|Home | About | Journals | Submit | Contact Us | Français|
The American College of Cardiology/American Heart Association (ACC/AHA) guidelines recommend warfarin for 3 months following bioprosthetic valve replacement (BVR), yet strong evidence supporting this recommendation is lacking, making process variation likely.
The ANSWER Registry enrolled 386 patients receiving Epic or Biocor BVRs from May 2007 to August 2008 at 40 centers. Patterns of discharge anticoagulation and outpatient INR values were collected. Mortality, embolic, and bleeding events were assessed to 6 months post-BVR.
The median patient age was 74 years (interquartile range [IQR], 67 to 80), 39% were female, and 65% were classified as high thromboembolic risk. Warfarin was prescribed in 38% of all BVR patients and 49% in those with high thromboembolic risk. The median time-to-therapeutic INR was 9 days (IQR, 1 to 18), with 20% never reaching therapeutic levels. Among those achieving a therapeutic INR, 78% and 57% had at least one subtherapeutic or supratheraputic INR during subsequent follow-up to 3 months. In follow-up, patients treated with warfarin had similar rates of embolic events (2.8% vs. 3.1%, p=0.884), but substantially higher bleeding incidence than those not treated with warfarin (12% vs. 3%, p=0.0012). Among patients who were anticoagulated, those with supratheraputic INRs had a 7-fold higher risk for overt bleeding events (26% vs. 3%).
Anticoagulation strategies following BVR are highly variable. In this population, challenges in achieving and maintaining therapeutic warfarin anticoagulation are common and associated with increased bleeding risk. Further work to clarify the optimal post-BVR anticoagulation strategy is needed.
Bioprostheses are rapidly becoming the treatment of choice for patients with valvular heart disease (1,2), and are the dominant strategy among elderly patients (3–5). Although the American College of Cardiology/American Heart Association (ACC/AHA) guidelines (6) recommend the use of warfarin(a vitamin K antagonist [VKA]) following bioprosthetic valve replacement (BVR)(Class IIa, Level of Evidence [LOE]: C), evidence supporting this recommendation is limited (7,8).
The extent of variation in antithrombotic strategies following BVR in contemporary practice, the practicalities of achieving and maintaining therapeutic levels of anticoagulation during the early post-operative period, and the association between post-BVR antithrombotic strategy and outcomes, are not well characterized. The ANticoagulation Strategy With Bioprosthetic Valves Post-Operative Event Registry (ANSWER) was prospectively designed to address these questions through the collection of the following information: 1) discharge anticoagulation strategy; 2) international normalized ratio (INR) logs; and 3) post-operative clinical events to 6 months post-BVR.
Sponsored by St. Jude Medical to compliment findings of the Acute Coronary Treatment and Intervention Outcomes Network (ACTION) Registry®, ANSWER was a multi-center study designed for investigation of antithrombotic strategies within the United States in patients receiving bioprosthetic Biocor, Biocor Supra, Epic, or Epic Supra valves. The data was collected, housed, and analyzed at the Duke Clinical Research Institute (DCRI), while St. Jude Medical provided material support for the operation of the data collection and infrastructure.
Three hundred and eighty-six patients from 40 hospital centers were enrolled over 16 months (May 2007 through August 2008), with a median enrollment of 7 patients per center (interquartile range [IQR], 3 to 14) (Appendix). Six additional sites were activated, but had no patient enrollment in ANSWER. All patients age ≥18 years and scheduled for a first-time aortic or mitral valve replacement with Biocor, Biocor Supra, Epic, or Epic Supra valves at participating centers, were eligible for inclusion. Exclusion criteria were active endocarditis, chronic hemodialysis, and emergency surgery. All patients provided informed consent prior to the qualifying surgery.
