The Incidence of Symptomatic Venous Thromboembolism During and After Prophylaxis With Enoxaparin:  A Multi-institutional Cohort Study of Patients Who Underwent Hip or Knee Arthroplasty FREE

VENOUS thromboembolism is an important complication of arthroplastic surgery of the lower extremity.¹ Without prophylaxis, approximately 50% of patients undergoing hip arthroplasty^2- 4 and 50% to 70% of those undergoing knee arthroplasty^5- 11 develop deep vein thrombosis. Therapy with low molecular weight heparins safely reduces the incidence of venous thromboembolism after major orthopedic surgery.^11- 26 Despite low molecular weight heparin prophylaxis, however, approximately 15% of patients undergoing hip arthroplasty and 30% of those undergoing knee arthroplasty have venographic evidence of deep vein thrombosis at the time of hospital discharge.^11- 26 In addition, 20% to 40% of patients develop new thrombi after hospital discharge, as shown by venography 3 weeks later.^27- 28

The clinical importance of postoperative venous thrombi that occur despite in-hospital prophylaxis is unknown. Most of these thrombi, whether they occur during low molecular weight heparin prophylaxis or after hospital discharge, remain asymptomatic. However, because they have the potential to grow and become symptomatic, it has been suggested that in-hospital prophylaxis alone may not provide adequate protection in patients undergoing major orthopedic procedures.^27- 28

Three additional approaches to in-hospital primary prophylaxis can be considered: (1) perform a predischarge venogram and treat patients who have venous thrombosis with full-dose anticoagulants; (2) perform predischarge compression ultrasonography to detect large proximal thrombi and treat the patient accordingly; and (3) continue outpatient prophylaxis for 4 to 6 weeks after hospital discharge.

All of these options have limitations. Routine predischarge venography is limited because it is invasive, it is associated with some complications, there is a high incidence of technically inadequate studies, and patients are often unwilling. Predischarge compression ultrasonography is noninvasive but is insensitive to some proximal and most calf thrombi. Continuation of prophylaxis as an outpatient is inconvenient and associated with adverse effects. Before a rational decision can be made about the need to perform routine predischarge compression ultrasonography or to continue prophylaxis after hospital discharge, it is important to obtain a reliable estimate of the incidence of large proximal thrombi detected by predischarge compression ultrasonography and the incidence of subsequent clinically important venous thromboembolism in patients with normal ultrasound results.

To determine the incidence of symptomatic venous thromboembolism during and after prophylaxis, we performed a prospective cohort study of 1984 patients who underwent hip or knee arthroplasty and received primary prophylaxis with enoxaparin. All patients underwent predischarge compression ultrasonography, and those with normal results were followed up until day 84 for the onset of symptomatic venous thromboembolism.

Between September 1993 and December 1994, 2294 consecutive adult patients who had unilateral hip or knee arthroplasty at 1 of 28 hospitals were evaluated for study enrollment. Of these, 209 patients were excluded for 1 or more of the following reasons: previous venous thromboembolism (133 patients), the need for oral anticoagulant therapy (38 patients), current malignancy (24 patients), bleeding diathesis (5 patients), uncontrolled hypertension (5 patients), participation in other thromboprophylaxis studies (9 patients), and scheduled for tibial osteotomy (1 patient). Of the 2085 eligible patients, 1984 (95%) gave written informed consent, including 1142 patients who underwent hip arthroplasty and 842 patients who underwent knee arthroplasty.

All enrolled patients received enoxaparin, 30 mg subcutaneously every 12 hours, starting postoperatively after hemostasis was achieved, generally 6 to 24 hours after surgery ended. Enoxaparin was given for a minimum of 5 days and was continued for a total of 14 days or until discharge, whichever came first. No other thromboprophylactic agents were allowed, except for antiembolic stockings. The use of nonsteroidal anti-inflammatory drugs was discouraged but allowed if clinically necessary.

