Concern about the potential risk for pulmonary embolism as a result of asymptomatic deep venous thrombosis has led clinicians to consider extending the duration of postoperative prophylaxis for up to 3 months or performing routine screening tests to diagnose asymptomatic venous thrombosis before hospital discharge [1, 4-7]. Compression ultrasonography is the most accurate noninvasive test for diagnosing proximal deep venous thrombosis [8]. Although compression ultrasonography is less sensitive than venography as a screening test for asymptomatic deep venous thrombosis, most of the proximal thrombi not detected by compression ultrasonography are very small and are of uncertain clinical importance [9, 10].

We performed a double-blind, randomized, controlled trial in patients receiving warfarin prophylaxis after total hip or knee arthroplasty to determine whether performing screening compression ultrasonography to diagnose and guide the treatment of asymptomatic proximal deep venous thrombosis influenced rates of subsequent symptomatic deep venous thrombosis or pulmonary embolism. We also determined the rates of symptomatic venous thromboembolic and major bleeding complications occurring within 90 days of hospital discharge.

Methods

Top Methods Results Discussion Author & Article Info References

This trial was conducted at the Victoria General Hospital, Halifax, Nova Scotia, between July 1993 and December 1995. The research protocol was approved by the hospital's research review committee. Informed consent was obtained from all patients.

Study Sample

Consecutive patients undergoing elective total hip or knee arthroplasty who received postoperative warfarin prophylaxis and were expected to be discharged from the hospital within 48 to 72 hours were eligible for the study. Exclusion criteria were 1) investigations for symptoms of deep venous thrombosis or pulmonary embolism during warfarin prophylaxis, 2) hospitalization for more than 30 days after surgery, 3) previous venous thromboembolic disease, 4) contraindications to anticoagulant therapy, 5) need for long-term anticoagulant therapy, 6) acute hip fracture as the indication for arthroplasty, 7) use of an anticoagulant other than warfarin for more than 24 hours after surgery, 8) lack of administration of warfarin within 72 hours after surgery, and 9) previous enrollment in the study. Within 7 months of the start of recruitment (after 265 patients had been enrolled), metastatic cancer was also made an exclusion criterion to ensure that only patients undergoing elective surgery were included.

Prophylaxis and Screening for Deep Venous Thrombosis

Patients received 5 mg of sodium warfarin (Coumadin, Dupont Pharma, Mississauga, Ontario, Canada) on the night of surgery. The warfarin dose was adjusted to increase the international normalized ratio (INR) to more than 1.7 by postoperative day 4 and to maintain it between 1.8 and 2.5 thereafter. Warfarin prophylaxis was continued until hospital discharge. All patients were evaluated daily, as per usual clinical practice by the orthopedic medical staff.

Consenting eligible patients were randomly assigned to undergo either bilateral screening compression ultrasonography or a sham ultrasonography procedure. The randomization scheme was determined by a computer algorithm. Assignments to the study groups were concealed in sealed envelopes that were opened by an ultrasonographic technologist only after consenting patients arrived in the ultrasonography department. Patients were stratified by type of arthroplasty. Compression ultrasonography was done by using a high-resolution color duplex-doppler scanner with an electronically focused linear array transducer and either 5- or 7.5-MHz probes. The entire proximal deep venous system between the proximal common femoral vein and the trifurcation of the popliteal vein in the calf was evaluated for compressibility at 1-cm intervals. The results of compression ultrasonography were considered to be positive if a vein or venous segment was not fully compressible [11]. Patients with positive results underwent confirmatory venography of the involved leg, which was done by using the method of Rabinov and Paulin [12]. If venography could not be done or if the results of venography were judged to be inadequate because of nonvisualization of the suspected area of thrombosis, the result of screening compression ultrasonography was considered to be true positive. Patients who received a diagnosis of asymptomatic proximal deep venous thrombosis were treated with unfractionated intravenous heparin for 5 days and had warfarin therapy continued for 3 months with the dose adjusted to maintain the INR between 2.0 and 3.0.

