Prevention of VTE in Orthopedic Surgery Patients

1.3.1 Baseline Risk for VTE:

We made considerable effort to determine the baseline risk of symptomatic VTE and bleeding in the absence of prophylaxis. For this purpose, we analyzed all controlled trials that had a placebo or no-treatment group extending back to 1959.⁶ This has obvious limitations because of important changes in surgical care, including changes in operative technique, earlier ambulation, and earlier discharge that have had an impact on rates of thrombosis and bleeding. For instance, although the average length of stay after HFS in the 1960s was 35 days,⁷ current averages of 3.2 days have been reported in a large cohort after arthroplasties,⁸ and early mobilization starts at 2 to 4 h after surgery.⁸ Randomized controlled trial (RCT) data typically showed a symptomatic VTE event rate of 15% to 30% without prophylaxis prior to 1980,^6,9‐12 and observational data suggest a further drop from around 5% to 1% to 2% in the years from 1989 to 2001.¹³

In recent years, there have been no large placebo controlled trials, and we did not identify any large, well-designed cohort studies to provide a baseline risk relevant to current practice. However, there are several large RCTs that have used low-molecular-weight heparin (LMWH), and we have estimated baseline risk by applying the observed risk of symptomatic VTE in patients treated with LMWH and adjusting it by the relative risk reduction in symptomatic VTE from prior randomized trials of LMWH compared with placebo.

First, we estimated contemporary average on-prophylaxis rates with LMWH for symptomatic DVT to be 0.8% and for PE to be 0.35% by averaging the LMWH event rates from trials enrolling > 16,000 patients since 2003.^4,14‐26 We selected the year 2003 because of a shift in surgical technique since that time to be less invasive and possibly less thrombogenic. Concerns that those rates could be too low given the sometimes highly selected nature of clinical trials, we compared this rate with older data from a large observational study.²⁷ The investigators identified 133 of 19,586 (0.7%) VTE events during the initial hospitalization for patients receiving prophylaxis (estimated prophylaxis compliance, 88%), suggesting that the symptomatic VTE rate of 1.15% we used is not too low.

Second, if we assume the effect of LMWH is similar in asymptomatic and symptomatic DVT, then the best evidence suggests that LWMH reduces the risk for DVT by 50% to 60% and PE by about two-thirds.³ Using this estimate, the contemporary off-prophylaxis rates are ∼1.8% for symptomatic DVT and 1% for PE for the first 7 to 14 days (the initial prophylaxis period most RCTs used and that correspond to the nonextended prophylaxis period).

The untreated baseline risk for the extended, out-of-hospital period, defined as the time period starting at around postoperative day 15 and extending up to 35 days, is likely to be somewhat lower because the VTE risk is highest close to surgery and the median time of diagnosis for thromboembolic events is 7 days after TKA and 17 days after THA.²⁷ We found only one trial that enrolled patients after 2003 that examined extended, out-of-hospital prophylaxis using a placebo group control to estimate the baseline risk for this time period.⁴ Extracting events from the time-to-event graph and from the text, 11 of 1,207 (0.91%) symptomatic VTE events were observed up to postoperative day 39, starting from the time enoxaparin was stopped at an average of 12 days postoperation. A trial that enrolled patients slightly before our cutoff years (2001 and 2002) found a higher rate in the placebo arm (symptomatic VTE, 8/330 [2.4%]).²⁸

In summary, we have estimated a symptomatic VTE rate that is about one-half the rate observed in the immediate postoperative period (1.5%; symptomatic DVT, 1%; PE, 0.5%). For this guideline, we therefore estimated a combined 35-day untreated baseline risk for symptomatic VTE of 4.3%.

Although epidemiologic data from the early 1990s suggest that the cumulative 90-day symptomatic VTE risk for THA is higher than that for TKA (2.8% vs 2.1%, respectively),²⁷ randomized trials fail to confirm this finding. Follow-up epidemiologic data from the mid-1990s also demonstrated that cumulative 90-day symptomatic VTE rates after HFS did not exceed those reported for arthroplasty (HFS, 1.9%; THA, 2.4%; TKA, 1.7).²⁹ We therefore concluded that a 4.3% combined symptomatic VTE untreated baseline risk for the first 35 days is the best approximation for all three major orthopedic surgeries. Table 2 and Figure 1 present a summary of the estimated symptomatic VTE rates for this guideline.

Because VTE-related deaths were rarely observed in trials since 2003, the data were insufficient to estimate current baseline risk. In addition, competing risks, such as cardiovascular and infectious causes of death, often outnumber the risk of death from VTE, particularly in HFS. When pooling study data, total mortality—because this outcome includes fatal bleeding—was selected to better represent the overall balance of fatal events. The majority of mortal events were seen in HFS populations that are elderly and experience considerable comorbidity.