The principal outcomes were symptomatic deep vein thrombosis and pulmonary embolism during and after prophylaxis, asymptomatic deep vein thrombosis detected by predischarge compression ultrasonography, and major hemorrhage. Patients with clinically suspected deep vein thrombosis underwent either serial compression ultrasonography,²⁹ serial impedance plethysmography,³⁰ or contrast venography according to the local preference at each clinical center. Deep vein thrombosis was diagnosed if the results were abnormal, and clinically important venous thrombosis was excluded if the results were normal. Patients with suspected pulmonary embolism underwent ventilation and perfusion lung scanning.³¹ Pulmonary embolism was excluded on the basis of normal perfusion scan results and confirmed by a high-probability scan; the latter was defined as showing 1 or more segmental perfusion defects with normal or near normal ventilation. Patients with abnormal lung scan results that were not classified as high probability underwent pulmonary angiography. Bilateral compression ultrasonography of the tibioperoneal trunk, popliteal vein, superficial femoral vein in at least 2 sites, and common femoral vein was routinely performed when prophylaxis was discontinued. Noncompressibility of a venous segment was considered diagnostic of venous thrombosis. Major hemorrhage was defined as overt bleeding, occurring after at least 1 dose of enoxaparin but within 48 hours of the last dose that (1) was associated with a decrease in the hemoglobin level of 20 g/L or more, (2) necessitated transfusion of 2 or more units of packed red blood cells, (3) was a hemarthrosis causing infection or requiring another operation, (4) was intraocular, (5) was retroperitoneal, or (6) involved the central nervous system. The results of objective tests for venous thromboembolism and bleeding episodes were adjudicated by a central adjudication committee. Patients with confirmed venous thromboembolism received anticoagulant treatment per local practice.

The study consisted of 2 parts, the in-hospital prophylaxis phase and the follow-up phase. Patients were assessed daily during the in-hospital prophylaxis phase for symptomatic deep venous thrombosis, pulmonary embolism, and bleeding; appropriate investigations were undertaken when symptoms occurred. As described above, all patients underwent bilateral compression ultrasonography of the leg veins 14 days after surgery or before discharge, whichever came first. Patients visited the clinic or were followed up by telephone at 6 and 12 weeks. Patients who developed symptoms suggestive of deep venous thrombosis or pulmonary embolism were investigated by objective testing.

The estimated incidences of symptomatic deep vein thrombosis, symptomatic pulmonary embolism, asymptomatic proximal vein thrombosis detected by routine ultrasonography, and major hemorrhage were based on the total cohort. The rates of venous thromboembolism and major hemorrhage in the 2 surgical groups were compared using the Fisher exact test. All 95% confidence intervals (CIs) were calculated from the binomial distribution. Data are given as mean ± SD.

The target sample of 2000 patients was based on a goal to achieve a 95% CI of approximately ±1.5% over a range of clinically acceptable rates of total venous thromboembolism (1.0%-6.0%).

The principal investigators independently designed the study, interpreted the results, and wrote this article. Research coordinators at each clinical center collected the data and forwarded them to the Hamilton Civic Hospitals Research Centre, McMaster University, Hamilton, Ontario, which functioned as the coordinating and methods center; monitored the quality of the trial; and performed the entry, cleaning, and analysis of data. The protocol was reviewed by the institutional review board of each participating site, and all patients gave written, informed consent, obtained by either the study coordinator or the investigator.

The Steering Committee convened after 717 patients were recruited because of 3 deaths caused by pulmonary embolism in patients who underwent knee arthroplasty. After a critical review of all 3 deaths by the full Steering Committee, a decision was made to continue the trial but to shorten the time to initiation of prophylaxis from 12 hours to 6 hours after the end of surgery.