In-hospital data collection included patient characteristics, risk factors for thromboembolic events, in-hospital medications, discharge anti-thrombotic and anti-platelet medications, dosage, intended duration, date of initiation, and target INR. Patients receiving warfarin were provided INR logs to record the dates and values of all INR tests and any change in warfarin dosage or (if applicable) warfarin discontinuation. Following discharge, scripted telephone interviews were conducted at 3- and 6-months by the registry coordinators at individual sites to collect information regarding the occurrence of rehospitalizations, emergency department visits, or urgent care visits. When medical encounters were reported, hospital records were reviewed with particular attention paid to the occurrence of adverse events. Per protocol, the DCRI conducted data audits on 10% of all events reported during the 6-month follow-up period through an independent abstraction of diagnostic and procedural billing codes associated with each encounter.
Rates of missing data in ANSWER were low (<1% for most variables); only left ventricular ejection fraction (LVEF, 6%) and New York Heart Association classification (35%) were missing in >5% of participants.
To evaluate the generalizability of the ANSWER cohort, we identified both a similar group of patients receiving Biocor, Biocor Supra, Epic, or Epic Supra valves from January 2007 to December 2008 in the Society of Thoracic Surgeons Adult Cardiac Surgery Database (STS ACSD) as well as the full STS ACSD cohort of patients receiving bioprosthetic valve prostheses at centers with participation in both ANSWER and the STS ACSD. In doing so, we established a base population for bioprosthetic AVR procedures both within the STS ACSD network and within the subset of ANSWER sites. A comparison of baseline characteristics and discharge antithrombotic strategies is presented.
To further delineate the extent of practice variation in antithrombotic treatment strategies at participating centers, the proportion of patients receiving each strategy at hospital discharge was described by valve type and across each of several important clinical characteristics, including the presence or absence of comorbidities known to impart additional thromboembolic risk as enumerated in the ACC/AHA valve guidelines (e.g., atrial fibrillation, prior thromboembolism, left ventricular dysfunction [ejection fraction, EF<30%], or the presence of a hypercoagulable state).
In addition, the use of Class I or IIa ACC/AHA guideline-recommended antithrombotic strategies (6) at hospital discharge was described overall, by valve type, and by the presence of thromboembolic risk factors. Patients with a contraindication to antiplatelet or anticoagulant medications (n=16), those with a pre-existing indication for warfarin (n=37), and those who died in-hospital (n=15) were excluded from this analysis. The remaining population (n=320) included patients with new indications for anticoagulation, based upon their BVR.
The target INR range was collected at discharge in 381 warfarin-treated patients (99%). In the remaining 5 patients without a specified INR range, the target range was conservatively set to 2.0–3.0. In 7 warfarin-treated patients who did not report a therapeutic INR during the 6 months of follow-up, the date of therapeutic INR was conservatively set as the date of the last INR measurement, plus one day.
International Normalized Ratio stability following the achievement of a therapeutic range INR was evaluated. The frequency of supra- and subtherapeutic INR ranges were described. The proportion of person-time spent with an INR <2.0 (subtherapeutic) and >3.0 (supratherapeutic) was calculated for each patient, with the last recorded INR level representing the end of person-time for each individual. For this analysis, the discharge target INR (as described above) was presumed to represent the target INR throughout the 6-month follow-up.
Post-operative adverse events were evaluated for the 6-month follow-up period in the overall cohort and separated according to warfarin use, maximum recorded INR, and proportion of person-time spent above or below an INR of 2.0–3.0. Endpoints included embolic or overt bleeding events, valve thrombosis, and death. Embolic events were classified as neurologic (e.g., stroke, TIA) and non-neurologic (not further subcategorized). Embolic and bleeding events within the first 48 post-operative hours were excluded from this analysis, as these events frequently represent operative complications.
All analyses were performed using SAS software (version 9.1, SAS Institute, Cary, North Carolina, USA).