1.3.2 Baseline Risk for Major Bleeding Events:

The risk for major bleeding with LMWH, and in particular without treatment, remains difficult to estimate because better operative techniques make deriving the untreated bleeding event rate from the placebo group of past RCTs in major orthopedic surgery problematic. To estimate untreated bleeding risk, we first determined the median major bleeding event rate from the placebo (or graduated compression stockings [GCS]) arm of LMWH trials and the Pulmonary Embolism Prevention (PEP) trial (subgroup that did not receive any heparin) because those trials were more recent.^30‐39 The median rate was 1.5%, but because of the low event rate in the LMWH trials, variability in the definitions of major bleeding across trials makes this estimate uncertain. This is, however, consistent with a systematic review that estimated the absolute untreated bleeding risk to be between 1% and 2%.³

Second, we selected the major bleeding event rate for LMWH from a recent review that examined the reporting definitions and event rates from the enoxaparin control arm of recent trials.⁴⁰ We chose a rate of 1.5%, which was slightly higher than the average (1.4%) and higher than 12 of 14 trials that enrolled > 16,000 patients since 2003 that we included in the estimate of baseline VTE risk (median, 0.91%; maximum, 1.9%).^4,14‐26 Recognizing the sometimes highly selective process of RCTs in enrolling patients with low bleeding risk, we believe that a selected bleeding rate that is somewhat higher than the median is therefore close to what would be observed in clinical practice. The baseline major bleeding rate of 1.5% (15 of 1,000) and that expected with LMWH are shown in Table 3 and Table S1 and are very close. (Tables and figures that contain an “S” before the number denote supplementary information not contained in the body of the article and available instead in an online data supplement; see the “Acknowledgments” for more information.) Intuitively, a greater bleeding rate might be expected with the use of LMWH, but this increased risk is likely within the large CI.

2.1.1 LMWH vs No Prophylaxis—Initial and Extended-Period Prophylaxis:

LMWH has become the thromboprophylaxis agent against which newer drugs are compared. Several studies published in the mid-1980s, during the 1990s, and as recently as 2008 have investigated LMWH compared with no prophylaxis in > 2,000 patients to test the hypothesis that LMWH decreases the incidence of VTE after arthroplasty^{31‐39,47,48} and HFS.^49,50 Our analysis included all studies of LMWH vs no prophylaxis whether GCS were used in both groups because this would not affect the relative risk observed for LMWH. This allowed us to make more-precise estimates for risk reduction of VTE and bleeding. We decided against pooling across other patient groups, such as nonorthopedic surgery patients, because of differences in risk and technique. In those trials, LMWH usually was continued for 6 to 14 days, which coincided with discharge from the hospital at the time those trials were conducted.

For THA or TKA, LMWH consistently reduces asymptomatic DVT by ∼50% (combined risk ratio [RR], 0.50; 95% CI, 0.43-0.59). Similar results were seen in two studies in HFS involving 218 patients.^49,50 Combining results from all relevant studies failed to demonstrate or to exclude a beneficial effect of LMWH on PE (RR, 0.58; 95% CI, 0.22-1.47). On the basis of moderate-quality evidence, the use of LMWH for the initial prophylaxis period (10-14 days) is expected to prevent 13 VTE per 1,000 patients undergoing major orthopedic surgery, assuming a baseline risk of 1% for PE and 1.8% for symptomatic DVT.

The definition and reporting of major bleeding was inconsistent across studies, and the results failed to demonstrate or to exclude a detrimental effect of LMWH on the occurrence of major bleeding (RR, 0.81; 95% CI, 0.38-1.72); the 95% CI was nine fewer to 11 more major bleeding events per 1,000. Few deaths occurred, and these were mainly seen in HFS patients; two VTE-associated deaths were seen in the placebo groups compared with one in the LMWH arm (Table 3, Figs S1-S4, Table S1).

2.1.2 LDUH vs No Prophylaxis—Initial Prophylaxis Period:

Numerous RCTs examined LDUH vs no prophylaxis throughout the 1970s and early 1980s. A systematic review involving close to 7,000 patients demonstrated a relative risk reduction of 58% (RR, 0.42; 95% CI, 0.36-0.50) in the incidence of asymptomatic DVT found by screening across 57 trials from surgical and nonsurgical populations.³ Only four of the 12 studies in orthopedic surgery used venography to confirm thrombotic events; the others used fibrinogen uptake. The relative effect estimates were similar for the eight studies involving > 500 patients undergoing elective hip replacement (RR, 0.53; 95% CI, 0.32-0.89) and six trials in HFS (RR, 0.56; 95% CI, 0.39-0.81) compared with the entire population.

A significant reduction in PE was observed by pooling all trials from surgical and nonsurgical populations (RR, 0.69; 95% CI, 0.49-0.99). Unfractionated heparin (UFH) was associated with a trend toward an increased risk of major bleeding (RR, 1.26; 95% CI, 0.99-1.6). Using our estimates of baseline risk, the relative effect translates into a reduction of 13 symptomatic VTEs per 1,000 with UFH, with an increase in major bleeding events of four per 1,000. Mortal events in major orthopedic surgery were only reported for HFS trials (RR, 0.96; 95% CI, 0.55-1.67), and across all patient groups, UFH appeared to have little or no effect on overall mortality (RR, 0.91; 95% CI, 0.8-1.04). The underlying quality of evidence was moderate (Table 6, Table S3).