The baseline characteristics of the study cohort, comprising 1142 patients with hip arthroplasty and 842 patients with knee arthroplasty, are shown in Table 1. The cohort was 68 years of age on average and consisted predominantly of women; most patients had surgery under general anesthesia. As shown in Table 2, the mean start of prophylaxis was 17.9±10.4 hours after surgery, and the mean number of enoxaparin injections was 18.0 ± 6.9, corresponding to a mean prophylaxis duration of 9 days. Antiembolism stockings were used by 22.4% of patients, and 16.0% were given nonsteroidal anti-inflammatory drugs during hospitalization. Ninety-two percent of patients completed their intended course of prophylaxis. The mean hospital stay was 12.2 ± 10.6 days; 96% of patients were weight bearing with support at the time of hospital discharge, and 5.2% received some sort of prophylaxis after discharge. These characteristics were consistently similar between the hip and knee cohorts, with the exception of "unable to bear weight," which was higher in patients who underwent hip arthroplasty. All patients were followed up until day 84 after surgery, and none were unavailable for follow-up.

There were 318 cases of clinically suspected deep vein thrombosis, 55 (17.3%) of which were subsequently confirmed by objective testing. In addition, 48 patients developed clinically suspected pulmonary embolism, and the diagnosis was confirmed in 24 (50.0%). Of these, 3 were fatal and confirmed by autopsy. Only 3 of 1936 patients who had predischarge ultrasonography (0.15%) had asymptomatic deep vein thrombosis; all of them had hip arthroplasty. Thus, 82 of 1984 patients in the cohort (4.1%; 95% CI, 3.3%-5.0%) developed objectively confirmed venous thromboembolism (Table 3): 49 of 1142 patients who had hip arthroplasty (4.3%; 95% CI, 3.1%-5.5%) compared with 33 of 842 patients who underwent knee arthroplasty (3.9%; 95% CI, 2.6%-5.2%; P=.73). Overall, 65 patients (3.3%; 95% CI, 2.5%-4.2%) developed symptomatic proximal vein thrombosis or pulmonary embolism, including 42 patients in the hip cohort (3.7%; 95% CI, 2.7%-4.9%) compared with 23 patients in the knee cohort (2.7%; 95% CI, 1.7%-4.0%; P=.25).

The distributions of venous thromboembolic events during and after prophylaxis are shown in Table 4. Approximately half of the events occurred during prophylaxis (42; 2.1%), and the remainder occurred after its discontinuation (40; 2.0%). Twenty-four of these 40 postprophylaxis events (14 in the hip cohort and 10 in the knee cohort) occurred before day 28 after surgery, 12 (9 in the hip cohort and 3 in the knee cohort) occurred between days 29 and 56, and the remaining 4 (2 in each subgroup) occurred between days 57 and 84.

Twelve patients (0.6%) in the entire cohort died. Three patients (0.2%), all of whom had undergone knee arthroplasty, were shown by autopsy to have died of pulmonary embolism. The first patient collapsed suddenly on the morning of the first postoperative day (day 1) before the start of prophylaxis; the second patient died suddenly on the second postoperative day after receiving 2 doses of enoxaparin; and the third patient developed a clinical picture of pulmonary embolism on postoperative day 15 when preparing for discharge and died the next day. This patient had received a 10-day course of enoxaparin prophylaxis, and results of routine ultrasound examination were normal. The causes of death in the remaining 9 patients were ischemic heart disease (3 patients on days 4, 12, and 28); pneumonia (2 patients on days 18 and 36); massive stroke (1 patient on day 33); sepsis (1 patient on day 81); multisystem organ failure (1 patient on day 5); and cause unknown but believed unlikely to be due to pulmonary embolism (1 patient on day 72).

Major hemorrhage occurred in 58 patients (2.9%; 95% CI, 2.2%-3.7%), including 34 patients in the hip cohort (3.0%; 95% CI, 2.0%-4.0%) and 24 patients in the knee cohort (2.9%; 95% CI, 1.7%-4.0%). The types of major bleeding episodes are shown in Table 4. The most common was wound hematoma, occurring in 2.4% of patients. One of the 58 bleeding episodes occurred in a patient who underwent hip arthroplasty on the day of surgery, 14 (10 in the hip cohort and 4 in the knee cohort) occurred on the first postoperative day, 28 (17 in the hip cohort and 11 in the knee cohort) occurred on day 2 or 3, and 15 (6 in the hip cohort and 9 in the knee cohort) occurred on days 4 to 13.

