Association of Warfarin Therapy Duration After Bioprosthetic Aortic Valve Replacement With Risk of Mortality, Thromboembolic Complications, and Bleeding FREE

Biological prostheses are preferred to mechanical valves for aortic valve replacement (AVR) surgery in elderly patients older than 65 years because of shorter life expectancy and lack of a need to use anticoagulation treatment in the long term. Especially in these patients, the trade-off between thromboembolic complications due to the valve implant and bleeding events as adverse effects from anticoagulation therapy must be balanced. Nevertheless, appropriate duration of anticoagulation treatment postoperatively is yet to be established because the risk of complications when the treatment is discontinued is unknown. Current guidelines^1- 3 of anticoagulation treatment after bioprosthetic AVR surgery recommends 3 months of warfarin treatment. However, this recommendation is primarily based on results from 1 retrospective study⁴ with limited number of events hampered by the observational design as well as low power like all other studies within this scope.

In this nationwide study, we investigated whether discontinuation of warfarin treatment within prespecified periods after bioprosthetic AVR surgery was associated with increased risk of thromboembolic complications, cardiovascular death, and bleeding incidents during a period of 13 years.

All patients having bioprosthetic AVR with or without coronary artery bypass graft (CABG) surgery performed between January 1, 1997, and December 31, 2009, in Denmark were identified through a search in the Danish National Patient Registry. Patients with prior cardiac surgery or other concomitant surgical procedures were excluded. The Danish Data Protection Agency approved the use of personal data (reference 2007-58-0015, internal reference GEH-2010-001). Retrospective studies do not require ethical approval in Denmark.

Hospital diagnoses, procedures, medication, age, sex, and causes of death were retrieved from the National Danish Registries. Due to a law-imposed demand of reporting these data, a high degree of completion is ensured. The Danish National Patient Registry contains information on all hospital admissions and diagnoses in Denmark since 1978. The diagnoses are classified according to the International Classification of Diseases, 8th Revision (ICD-8) and International Statistical Classification of Diseases, 10th Revision (ICD-10) (ie, patients with stroke were identified by stroke diagnoses recorded in the National Hospital Registry as codes I61, I62, I63, and I64). Previously, these diagnoses have proven valid with positive predictive values of 74% to 97%.^5- 6 Similarly, to retrieve information on comorbidity, we identified prespecified discharge diagnoses (for ICD-8 and ICD-10 codes, see eTable 1) in the Danish National Patient Registry dating back 18 years preoperatively. Patients with diabetes were identified through their use of glucose-lowering medication.

Primary events studied included stroke, thromboembolic complications (ie, ischemic strokes, myocardial infarctions, and peripheral arterial emboli), bleeding incidents (ie, gastrointestinal, intracranial, urinary tract, and airway bleedings⁷), and cardiovascular death (eTable 2). The term cardiovascular death included all deaths, with a cardiovascular cause registered on the death certificates forming the basis of the Danish National Death Registry.

Use of medication was obtained from the Danish Registry of Medicinal Product Statistics where all prescriptions dispensed from Danish pharmacies since 1995 are recorded. Registration is complete due to linkage to reimbursement from the state, and medication from other sources is limited.⁸ The information for each medication coded according to the Anatomical Therapeutic Chemical (ATC) System included number and strength of tablets. Calculations regarding the daily warfarin dosage were restricted by a minimal dose of warfarin of 0.625 mg and a maximal dose of 17.5 mg. A patient starting warfarin therapy was assumed to start on a default dose of 5 mg. Whenever a new prescription was claimed, the average dose of up to 3 previous prescription periods was calculated. Those prior prescription periods, which could represent a continuous treatment with at least the minimal dose, were included in that calculation. To avoid conditioning on the future, later prescriptions were not included in calculations. Patients were assumed to continue with the dose below this dose that is a multiple of 0.625 mg (one quarter of a tablet). The minimal and maximal dose set limits to this calculation. Discontinuance of treatment was assumed when there were no tablets left according to the calculations. The value of these calculations are dependent on the calculations not being highly sensitive to assumptions. Sensitivity analyses were performed by using a minimal dose of 5 mg and a default dose of 7.5 mg, and by using an automatic lengthening of any treatment period of either 14 or 30 days.

Information obtained on other medication at the time of surgery required that each patient claimed at least 1 prescription on prespecified medication within 3 months before surgery.

