Prolonged Thromboprophylaxis With Oral Anticoagulants After Total Hip Arthroplasty:  A Prospective Controlled Randomized Study FREE

DESPITE ADEQUATE antithrombotic prophylaxis, more than 15% of patients who undergo total hip arthroplasty still develop (mostly asymptomatic) postoperative deep venous thrombosis (DVT) of the legs, as shown by bilateral ascending phlebography performed at the time of hospital discharge.¹ Although most of these thrombi are confined to the calf vein system, approximately 10% involve the proximal veins and therefore carry the potential risk of later complications.²

Concern about the risk of pulmonary embolism (PE) developing as a result of asymptomatic DVT has led investigators and clinicians to consider extending the duration of postoperative prophylaxis for up to 6 weeks after hospital discharge in all patients who undergo hip arthroplasty.^3- 9 Despite minor variations in design and results, the findings of these randomized studies are consistent in showing that prolonged thromboprophylaxis with unfractionated⁹ or low-molecular-weight heparin (LMWH)^3- 8 decreases the frequency of venographically detected DVT by approximately 50% without enhancing the hemorrhagic risk.

Despite the increasing availability of LMWHs, oral anticoagulants are still widely used for the prevention of postoperative DVT in patients who have undergone major orthopedic surgery.¹ However, unlike LMWHs, which have been studied extensively, oral anticoagulants have not been the subject of controlled studies to assess the benefit of prolonging anticoagulation beyond the hospital stay. We therefore undertook a prospective controlled study of patients who received warfarin sodium prophylaxis after undergoing total hip arthroplasty. Patients free from proximal vein thrombosis, as assessed by an ultrasound examination at the time of hospital discharge, were randomized to discontinue taking the drug or to continue taking it for 4 more weeks. The main study outcome was to compare the rate of symptomatic and asymptomatic thromboembolic events that occurred during this period. Also, we tried to identify clinical parameters associated with the development of late postoperative thromboembolism. The study was prematurely terminated after the inclusion of 360 patients because a statistically significant and clinically relevant superiority of extended thromboprophylaxis over short-term thromboprophylaxis was observed.

This trial was conducted at the University Hospital of Padua, Padua, Italy, from September 1998 to December 2000. The research protocol was approved by the local ethics board.

Consecutive patients who underwent elective total hip arthroplasty and received warfarin prophylaxis during hospitalization were potentially eligible for the study provided they had not undergone previous hip surgery on the same side or did not have a history of thromboembolic disorders. Eligible patients received 5 mg/d of sodium warfarin starting on the second preoperative day; after the intervention, the dosage was adjusted to increase the international normalized ratio (INR) between 2.0 and 3.0. Eligible patients were excluded from the study if they developed venous thromboembolic complications or major bleeding during hospitalization. Patients with asymptomatic proximal DVT, as shown by a bilateral compression ultrasound (CUS) examination performed before hospital discharge, were also excluded, as were those who needed long-term anticoagulation, were unavailable for long-term follow-up, or refused to give their written informed consent.

Compression ultrasound examination was performed with a high-resolution echo-color Doppler ultrasound system (Acuson XP 128; Ambassador Medical Inc, Carmel, Ind) with 7.5- and 5-MHz probes. Of the proximal deep venous system, the common femoral vein, the superficial femoral vein at the mid thigh, and the popliteal vein up to its trifurcation into the calf veins were evaluated for compressibility. The results of CUS were classified as diagnostic of proximal vein thrombosis in case of vein incompressibility. In patients with positive test results, anticoagulant treatment was continued for 3 months.

A careful history detailing the presence of risk factors for DVT was obtained from all consenting patients with a normal CUS test result at baseline; then, according to a list generated by a computer, the patients were randomly assigned to discontinue oral anticoagulant therapy or to continue it for 4 more weeks, with the dosage adjusted to maintain the INR between 2.0 and 3.0.

