiogram was 28.5 ± 10.9 days. There were no significant differences in interval follow-up times between the anticoagulated and non-anticoagulated subgroups.

Warfarin was used in 95% of anticoagulated subjects; the remaining 5% were anticoagulated with rivaroxaban, apixaban, or enoxaparin. Antiplatelet use was more common in non-anticoagulated subjects. After TAVR, 73% of patients on anticoagulation were taking low-dose aspirin vs 92% of non-anticoagulated patients (P<.01). Similarly, 25% of anticoagulated patients were on clopidogrel vs 90% of non-anticoagulated patients (P<.01). Only 5.7% of anticoagulated patients received DAPT, representing 1.7% of the entire study cohort.

Pre-TAVR echocardiograms demonstrated no difference in mean AV gradient (44.8 mm Hg in the non-anticoagulated group vs 43.7 mm Hg in the anticoagulated group; P=.65) or pre-TAVR LVEF (58.6% vs 59.4%; P=.74) Prosthetic valve characteristics are described in Table 1. The choice of prosthetic valve was relatively evenly distributed between Sapien XT, Sapien 3, and CoreValve, whereas Sapien (8.7%), Evolut R (3.2%), and Lotus valves (1.5%) were less commonly used. There were no significant differences in valve type or size between anticoagulation groups (Table 1).

Following TAVR, the average mean prosthetic valve gradient was 11.8 ± 5.6 mm Hg, which was stable at 11.1 ± 5.5 mm Hg at the 4-week visit. The average peak velocity was 2.33 ± 0.52 m/s, which also remained stable. There were no significant differences between anticoagulated and non-anticoagulated groups in the mean or peak gradients or velocity initially or at 4 weeks (Table 2) (Figure 1). After adjusting for age, gender, LVEF, valve type, and size, there was still no significant difference in the between-echo changes in valve performance by gradients or velocity (D mean gradient, 0.09 mm Hg higher in the anticoagulated group; 95% confidence interval (CI), -0.81 to 0.997 [P=.80]; D peak velocity, 0.004 m/s lower in the anticoagulated group; 95% CI, -0.086 to 0.077 [P=.91]). 

The change in mean AV gradient from implant to 1 month was approximately normally distributed for all patients (Figure 2). Overall, few patients had clinically significant changes in their prosthesis gradient at 4 weeks. Valve gradients did not increase over 20 mm Hg in the initial month for any patient. Only 4 patients had 4-week gradients that increased over 10 mm Hg; of these, 2 patients were anticoagulated and 2 patients were not. In total, 27 patients had 4-week AV gradients that decreased by more than 5 mm Hg; of these, 18 patients were anticoagulated and 9 patients were not anticoagulated.

Discussion

Our data suggest that therapeutic anticoagulation, when used for pre-existing indications, does not appear to affect early clinical performance of TAVR prostheses in native annuli. This topic is of particular relevance in light of prior studies demonstrating that pre-existing anticoagulation is associated with reduced leaflet thrombosis by CT scan,² that leaflet thrombosis is reduced following initiation of therapeutic anticoagulation,^2,3 and that the absence of post-TAVR anticoagulation is an independent predicting factor for valve thrombosis.¹ Nevertheless, anticoagulation following TAVR in a native annulus is not routinely implemented at most centers and data on the attendant bleeding risks relative to the potential benefits of routine anticoagulation with or without antiplatelet agents following TAVR remain unknown.

Of course, not all valve thrombosis is clinically detectable. An increase in mean gradient by 10 mm Hg and an absolute mean gradient of >20 mm Hg was associated with leaflet thickening in a recent large cohort of TAVR patients,³ but the majority of patients with restricted leaflet motion maintained a normal mean gradient <20 mm Hg and the association of hemodynamically detectable differences in gradient and imaging findings of restricted leaflet motion or thrombosis has been variable.^2,4-6 Nevertheless, our data demonstrate no differences in the change in mean AV gradient or peak velocity in the first month after TAVR between patients on anticoagulation therapy as compared with patients on no anticoagulation, adjusting for age, sex, valve size and type, and LVEF. Of all patients, the 4 outliers who demonstrated an increase in mean valve gradient above 10 mm Hg between implant and 1 month were evenly split between anticoagulated and non-anticoagulated groups. The vast majority of non-anticoagulated patients were on DAPT (predominantly aspirin and clopidogrel). 