The sham ultrasonography procedure was designed to mimic the technique of the genuine compression ultrasonography. For both the genuine and the sham procedures, the ultrasonography screen was shielded from patients. Thus, patients, nurses, and physicians did not know whether a patient had had a genuine or a sham procedure. For sham procedures, the ultrasonography screen was kept blank; thus, neither the ultrasonographic technologists nor the radiologists knew whether deep venous thrombosis was present. A report of the results of screening ultrasonography was included in each patient's medical record. If the result was positive, it was reported as such. All other reports (negative results of genuine procedures and "results" of sham procedures) were reported as "an ultrasound was performed in accordance with the Post-Arthroplasty Screening Study protocol." In these cases, the report did not include any interpretation and did not indicate whether the patient had undergone a genuine or a sham procedure. Thus, physicians were blinded to study group assignment unless the results of ultrasonography were positive.

All patients were given an information sheet outlining the symptoms of deep venous thrombosis and pulmonary embolism, and they were asked to present immediately for medical attention if these developed. Patients were seen routinely in a follow-up clinic or were contacted by telephone 42 and 90 days after randomization.

Outcome Measures

The primary outcome cluster in this study was defined a priori as the development of one of the following after randomization: symptomatic deep venous thrombosis involving the proximal venous system, confirmed by compression ultrasonography or venography; symptomatic pulmonary embolism, confirmed by ventilation-perfusion lung scanning, pulmonary angiography, or autopsy; or major bleeding after the institution of anticoagulant therapy for patients in whom asymptomatic deep venous thrombosis was diagnosed by screening compression ultrasonography.

Proximal deep venous thrombosis in symptomatic patients was diagnosed by using either compression ultrasonography or venography, as described above [11, 12]. Pulmonary embolism was diagnosed in symptomatic patients on the basis of the following test results: a high-probability ventilation-perfusion lung scan [13], an intraluminal filling defect on more than one view on pulmonary angiography [13], or a non-high-probability ventilation-perfusion lung scan and concomitant deep venous thrombosis shown on compression ultrasonography or venography. Bleeding complications were classified by using previously described criteria [14]. In brief, bleeding was classified as major with major morbidity; major, requiring acute medical or surgical intervention; moderate; minor; or not significant.

Compression ultrasonograms, venograms, ventilation-perfusion lung scans, and pulmonary angiograms from all suspected screening and symptomatic thromboembolic events and from all bleeding episodes and deaths were reviewed by a panel of experts who were blinded to the study group of the patient.

Follow-up of Excluded Patients

To maximize the generalizability of our study findings, we followed several subgroups of excluded patients who underwent elective joint arthroplasty for 90 days after hospital discharge for the development of symptomatic venous thromboembolism. Patients were excluded from this secondary analysis if they had metastatic cancer, had undergone joint arthroplasty within 3 months of a hip fracture, or refused to give informed consent.

Sample Size and Statistical Analysis

In designing this study, we projected that the primary outcome event rates in the screening ultrasonography and the sham ultrasonography groups would be 2.5% and 6.7%, respectively. On the basis of these event rates, our target sample size was 338 patients per group (which gave us an 80% power with an error of 5%) to detect this 4.2% absolute reduction. A blinded interim analysis done after 549 patients had been enrolled showed that our overall event rate was lower than we had originally projected; therefore, we increased the target sample size to 1000 patients total.

The principal analysis was a comparison of the rates of primary outcome clusters in the two study groups. The difference in these absolute rates, as well as the 95% CIs surrounding this difference, were calculated. Proportions were calculated directly from our data; the CIs were derived by using exact binomial probabilities.

