Prevention of VTE in Nonorthopedic Surgical Patients

2.1 ES vs No Prophylaxis

A Cochrane review summarized results of eight older trials of ES vs no prophylaxis, including four trials in general surgery and one trial each in orthopedic, cardiac, gynecologic, and neurosurgery.²⁷ The studies had many limitations, including small samples, incomplete blinding, uncertain concealment of treatment allocation, and use of fibrinogen leg scanning to identify asymptomatic DVT. Across all trials, ES reduced the odds of DVT (including distal and asymptomatic DVT) by 65%. A previous meta-analysis reported similar results for all DVT, but reductions in proximal DVT and PE were neither confirmed nor excluded.²⁴

More recently, a large, multicenter, randomized controlled trial in patients with acute stroke provided additional, indirect evidence by comparing thigh-length ES plus routine care with routine care alone (including the use of heparin, warfarin, or alteplase in 12% of participants). Reductions in the risk of fatal or nonfatal PE (OR, 0.65; 95% CI, 0.32-1.31) and symptomatic proximal DVT (OR, 0.84; 95% CI, 0.53-1.31) were neither confirmed nor excluded, but use of ES was associated with a fourfold increase in the risk of skin complications (5.1% vs 1.3%), including breaks, ulcers, blisters, and necrosis.²⁶ A subsequently published trial of thigh-length stockings vs calf-length stockings found that thigh-length stockings reduced the risk of symptomatic or asymptomatic proximal DVT by 31%, an absolute difference of 2.5 percentage points.²⁹ In this study, skin complications were observed in 3.9% of patients in the thigh-length ES group. The incidence of skin complications with ES in nonorthopedic surgical patients is likely to be lower than that observed in these trials of elderly patients with stroke who wore stockings for up to 30 days.

2.2 IPC vs No Prophylaxis

Several meta-analyses have compared IPC and no prophylaxis in mixed surgical populations.^23‐25 Urbankova et al²⁵ identified 15 trials, including five in orthopedics, four in general surgery, three in oncologic surgery, three in neurosurgery, and one in urology. Roderick et al²⁴ identified 19 trials, including five in general surgery, five in orthopedics, five in neurosurgery, two in gynecology, one in urology, and one in trauma. Many studies were limited by small samples, lack of blinding, unclear concealment of allocation sequence, and use of fibrinogen leg scanning (or less commonly, ultrasound or venography) to identify asymptomatic DVT, although DVT was subsequently confirmed by venography in most of the studies that used fibrinogen scanning. Both analyses found that compared with no prophylaxis, IPC reduced the risk of DVT (including asymptomatic and distal DVT) by 60%. In the analysis that examined proximal DVT, IPC reduced the odds by 50%.²⁴ Results failed to demonstrate or to exclude an effect on PE.²⁵ Other outcomes (fatal PE, skin complications) were not reported.

Adherence with IPC often is less than optimal. However, in one randomized trial of patients with acute spinal cord injury, 90% of participants were noted to use IPC for at least 75% of the recommended 22 h per day.³⁰ In another study, adherence with IPC was assessed at six times over a 24-h period in 227 nonambulatory trauma patients.³¹ Although full adherence was noted in only 19% of patients, overall adherence across all six measurements was 53%.

2.3 Unfractionated Heparin vs No Prophylaxis

Low doses (10,000-15,000 units/d) of subcutaneously administered unfractionated heparin have been evaluated in numerous randomized controlled studies in heterogeneous surgical populations. Moderate- to high-quality evidence comes from a meta-analysis that analyzed data from 69 studies of LDUH prophylaxis in general surgery, urological surgery, and orthopedic surgery.⁹ Many of the studies were limited by lack of blinding, unclear concealment of treatment allocation, and use of fibrinogen leg scanning to identify asymptomatic DVT. However, results were consistent with those from the International Multicenter Trial, a large randomized controlled trial with a low risk of bias.⁸ In our reanalysis of data from this meta-analysis, we found that LDUH was associated with an 18% reduction in the odds of death from any cause, a 47% reduction in the odds of fatal PE, and a 41% reduction in the odds of nonfatal PE, along with a 57% increase in the odds of nonfatal major bleeding (Figs S1-S5).