After randomization, the patients were discharged and were invited to return for clinical and ultrasound examinations after 1, 2, and 4 weeks. During each follow-up examination, bilateral CUS was performed and the results were interpreted using the criteria that had been adopted at baseline by an operator who was totally blind to the patients' details and study arm. After discharge, the patients who were randomized to the extended-therapy group were encouraged to have their INR blood testing performed once a week, or more frequently in case of inadequate anticoagulation, at the local anticoagulation surveillance center, where dedicated and trained physicians, with the help of computer-assisted programs, could provide a high standard of care. After the completion of the first 4 weeks of therapy, the patients were clinically followed up for 2 more months.

The patients were asked to present immediately for medical attention if signs or symptoms of venous thromboembolism developed. Venous thromboembolism was considered present if (1) symptomatic DVT was confirmed by new abnormalities observed on CUS or by an intraluminal filling defect on ascending phlebography; (2) symptomatic PE was confirmed by a high-probability ventilation-perfusion lung scan, a spiral computed tomographic scan, or an abnormal finding on angiography; or (3) (in case of death) PE was confirmed by autopsy or could not be ruled out. Any major bleeding event occurring in the study patients was recorded. Bleeding was defined as major if it (1) was clinically overt and associated with either a decrease in hemoglobin of at least 2.0 g/dL or a need for a transfusion of 2 or more units of red blood cells; (2) was intracranial or retroperitoneal; or (3) resulted in the permanent discontinuation of anticoagulation. In case of death, the cause of death was either investigated by autopsy or adjudicated according to the opinion of a physician who was unaware of the study aims. Information on all suspected (both asymptomatic and symptomatic) outcome events was reviewed and classified by an independent adjudication committee whose members were unaware of the treatment assignment.

The main aim of the study was to compare the efficacy of the 2 treatment strategies regarding a composite outcome of symptomatic venous thromboembolic complications and asymptomatic proximal DVT occurring during the first 4 weeks of follow-up. Furthermore, we determined the efficacy of the 2 treatment strategies during the complete 3-month follow-up period. Finally, we assessed the association of a number of clinical parameters (eg, age, sex, obesity, prolonged immobilization, length of hospital stay, quality of in-hospital anticoagulation, and presence of varicose veins, cancer, heart/lung failure, and estrogen therapy) with the development of late postoperative thromboembolism.

Based on the results of a previous study that suggested that withholding extended thromboprophylaxis from patients who received warfarin treatment during hospitalization was safe provided they had negative results on bilateral ultrasound examination before discharge,¹⁰ the present study was designed to demonstrate the equivalence between the 2 study groups. We calculated that approximately 600 patients per group would be required to demonstrate equivalence between the 2 study groups. The study was prematurely terminated by the steering committee after the inclusion of the first 360 patients because of the finding of an unexpected statistically significant and clinically relevant superiority of extended thromboprophylaxis over short-term thromboprophylaxis.

The efficacy analysis was intention-to-treat and included all patients who were randomly assigned to either strategy. We determined the proportion of patients who developed DVT in each group, as well as the absolute difference between proportions and its 95% confidence intervals (CIs). We calculated the relative risk (RR) for the development of thromboembolic complications (and its 95% CIs) by dividing the incidence rate in the control group by the incidence rate in the warfarin group. The RR was considered to be statistically significant when the lower limit of the 95% CI was greater than 1.0. Also, we calculated the number needed to treat to prevent 1 thromboembolic event. Odds ratios (ORs) and their 95% CIs were used to describe the association between clinical variables and postoperative thromboembolism. An OR was considered to be statistically significant when the lower limit of the 95% CI was greater than 1.0. For the comparison of patient characteristics at the time of study entry, we used the χ² test for qualitative variables and the t test for quantitative variables. The 95% CIs and P values were calculated according to the normal approximation of the binomial distribution. The individual quality of warfarin anticoagulation was considered high if the INR was within or above the therapeutic range in more than 70% of determinations.