2.1.3 Vitamin K Antagonist vs No Prophylaxis—Initial Prophylaxis Period:

Evidence for use of vitamin K antagonists (VKAs) comes from eight RCTs involving 703 patients, most with hip fracture, that demonstrated a 55% relative risk reduction in primarily asymptomatic DVT (RR, 0.45; 95% CI, 0.32-0.62).³ PEs were reduced by almost 80% (RR, 0.21; 95% CI, 0.08-0.53), although this result is based on only 32 events. Although patients and clinicians in those trials were not blinded, two trials blinded the thrombosis outcome adjudicators. VKA use was associated with a trend toward increased bleeding (RR, 1.50; 95% CI, 0.92-2.43), which was described as wound hematomas, wound bleeding, wound leakage, hematuria, and hematemesis. There was also more blood transfused and one intracerebral hemorrhage in the VKA group.⁶¹ Results showed a trend toward a mortality reduction (RR, 0.76; 95% CI, 0.54-1.07). Based on moderate-quality evidence, VKA prophylaxis for 10 to 14 days would result in 18 fewer VTEs and seven more major bleeding events per 1,000 (Table 7, Table S4).

2.1.4 Aspirin vs No Prophylaxis—Initial Plus Extended Prophylaxis Period:

Aspirin is inexpensive, orally administered, and widely available. In the 1970s and 1980s, a number of studies investigated the use of aspirin in THA,^10,62‐65 TKA,⁶⁶ and HFS.^67‐72 Those studies used high doses of aspirin of up to 3.8 g daily. They suffer from serious methodologic limitations, including the use of an unreliable method for DVT screening, such as fibrinogen uptake; lack of blinding; and lack of allocation concealment. Additionally, there was strong evidence of reporting and publication bias.

Because of this low quality of evidence, a subsequent trial, PEP, was initiated to study the effects of 160 mg of aspirin given for 35 days against placebo in a routine practice setting that allowed for additional antithrombotic intervention if deemed necessary.³⁰ This multicenter trial enrolled 17,444 patients predominantly after HFS in the mid-1990s and included patients after hip arthroplasty. This study has been criticized because of perceived changes in the primary outcome and adjustments of sample size. There were additional problems with the presentation of the results that made evaluation of the bleeding end point difficult. The PEP study, however, had considerable strengths, including concealment of allocation through remote randomization; blinding of patients, caregivers, and investigators; and an independent, blinded adjudication committee that interpreted objectively confirmed end points, such as venographically or DUS-confirmed DVT, high probability ventilation/perfusion scans, or pulmonary angiograms. In addition, there was near-complete follow-up (99.6%).

Although the combined results (arthroplasty and HFS) failed to demonstrate or exclude a beneficial effect of aspirin on nonfatal PE, there was a modest 28% relative risk reduction in symptomatic DVT (RR, 0.72; 95% CI, 0.53-0.96). The upper boundary of the CI crosses a threshold of 10% that clinicians consider the desirable minimum clinical effect, and the CI of the absolute effect includes as few as one less DVT in 1,000. The results, therefore, are imprecise, despite the large number of patients enrolled. Although there were 19 VTE-associated deaths in the aspirin group compared with 45 in the placebo group, the RR for overall mortality was 0.96 (95% CI, 0.85-1.09). There was a trend toward more major nonfatal bleeding associated with aspirin (RR, 1.12; 95% CI, 0.94-1.34), but there were no difference in bleeding requiring reoperation or bleeding deaths. In addition, the investigators reported no difference in major bleeding in the subgroup that did not receive additional heparin (aspirin alone, 95 of 3,711; placebo alone, 94 of 3,789). Perioperative aspirin use was associated with a trend toward more nonfatal myocardial infarctions (RR, 1.59; 95% CI, 0.98-2.57).

In summary, given the moderate-quality evidence, it appears that low-dose aspirin given before major orthopedic surgery and continued for 35 days will result in seven fewer symptomatic VTEs per 1,000 but at the expense of a possible three more major bleeding episodes and two additional nonfatal myocardial infarctions per 1,000, thus resulting in a close balance between desirable and undesirable effects (Table 8, Figs S9-S14, Table S5).

When considering aspirin vs anticoagulants, the impact of anticoagulants on myocardial infarction has not been studied. The relative effects of aspirin are likely similar whether other additional thromboprophylaxis, including heparins or mechanical interventions, are used. The absolute reduction in thrombosis, however, will be greater in the absence of anticoagulants than in their presence, and the absolute increase in bleeding, if present, is likely to be less in the absence of anticoagulants than in their presence.

2.1.5 Fondaparinux vs No Prophylaxis—Extended Prophylaxis Period:

We did not identify trials examining fondaparinux vs placebo for the initial prophylaxis period. However, one trial that used fondaparinux for 6 to 8 days in HFS randomized 656 patients on postoperative days 6 to 8 to either extended fondaparinux for an additional 19 to 23 days or placebo.²⁸ No PE was observed in the fondaparinux group compared with two of 330 in the placebo group. The results for symptomatic DVT failed to demonstrate or to exclude a beneficial effect (RR, 0.17; 95% CI, 0.02-1.39). Six major bleeding events occurred in the fondaparinux group compared with none in the placebo group (RR, 13; 95% CI, 0.74-231), and results failed to exclude a beneficial or detrimental effect of fondaparinux on total mortality (RR, 0.76; 95% CI, 0.27-2.16) (Table 9, Figs S15-S20; Table S6).

Based on moderate-quality evidence, 12 fewer symptomatic VTE per 1,000 would be expected with the use of fondaparinux, but this beneficial effect would be offset by an increase of at least 12 major bleeds per 1,000. The close balance between desirable and undesirable effects makes the use of fondaparinux for extended thromboprophylaxis less appealing, particularly compared with LMWH.