2.4 LMWH vs No Prophylaxis

A meta-analysis summarized data from eight trials of five different preparations of LMWH vs no prophylaxis in general or abdominal surgery.¹³ Two of eight studies were open label, and of three studies that reported symptomatic VTE, only one was potentially biased by the routine use of fibrinogen leg scanning to identify asymptomatic DVT. In the control groups, the pooled (baseline) risks of clinical PE, clinical VTE, and death were 0.5%, 0.9%, and 0.9%, respectively. Compared with no prophylaxis, LMWH reduced the risk of clinical PE and clinical VTE by ∼70%. In addition, LMWH was associated with a possible reduction in the risk of death from any cause (risk ratio [RR], 0.54; 95% CI, 0.27-1.10). LMWH led to an approximate doubling of the risks of major bleeding (RR, 2.03; 95% CI, 1.37-3.01) and wound hematoma (RR, 1.88; 95% CI, 1.54-2.28). Similar results were reported in the more recent meta-analysis by the British National Collaborating Centre for Acute Care, which included studies of GI, gynecologic, urological, and thoracic surgery.³²

2.5 LMWH vs LDUH

A meta-analysis of 51 randomized controlled trials compared LMWH and LDUH in > 48,000 general and abdominal surgery patients.¹³ About one-third of the studies were open label, and a majority used fibrinogen uptake scanning (with or without confirmatory venography) to identify asymptomatic DVT. In most studies, follow-up was for either 7 days or 1 month. Across all studies that reported clinical VTE events, the risk was ∼30% lower in the LMWH groups. However, this difference was not apparent when the analysis was restricted to blinded, placebo-controlled trials. In addition, results failed to demonstrate or to exclude a beneficial effect of LMWH vs LDUH on clinical PE, death from any cause, major bleeding, and wound hematoma. Similar results were reported in the meta-analysis by the British National Collaborating Centre for Acute Care.³²

2.6 Extended- vs Limited-Duration LMWH

The risk of VTE remains elevated for at least 12 weeks following surgery. A population-based, prospective study from the United Kingdom reported that compared with no surgery, the risk of VTE remained 10 to 50 times higher in weeks 7 to 12 following inpatient surgery.³³ In another study, the median time to postoperative VTE was 65 days.³⁴ Several studies compared extended-duration prophylaxis with LMWH (typically for 4 weeks) with limited-duration prophylaxis. Three systematic reviews summarized the results of these studies.^15‐17 Study limitations include an open-label design in two studies and measurement of asymptomatic DVT by venography as a surrogate outcome. All three analyses concluded that extended-duration prophylaxis reduced the risk of symptomatic or asymptomatic DVT by at least 50%, and two reported that proximal DVT was reduced by 75%. Results failed to demonstrate or exclude differences between groups in other outcomes, including major bleeding and death.

More recently, a multicenter, randomized, blinded, placebo-controlled trial compared an additional 3 weeks of pharmacoprophylaxis with bemiparin with no additional prophylaxis in 626 patients who underwent abdominal or pelvic surgery for cancer, all of whom received ∼1 week of prophylaxis with once-daily bemiparin.¹⁸ Surveillance venography was performed after 3 weeks, and patients were followed for clinical events for as long as 3 months. Approximately 20% of patients were excluded from assessment of the primary end point because venography was inadequate or not performed. The primary outcome was a composite of any DVT (including asymptomatic and distal events), nonfatal PE, and death from any cause. Although the risk of the composite outcome was 24% lower and the risk of proximal DVT was 88% lower in the extended-duration prophylaxis group, there were no symptomatic, nonfatal VTE events in either group. Although results failed to demonstrate or exclude a difference in bleeding, major bleeding was very uncommon, suggesting that any true underlying absolute differences will be small.

2.7 Fondaparinux vs LMWH

Fondaparinux was compared with the LMWH dalteparin in a blinded, randomized controlled trial of 2,927 patients at high risk for VTE who underwent abdominal (primarily GI) surgery.²⁰ Fondaparinux was associated with a possible reduction in asymptomatic or symptomatic DVT (RR, 0.75; 95% CI, 0.52-1.09), but results failed to demonstrate or exclude differences in the risks of fatal PE and nonfatal symptomatic VTE. There was a possible increase in the risk of nonfatal major bleeding with fondaparinux (RR, 1.43; 95% CI, 0.93-2.21), but differences in the risks of fatal bleeding and bleeding requiring reoperation were neither confirmed nor excluded.