Over the 27-month recruitment period, 383 eligible patients underwent hip arthroplasty at St Anthony Hospital, Padua. Eighteen patients were excluded from enrollment because of the development of venous thromboembolic complications² or major bleeding¹ during hospitalization, detection of an asymptomatic proximal DVT before hospital discharge,¹⁰ need for long-term anticoagulation,¹ and unavailability for long-term follow-up.⁴ Of the remaining 365 patients, 5 refused to give their informed consent. Therefore, 360 patients were randomized to discontinue anticoagulant therapy (n = 176) or to continue it for 4 more weeks (n = 184). The main demographic and clinical characteristics of the study patients are shown in Table 1. No appreciable differences were observed between the warfarin group and the control group.

No patient was unavailable for clinical follow-up. Three patients per group violated the protocol. In the control group, 2 patients resumed anticoagulant therapy (because of a reintervention and the identification of atrial fibrillation, respectively), and the third refused to undergo the serial ultrasound examinations. In the oral anticoagulant group, 2 patients stopped taking warfarin after a short period (one of them without valid reasons, and the other because of a reintervention), and the third patient did not undergo the serial ultrasound examinations. In the group of patients who were randomized to continue warfarin treatment, the quality of oral anticoagulation was satisfactory, being high in 131 (72%) of the 182 patients who completed the 4-week period of extended prophylaxis.

During the first 4 weeks of follow-up, objectively confirmed venous thromboembolic complications were recorded in 10 patients: 9 (5.1%) of the 176 control patients, and 1 (0.5%) of the 184 patients who continued the warfarin treatment. The absolute difference in the incidence of events was 4.57% (95% CI, 1.15-7.99). The RR of venous thromboembolism developing in control patients compared with patients assigned to extended thromboprophylaxis was 9.4 (95% CI, 1.2-73.5). The number needed to treat was 22. Of the 9 events observed in the control group, 4 were symptomatic (proximal DVT in 3 patients, and PE in 1 patient), while the remaining 5 were asymptomatic (Table 2). The only event observed in the warfarin group was an asymptomatic proximal DVT recorded after 4 weeks.

In the following 2 months, 2 more symptomatic events were recorded, both of them in patients who had received prolonged thromboprophylaxis. The RR of objectively confirmed venous thromboembolism developing in control patients compared with patients assigned to extended thromboprophylaxis during the whole 3-month follow-up period was still substantial, albeit nonsignificant (RR, 3.1; 95% CI, 0.9-11.4).

Of the 184 patients randomized to extended prophylaxis, 1 developed major (retroperitoneal) bleeding after 3 weeks of warfarin therapy. The INR at the time of the bleeding exceeded by far the upper limit of the targeted range (INR, 5.9). Therefore, the frequency of major bleeding in the group of patients who were randomized to prolonged prophylaxis was 0.5% (95% CI, 0.02-3.0). No control patients developed hemorrhagic disorders during the study period.

No patient died during the follow-up period or experienced any other adverse event potentially related to the study drug.

Of the tested parameters, only old age (8 of the 9 patients who developed venous thromboembolism compared with 77 of the 167 patients who did not [OR, 9.35; 95% CI, 1.14-76.4]) and varicose veins (4 of the 9 patients who developed venous thromboembolism compared with 10 of the 167 patients who did not [OR, 12.6; 95% CI, 2.9 -54.2]) were found to be associated with the development of late postoperative thromboembolism in the control patients, whereas the others were not.

The optimal duration of thromboprophylaxis after major orthopedic surgery is controversial.¹ Although recent clinical trials have consistently demonstrated that, at least in hip replacement surgery, prolonging prophylaxis with LMWH for a few more weeks after hospital discharge considerably reduces the incidence of venographically detected DVT,^3- 9 several cohort and randomized studies have shown that the rate of late symptomatic venous thromboembolic complications in patients who receive proper LMWH prophylaxis during hospital stay after major orthopedic surgery is acceptably low.^11- 15 It should be noted, however, that according to the results of 2 recent meta-analyses of comparative studies, prolonging thromboprophylaxis is also likely to lower the rate of late symptomatic clinical events.^16- 17 As this strategy does not appreciably enhance the risk of major bleeding,^3- 9 it has the potential to offer a better perspective than prophylaxis administered only during hospital stay.