We performed this study to assess the need for possible additional measures after in-hospital prophylaxis with enoxaparin in high-risk orthopedic patients. In a large cohort of almost 2000 patients undergoing major hip or knee surgery, there was only 1 death (0.05%) and 39 (2.0%) clinical events that might have been prevented by continuing prophylaxis after hospital discharge.

Patients received enoxaparin in the hospital for a mean of 9 days and had compression ultrasonography before discharge. The 1 patient who died of massive pulmonary embolism after prophylaxis, on day 15, had received enoxaparin prophylaxis for 10 days and had normal results from routine ultrasound examination.

The expected incidence of silent deep vein thombosis at the time of hospital discharge in our patient population was approximately 15% in the hip cohort and 30% in the knee cohort, with at least one quarter of them involving the proximal veins. We used a limited ultrasonography method that has been shown to be sensitive to proximal vein thrombosis in symptomatic patients.²⁹ However, results were positive in only 0.15% of patients, presumably because most thrombi present at the time of hospital discharge were small and nonocclusive.³² Had we used a more sensitive method for detecting venous thrombosis, such as venography or a more detailed ultrasonographic examination, we would have identified more thrombi and treated more patients with anticoagulants. Overall, this may have resulted in preventing some cases of symptomatic deep vein thrombosis after discharge. However, the fact that few patients developed subsequent symptomatic disease, despite our low rate of detection of asymptomatic venous thrombosis, considerably limits the potential gains. This must be balanced with the cost and adverse effects of treating most patients unnecessarily.

Consistent with the hypothesis that most thrombi occurring despite low molecular weight heparin prophylaxis are not clinically important, only 2.1% of our patients developed symptomatic venous thromboembolism during their hospital stay. Two patients died of pulmonary embolism in the hospital: 1 on the first postoperative day, before the initiation of enoxaparin therapy, and the other on the second postoperative day, after 2 doses of enoxaparin had been given. These deaths were unlikely to have been preventable, except possibly by preoperative prophylaxis.

We compared our results with those reported by Planes and associates²⁷ and Bergqvist and coworkers²⁸ in recent studies that evaluated the effectiveness of outpatient prophylaxis in patients undergoing hip replacement. In both studies, all patients had received low molecular weight heparin in the hospital for approximately 10 to 14 days. In the study by Planes et al,²⁷ patients underwent predischarge venography. Patients with a normal predischarge venogram were randomly allocated to continue low molecular weight heparin prophylaxis or placebo, and venography was repeated 21 days later. The incidence of new thrombosis detected by the second venogram was 19% (8% proximal) in the placebo group and 7% (6% proximal) in the low molecular weight heparin group; most of these thrombi were clinically silent. The design of the study by Bergqvist et al²⁸ was essentially the same as that of Planes et al²⁷ except that predischarge venography was not performed. The incidence of thrombosis at the 3-week venogram was 39% (24% proximal) in placebo patients and 18% (7% proximal) in those who received low molecular weight heparin; most thrombi were asymptomatic. When interpreted together, the results of our study and those reported by Planes et al²⁷ and Bergqvist et al²⁸ suggest that most thrombi that develop after initial low molecular weight heparin prophylaxis are silent and clinically nonsignificant.

Our results are likely to be valid and generalizable. The large cohort in our study included patients from 28 institutions with a wide range of disease severity, comorbid conditions, and type of surgical care, thus increasing the external validity of our findings. The large sample size yielded narrow CIs around the estimated rates of both thromboembolic and major hemorrhagic events. On the other hand, our results may not be applicable to patients with previous venous thromboembolism because such patients were excluded. It is uncertain whether our findings can be extrapolated to other low molecular weight heparins.