Results

Top Methods Results Discussion Author & Article Info References

Study Sample

Over the 27-month recruitment period, 1284 patients underwent total hip or knee arthroplasty at our institution (Figure 1). Three postoperative deaths occurred before patients were screened for study eligibility. A 72-year-old man with a history of ischemic heart disease died suddenly on the evening of his surgical procedure; no autopsy was done. A second patient died of aspiration pneumonia, and a third died of gangrenous cholecystitis. One hundred ninety-one patients (14.9%) met at least one exclusion criterion and were ineligible for the study: Seventy-six had a history of documented venous thromboembolism, 34 were investigated for symptomatic venous thromboembolism before hospital discharge, 32 underwent joint replacement for acute fracture, 12 had metastatic cancer, 11 required long-term anticoagulant therapy, 9 had a history of bleeding diathesis, 8 received anticoagulants other than warfarin in the postoperative period, 6 were hospitalized for more than 30 days, and 3 had no warfarin ordered for more than 3 days after surgery. Three of the patients who were excluded because they received alternate anticoagulant therapy in the postoperative period died of complications of ischemic heart disease.

View larger version (23K): [in this window] [in a new window]   Figure 1. Management allocations and outcomes of potentially eligible patients.

Thus, 1090 patients met the eligibility criteria. Of these, 31 patients refused to give informed consent, 6 were confused and unable to give informed consent, 26 were discharged unexpectedly before randomization, and 3 could not be enrolled because of scheduling difficulties in the ultrasound department. We therefore enrolled 1024 (93.9%) of the 1090 eligible patients in the study; of these, 508 (49.4%) underwent total hip arthroplasty and 518 underwent total knee arthroplasty. Five hundred eighteen patients underwent screening compression ultrasonography, and 506 patients underwent the sham procedure. The two groups were similar with respect to age, sex, type of anesthetic used, whether surgery was primary or revision surgery, blood loss, duration of surgery, use of compression stockings, level of ambulation before discharge, and length of hospital stay (Table 1).

Of the 518 patients who underwent screening compression ultrasonography, 19 (3.7%) had a positive result. Six of the 19 had proximal deep venous thrombosis excluded by venography. Of the remaining 13 patients, proximal deep venous thrombosis was confirmed by venography in 8, contraindications to venography existed in 2, and venography was inadequate for interpretation in 3. These 13 patients received heparin for 5 days and warfarin for 3 months.

Outcomes in Randomly Assigned Patients

No patient was lost to follow-up. Of the 518 patients randomly assigned to receive screening compression ultrasonography, 4 (0.8%) developed symptomatic proximal deep venous thrombosis on postoperative days 7, 14, 39, and 64 (Table 2). All 4 had had normal results on screening compression ultrasonography before hospital discharge. One patient who received treatment for asymptomatic proximal deep venous thrombosis that was diagnosed before discharge by screening compression ultrasonography had a major bleeding event during treatment. Thus, the overall primary outcome cluster event rate in the screening compression ultrasonography group was 1.0% (95% CI, 0.3% to 2.2%) (5 of 518 patients).

Of the 506 patients randomly assigned to the sham ultrasonography group, 3 developed symptomatic deep venous thrombosis on postoperative days 13, 24, and 35 and 2 patients had nonfatal symptomatic pulmonary embolism on postoperative days 19 and 41, for a total primary outcome cluster event rate of 1.0% (CI, 0.3% to 2.2%) (5 of 506 patients). The observed difference in the primary outcome cluster between the screening compression ultrasonography group and the sham ultrasonography group was 0 percentage points (CI, –1.2 to 1.2 percentage points).

Two deaths occurred during the 90-day follow-up period; both were in the screening compression ultrasonography group. A 59-year-old woman with metastatic breast cancer died of complications of cancer 38 days after undergoing total hip arthroplasty. A 77-year-old woman with a history of ischemic heart disease had an unwitnessed death 36 days after undergoing total knee arthroplasty; no autopsy was performed.

Bleeding Complications

A total of 16 bleeding complications (1.6% [CI, 0.8% to 2.4%]), including two major bleeding episodes (0.2% [CI, 0% to 0.5%]), occurred within the total group of 1024 randomly assigned patients. Only five of these bleeding events, including one major bleeding episode, occurred after randomization. The major bleeding events were wound hematomas that occurred in 2 patients who underwent total hip arthroplasty.