Moderate-quality evidence from studies of patients undergoing elective hip replacement, elective knee replacement, and hip fracture surgery, when pooled with results of the previous study²⁰ in abdominal surgery, suggests that when compared with LMWH, fondaparinux does not reduce patient-important VTE events but leads to more major bleeding events.³⁵

2.8 Fondaparinux Plus IPC vs IPC Alone

Another placebo-controlled study compared fondaparinux plus IPC with IPC alone in 1,309 patients who underwent major GI, gynecologic, urological, or other abdominal surgery.¹⁹ In this study, the risk of any VTE (including asymptomatic DVT) was 69% lower in the fondaparinux group, and fondaparinux was associated with a possible reduction in the risk of proximal DVT (RR, 0.14; 95% CI, 0.02-1.14), but there was only one case of symptomatic VTE in each of the treatment groups. Major bleeding was more common among those who received fondaparinux (RR, 10.2; 95% CI, 1.31-79.7), but differences between the groups in fatal bleeding and bleeding requiring reoperation were neither confirmed nor excluded.

2.9 Low-Dose Aspirin (160 mg) vs No Prophylaxis

Perioperative use of low-dose aspirin was studied in orthopedic surgical patients in the PEP (Pulmonary Embolism Prevention) trial, a blinded, placebo-controlled study of > 13,000 patients undergoing hip fracture surgery and almost 4,100 patients undergoing elective arthoplasty.²¹ The treatment group received aspirin 160 mg/d for 35 days, with the first dose chewed prior to surgery. In our reanalysis of data from both hip fracture and elective arthroscopy patients (Figs S6-S11), benefits included a 28% reduction in the risk of nonfatal symptomatic DVT (RR, 0.72; 95% CI, 0.53-0.96) and a 58% reduction in the risk of fatal PE (RR, 0.42; 95% CI, 0.25-0.72), whereas harms included a possible increase in the risk of nonfatal myocardial infarction (RR, 1.59; 95% CI, 0.98-2.57). Differences between aspirin and placebo were neither confirmed nor excluded for other outcomes.

Strengths of the PEP trial include the very large sample, adequate blinding of patients and outcome adjudicators, adequate concealment of the allocation sequence, complete follow-up, and reporting of well-defined clinically important outcomes. However, although several types of nonfatal bleeding complications were reported, it is somewhat difficult to assess their severity. A potentially more important limitation is uncertainty about whether the results are applicable to nonorthopedic surgical patients. There have been no studies of low-dose aspirin in nonorthopedic surgical patients, and we consider higher doses of aspirin to be a distinct intervention with uncertain risks and benefits (Figs S12-S23). Because of concerns about indirectness, attendees at the AT9 final conference voted that low-dose aspirin should not be an alternative for pharmacologic prophylaxis in most nonorthopedic surgical patients. Our recommendations for low-dose aspirin, therefore, apply only in circumstances in which LDUH and LMWH are contraindicated or not available.

2.10 Mechanical vs Pharmacologic Prophylaxis

A meta-analysis identified 16 studies that compared mechanical prophylaxis with either LDUH or LMWH, including seven studies in general or abdominal-pelvic surgery, six in orthopedics, and three in trauma.²⁸ Studies compared heparin with IPC (nine studies), foot pump (four studies), or ES (three studies). Sample sizes ranged from 51 to > 2,000 participants. Patients and treating physicians were not blinded to treatment assignment, and radiologists were blinded in only six studies. Follow-up ranged between 3 and 6 weeks in most studies. When results from all studies were pooled, a difference in the risk of DVT (including asymptomatic and distal DVT) was neither confirmed nor excluded (RR, 1.07; 95% CI, 0.72-1.61). However, when the analysis was restricted to eight studies that compared mechanical prophylaxis with LMWH, the risk of DVT was 80% higher in the mechanical prophylaxis group (RR, 1.80; 95% CI, 1.16-2.79). The risk of major bleeding complications was 57% lower in those who received mechanical prophylaxis, with no difference in the relative risk of bleeding between studies of LDUH and LMWH.

2.11 Mechanical Prophylaxis Plus Pharmacologic Prophylaxis vs Pharmacologic Prophylaxis Alone

Ten studies compared ES plus pharmacologic prophylaxis with pharmacologic prophylaxis alone, including six in general or abdominal surgery^36‐41 and four in orthopedics.^42‐45 Background (pharmacologic) prophylaxis included LDUH in five studies, dextran in three studies, LMWH in one study, and aspirin in one study. Many of the studies were limited by small samples, incomplete blinding, uncertain concealment of the allocation sequence, and measurement of surrogate outcomes (Table S4). Pooling the results of these studies, we found that the addition of ES resulted in a 60% reduction in DVT (including asymptomatic and distal DVT) and a 72% reduction in proximal DVT, but a difference in the risk of PE was neither confirmed nor excluded (OR, 0.43; 95% CI, 0.16-1.18) (Figs S24-S27).