Despite the growing availability of LMWHs, oral anticoagulants are still widely used for prophylaxis of postoperative DVT in patients who have undergone major orthopedic surgery.^1,18 Although they have been reported to be less effective than LMWHs for the prevention of DVT as detected by venography,^7,19- 22 no difference has been observed between the 2 therapeutic strategies in terms of clinically symptomatic end points in either the hospital stay or the long-term follow-up after hospitalization.^7,13,19- 23 Also, oral anticoagulants are associated with a lower hemorrhagic risk^7,13,19- 23 and are likely to be more cost-effective than LMWHs for this indication, at least in the United States.²⁴ Finally, in a recent retrospective case-control survey of a wide series of patients drawn from the California Medicare records, the use of warfarin after discharge from the hospital was found to be independently associated with a significant protection against rehospitalization for symptomatic thromboembolism after total hip surgery.²⁵ Unlike LMWHs, which have been studied extensively, however, oral anticoagulants have not been the subject of controlled studies assessing the benefit of prolonging anticoagulation beyond the hospital stay.

The results of our prospective randomized study strongly suggest that in patients who undergo hip surgery and who are treated with warfarin during hospitalization (9 days on average), prolonging the administration of the drug for 4 weeks beyond hospital discharge significantly reduces the incidence of late venous thromboembolic complications without enhancing the hemorrhagic risk compared with prophylaxis that is administered only during the hospital stay. Although patients with asymptomatic proximal DVT, as assessed by ultrasonography before discharge, were excluded from randomization, the rate of venous thromboembolism in the control group during the first 4 weeks (5.1%) was approximately 10 times as high as that observed in the warfarin group (0.5%), accounting for a statistically significant RR (9.4; 95% CI, 1.2-73.5). According to these results, only 22 patients need to be treated to prevent 1 episode of venous thromboembolism. It is worthy of mention that of the 9 observed events in the control group, 4 were symptomatic (including an episode of PE), whereas only 1 episode of asymptomatic DVT was observed in the warfarin group. Also, only 1 patient in the warfarin group (0.5%) experienced major bleeding.

Because the clinical follow-up evaluation after the first 4 weeks of observation disclosed 2 additional episodes of symptomatic thromboembolism, both of them in the group of patients who had received 1 month of warfarin therapy (Table 2), we cannot exclude the possibility that anticoagulant therapy for an even longer period (for 8-12 weeks after surgery) might further improve the clinical outcome of patients after hip surgery. The benefit-risk ratio of such an extensive regimen of prophylaxis should be properly investigated.

The results of the present study are apparently in striking contrast to those recently demonstrated in a cohort of patients who were followed up prospectively at our institution after receiving a warfarin thromboprophylaxis during their hospital stay.¹⁰ After excluding from follow-up patients with a negative ultrasound test result before discharge, we recorded only 2 asymptomatic cases of proximal DVT in a series of 193 (1.0%) consecutive patients who underwent hip surgery and who were treated with warfarin during hospitalization. It should be noted, however, that the length of hospital stay in the current investigation (9 days on average) was considerably shorter than that observed in our previous study (18 days on average). The unusually long hospital stay and duration of anticoagulation in the patients who were enrolled in our previous study most likely accounted for the low rate of subsequent cases of thromboembolism, a finding that is fully consistent with that observed in the patients who were randomized to the extended-prophylaxis group in the current investigation.

The low rate of venous thromboembolism in patients who were randomized to the warfarin group is at variance with that observed by Caprini et al²⁶ in their recent cohort study. Of 125 patients who underwent a total hip replacement and received warfarin therapy until 1 month after surgery, more than 10% developed proximal DVT (which was mostly asymptomatic), as assessed by duplex ultrasonography. However, in their study, a remarkably high proportion of patients failed to reach a therapeutic INR, which is in contrast to our results. Therefore, effective monitoring and dosage adjustment of warfarin therapy constitute an important component of the success of prolonged warfarin prophylaxis in major orthopedic surgery.