The studied cohort originates from a fixed population defined by all individuals in Denmark alive and aged older than 18 years on or after January 1, 1997, and patients can only be lost to follow-up by emigrating. In the current study period, 3 patients (0.07%) emigrated and were censored at time of emigration. We determined the occurrence of events 30 to 89 days, 90 to 179 days, 180 to 364 days, 365 to 729 days, and at least 730 days after surgery. At these time points, analysis time was split and event rates for patients with and without warfarin treatment were calculated as the number of events divided by the sum of person-time. Patients were censored at event, emigration, or end of follow-up. Poisson regression models were used for survival type analysis to calculate incidence rate ratios as an indicator of relative risk (RR). With an exposure ratio of warfarin treatment vs nontreatment of 3.6, a risk of cardiovascular death of 7.5% among patients without warfarin and a type 1 error risk of .05, we obtained a power of 66% to detect a difference in RR of 1.3, 86% power for an RR of 1.4, and 96% for an RR of 1.5. Subgroup analyses were preplanned and intended to challenge the results by removing patients with a variety of profiles.

All P values reported were 2-sided and considered statistically significant if P < .05. Statistical analyses were performed by using SAS version 9.2 (SAS Institute) and Stata version 11.1 (StataCorp LP).

A total of 4075 patients were identified who had bioprosthetic AVR surgery performed after excluding patients in warfarin treatment before surgery (n = 684) and patients with a diagnosis of atrial fibrillation within 30 days after surgery (n = 1215). A total of 301 patients were both in warfarin treatment before surgery and had atrial fibrillation within 30 days after surgery. In addition, 8 patients discontinued warfarin treatment within the first 30 days after surgery. This period was omitted from the analyses ensuring all patients had an equal chance to claim a prescription for warfarin. A flow diagram of patients included in our study is shown in Figure 1. Mean age was 74.6 years (range, 18-95 years) and 1670 patients (41%) were women. Baseline characteristics by warfarin treatment are shown in Table 1.

To determine the importance of warfarin in antithrombotic treatment after bioprosthetic AVR surgery, we compared the effects of warfarin, aspirin, the combination of warfarin and aspirin, and no treatment on cardiovascular death after surgery (Figure 2 and eTable 3). Monotherapy with aspirin was used infrequently and rates of cardiovascular death did not differ significantly from rates among patients with no treatment; therefore, to provide comparative estimates for the effect of anticoagulation, the remaining analyses were made between warfarin vs no warfarin treatment independent of aspirin treatment.

Overall, 361 patients (8.9%) experienced a stroke, 615 (15.1%) had a thromboembolic event, and 364 (8.9%) encountered a bleeding incident after the date of surgery. During an observation period of 12 557 person-years (median duration of follow-up, 6.57 person-years), 1156 patients (28.4%) died, with 879 (76.0%) of these deaths being related to cardiovascular disease. Within 30 days after surgery, 109 patients (2.7%) experienced a stroke, 201 patients (5.0%) experienced thromboembolic events, and 75 patients (1.8%) had bleeding episodes. Thirty-day mortality was 6.0% and 1-year mortality was 10.9%.

The incidence rates of stroke, thromboembolism, bleeding incident, and cardiovascular death within prespecified periods after surgery with and without warfarin treatment are shown in Figure 3 (for overall mortality, see eFigure and eTable 4). The first postoperative month was omitted from these analyses ensuring all patients had an equal chance to claim a prescription. Table 2 shows the risk of strokes, thromboembolic events, cardiovascular deaths, and bleeding incidents over time unadjusted and adjusted for age, sex, concomitant CABG surgery, comorbidity, and calendar year.

Results from subgroup and sensitivity analyses are shown in eTable 5. Apart from the association found 180 to 364 days after surgery, changes in the estimated risks of cardiovascular death remained the same, except for the analysis excluding patients with concomitant CABG surgery.

Calculating the number needed to treat/number needed to harm within 90 to 180 days after surgery, we found that for every 23 (95% CI, 14-54) patients not being treated with warfarin, 1 patient died from cardiovascular cause; and for every 74 (95% CI, 27-95) patients being treated with warfarin, 1 patient experienced bleeding complications requiring hospital admission.