There are several possible explanations for the variable association of valve gradient with thrombus identified with dedicated CT. While subclinical thrombosis prevalence is estimated at 7%-40% across various cohorts,^2-5 severe leaflet restriction resulting in significant valve obstruction is likely much less common, estimated at 14%-18%^3,5 of those with an identified thrombus. Thus, while thrombus formation may be more common in patients only taking antiplatelet therapy after TAVR, and leaflet restriction on CT is a more sensitive means of recognizing the pathology, the true incidence of clinically significant obstructive disease deserves further study. Evidence of clinically significant bioprosthetic valve thrombosis has been previously linked to echocardiographic findings of an increase in mean gradient above 50% from baseline as a validated independent predictor of bioprosthetic valve thrombosis.⁷ Although the burden of thrombus identified using modern CT technology required to cause valve obstruction is not defined, obstruction may require thrombotic involvement of at least two leaflets based off in vitro data.²

Additionally, valve thrombosis is predicated upon multiple variables including surface clotting factors, hemostasis, flow turbulence, and surrounding tissue damage.⁸ How these factors influence different patient populations is currently unknown. Valve-specific factors have also been examined, and large valve size was previously found to be an independent predictor of valve thrombosis.⁵ Some speculate the risk of thrombosis is higher in balloon-expandable valves as compared with self-expanding valves due to more aggressive ballooning resulting in endothelial disruption,⁸ but recent studies representing both valve classes have not observed a difference³ and trauma has been observed with both transcatheter valve types in ex vivo studies.⁹ Our analysis similarly did not demonstrate any significant differences in AV gradients between different valve types, overall or within groups based on anticoagulation status.

The early post-TAVR period is a relevant point of observation for thrombosis-related valve changes, as transcatheter bioprosthetic valves demonstrate the highest risk of thromboembolism in the first 3 months following implantation,¹⁰ after which the risk returns to a baseline consistent with the general population. This 3-month period also coincides with the highest-risk period of surgical bioprosthetic valve thrombosis requiring re-operation¹¹ and valve endothelialization. Our study evaluated for change in valve function at approximately 1 month after valve implantation, which is in the range where valve thrombosis has been observed on CT in other studies. Our time frame from TAVR to CT for thrombosis detection is near the median time in one trial⁵ and within the first quartile in two others,^2,3 with median time of approximately 80 days.

Recent evidence led to a change in the 2017 update to the American Heart Association/American College of Cardiology valvular heart disease guidelines to consider 3 months of anticoagulation following TAVR in patients at low risk of bleeding.¹² Many patients have another indication for anticoagulation in the TAVR population. For instance, atrial fibrillation alone has been estimated to be present in 16%-40% of TAVR patients.¹³ Similarly, atrial fibrillation was reported in 50% of our cohort. Only 30% were on anticoagulation, however, and it may be that the 38% with a history of atrial fibrillation who were not on anticoagulation may be an important target based on mean age alone (81 ± 10 years). A recent study comparing VKA alone vs VKA plus antiplatelet therapy after TAVR showed no benefit with respect to stroke, major adverse cardiac event, or death, while major bleeding was more common (hazard ratio, 1.85; 95% CI, 1.05 to 3.28; P=.04),¹⁴ indicating that anticoagulation alone is sufficient in the absence of another absolute indication for antiplatelet therapy such as cardiac stents. 

Conversely, the decision to routinely start patients without a pre-existing indication on anticoagulation remains challenging. The typical elderly or otherwise comorbid TAVR patient is rarely at low risk of bleeding, and evidence regarding risk of thromboembolic events after TAVR related to valve thrombosis remains weak. Chakravarty et al reported an increased risk of transient ischemic attack but not stroke or myocardial infarction in patients with reduced leaflet motion in a 12-center study of over 700 post-TAVR patients.³ Whether or not treatment with anticoagulation in the setting of reduced leaflet motion improves clinical outcomes remains to be seen. Similar to atrial fibrillation, baseline risk of valve thrombosis and stroke is likely impacted by other comorbidities such as LV function,⁸ and the absolute benefit of anticoagulation for reducing the risk of thromboembolic events and preserving valve function after TAVR may weigh heavily on other risk factors. 