Thromboembolic Complications in Excluded Patients

Of the 260 patients excluded from the study, 44 did not undergo elective joint arthroplasty (32 had a fracture, and 12 had metastatic disease). Thirty-two refused to give informed consent, and 6 were confused and could not give informed consent. Six patients died before hospital discharge. The remaining 172 patients, all of whom underwent elective joint arthroplasty and most of whom received postoperative warfarin prophylaxis, were followed during hospitalization and for 90 days after hospital discharge. Three of these patients had symptomatic proximal deep venous thrombosis, and 2 had nonfatal pulmonary embolism (Table 3). All 5 of these patients were investigated for suspected venous thromboembolism before hospital discharge. A 93-year-old woman who did not receive anticoagulant therapy in the postoperative period died of a myocardial infarction 82 days after total hip arthroplasty.

Thus, if the results of the 172 excluded patients are included, 14 symptomatic venous thromboembolic complications, 4 nonfatal pulmonary emboli, and 10 proximal deep venous thrombi occurred during hospitalization or within 90 days of hospital discharge in the cohort of 1196 patients (1.2% [CI, 0.6% to 2.0%]) who underwent elective total hip or knee arthroplasty. Eight of these complications occurred in 589 patients (1.4% [CI, 0.6% to 2.3%]) after total hip arthroplasty, and six occurred in 607 patients (1.0% [CI, 0.4% to 2.1%]) after total knee arthroplasty.

Discussion

Top Methods Results Discussion Author & Article Info References

We show that after total hip or knee arthroplasty, the routine use of screening compression ultrasonography before hospital discharge to identify and guide the treatment of asymptomatic proximal deep venous thrombosis did not reduce the rate of subsequent symptomatic thromboembolic complications. Although 13 of 518 truly screened patients (2.5%) were identified as having asymptomatic proximal deep venous thrombosis, rates of subsequent thromboembolic complications over a 90-day follow-up period were similar in patients who had normal bilateral compression ultrasonography at discharge and patients who underwent sham screening. In addition, 1 patient developed a major bleeding complication while being treated for an asymptomatic deep venous thrombosis.

The failure of screening compression ultrasonography to reduce the rate of venous thromboembolism after hospital discharge probably relates to the very low rate of symptomatic thromboembolic complications (1.0%) seen in patients receiving warfarin prophylaxis who did not undergo genuine screening ultrasonography for deep venous thrombosis before discharge. The low rate of symptomatic venous thromboembolic complications was surprising given the high rates of asymptomatic deep venous thrombosis found by screening venography in patients receiving warfarin prophylaxis after total joint arthroplasty. A pooled analysis of randomized clinical trials that used warfarin prophylaxis reported rates of venographically confirmed proximal deep venous thrombosis of more than 10% after total hip or knee arthroplasty [3]. The rates of symptomatic deep venous thrombosis and pulmonary embolism observed in our study demonstrate that most deep venous thrombi that develop after joint arthroplasty in patients receiving warfarin prophylaxis do not have clinically important consequences.

It is unlikely that bias accounts for the failure of screening compression ultrasonography to prevent subsequent thromboembolic complications in our study. We elected to perform sham compression ultrasonographic procedures in the control group to minimize diagnostic suspicion bias on the part of physicians, nurses, and patients; this bias may have made patients in the control group more likely to be investigated for symptoms of deep venous thrombosis and pulmonary embolism. All suspected outcome events were adjudicated by a panel that was unaware of study group assignments. The positive predictive value of screening ultrasonography and the incidence of asymptomatic proximal deep venous thrombosis detected in our study are consistent with previous reports in the literature [8], making it unlikely that the quality of ultrasonography influenced our results. Our study design could not account for the low rate of observed symptomatic thromboembolic complications. Complete follow-up of all randomly assigned patients was achieved, and the 90-day follow-up period after discharge was of sufficient duration for thromboembolic complications related to total hip or knee arthroplasty to occur. To ensure the generalizability of our findings, all consenting patients who underwent elective total hip or knee arthroplasty but were excluded for other reasons were followed for 90 days after discharge. Only 5 of these 172 excluded patients developed symptomatic venous thromboembolic complications. The pooled rate of symptomatic thromboembolic complications in patients who underwent elective total hip or knee arthroplasty was 1.2% (CI, 0.6% to 2.0%) (14 of 1196 patients).