To exclude other factors explaining the increased risk of cardiovascular death 3 to 6 months after surgery other than discontinuing warfarin treatment, we reviewed diagnoses from hospital admissions within 1 month before cardiovascular death among patients with and without warfarin treatment dying within 3 to 6 months after surgery. We found no specific pattern in diagnoses explaining this finding. Similarly, we retrieved information on medication use within the first 6 months after surgery to investigate whether drugs prescribed could influence cardiovascular death 3 to 6 months after surgery and found no indication of changes in phamacotherapy.

This is to our knowledge the largest study examining thromboembolic complications with or without warfarin treatment after bioprosthetic AVR surgery. We demonstrated a clear benefit associated with warfarin during the initial 3 months after surgery, and furthermore our data suggest that extension of warfarin treatment to 6 months postoperatively may reduce the risk of cardiovascular death.

The guidelines regarding anticoagulation treatment after bioprosthetic AVR surgery^1- 3 suggesting duration of warfarin treatment of 3 months postoperatively were all primarily based on a single observational study⁴ from 1995, including 424 patients with an antithrombotic regimen involving warfarin, aspirin, dipyridamole alone, or in combination. This study⁴ found the rate of thromboembolism decreasing significantly over time after surgery from 41% per year (1-10 days postoperatively) over 3.6% per year (11-90 days postoperatively) to 1.9% per year (>90 days after surgery). Furthermore, the authors found that warfarin reduced the risk of thromboembolic complications significantly. Accordingly, due to high risk of thromboembolic events within 90 days after surgery, the authors concluded that anticoagulation was indicated 3 months postoperatively—a finding which was confirmed by other similar small studies.^9- 10

The aforementioned rate of 41% per year represented 5 thromboembolic events, with the rate of 41% being an extrapolation from the first 10 days assuming a constant thromboembolic risk throughout the following year. This assumption could be questioned because the release of microemboli during surgery¹¹ would induce an initial high risk of thromboembolic complications immediately after surgery, which would decline rapidly the following month. This was reflected by the decrease in the thromboembolic rate reported 11 to 90 days postoperatively of 3.6% per year, even though this figure in fact only represented 3 thromboembolic events.

We found similar trends concerning thromboembolic events with initially higher rates decreasing the following year after surgery. However, our results were based on 615 thromboembolic events during the study period, with 40 thromboembolic events occurring within 30 to 89 days after surgery. Furthermore, we studied the effect of warfarin treatment on cardiovascular death, which has not previously been the subject of investigation.

In recent years, the benefit of treatment with anticoagulation after bioprosthetic AVR surgery has been challenged. An observational study¹² analyzing data on 1151 patients undergoing bioprosthetic AVR surgery with or without CABG surgery found an incidence of cerebral vascular accidents of 2.4% among patients treated with warfarin compared with 1.9% among patients with no anticoagulation. The authors concluded that warfarin had no protective effect on neurological events but 23% of the patients were treated with warfarin preoperatively and, therefore, could be susceptible to stroke due to atrial fibrillation. A retrospective study¹³ of 500 patients with pericardial bioprostheses having 48 thromboembolic events found an increased risk of thromboembolism among patients treated with warfarin therapy compared with patients with no treatment, with a risk ratio of 3.0 after a 4-year follow-up period. However, this study population included patients with concomitant surgery on the ascending aorta and mitral annuloplasty, making findings difficult to interpret. Our study focused solely on patients with bioprosthetic AVR surgery and included a much larger population. Nevertheless, we did observe a minor increase in rates of thromboembolic events more than 1 year after surgery with warfarin therapy being associated with a worse outcome.

An observational study¹⁴ on 861 patients undergoing isolated bioprosthetic AVR surgery sought to determine whether aspirin or warfarin treatment was protective on the risk of thromboembolic complications during the first 90 days after surgery in an institution where antithrombotic treatment was administered according to the discretion of surgeons. The authors from that study¹⁴ concluded that early anticoagulation only seemed to reduce thromboembolic risk in high-risk groups (ie, women, highly symptomatic patients, and patients with small aortic prostheses). However, they did not attempt to address confounding by silent postoperative atrial fibrillation. Aspirin was used infrequently as monotherapy in our patients and consequently our study does not contribute valuable data for the use of aspirin as an alternative to warfarin.

Some guidelines recommend aspirin as an alternative to postsurgical warfarin therapy in patients without risk factors.^1- 2,15 However, these recommendations are widely based on consensus opinion of experts, case studies, standard of care studies, or small prospective studies with insufficient power to substantiate findings. No prospective study has yet proved the safety of omitting antithrombotic treatment in patients with bioprosthetic AVR surgery in the first 3 months after surgery. Our study is no exception. No conclusions could be drawn due to paucity in patients receiving no treatment.