Five studies compared IPC plus pharmacologic prophylaxis with pharmacologic prophylaxis alone, including four studies in orthopedics and one study in general surgery.^46‐50 Background prophylaxis included LMWH (two studies), LDUH (one study), dextran (one study), and aspirin (one study). Once again, most studies were limited by small samples, incomplete blinding, unclear concealment of the allocation sequence, and measurement of surrogate outcomes. Pooled results across all five studies revealed a possible reduction in symptomatic or asymptomatic DVT (OR, 0.45; 95% CI, 0.20-1.03), but differences in proximal DVT or PE were neither confirmed nor excluded (Figs S24-S27).

For studies of both ES and IPC, reductions in symptomatic or asymptomatic DVT were similar across subgroups defined by surgical population (general and abdominal vs orthopedic), background agent used for pharmacoprophylaxis, whether the allocation sequence was adequately concealed, and whether there was blinded assessment of outcomes. Reductions in symptomatic or asymptomatic DVT differed depending on the test or tests used to identify and confirm DVT, with greater magnitudes of benefit observed in studies that used ultrasound (with or without confirmatory venography) or fibrinogen uptake with confirmatory venography than in those that used fibrinogen uptake or venography alone (Figs S28-S47).

2.12 IVC Filter vs No IVC Filter

The highest-quality evidence regarding the effectiveness of IVC filters is indirect, coming from a randomized controlled trial that compared IVC filter placement to no filter placement in patients with objectively confirmed, symptomatic, proximal DVT. In this study, filter placement was associated with a 78% reduction in the odds of symptomatic or asymptomatic PE at day 12, but after 2 years, there was an 87% increase in the odds of DVT, and a difference in PE was neither confirmed nor excluded.⁵¹ After 8 years of follow-up, a 9% absolute reduction in the risk of PE was offset by a 10% absolute increase in the risk of DVT.⁵²

More direct, but lower-quality evidence comes from a large, prospective cohort study that used propensity scoring methods to compare VTE outcomes among bariatric surgery patients with and without IVC filters.⁵³ Before propensity adjustment, patients with IVC filters had higher rates of postoperative VTE and death or serious disability. Following propensity adjustment, the difference in postoperative VTE was no longer statistically significant, but the risk of death or serious disability remained 2.5 times higher in the filter group.

A systematic review of seven nonrandomized studies in trauma reported that the pooled odds of PE were 79% lower (OR, 0.21; 95% CI, 0.09-0.49) among patients who received an IVC filter compared with historical control subjects who were variably matched for type of injury, age, sex, injury severity, and VTE risk.⁵⁴ A previous systematic review of 16 case series reported the following pooled risks after IVC filter placement: PE, 0.6%; DVT, 9.3%; insertion site thrombosis, 2%; IVC occlusion or thrombosis, 1.6%; placement complications, 1.4%; and filter migration, 0.4%.⁵⁵ Thus, although placement of an IVC filter probably reduces the risk of PE over the short term, complications appear to be frequent, and long-term benefits are unclear. Although retrievable filters have the potential to reduce long-term complications, they often are not removed.

2.13 VCU vs No VCU

Most studies of surveillance VCU have been performed in trauma patients. These patients often have contraindications to pharmacologic and mechanical prophylaxis, and the risk of VTE may be high even when prophylaxis is used.^56‐59 However, it is not clear that using VCU to detect and treat asymptomatic DVT reduces the risk of PE or fatal PE. Some studies have demonstrated that PE can occur even when VCU is negative.^60,61 A large retrospective study from a single center reported that over a 6-year period ending in 2000, the frequency of surveillance VCU decreased from 32% to 3.4%, with no increase in the incidence of PE.⁶¹ Furthermore, compared with venography, > 50% of the apparently positive findings on surveillance VCU may be false positives,³⁰ and the potential risks associated with treating false-positive findings are substantial.