2.1.6 Mechanical Interventions vs No Prophylaxis—Initial Prophylaxis:

There are few data regarding the use of GCS compared with no prophylaxis in major orthopedic surgery, although they are used frequently in conjunction with other thromboprophylaxis. A systematic review identified nine trials in a variety of patient populations,³ but only one small trial included orthopedic surgery patients.⁷³ The pooled results from all trials failed to demonstrate or to exclude a beneficial or detrimental effect of GCS on PE (RR, 0.63; 95% CI, 0.32-1.25). Although GCS showed a beneficial effect on asymptomatic, venographically confirmed DVT overall (RR, 0.51; 95% CI, 0.36-0.73), evidence from a higher-quality large trial in patients with stroke^74,75 only showed a trend toward reduced symptomatic DVT (RR, 0.92; 95% CI, 0.77-1.09), and this was offset by a fourfold increase in skin complications (Table 10, Table S7).

Mechanical approaches to perioperative thromboprophylaxis with pneumatic compression devices have the potential advantage of reducing the incidence of VTE but without the risk for increased bleeding. In addition, an intermittent pneumatic device (IPCD) can be used in the contralateral leg even during surgery and the immediate postoperative period.

Seven RCTs that included > 900 patients undergoing arthroplasty or HFS compared mechanical compression to no thromboprophylaxis.^{31,66,76‐79} Six used an IPCD, and one a venous foot pump (VFP).⁷⁷ The risk of bias varied. For instance, in most trials, it was unclear whether allocation was concealed. Blinding of patients and caregivers is not possible in such studies, and not all provided blinded VTE adjudication. In addition, a systematic review indicated funnel plot asymmetry, raising the possibility of publication bias.⁸⁰ Variation in design and performance of the devices as well as information about compliance, which was rarely reported in older trials, introduce uncertainty in how to apply the evidence.

Taken together, the evidence is of low quality. Nevertheless, a relative risk reduction of > 50% was observed for both DVT and PE in THA, TKA, and HFS (PE RR, 0.4; 95% CI, 0.17-0.92; DVT RR, 0.46; 95% CI, 0.35-0.61). The corresponding estimated absolute risk difference is 16 fewer symptomatic VTE per 1,000. The results failed to demonstrate or to exclude a beneficial effect on mortality (Table 11, Figs S21-S23, Table S8).

Compliance remains the biggest challenge associated with the use of IPCDs. Most devices currently in use require an external power source, and they often are found not functioning when patients are getting out of bed or being transported. Properly functioning IPCDs were encountered in < 50% in one study⁸¹ and as low as 19% in another.⁸² In addition, those studies reported no significant improvement in compliance rates with systematic education and training of nursing and other staff. However, because the low compliance is presumably largely due to the IPCD requiring a power outlet, newer battery-powered portable devices are now available, and a recent study reported increased compliance with those devices (77.7% vs 58.9%).⁸³

Other disadvantages of IPCDs are logistical and include having enough units available and keeping them in good working condition. Additionally, there are multiple devices available that have differing properties, and this makes comparison of benefits difficult.

In summary, use of an IPCD for thromboprophylaxis is attractive because of its possible effectiveness and likelihood of no increase in bleeding events. However, suboptimal compliance with the use of an IPCD while in the hospital and the inability to continue this treatment at home for most patients may limit their use. Newer battery-powered IPCDs that monitor compliance might be successfully used after discharge.

2.1.7 Other Modalities vs No Thromboprophylaxis:

Few recent orthopedic trials have compared other thromboprophylaxis agents against placebo.⁸⁴ However, large, well-done trials with direct comparisons against LMWH are available for newer antithrombotic agents, and their similar effects attest to their benefits compared with no prophylaxis. Examples include fondaparinux, apixaban, dabigatran, and rivaroxaban. The latter three have been evaluated in THA and TKA but not in HFS.

Recommendations

2.1.1. In patients undergoing THA or TKA, we recommend use of one of the following for a minimum of 10 to 14 days rather than no antithrombotic prophylaxis: LMWH, fondaparinux, apixaban, dabigatran, rivaroxaban, LDUH, adjusted-dose VKA, aspirin (all Grade 1B), or an IPCD (Grade 1C).

Remarks: We recommend the use of only portable, battery-powered IPCDs capable of recording and reporting proper wear time on a daily basis for inpatients and outpatients. Efforts should be made to achieve 18 h of daily compliance. One panel member believed strongly that aspirin alone should not be included as an option.

2.1.2. In patients undergoing HFS, we recommend use of one of the following rather than no antithrombotic prophylaxis for a minimum of 10 to 14 days: LMWH, fondaparinux, LDUH, adjusted-dose VKA, aspirin (all Grade 1B), or an IPCD (Grade 1C).

Remarks: We recommend the use of only portable, battery-powered IPCDs capable of recording and reporting proper wear time on a daily basis for inpatients and outpatients. Efforts should be made to achieve 18 h of daily compliance. One panel member believed strongly that aspirin alone should not be included as an option.