For the purpose of our study, we used a composite outcome of symptomatic and asymptomatic venous thromboembolic complications. The introduction in the last 2 decades of the use of ascending phlebography as a diagnostic tool for the detection of postoperative DVT in high-risk surgical patients has led to the demonstration of (asymptomatic) thrombosis in up to 15% of patients who undergo total hip arthroplasty.¹ Most of those thrombi probably carry a low risk of complications, as several clinical studies have consistently reported an acceptably low rate of symptomatic thromboembolic events in the follow-up of patients who received their antithrombotic protection only during hospitalization.^11- 15 If this is likely to be true for isolated calf vein thrombi, which form the large majority of postoperative thromboses, we suspect that this is not the case for proximal vein thrombosis, which was the target of our investigation. We thought it important to document the asymptomatic involvement of the proximal vein system, because in patients with previously unaffected deep veins (as in our investigation) the involvement of the popliteal and particularly the common femoral veins is an essential requirement for the development of more serious complications, such as fatal or nonfatal PE. It is interesting to note that in the group of patients who were treated with prolonged warfarin therapy, in whom no symptomatic events developed during the 4-week follow-up period, only 1 case of asymptomatic proximal DVT was recorded, which contrasts with findings observed in the control group (4 symptomatic events and 5 cases of asymptomatic proximal DVT). We believe that every future study that addresses the prevention of DVT in high-risk surgical patients should rely on either systematic bilateral phlebography or a combination of clinically symptomatic events and bilateral assessment of the proximal vein system.

Despite the early termination of the study and the relatively low number of patients who were enrolled in the current investigation, old age and varicose veins were found to be significantly associated with the development of late venous thromboembolism. These findings, along with those detected by White et al²⁵ (ie, obesity and previous thromboembolism) in their retrospective study of a large number of patients, may help identify patients who could benefit from prolonged anticoagulation to reduce the incidence of thromboembolism after hip arthroplasty.

We are aware of at least 4 potential limitations of our study. First, our study, as well as others involving oral anticoagulants, was not a placebo-controlled double-blind trial. However, our findings are likely to be valid because a number of measures were taken to avoid bias: (1) inclusion of consecutive patients; (2) follow-up of all randomized patients; (3) central adjudication of all outcome events by a committee whose members were unaware of the treatment assigned; (4) assessment of venous thromboembolism and episodes of bleedings on the basis of objective predetermined criteria; and (5) inclusion of all randomized patients in the study analysis. Second, the sensitivity of CUS for detection of postoperative symptomless DVT is generally considered low.^27- 28 The use of this technique might have resulted in the recruitment of patients with (asymptomatic) DVT not detected by CUS, as well as in the underestimation of thrombotic episodes in the 4 weeks of follow-up in each patient group. However, the sensitivity of this diagnostic technique for the development of thrombosis involving the proximal vein system in orthopedic patients has been found to be acceptably high.^27,29 Also, as we used the same follow-up procedure in both patient groups, the RR of developing thrombotic complications (as predefined by us) in the control group compared with the warfarin group is likely to be valid. Third, the complete identification of clinical parameters associated with the development of late postoperative thromboembolism after total hip replacement was hindered by the early termination of the study and the relatively small number of patients available for this potentially important calculation. Fourth, as we did not enroll patients who were candidates for major knee surgery, we do not know whether our conclusions would also be applicable to this category of patients, who might differ in this regard from patients who undergo hip surgery.^8,18

In conclusion, extending prophylaxis with warfarin for a few weeks beyond the hospital stay has the potential to considerably improve the outcome of patients who undergo hip arthroplasty. Further studies are needed to better identify predictors of late postoperative thromboembolism after major orthopedic surgery, which might enable physicians to target those patients who would benefit the most from extended prophylaxis.

Accepted for publication January 31, 2002.