2.14 Economic Evaluations of Interventions for Thromboprophylaxis

At least seven studies have examined economic outcomes associated with thromboprophylaxis in nonorthopedic surgical patients (Tables S5, S6). Most used a decision analysis approach and assumed a societal perspective in which all costs were considered. None of the results met prespecified criteria for upgrading or downgrading recommendations on the basis of resource use considerations.³

One study compared ES, IPC, LDUH, and no prophylaxis. Compared with no prophylaxis, ES saved 28 lives and reduced costs by $335,000 per 10,000 patients treated. Compared with ES, IPC saved six additional lives and cost an additional $413,000 per 10,000 patients treated, whereas LDUH saved seven additional lives and cost an additional $568,000 per 10,000 patients treated.⁶²

Four studies compared LMWH with LDUH in different surgical populations (colorectal, general, gynecologic, and abdominal surgery) within different health-care systems (Ontario, Canada; Germany; US Medicare).^63‐66 In two of these studies,^63,65 total costs associated with LMWH treatment were marginally higher than those for LDUH. In contrast, in a study of general surgical patients in Germany,⁶⁴ LMWH was more effective than LDUH by 0.01 quality-adjusted life years (QALYs) and was less expensive by $160 per patient treated. In another study in abdominal surgery patients that used Medicare reimbursement as a proxy for costs,⁶⁶ LMWH prophylaxis with dalteparin 5,000 units/d cost $21,800 per QALY gained relative to LDUH. One study compared LMWH plus IPC with IPC alone in gynecologic surgery patients and found that LMWH plus IPC cost between $7,200 and $20,000 per QALY gained.⁶⁷ Finally, one study compared enoxaparin and fondaparinux and reported that total hospital charges were higher for patients treated with enoxaparin.⁶⁸

3.1 Target Population: General and Abdominal-Pelvic Surgery, Including GI Surgery, Gynecologic Surgery, and Urological Surgery

This section covers general and abdominal-pelvic surgery. This group includes patients undergoing GI, urological, and gynecologic surgery as well as other general surgery patients (including those having operations on the breast and thyroid and parathyroid glands).

3.2 Target Population: Bariatric Surgery

Despite the explosion in the number of bariatric surgical procedures over the past 2 decades, few randomized controlled trials have evaluated interventions for VTE prophylaxis in these patients. Low-quality evidence comes from a number of uncontrolled and nonrandomized controlled studies (Table S9). We elected to apply higher-quality evidence about relative risks from randomized controlled trials in patients undergoing abdominal and pelvic surgery (section 3.1.2) when making recommendations for bariatric surgery patients, most of whom are at high risk for VTE.

3.3 Target Population: Vascular Surgery

Eight small randomized controlled trials of thromboprophylaxis have been performed in vascular surgery (Tables S2, S3).^109‐116 Most enrolled patients undergoing diverse vascular procedures, but two studied patients undergoing aortic surgery,^109,115 and one enrolled patients undergoing lower-extremity amputation.¹¹⁶ Three studies compared LDUH (with or without IPC) to no prophylaxis, one compared aspirin to no prophylaxis, and three compared LDUH to LMWH. One study compared LDUH, LDUH plus ergotamine, and dextran.¹¹² Studies were limited by small samples, incomplete blinding, unclear concealment of treatment allocation, and inconclusive results (Figs S48-S51). Because of these limitations, we apply more-precise estimates of relative risk from higher-quality studies in general and abdominal-pelvic surgery when making recommendations for vascular surgery patients.

3.4 Target Population: Plastic and Reconstructive Surgery

Because there have been no randomized controlled trials of thromboprophylaxis in plastic and reconstructive surgery, we applied indirect evidence about relative risks from trials in general and mixed surgical patients when making recommendations.

3.5 Explanation of Evidence Profiles and Rationale for Recommendations

We believe that the risk stratification scheme described in Table 5 is appropriate for use in general, GI, urological, gynecologic, bariatric, and vascular surgery patients. In addition, the Caprini score can be used in plastic and reconstructive surgery patients, although the baseline risk of VTE appears to be lower among these patients with any given Caprini score (Table 5). For example, although a Caprini score of 3 to 4 is associated with a moderate risk of VTE (∼3.0%) in general or abdominal-pelvic surgery, this same score is associated with a low risk of VTE (∼1.5%) in plastic and reconstructive surgery.