2.3.1 LMWH vs LDUH—Initial Prophylaxis:

A systematic review of comparisons between LMWH and LDUH included > 23,000 patients from 64 trials across surgical and nonsurgical patient groups; 2,800 patients were included in arthroplasty or HFS trials.³ Pooled estimates showed a 20% relative risk reduction of primarily asymptomatic DVT in favor of LMWH (RR, 0.80; 95% CI, 0.73-0.88), with similar effects seen in the subgroups of THA, TKA, and HFS. LMWH was associated with a trend toward reduced PE in THA, although the pooled results from all groups failed to demonstrate or exclude a beneficial effect of LMWH on PE (RR, 0.78; 95% CI, 0.49-1.24). There was a trend toward less major bleeding with LMWH after THA (RR, 0.59; 95% CI, 0.34-1.01) but not across all trials (RR, 0.91; 95% CI, 0.75-1.09). These results suggest that LMWH may reduce symptomatic VTE from 16 per 1,000 with LDUH to 13 per 1,000 without an increase in major bleeding (Table 12, Table S9).

There have been no trials directly comparing the effectiveness of LDUH every 12 h vs LDUH every 8 h. In 1988, two separate meta-analyses were published that commented on UFH dosing schedules.^87,88 Collins et al⁸⁷ included studies in orthopedic, urologic, and general surgery. Overall, a 72% odds reduction was found for the 8-h regimen and a 63% odds reduction for the 12-h regimen, which was not a significant difference. In orthopedic surgery studies only, the odds reduction was 68% for both regimens. In contrast, the meta-analysis by Clagett et al⁸⁸ was confined to general surgery studies and reported DVT rates in pooled analysis of 11.8% with the 12-h regimen compared with 7.5% using the 8-h regimen. The authors concluded that the 8-h regimen was superior. Neither meta-analysis reported differences in major bleeding between these regimens. These indirect comparisons provide only low-quality, or perhaps very-low-quality, evidence for the alternate regimens.

2.3.2 LMWH vs VKAs—Initial and Extended Prophylaxis:

Several RCTs in THA and TKA^85,89‐95 but not HFS have compared LMWH to VKA (mainly warfarin) in > 9,000 patients for the initial prophylaxis. The results failed to establish or refute a difference in PE (RR, 0.68; 95% CI, 0.22-2.1), but LMWH use was associated with significantly less asymptomatic DVT (RR, 0.68; 95 % CI, 0.6-0.78) at the cost of an increase in major bleeding events (RR, 1.56; 95% CI, 1.23-2.0). Most of these trials, however, started LMWH shortly before surgery, which as we have discussed, likely increases the risk of bleeding substantially. Our sensitivity analysis, excluding trials that administered LMWH close to the operation (< 12 h perioperatively),^85,89‐91 still shows a trend in increased bleeding events, but the magnitude of the effect is smaller (RR, 1.36; 95% CI, 0.95-1.96). We used this RR in our evidence summaries for the initial thromboprophylaxis period with VKA vs LMWH.

Based on those considerations, we estimate that there will be three fewer symptomatic VTE events per 1,000 with the use of LMWH compared with warfarin, but this benefit is closely balanced by a possible increase of four major bleeding events per 1,000. However, given the two fatal bleeding events with the use of VKA (vs none in the LMWH group), safety concerns with warfarin remain (Table 13, Figs S24-S28, Table S10). Furthermore, the evidence regarding extended prophylaxis, presented next, favors LMWH.

2.3.3 LMWH vs Aspirin—Initial and Extended Prophylaxis:

Two trials compared LMWH against aspirin, with one trial using aspirin 325 mg bid⁹⁷ and the other 650 mg bid (only the abstract was available).⁹⁸ The pooled results showed more asymptomatic DVT in the aspirin group (RR, 1.87; 95% CI, 1.3-2.7), but PEs were too few to provide a meaningful estimate. No major bleeding events or deaths were reported. Overall, the evidence from a head-to-head comparison of LMWH compared with aspirin is sparse and of low quality. However, indirect evidence from trials of LMWH and aspirin against placebo also shows greater relative efficacy of LMWH (Table 15, Figs S33, S34, Table S12).

2.3.4 LMWH vs Fondaparinux—Initial Prophylaxis:

Several large trials compared fondaparinux 2.5 mg started 6 to 8 h after wound closure with LMWH (started either 12 h before or after surgery) in THA,^99,100 TKA,¹⁰¹ and HFS.¹⁰² Because the relative effects across outcomes were similar, we included a trial in abdominal surgery patients,¹⁰³ thus including > 10,000 patients. In addition, we included all trials, whether GCS were used in all or only in a portion of patients, as long as it was used equally in both arms.

The pooled results failed to demonstrate or exclude a beneficial or detrimental effect of fondaparinux on symptomatic DVT and PE despite a substantial reduction in asymptomatic DVT. There was a substantial increase in bleeding requiring reoperation associated with the use of fondaparinux (RR, 1.85; 95 % CI, 1.1-3.11), but the results failed to demonstrate a difference in nonfatal major bleeding (RR, 1.35; 95 % CI, 0.89-2.05). VTE deaths were rare and similar in both groups (fondaparinux 5/5,049 vs LMWH 6/5,046). There were two fatal bleeds with fondaparinux and three with LMWH. Caution is advised with fondaparinux in patients weighing < 50 kg (110 lbs) and elderly and frail patients because bleeding complications may be increased. In summary, based on moderate-quality evidence, the use of fondaparinux compared with LMWH does not appear to reduce patient-important VTE events but may increase major bleeding events by nine per 1,000 (Table 16, Figs S35-S40, Table S13).