Table 1 provides baseline clinical characteristics for the ANSWER cohort. Characteristics of patients enrolled in ANSWER were typical of patients receiving a Biocor or Epic valve during this time period in the STS ACSD(n=6,883), including similar demographics (median age, 74 vs. 75 years; female 39% vs. 42%) and comorbidities (diabetes, 26% vs. 29%; hypertension, 79% vs. 80%). Patients from the ANSWER cohort were slightly less likely than the STS ACSD cohort to receive aspirin (79% vs. 83%) or warfarin (38% vs. 42%) at discharge.
Thirty-eight of 40 ANSWER sites also participated in the STS ACSD. Within these sites, 9% (340 of 3639) of patients receiving aortic valve bioprostheses were enrolled in ANSWER, including 47% of all patients receiving the pre-specified St. Jude valve models. Patients enrolled in ANSWER were representative of the overall cohort receiving aortic valve bioprostheses at these sites, including similar demographics (median age, 74 years; female 39% vs. 41%) and comorbidities (diabetes, 26% vs. 30%; hypertension, 81% vs. 79%). Discharge medications were similar for the ANSWER cohort, as for that of the base population of patients at participating sites (aspirin: 86% vs. 84%; warfarin: 37% vs. 36%).
Within the ANSWER cohort, 85 patients (22%) had a history of atrial fibrillation, 135 (35%) had a decreased EF (<30%), 6 (2%) had a hypercoagulable state, and 46 (12%) had a prior thromboembolic event. Combined, 251 patients (65%) had risk factors for thrombo-embolism, including 154 (40%) with 1risk factor, 86 (22%) with 2risk factors, and 11 (3%) with 3 risk factors.
Warfarin was prescribed in 38.1% of all ANSWER patients and 37% of those without pre-operative indications or contra-indications (Table 2). Patients receiving warfarin at discharge were slightly older than those not receiving warfarin (74 vs. 72 years), less often female (37% vs. 38%), and more often with a history of New York Heart Association Class III or IV heart failure (43% vs. 22%), hypercoagulable state (4% vs. 0%), pre- or post-operative atrial fibrillation (65% vs. 30%), or prior thrombosis (7% vs. 4%) (Table 1). Warfarin and antiplatelet therapies were used in combination more commonly among those patients with thromboembolic risk factors than among those without (34% vs. 13%) (Table 2). A higher proportion of patients receiving isolated mitral versus aortic bioprostheses were discharged on warfarin (61% vs. 34%). Among patients undergoing combined coronary artery bypass grafting plus valve procedures, 12% were discharged on both warfarin and an antiplatelet agent.
An antiplatelet agent was prescribed in 84% of all BVR patients, including 82% of those with, and 87% of those without, additional thromboembolic risk factors. Antiplatelet agents were used exclusively in 59% of all patients, including 49% of those with additional thromboembolic risk factors.
Among patients without a pre-operative indication for, or contra-indication to warfarin, 50% were discharged with either an ACC/AHA Class I- or IIa-recommended antithrombotic strategy. The use of guideline recommended therapies was lowest among those patients with (versus without) risk factors for thromboembolism (34% vs. 91%), patients receiving mitral valves versus aortic valves (39% vs. 51%), women versus men (42% vs. 55%), and self-pay patients versus the insured (32% vs. 51%). No difference was observed between patients <75 years of age versus those ≥75 years of age (51% vs. 49%).
The median time from surgery to first warfarin dose was 3 days, and the intended INR range was 2.0–3.0 in most patients (83%). The intended duration of therapy was ≥6 months in 48% of patients, 1–3 months in 49%, and ≤1 month in 4%.
During the 6-month follow-up period, patients had an average of 10 (range 2–24) outpatient INR tests performed. The majority of INR testing took place in physician offices (69%), as opposed to dedicated warfarin clinics (12%) or hospital-based testing (19%). The median time from the first warfarin dose to a therapeutic INR was 9 days (IQR, 18 days), with 43% of patients achieving a therapeutic INR prior to hospital discharge. A therapeutic INR was achieved beyond 4 weeks in 19.6% of patients. Seven patients (7%) never achieved a therapeutic INR during the 6-month follow-up period (Figure). No difference in median post-procedural length of stay (8 days) was observed between those discharged with (versus without) a therapeutic INR.