Information presented in the Table 8 can be used as a guide to help to identify patients in whom the risk of bleeding is high or the consequences of bleeding are especially severe. Statements about the quality of evidence refer to recommendations for patients undergoing general or abdominal-pelvic surgery. Because of indirectness, the quality of evidence should be rated down in other surgical populations.

Among patients with a very low risk of symptomatic VTE (< 0.5%), there is moderate-quality evidence that the harms of pharmacologic prophylaxis with LDUH or LMWH outweigh the benefits. Compared with no prophylaxis, one can expect zero to three fewer nonfatal VTE events and four to 10 more nonfatal major bleeding complications per 1,000 patients treated with LDUH. Trade-offs are similar for LMWH and no prophylaxis. There is low-quality evidence that compared with no prophylaxis, mechanical prophylaxis with IPC or ES can also be expected to prevent zero to three nonfatal VTE events at the expense of inconvenience, cost, and an uncertain number of skin complications, including breaks, blisters, ulcers, and necrosis, suggesting that the harms of mechanical prophylaxis probably outweigh the benefits in this very-low-risk group.

Among patients with a low risk of VTE (∼1.5%), moderate-quality evidence suggests that, compared with no prophylaxis, pharmacologic prophylaxis with either LDUH (Table 9) or LMWH (Table 10) can be expected to result in similar numbers of nonfatal VTE events prevented and nonfatal major bleeding events caused, and there is no important reduction in fatal PE. Low-quality evidence suggests that mechanical prophylaxis with either IPC (Table 11) or ES (Table 12) can be expected to prevent about eight to 10 nonfatal VTE events per 1,000 patients treated at the expense of an uncertain number of skin complications. Although direct high-quality evidence is lacking, we favor IPC over ES primarily on basis of indirect evidence from the Clots in Legs or Stockings after Stroke (CLOTS1) trial in patients with stroke that ES increased the risk of skin complications without reducing the risk of VTE.²⁶ Low-quality evidence favors mechanical prophylaxis over pharmacologic prophylaxis with either LMWH (Table 13) or LDUH (Table 14) in this group of patients.

Among patients with a moderate risk of VTE (∼3.0%) who are not at high risk for major bleeding complications, moderate-quality evidence indicates that compared with no prophylaxis, pharmacologic prophylaxis with either LDUH (Table 9) or LMWH (Table 10) will result in approximately twice as many nonfatal VTE events prevented as nonfatal major bleeding events caused. In addition, one can expect zero to three fewer deaths from PE per 1,000 patients treated. Low-quality evidence suggests that mechanical prophylaxis with either IPC (Table 11) or ES (Table 12) can be expected to prevent 13 to 17 nonfatal VTE events per 1,000 patients treated at the expense of an uncertain number of skin complications. Although low-quality evidence for the direct comparisons between mechanical prophylaxis and LMWH (Table 13) or LDUH (Table 14) seems to favor mechanical prophylaxis in this group of patients, three of the seven authors of this article placed more value on the higher-quality evidence favoring pharmacologic prophylaxis and, therefore, preferred pharmacologic prophylaxis over mechanical prophylaxis in this group. There is moderate-quality evidence that LMWH is at least as safe and effective as LDUH (Table 15).

Among patients with a moderate risk of VTE (∼3.0%) who are at high risk for major bleeding complications, moderate-quality evidence indicates that compared with no prophylaxis, pharmacologic prophylaxis with either LDUH (Table 9) or LMWH (Table 10) can be expected to result in similar numbers of nonfatal bleeding events caused and nonfatal VTE events averted. Although the quality of the evidence is low, the balance between desirable and undesirable outcomes appears to be more favorable with mechanical prophylaxis (Tables 11, 12), particularly IPC, which is expected to result in seven to 20 fewer nonfatal VTE events per 1,000 patients at the expense of an uncertain number of skin complications.

Among patients who are at high risk for VTE (∼6.0%) but not at high risk for major bleeding complications, there is high-quality evidence that compared with no prophylaxis, LDUH will result in one to eight fewer deaths from PE (Table 9), and there is moderate-quality evidence that LMWH prophylaxis may result in six fewer (95% CI, nine fewer to one more) deaths from PE (Table 10). In addition, there is moderate-quality evidence that both LDUH and LMWH will result in substantially more nonfatal VTE events prevented than nonfatal major bleeding events caused. Only low-quality evidence supports the use of mechanical prophylaxis with IPC or ES, which can be expected to result in 30 to 40 fewer nonfatal VTE events per 1,000 patients and an uncertain number of skin complications (Tables 11, 12). However, there is low-quality evidence that in this group of patients, use of mechanical prophylaxis compared with LMWH results in a similar number of major bleeding events averted and VTE events not prevented (Table 13). Nevertheless, we favor LDUH or LMWH over mechanical methods in this group because of the higher quality of evidence and the expected reductions in fatal PE.