2.3.5 LMWH vs Rivaroxaban—Initial and Extended Prophylaxis:

Rivaroxaban, an oral direct factor Xa inhibitor, is approved in the United States, Canada, and Europe for the prevention of VTE after THA and TKA, but it has not been evaluated in HFS. Seven RCTs enrolling > 10,000 patients after THA^18,19,104 and TKA^20‐22 examined the efficacy of rivaroxaban 10 mg/d (started 6-8 h postoperatively) against enoxaparin 40 mg/d. Enoxaparin was usually started the evening before surgery and continued 6 to 8 h postoperatively, but two studies used 30 mg bid dosing rather than 40 mg once daily and started 12 h postoperation. For TKA patients, rivaroxaban usually was given for 10 to 15 days, and earlier trials in THA had similarly short treatment durations, but one later trial treated patients for 31 to 39 days.¹⁹ Because the relative effects of extended prophylaxis were similar to shorter-term trials, we estimated pooled effects across all rivaroxaban trials to increase precision, as long as rivaroxaban and control treatment were given for the same duration.

Rivaroxaban reduced symptomatic DVT by > 50% (RR, 0.41; 95% CI, 0.20-0.83). There was a trend toward increased major bleeding and bleeding requiring reoperation (major bleeding: RR, 1.58; 95% CI, 0.84-2.97; bleeding requiring reoperation: RR, 2.0; 95% CI, 0.86-4.83; combined: RR, 1.73; 95% CI, 0.94-3.17). The absolute rates for major bleeding were low in both arms, and the rates were lower than one would expect from other large trials using similar enoxaparin controls. Unlike other trials, the two major THA studies (RECORD 1 and 2) did not include surgical site bleeding (other than bleeding requiring reoperation), and drop in hemoglobin level was calculated compared with the postoperative instead of the preoperative baseline value.⁴⁰

The evidence summaries therefore include the alternate major bleeding rate of 1.5% to better illustrate the trade-offs between VTE and bleeding with rivaroxaban: The best estimates suggest that five fewer symptomatic DVT per 1,000 achieved with rivaroxaban over LMWH will be offset by nine more major bleeding events. In summary, based on moderate-quality evidence, both the possibility of increased major bleeding events and the availability of long-term safety data for LMWH makes LMWH more appealing than rivaroxaban in spite of the inconvenience of subcutaneous administration (Table 17, Figs S41-S47, Table S14).

Extended Prophylaxis With Rivaroxaban:

The extended use of rivaroxaban was studied in one trial enrolling > 2,400 patients after THA.⁴ The control group received short-term LMWH for the first 12 days followed by placebo for an additional 22 days. Rivaroxaban significantly reduced symptomatic VTE (symptomatic DVT: RR, 0.18; 95 % CI, 0.04-0.82; PE: RR, 0.25; 95 % CI, 0.02-2.2). There was only one major bleeding event in both groups. However, in contrast to most other studies, the major bleeding definition in this study excluded surgical site bleeding, and the baseline used for change in hemoglobin level was postoperative day 1. The result was a major bleeding rate of only one-10th of comparable studies using the same control agent.⁴⁰ Bleeding requiring reoperation was recorded.

Based on moderate-quality evidence, 12 fewer symptomatic VTE would be expected. However, because of the uncertainty about the major bleeding rate, it is unknown whether some of the benefit would be offset by a higher bleeding rate of rivaroxaban compared with placebo (Table 18, Figs S41-S53, Table S15).

2.3.6 LMWH vs Dabigatran—Initial and Extended Prophylaxis:

Dabigatran, a new oral direct thrombin inhibitor, has been approved by the US Food and Drug Administration since 2010 for stroke prevention in atrial fibrillation, and European and Canadian agencies have granted marketing authorization for the prevention of VTE after total hip and knee arthroplasty. Four RCTs examined the use of dabigatran in > 10,000 patients undergoing THA^23,24 and TKA^25,26 at doses of 220 and 150 mg taken orally once daily (usually started within 4 h postoperatively at half the dose) compared with enoxaparin (mainly at doses of 40 mg once daily started the evening before surgery, although one study used the 30 mg bid dosing schedule that commenced 12 h postoperatively). Treatment duration ranged from 10 to 15 days (for TKA) to 28 to 35 days for THA. Again, relative effects were similar to the shorter-term TKA trials, facilitating pooled effects across all dabigatran trials.

The studies using the 220 mg dose of dabigatran failed to demonstrate or exclude a difference in the number of symptomatic VTEs (PE: RR, 1.22; 95% CI, 0.52-2.85; DVT: RR, 0.7; 95% CI, 0.12-3.91) or major bleeding events (RR, 1.06; 95% CI, 0.66-1.72). Point estimates of absolute differences between thrombotic and bleeding events were closely balanced to within one event per 1,000 (Table 19, Figs S54-S59, Table S16).