The median time-to-therapeutic INR was slightly longer in women than men (9.5 vs. 8.5 days), patients ≥75 versus <75 years old (10 vs. 9 days), and those patients followed in anticoagulation clinics versus those followed in hospital or physician labs (10, 8, and 8 days, respectively). The shortest median time-to-therapeutic INR was achieved among those with Veterans Affairs/military and Medicare insurance (5.5 and 6.0 days).
After achieving an initial therapeutic INR, 78% of patients had at least one subsequent subtherapeutic INR (INR <2.0), while 57% had at least one subsequent supratherapeutic INR (INR >3.0) over the course of the 6-month follow-up. Considering all INR values that were obtained in warfarin-treated patients, less than half (41%) of the person-time was spent in the therapeutic range, with a nearly equal amount of time (39%) spent subtherapeutic, and an additional 20% of time spent supratherapeutic.
Patients monitored in anticoagulation clinics were more likely to have supratherapeutic INRs, while those monitored in physician offices and hospital labs were more likely to have subtherapeutic INRs. Forty-two percent of patients monitored in anticoagulation clinics, 25% of those monitored at physician offices, and 22% of those monitored at hospital labs spent >30% of person-time with an INR >3.0. Conversely, 61% of patients monitored in both physician offices and hospital labs, and 33% of patients monitored in anticoagulation clinics, spent >30% of person-time with an INR <3.0.
Overall, 6-month mortality occurred in 9% of patients, embolic events occurred in 4%, and overt bleeding occurred in 11% (Table 3). Two of the overt bleeding events involved intracranial hemorrhages, including 1 event at 3 days post-op and 1 event at 71 days post-op in a warfarin-treated patient.
Among warfarin-treated patients, the probability of bleeding was substantially elevated with increased time spent supra-therapeutic, such that patients spending 0%, 1–30%, and >30% of time with an INR >3.0 had a 3%, 11%, and 26% probability of bleeding by 6 months, respectively. Additionally, the probability of bleeding was substantially lower among those with a maximum INR ≤3.0 versus those with a maximum INR >3.0 (3% vs. 16%). Patients discharged on both an antiplatelet agent and warfarin did not have an increased incidence of bleeding versus those discharged on warfarin alone (11% vs. 15%).
Of the 11 embolic events, 82% were neurologic (5 strokes, 4 transient ischemic attacks/reversible ischemic neurological deficits) and 18% were non-neurologic. No valve thrombosis was observed. Over the 6-month follow-up period, patients taking warfarin had a similar incidence of thromboembolism as those not taking warfarin (2.8% vs. 3.1%; p=0.884).
This analysis provides important observations regarding the challenges of post-BVR thromboprophylaxis in community practice. First, there is considerable variation in post-BVR antithrombotic strategies, and many patients at higher risk for thromboembolic events are not receiving anticoagulation with VKA at hospital discharge. Second, when discharge warfarin is instituted, its dosing is suboptimal, with a large proportion of patients remain subtherapeutic for a prolonged period following surgery. Third, once a therapeutic INR has been achieved, INR maintenance remains problematic, and out of range values are associated with increased clinical events.
Based primarily on consensus opinion and case series (LOE: C), the ACC/AHA guidelines for the management of patients with valvular heart disease recommend warfarin indefinitely for BVR patients at high-risk for thromboembolism (Class I) and for 3 months following BVR for those without risk factors (Class IIa) (6). In the ANSWER population, 50% of patients were discharged without either Class I- or IIa-recommended therapies, including 68% of patients at high-risk for thromboembolism. Much of this discrepancy involves the Class I recommendation (LOE: C) for warfarin among aortic valve replacement patients with additional thromboembolic risk factors. These patients make up 61% of the ANSWER aortic valve replacement population, but only 49% were treated with warfarin. Additionally, nearly 5% of all ANSWER BVR patients were discharged without any antithrombotic medications. Although most physicians prescribe some form of antithrombotic medication at hospital discharge following BVR, the isolated use of anti-platelet medications is common, especially among patients with isolated aortic valve procedures and among those without additional risk factors. In contrast, the use of warfarin was most common among patients with mitral valve prostheses and those with additional thromboembolic risk.