Among patients with a high risk of VTE (∼6.0%) who are at high risk for major bleeding complications, moderate-quality evidence indicates that the trade-offs still favor pharmacologic prophylaxis with either LDUH (six fewer fatal PE, 33 fewer nonfatal VTE events, and 12 more nonfatal major bleeding events per 1,000 patients) or LMWH (six fewer deaths from any cause, 41 fewer nonfatal VTE events, and 23 more nonfatal major bleeding events per 1,000 patients) over no prophylaxis (Tables 9, 10). However, as the baseline risk of major bleeding approaches 4%, the harms of pharmacologic prophylaxis begin to outweigh the benefits, suggesting that mechanical prophylaxis with IPC (Table 11) or ES (Table 13) should be chosen when the risk of bleeding is judged to be very high or the consequences of major bleeding are believed to be particularly severe.

Among high-VTE risk patients, there is low-quality evidence (Tables 16, 17) that the absolute number of nonfatal VTE events can be further reduced by the addition of either IPC (10 fewer events per 1,000) or ES (11 fewer events per 1,000) to pharmacologic prophylaxis at the expense of an uncertain number of skin complications. The additional reduction in VTE applies to lower-risk groups as well, but the absolute number of events prevented is fewer.

Among patients at high risk for VTE undergoing abdominal surgery for cancer, there is moderate-quality evidence that compared with limited-duration prophylaxis (1 week), extended-duration prophylaxis (4 weeks) with LMWH provides additional protection from nonfatal VTE (13 fewer events per 1,000), without an important increase in the risk of nonfatal major bleeding complications (Table 18). The additional reduction in VTE applies to lower-risk groups as well, but the absolute number of events prevented is smaller. In addition, the quality of evidence is lower in noncancer surgery patients and patients at lower risk for VTE because of indirectness.

Some patients who would otherwise benefit from anticoagulant prophylaxis are not eligible to receive LDUH or LMWH primarily because of heparin allergy or a history of heparin-induced thrombocytopenia. Among such patients who are at high risk for VTE but not at increased risk for perioperative bleeding complications, low-quality evidence supports the use of either fondaparinux (Table 19), low-dose aspirin (Table 20), or mechanical prophylaxis (Tables 11, 12) over no prophylaxis. Among such patients at high risk for VTE who are at high risk for major bleeding, trade-offs favor mechanical prophylaxis. Because of the very low quality of the evidence and the availability of preferable alternatives, we do not recommend the use of high-dose aspirin for VTE prevention in any group of patients (Table 21).

For patients at high risk for VTE who are not candidates for either mechanical or pharmacologic prophylaxis, very-low-quality evidence suggests that IVC filter placement will probably cause at least as many DVT events as PE events prevented and that additional serious complications may occur in as many as 5% of patients (Table 22). Likewise, it is not clear that using VCU to detect and treat asymptomatic DVT reduces the risk of PE or fatal PE in patients at high risk for VTE. Furthermore, false-positive findings are common, and the potential risks associated with treating false-positive findings are substantial.

3.6 Recommendations

The following recommendations apply to patients undergoing general surgery, GI surgery, urological surgery, gynecologic surgery, bariatric surgery, vascular surgery, and plastic and reconstructive surgery (Table 23).

3.6.1. For general and abdominal-pelvic surgery patients at very low risk for VTE (< 0.5%; Rogers score, < 7; Caprini score, 0), we recommend that no specific pharmacologic (Grade 1B) or mechanical (Grade 2C) prophylaxis be used other than early ambulation.

3.6.2. For general and abdominal-pelvic surgery patients at low risk for VTE (∼1.5%; Rogers score, 7-10; Caprini score, 1-2), we suggest mechanical prophylaxis, preferably with IPC, over no prophylaxis (Grade 2C).

3.6.3. For general and abdominal-pelvic surgery patients at moderate risk for VTE (∼3.0%; Rogers score, > 10; Caprini score, 3-4) who are not at high risk for major bleeding complications, we suggest LMWH (Grade 2B), LDUH (Grade 2B), or mechanical prophylaxis, preferably with IPC (Grade 2C), over no prophylaxis.