A potentially important limitation of the results of our study is their generalizability to current practice in many hospitals, which is to discharge patients within 5 to 7 days of hip or knee arthroplasty. In our study, patients received prophylaxis for a mean of 9 days; therefore, it is possible that the incidence of symptomatic venous thrombosis would have been higher if patients had been discharged earlier. However, our findings indicate that there is little to gain from extending postoperative prophylaxis with low molecular weight heparin beyond 9 days.

Accepted for publication September 25, 1997.

Supported by a research grant from Rhône-Poulenc Rorer, Montreal, Quebec. Dr Leclerc was a Chercheur Boursier from the Fonds de la Recherche en Santé du Québec; Dr Hirsh is a recipient of a Trillium Award from the Ontario Ministry of Health; Dr Ginsberg is a recipient of a Research Scholarship from the Heart and Stroke Foundation of Canada.

Presented at the Scientific Meeting of the Mediterranean League Against Thrombosis, Montpellier, France, October 25, 1996.

Members of the Canadian Collaborative Group are as follows: Study Investigators: W. H. Bailey and M. Kovacs, Victoria Hospital, London, Ontario; R. Bourne, M. Cruickshank, and C. Rorabeck, University Hospital, London; W. Brien, R. W. Grainger, and D. Chess, St Joseph's Health Science Centre, London; M. Brigden, B. Barry, and H. Wass, Royal Jubilee Hospital, Victoria, British Columbia; E. Brooks, The Montreal General Hospital, Montreal, Quebec; L. Desjardins, F. Morin, and J. Villeneuve, Centre Hospitalier de l'Université Laval, Ste Foy, Quebec; R. Devarajalu and W. A. Silver, Regina General Hospital, Regina, Saskatchewan; S. Dolan, St John Regional Hospital, St John, New Brunswick; J. Drouin and A. Giachino, Ottawa General Hospital, Ottawa, Ontario; R. Dumais, P. Charron, and M. Lepine-Martin, Centre Hospitalier de l'Université de Sherbrooke, Sherbrooke, Quebec; W. Dust, Royal University Hospital, Saskatoon, Saskatchewan; D. Harder, Royal Columbian Hospital, New Westminster, British Columbia; V. Ing, Halifax Infirmary Camphill Hospital, Halifax, Nova Scotia; J. Kassis, Hôpital Maisonneuve-Rosemont, Montreal; G. Laflamme and B. l'Espérance, Hôpital du Sacré-Couer, Montreal; R. Lapp, Victoria General Hospital, British Columbia; G. Levesque, J. Lepine, and G. Poirier, Centre Hospitalier Regional de Rimouski, Rimouski, Quebec; M. Mant, University of Alberta Hospital, Edmonton; J. Oliver, Kelowna Hospital, Kelowna, British Columbia; W. Pisesky, Vernon Jubilee Hospital, Vernon, British Columbia; P. Powers, St Joseph Hospital, Hamilton, Ontario; J. Schatzker, M. Tile, R. Jay, and G. Hunter, Sunnybrook Health Science Center, Toronto, Ontario; A. G. G. Turpie, Hamilton General Hospital; H. Weinberg and J. Wilson, Humber Memorial Hospital, Toronto; J. Weitz, Henderson General Hospital, Hamilton; P. Weller, Toronto East General Hospital; D. Wilson, Mississauga General Hospital, Mississauga, Ontario; and D. Yen, Kingston General Hospital, Kingston, Ontario; Steering Committee: M. Gent, chairman, W. H. Geerts, J. Hirsh, L. Laperrière, J. R. Leclerc, and A. G. G. Turpie; Adjudication Committee: M.Gent, J. S. Ginsberg, and J. R. Leclerc; and Study Coordinators: S. Haley, and P. Hudoba.

Reprints: Jacques R. Leclerc, MD, Cardiovascular Research, Lilly Research Laboratories, Lilly Corporate Center, Indianapolis, IN 46285-2146 (e-mail: Leclerc_Jacques_R@Lilly.com).