However, few observational studies have been centered on whether antiplatelet therapy is a sufficient protection against thromboembolic complications. For instance, 1 study¹⁶ analyzing 251 patients undergoing bioprosthetic AVR surgery, where treatment was determined by surgeons adherence with guidelines, compared warfarin treatment with aspirin in protecting against neurological events. The authors found no advantages of early anticoagulation but mean age and mean EUROSCORE were significantly higher in patients treated with warfarin. A prospective study¹⁷ assessing the efficacy of ticlopidine on thromboembolism included 235 patients with valve repair or bioprosthetic valve replacement in aortic, mitral, or tricuspid position. The authors concluded that ticlopidine seemed to prevent thromboembolism better than other therapy based on 2 and 4 thromboembolic events, respectively. Due to the relatively rare presentation of thromboembolic events and the devastating effects of this complication, large randomized studies comparing warfarin with antiplatelet therapy remains to be seen.^18- 19

In addition, in our study a significant association was found between bleeding incidents and discontinuing warfarin 30 to 89 days postoperatively. A possible explanation may be that initiation of warfarin can be challenging to obtain the recommended international normalized ratio in some patients. Furthermore, it is not known whether these patients discontinued the warfarin treatment due to bleeding incidents or discontinued warfarin treatment and had a bleeding incident afterwards. However, considering the calculated number needed to treat/number needed to harm analysis previously shown, the association of bleeding complications with warfarin use seem to be less worrisome as opposed to a possible benefit from the association between warfarin use and decreased cardiovascular death.

With no randomized trials to guide the length of warfarin treatment, our results call for a review of guidelines in the field to consider an extension of the treatment to 6 months after surgery, especially in patients with an increased risk of cardiovascular death. Increased follow-up procedures may be instituted to monitor, especially in the oldest patients, bleeding complications, although the increased risk of bleeding found among patients treated with warfarin from 90 to 179 days was nonsignificant.^20- 26

We did not have access to information regarding the international normalized ratio. Thus, we were not able to assess time in therapeutic range, which most likely is more influential on the risk of thromboembolic events, and especially bleeding events, than prescription of warfarin alone.²⁷ Similarly, we were not able to monitor adherence to treatment assignment. The fact that we had no data on the reasons for omitting, discontinuing, or extending warfarin treatment may introduce hidden confounders, including confounding by indication. However, warfarin discontinuation according to our estimation of use occurred when the patient had no tablets left. Thus, for event driven discontinuation to occur, the patient should run out of tablets simultaneously with the occurrence of an event; in which case, the event could be attributed to warfarin treatment inducing an underestimation of warfarin effect on thromboembolic events and possibly an overestimation of the effect on bleeding events. However, we may have underestimated the risk of bleeding in warfarin treatment due to lack of information on bleeding episodes not resulting in hospitalization.

The main limitation of the databases is the lack of detailed clinical information related to important risk factors (eg, smoking), which could influence outcome. Although this is accounted for by including comorbidity, we cannot exclude the influence of some degree of residual confounding. The prescription register contains only data on prescribed drugs subject to reimbursement, making it vulnerable to changes in reimbursement policies. Drugs sold over-the-counter are not registered, introducing a problem regarding aspirin. However, the financial incentive for chronic patients to have aspirin prescribed and be reimbursed instead of buying the drug over-the-counter presumably reduces the sale of aspirin over-the-counter. Furthermore, reliance on administrative databases induces a risk of time-varying measurement errors of exposure as well as biases from survival effects.

Our study demonstrates that discontinuing warfarin therapy within the first 3 months after surgery is associated with a significant increase in the risk of stroke, thromboembolic complications, and cardiovascular death. The novelty of our study is the finding that discontinuing warfarin therapy within 90 to 179 days after surgery is associated with a significant increase in the risk of cardiovascular death.

International guidelines on anticoagulation after a bioprosthetic value AVR have been written with limited data on the appropriate duration of warfarin treatment after surgery. Consequently, our study challenges current guidelines on the duration of antithrombotic treatment after AVR surgery with biological valves by presenting results suggesting that these patients will gain from an additional 3 months of warfarin treatment in terms of reduced cardiovascular death without risking a significant increase in bleeding events.