Two recent cohort studies support our observation that rates of symptomatic venous thromboembolic complications after total joint arthroplasty are far lower than suggested by clinical trials that use venographically detected deep venous thrombosis as a study end point. In a retrospective cohort study, Warwick and associates [15] reported a 0.34% rate of fatal pulmonary embolism, a 1.2% rate of nonfatal pulmonary embolism, and a 1.9% rate of deep venous thrombosis in 1162 patients who wore antiembolic stockings over a 6-month follow-up period after elective total hip replacement done at a single institution. In a prospective cohort study, Khaw and colleagues [16] reported a 0.2% rate of fatal pulmonary embolism and a 1.3% rate of nonfatal pulmonary embolism over a 3-month follow-up interval in 499 patients who underwent total knee arthroplasty and wore postoperative compression stockings at a single institution [16]. No screening tests for deep venous thrombosis were performed in these two studies.

We elected to use warfarin prophylaxis. At the beginning of our trial, warfarin was the drug most commonly used for anticoagulant prophylaxis after total joint arthroplasty in North America. Not only was a very low rate of symptomatic thromboembolic complications seen with warfarin prophylaxis, but the rate of major bleeding was low and clinically acceptable.

Our study has several potential limitations. First, although we saw no difference in the rates of clinically important venous thromboembolic complications after hospital discharge in the genuine and sham ultrasonography groups, our study may have had insufficient power to demonstrate a protective effect of ultrasonography. However, given the narrow 95% CIs around the estimated true difference in event rates between the two groups ( –1.2% to 1.2 percentage points), it is unlikely that such a benefit, even if it truly existed, would be considered clinically important. Second, even though compression ultrasonography is the most accurate noninvasive test, its limitations in sensitivity and specificity as a screening test for postoperative deep venous thrombosis have been well described [8]. Had we used a more sensitive screening test (such as venography) in our intervention group and treated all patients found to have deep venous thrombosis with anticoagulant therapy, it is possible that the rate of symptomatic venous thromboembolic complications would have been lower. However, given the low rate of complications in our control group, the potential for increased benefit would be small and this venography strategy would result in a more than 10-fold increase in the number of patients requiring treatment for deep venous thrombosis, making the likelihood of bleeding complications much higher. Symptomatic venous thromboembolic complications have also been described in long-term follow-up of cohorts of patients undergoing screening venography, including some patients with negative venograms [17]. Third, the length of hospitalization after joint arthroplasty has been reduced in many orthopedic centers in North America since the completion of our study. Whether in-hospital prophylaxis is sufficient for patients discharged 4 or 5 days after joint arthroplasty must be addressed in future clinical trials.

Our results suggest that continuing warfarin prophylaxis beyond an average of 9 days after total hip or knee arthroplasty would be of little value, given the low rate of symptomatic venous thromboembolic complications. Our conclusion contrasts with the results of two recent clinical trials, which demonstrated that continuing prophylaxis 3 weeks beyond hospital discharge reduced the rate of venographically confirmed deep venous thrombosis after total hip arthroplasty [6, 7]. The discrepancy between the conclusions of our trial and those of the other two studies is probably due to the use of a surrogate outcome of asymptomatic deep venous thrombosis (diagnosis by screening venography) in those studies. The potential difficulty of relying on studies based on surrogate outcomes to influence clinical practice was described recently [18]. The rate of symptomatic venous thromboembolism after joint arthroplasty is low and is far less than the rate of asymptomatic deep venous thrombosis diagnosed by mandatory screening venography, showing that most (but probably not all) cases of postoperative deep venous thrombosis diagnosed by venography are self-limited and resolve without causing clinically important complications, at least in patients who have received postoperative warfarin prophylaxis.

Dr. Gent: Hamilton Civic Hospitals Research Centre, 711 Concession Street, Hamilton, Ontario L8V 1C3, Canada.