These data suggest a greater reticence of United States physicians to the use of VKA, when compared with the 60% VKA use reported by their “most-of-world” colleagues (9). However, these data do highlight a trend in the United States, which is consistent with a high degree of practice variation observed in the international community(9,10). The discrepancy between clinical practice and guideline recommendations observed both here and abroad(9)may reflect: 1) the limited information on the risk of thromboembolic events in post-BVR patients (11); 2) the lack of randomized trial data supporting the use of VKA to reduce these thromboembolic events in post-BVR populations; 3) an unfamiliarity with the guideline recommendations among treating physicians (7); 4) a hesitancy of treating physicians to use VKA post-BVR, due to an associated increase in bleeding events; and/or 5) a manifestation of the difficulties in clinical coordination required to manage warfarin and other VKA in an appropriate therapeutic range. The observation that those at highest risk for bleeding (e.g., women and the elderly) are among the least likely to receive warfarin supports the hypothesis that the contrast between practice and guidelines stems from a perceived risk/benefit trade-off. In contrast, the observation that those patients without insurance are less likely to be treated warfarin may reflect socioeconomic challenges of follow-up care for this therapy (12,13).
In addition to highlighting the extent of practice variation in the use of antithrombotics following BVR, our study also reaffirms real-world challenges in establishing and maintaining a therapeutic INR range following warfarin initiation in clinical practice. Warfarin initiation is especially difficult in the post-operative period due to varying pharmacokinetics (14), concomitant medications, and dietary factors. These difficulties are reflected in the ANSWER population in both prolonged time-to-therapeutic INRs and a high incidence of subsequent out-of-range INR values following the achievement of a therapeutic INR. As in other populations (15), the median time-to-therapeutic INR in the ANSWER population was >1 week (median = 9 days), with 20% of warfarin-treated patients remaining subtherapeutic beyond 4 weeks.
Patients monitored at dedicated anticoagulation clinics did not achieve more rapid therapeutic anticoagulation than those monitored in physician offices or hospital labs, despite slightly more frequent INR checks at anticoagulation clinics. However, a difference in the intensity of warfarin therapy was noted according to the type of INR monitoring facility. Patients monitored in anticoagulation clinics were more likely to maintain supratherapeutic INRs, while those monitored in physician offices or hospital labs were more likely to maintain subtherapeutic INRs.
The highest risk for thromboembolism and overt bleeding was observed during the 10 days following BVR (15). In the ANSWER population, between one-half and two-thirds of all thromboembolic and bleeding events occurred prior to discharge, emphasizing the early risk which exists in this population. Yet the comparative effectiveness of VKA versus aspirin therapy in preventing such thromboembolic events in the early, high-risk period following BVR is poorly understood. Case series have demonstrated a low incidence of early thromboembolic events following BVR in patients without additional thromboembolic risk factors in whom only antiplatelet medications were used (16), and observational comparisons of antiplatelet versus VKA have revealed no differences in thrombotic or bleeding events (17,18). Yet to date, only 2 small randomized controlled pilot trials have been carried out in this area, including 1 comparing aspirin (100mg/day) versus warfarin (n=75) (19) and 1 comparing triflusal (an antiplatelet agent) versus acenocoumarol (a VKA, n=193) (20). Thromboembolic events were rare in each of these studies, and neither was adequately powered to demonstrate a significant difference across this endpoint. In each of these 2 trials, patients receiving VKA demonstrated an increased risk of hemorrhage compared with those receiving antiplatelet agents (aspirin or triflusal). Nevertheless, it is unclear the extent to which this observation reflects difficulties in post-operative warfarin management, rather than the efficacy of antiplatelet medications (versus VKA) in the prevention of thromboembolism. Certainly, warfarin has proven to be superior to aspirin for the prevention of thromboembolism in other clinical settings (21,22). In the ANSWER cohort, thromboemolic events were rare. The risk of events was similar among those with warfarin therapy at discharge versus without; however, this result may reflect a mitigation of thromboembolic risk among the higher risk warfarin-treated population, rather than therapeutic equivalence. Low thromboembolic event rates precluded risk-adjustment to adequately account for this potential bias.