Although dabigatran at the 150-mg dose reduced asymptomatic DVT less than enoxaparin (RR, 1.2; 95% CI, 1.05-1.37), one trial that used a 50% higher dosing schedule (enoxaparin 30 mg bid) contributed the majority of the excess asymptomatic events. Symptomatic VTE results, however, failed to demonstrate or to exclude a beneficial effect of dabigatran compared with LMWH (PE: RR, 0.31; 95% CI, 0.04-2.48; symptomatic DVT: RR, 1.52; 95% CI, 0.45-5.05). Overall, the additional two symptomatic VTE events per 1,000 observed with the lower dose of dabigatran are offset by four additional major bleeding events per 1,000 in the enoxaparin group, although this increased bleeding is more likely with the higher enoxaparin dose of 30 mg bid (Table 20, Figs S60-S65, Table S17).

In summary, dabigatran is similar to LMWH in terms of efficacy and propensity to cause bleeding, based on moderate-quality evidence. Greater long-term experience with LMWH still favors its use.

2.3.7 LMWH vs Apixaban—Initial and Extended Prophylaxis:

Apixaban, an oral direct factor Xa inhibitor, is approved in Europe for the prevention of VTE after THA and TKA but similar to the other newer agents, has not been evaluated in HFS. Four RCTs enrolling close to 12,000 patients after THA¹⁴ and TKA^15‐17 examined the efficacy of apixaban 2.5 mg bid taken orally (started 12-24 h postoperatively) against enoxaparin. Enoxaparin at the 40-mg dosing schedule was started the evening before surgery and continued after surgery according to the investigators’ standard of care (usually 12 h postoperatively). Two studies used 30 mg bid dosing rather than 40 mg once daily and started 12 h postoperatively. For TKA patients, apixaban usually was given for 10 to 14 days, and the single trial in THA used an extended protocol of 32 to 38 days.

Apixaban reduced symptomatic DVT by 59% (RR, 0.41; 95% CI, 0.18-0.95) and appeared to have little or no effect on major nonfatal bleeding (RR, 0.76; 95% CI, 0.44-1.32) or bleeding requiring reoperation (RR, 0.82; 95% CI, 0.15-4.58) compared with enoxaparin. However, similar to the two major rivaroxaban trials, drop in hemoglobin level was calculated compared with the postoperative instead of the preoperative baseline value for the ADVANCE (Apixaban Dosed Orally vs Anticoagulation with Enoxaparin) 2 and 3 trials, which may underestimate the true major bleeding event rate.⁴⁰ Results failed to demonstrate a beneficial or detrimental effect of apixaban on nonfatal PE (RR, 1.09; 95% CI, 0.31-3.88) and total mortality (RR, 1.87; 95% CI, 0.61-5.74), and the only five deaths from VTE were found in the apixaban group.

Best estimates suggest that seven fewer symptomatic DVT per 1,000 could be achieved with apixaban over LMWH without an appreciable increase in major bleeding events (from eight fewer to five more per 1,000), although results failed to demonstrate a difference when all nonfatal and fatal VTE were combined (Fig S72).

In summary, based on moderate-quality evidence, apixaban is similar to LMWH in terms of efficacy based on all symptomatic VTEs (including DVT, non-fatal and fatal PE) (see Fig S72) and showed a comparable low risk for major bleeding events. However, the lack of long-term postmarketing safety data (eg, the confirmation of bleeding-related safety) for apixaban currently makes LMWH still the agent of choice (Table 21, Figs S66-S72, Table S18).

2.3.8 IPCDs vs Pharmacologic Thromboprophylaxis—Initial Prophylaxis:

Compression devices are attractive because they do not increase bleeding. IPCDs were compared against VKAs in > 500 patients from four trials: three in patients undergoing THA^105‐107 and one with both THA and TKA.¹⁰⁸ Because of the small sample sizes, no PE was observed. The results for asymptomatic DVT failed to demonstrate or to exclude a beneficial effect of IPCDs over VKAs (RR, 0.79; 95% CI, 0.5-1.25). All major bleeding events were reported in one study¹⁰⁶ in which warfarin was started 1 week prior to the operation and the INR was kept initially at ≤ 1.5 during the operation. In this trial, eight patients required ≥ 4 units of blood transfusion, and two had higher intraoperative blood loss. Because the usual practice is to give warfarin the night before surgery and adequate anticoagulation levels will not be achieved for several days, those bleeding events may not be applicable to current practice. Using a more-precise estimate of 2% (90 major bleeds observed in 4,547 patients) as seen in the VKA arm of RCTs vs LMWH, it is likely that 19 more bleeds will occur per 1,000, offsetting the two fewer DVT seen with warfarin (Table 22, Figs S73-S75, Table S19).

Pneumatic compression devices were compared with LMWH in > 1,000 patients scheduled for THA^109‐111 and TKA^31,112: five studies used a VFP, and two used an IPCD. We included studies in our analysis whether GCS were used in both treatment arms. A single nonfatal PE was observed in the IPCD/VFP group. Use of a compression device was associated with a trend toward an increase in asymptomatic DVT (RR, 1.38; 95% CI, 0.92-2.06). Less major bleeding occurred in the IPCD group (RR, 0.32; 95% CI, 0.12-0.89). In these studies, bleeding event adjudication was not blinded, and bleeding events were inconsistently reported (eg, bleeding requiring reoperation remained unreported despite the sample size of > 1,000). Three deaths from VTE occurred with the compression device vs none in the LMWH group.