Study limitations. Several study limitations warrant discussion. We report on the short-term impact of anticoagulation on valve function as determined by echocardiography, and our study is not powered to detect differences in clinical outcomes. Furthermore, as this was a retrospective analysis, information regarding subclinical thrombosis is not available. We also chose to focus on the short-term implications of anticoagulation because this seemed to best correspond to the time frame in which thrombosis may occur and was less confounded by changes in anticoagulation status that may happen in the year following TAVR. However, 4 weeks of follow-up may be insufficient to detect a difference in valve gradients. This was also a single-center study, and our data may not be applicable broadly, although our population is likely similar to many other referral centers for TAVR. Furthermore, the use of anticoagulation and antiplatelet therapies was determined by report at hospital discharge, but could not be verified by pharmacy records or patient reporting. 

Conclusion

TAVR valve function as measured by echocardiographic parameters showed no difference based on anticoagulation status following hospital discharge when implanted in the native aortic annulus. These data suggest that routine anticoagulation following native-valve TAVR may be of limited clinical benefit. Further study from ongoing clinical trials, aimed specifically at addressing questions of anticoagulation after TAVR, will be required to change current practice patterns.

References

1.    Nishimura RA, Otto CM, Bonow RO, et al. 2017 AHA/ACC focused update of the 2014 AHA/ACC guideline for the management of patients with valvular heart disease. J Am Coll Cardiol. 2017;70:252-289. Epub 2017 Mar 15.

2.    Makkar RR, Fontana G, Jilaihawi H, et al. Possible subclinical leaflet thrombosis in bioprosthetic aortic valves. N Engl J Med. 2015;373:2015-2024. 

3.    Chakravarty T, Søndergaard L, Friedman J, et al. Subclinical leaflet thrombosis in surgical and transcatheter bioprosthetic aortic valves: an observational study. Lancet. 2017;389:2383-2392. Epub 2017 Mar 19.

4.    Hansson NC, Grove EL, Andersen HR, et al. Transcatheter aortic valve thrombosis. J Am Coll Cardiol. 2016;68:2059-2069.

5.    Pache G, Schoechlin S, Blanke P, et al. Early hypo-attenuated leaflet thickening in balloon-expandable transcatheter aortic heart valves. Eur Heart J. 2016;37:2263-2271.

6.    Leetmaa T, Hansson NC, Leipsic J, et al. Early aortic transcatheter heart valve thrombosis. Circ Cardiovasc Interv. 2015;8:e001596.

7.    Egbe AC, Pislaru SV, Pellikka PA, et al. Bioprosthetic valve thrombosis versus structural failure: clinical and echocardiographic predictors. J Am Coll Cardiol. 2015;66:2285-2294. 

8.    Dangas GD, Weitz JI, Giustino G, Makkar R, Mehran R. Prosthetic heart valve thrombosis. J Am Coll Cardiol. 2016;68:2670-2689.

9.    Convelbo C, Guetat P, Cambillau M, et al. Crimping and deployment of balloon-expandable valved stents are responsible for the increase in the hydraulic conductance of leaflets. Eur J Cardiothorac Surg. 2013;44:1045-1050. 

10.    Stortecky S, Windecker S. Stroke: an infrequent but devastating complication in cardiovascular interventions. Circulation. 2012;126:2921-2924.

11.    Brown ML, Park SJ, Sundt TM, Schaff HV. Early thrombosis risk in patients with biologic valves in the aortic position. J Thorac Cardiovasc Surg. 2012;144:108-111. 

12.    Nishimura RA, Otto CM, Bonow RO, et al. 2014 AHA/ACC guideline for the management of patients with valvular heart disease. J Am Coll Cardiol. 2014;63:e57-e185.

13.    Mok M, Urena M, Nombela-Franco L, et al. Clinical and prognostic implications of existing and new-onset atrial fibrillation in patients undergoing transcatheter aortic valve implantation. J Thromb Thrombolysis. 2013;35:450-455.

14.     Abdul-Jawad Altisent O, Durand E, Muñoz-García AJ, et al. Warfarin and antiplatelet therapy versus warfarin alone for treating patients with atrial fibrillation undergoing transcatheter aortic valve replacement. JACC Cardiovasc Interv. 2016;9:1706-1717.

From the ¹Department of Internal Medicine and ²Division of Cardiology, Seattle University of Washington, Seattle, Washington.

Disclosure: The authors have completed and returned the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr McCabe reports honoraria from Edwards Lifesciences and Boston Scientific. The remaining authors report no conflicts of interest regarding the content herein.

Manuscript submitted August 7, 2017, provisional acceptance given August 15, 2017, final version accepted August 24, 2017.

Address for correspondence: James M. McCabe, MD, University of Washington, 1959 NE Pacific St, Box 356422, Seattle, WA 98195-6422. Email: jmmccabe@uw.edu