In the ANSWER population, the greatest risk of bleeding was observed among those patients with supra-therapeutic INRs, especially among those spending more than 30% of their patient-time with an INR above 3.0. When compared to patients in the non-warfarin treated cohort, patients whose INR remained below 3.0 did not experience an elevated bleeding risk. This fact further substantiates the assertion that supratherapeutic anticoagulation (rather than therapeutic anticoagulation) is a major risk factor for bleeding in post-BVR patients. Accordingly, measures to prevent supratherapeutic anticoagulation in these patients can be expected to yield an improved risk-benefit profile.
These results highlight the potential role for alternative treatment strategies for BVR patients. If anticoagulants are to be used following BVR, further efforts are needed to maximize their risk-benefit profiles, including the use of alternative dosing (i.e., gene-based warfarin dosing) and monitoring (i.e., home INR monitoring) systems or alternative vitamin K antagonists (e.g., acenocoumarol). Although it is not currently standard practice in the United States, home monitoring may result in increased time in therapeutic range (23). Additionally, the effectiveness of anti-platelet only treatment strategies for the optimization of outcomes following BVR may warrant further randomized investigation, as these are currently the dominant strategies in many institutions around the globe. Finally, newer anticoagulants (e.g., direct thrombin inhibitors) may offer an alternative to VKA in BVR patients. Indeed, recently published randomized comparisons of Dabigatran (a direct thrombin inhibitor) versus warfarin in patients with atrial fibrillation (24) or acute venous thromboembolism (25) have demonstrated an equivalent and potentially superior (dose-dependent) risk-benefit profile for this novel therapeutic. To date, no studies have tested the effectiveness of direct thrombin inhibitor agents in BVR patients, and given the clinical equipoise demonstrated in this and other registry analyses, any such study should ideally include separate comparisons with both VKA and antiplatelet agents.
While ANSWER presents an important analysis of post-BVR anti-coagulation, the registry has several limitations. First, ANSWER data were drawn from a select group of 42 centers and only from patients treated with Biocor and Epic valves. Furthermore, given the number of patients enrolled in this cohort, it is unlikely that all BVR patients at these centers received a Biocor or Epic valve; it is unclear what, if any, bias this may have introduced. However, we did not find any meaningful differences between the patients enrolled in ANSWER and those who received similar procedures in the STS ACSD, suggesting that the results may be generalizable within the United States. Second, the size of the sample enrolled in ANSWER and the frequency of clinical events limits our ability to draw conclusions about risk factors for observed clinical events. Third, outpatient INR values were dependent upon patient reporting; therefore, these values may be subject to certain biases and inaccuracies. Finally, ANSWER provides primarily a descriptive analysis. The lack of randomization to antithrombotic therapies prevents an unbiased comparison of the effectiveness of warfarin versus aspirin therapy for the prevention of thromboembolism.
In conclusion, by examining this multi-center registry of BVR patients, we found that there is substantial variation in antithrombotic strategies following hospital discharge, and that the incidence of thromboembolic and bleeding events remains high in this population. Clearly, important issues linger in the maintenance of therapeutic anticoagulation post-BVR, resulting in a continued need for clarification of patient risk, as well as optimal processes of care.
The authors would like to acknowledge Erin LoFrese for her editorial contributions to this manuscript. Ms. LoFrese did not receive compensation for her contributions, apart from her employment at the institution where this study was conducted.