Overall, 10 fewer symptomatic VTE events per 1,000 can be expected with the use of LMWH compared with a compression device at the expense of 10 additional major bleeds per 1,000. This closely balanced estimate is sensitive to the baseline bleeding risk, which was set to 1.5% for LMWH as observed in trials since 2003. Although the actual observed bleeding rate was 2.6%, these trials were performed before our cutoff for contemporary surgical technique and may not be representative of current practice. Additionally, there was no blinding, and this could result in overestimating the number of major bleeds associated with LMWH (Table 23, Figs S76-S79, Table S20). In summary, low-quality evidence, mostly because of imprecision and risk of bias, reduces our confidence in the estimate of the true effect of an IPCD against LMWH and tilts our judgment in favor of LMWH.

Newer-generation IPCDs have the advantage of being portable and able to record effective use time. Two trials compared these IPCDs in combination with low-dose aspirin (81-100 mg) to LMWH in THA¹¹³ and both THA and TKA¹¹⁴ enrolling > 500 patients. Results failed to demonstrate or to exclude a beneficial effect of the IPCD on PE due to the low number of events observed, but fewer asymptomatic DVT were seen in one of the two trials (pooled RR, 0.47; 95% CI, 0.24-0.91). Fewer major bleeding events occurred with IPCD in one of the trials with LMWH but not in the other (pooled estimate RR, 0.04; 95% CI, 0-0.7). However, all results were imprecise because of low numbers of events (total of 42 VTE and 11 bleeding events), and the definition of bleeding differed from other trials, making a direct comparison difficult (Table 24, Figs S80-S82, Table S21).

Overall, there are significant methodologic limitations in the trials of new- and prior-generation IPCD vs LMWH, which include lack of concealment of allocation, an unblinded adjudication process for bleeding, the uncertainty generated by the lack of a standard definition of major bleeding, and a generally small sample size and variation in the properties of pneumatic compression devices. These limitations make it difficult to accept the apparent benefit of new-generation IPCD in combination with aspirin over LMWH based on a simple trade-off of thrombotic events against patient-important bleeding.

2.3.9 Summary—Choice of Thromboprophylaxis:

Selecting from the range of pharmacologic and mechanical interventions in major orthopedic surgery, the agent that has similar or superior properties of effective thromboprophylaxis combined with little risk of bleeding and extensive clinical experience is LMWH; extending thromboprophylaxis up to 35 days compared with 10 to 14 days results in an additional reduction of symptomatic VTE with a similar safety profile.

In situations where LMWH is unavailable (eg, formulary restrictions) or the patient has a history of heparin-induced thrombocytopenia, reasonable alternate choices include apixaban, dabigatran, rivaroxaban, VKA, fondaparinux, IPCD, or IPCD in combination with low-dose aspirin. The choice of a second-line strategy should be guided by its relative effectiveness, propensity to cause major bleeding (fondaparinux, rivaroxaban, VKA), and challenges with logistics and expected compliance (mechanical devices, VKA, and any drug that requires injections during the out-of-hospital period). Apixaban 2.5 mg bid taken orally as well as dabigatran 220 mg (with the availability of an alternate lower dose of 150 mg) once daily combined with no monitoring requirement appear to have the most of these desirable properties. However, long-term safety data (eg, the absence of clinically relevant liver toxicity) will be important when using these new oral antithrombotic agents.

Recommendations

2.3.1. In patients undergoing THA or TKA, irrespective of the concomitant use of an IPCD or length of treatment, we suggest the use of LMWH in preference to the other agents we have recommended as alternatives: fondaparinux, apixaban, dabigatran, rivaroxaban, LDUH (all Grade 2B), adjusted-dose VKA, or aspirin (all Grade 2C).

Remarks: If started preoperatively, we suggest administering LMWH ≥ 12 h before surgery. Patients who place a high value on avoiding the inconvenience of daily injections with LMWH and a low value on the limitations of alternative agents are likely to choose an alternative agent. Limitations of alternative agents include the possibility of increased bleeding (which may occur with fondaparinux, rivaroxaban, and VKA), possible decreased efficacy (LDUH, VKA, aspirin, and IPCD alone), and lack of long-term safety data (apixaban, dabigatran, and rivaroxaban). Furthermore, patients who place a high value on avoiding bleeding complications and a low value on its inconvenience are likely to choose an IPCD over the drug options.

2.3.2. In patients undergoing HFS, irrespective of the concomitant use of an IPCD or length of treatment, we suggest the use of LMWH in preference to the other agents we have recommended as alternatives: fondaparinux, LDUH (Grade 2B), adjusted-dose VKA, or aspirin (all Grade 2C).

Remarks: For patients in whom surgery is likely to be delayed, we suggest that LMWH be initiated during the time between hospital admission and surgery but suggest administering LMWH at least 12 h before surgery. Patients who place a high value on avoiding the inconvenience of daily injections with LMWH and a low value on the limitations of alternative agents are likely to choose an alternative agent. Limitations of alternative agents include the possibility of increased bleeding (which may occur with fondaparinux) or possible decreased efficacy (LDUH, VKA, aspirin, and IPCD alone). Furthermore, patients who place a high value on avoiding bleeding complications and a low value on its inconvenience are likely to choose an IPCD over the drug options.

2.4. For patients undergoing major orthopedic surgery, we suggest extending thromboprophylaxis in the outpatient period for up to 35 days from the day of surgery rather than for only 10 to 14 days (Grade 2B).