Remarks: Three of the seven authors favored a strong (Grade 1B) recommendation in favor of LMWH or LDUH over no prophylaxis in this group.

3.6.4. For general and abdominal-pelvic surgery patients at moderate risk for VTE (∼3.0%; Rogers score, > 10; Caprini score, 3-4) who are at high risk for major bleeding complications or those in whom the consequences of bleeding are thought to be particularly severe, we suggest mechanical prophylaxis, preferably with IPC, over no prophylaxis (Grade 2C).

3.6.5. For general and abdominal-pelvic surgery patients at high risk for VTE (∼6.0%; Caprini score, ≥ 5) who are not at high risk for major bleeding complications, we recommend pharmacologic prophylaxis with LMWH (Grade 1B) or LDUH (Grade 1B) over no prophylaxis. We suggest that mechanical prophylaxis with ES or IPC should be added to pharmacologic prophylaxis (Grade 2C).

3.6.6. For high-VTE-risk patients undergoing abdominal or pelvic surgery for cancer who are not otherwise at high risk for major bleeding complications, we recommend extended-duration pharmacologic prophylaxis (4 weeks) with LMWH over limited-duration prophylaxis (Grade 1B).

Remarks: Patients who place a high value on minimizing out-of-pocket health-care costs might prefer limited-duration over extended-duration prophylaxis in settings where the cost of extended-duration prophylaxis is borne by the patient.

3.6.7. For high-VTE-risk general and abdominal-pelvic surgery patients who are at high risk for major bleeding complications or those in whom the consequences of bleeding are thought to be particularly severe, we suggest use of mechanical prophylaxis, preferably with IPC, over no prophylaxis until the risk of bleeding diminishes and pharmacologic prophylaxis may be initiated (Grade 2C).

3.6.8. For general and abdominal-pelvic surgery patients at high risk for VTE (∼6%; Caprini score, ≥ 5) in whom both LMWH and unfractionated heparin are contraindicated or unavailable and who are not at high risk for major bleeding complications, we suggest low-dose aspirin (Grade 2C), fondaparinux (Grade 2C), or mechanical prophylaxis, preferably with IPC (Grade 2C), over no prophylaxis.

3.6.9. For general and abdominal-pelvic surgery patients, we suggest that an IVC filter should not be used for primary VTE prevention (Grade 2C).

3.6.10. For general and abdominal-pelvic surgery patients, we suggest that periodic surveillance with VCU should not be performed (Grade 2C).

4.1 Baseline Risk, Risk Factors, and Risk Stratification for VTE

The risk of VTE following cardiac surgery is uncertain. Predisposing factors include perioperative stasis, inflammation, and activation of the coagulation system, but these are mitigated by early mobilization and use of anticoagulants, aspirin, and other antiplatelet drugs.

Relatively precise, but possibly dated estimates of the risk of VTE following cardiac surgery come from the California Patient Discharge Data Set for the years 1992 to 1996.⁷⁵ In this large data set, the risks of VTE in the 91 days after coronary artery bypass graft (CABG) and valve replacement surgery were 1.1% and 0.5%, respectively. Similarly, in an analysis of registry data from New York State in 1999, 133 of 16,325 (0.8%) patients were readmitted for VTE within 30 days after CABG.¹²⁶ Unfortunately, information about the use of prophylaxis was not reported in either study.

Based on results of these and other studies, we believe that most cardiac surgery patients are at moderate risk for VTE.^127‐131 Possible factors that increase the risk of VTE in cardiac surgery include older age,¹³⁰ postoperative complications,^127,130 prolonged preoperative hospitalization or postoperative recovery,^128,129 CABG surgery compared with valve surgery,¹²⁹ and off-pump CABG compared with cardiopulmonary bypass.¹³²

4.2 Baseline Risk, Risk Factors, and Risk Stratification for Major Bleeding Complications

A systematic review of English-language studies of surgical bleeding complications published between 1997 and 2007 identified six studies in cardiac surgery patients, including four retrospective cohort studies and two randomized trials.¹³³ In five studies, major bleeding was defined as bleeding requiring reexploration^90‐92,134; across these studies, the risk of major bleeding was remarkably consistent, with a median risk of 4.7% (range, 3.1%-5.9%). Thus, we classify most cardiac surgery patients as being at high risk for anticoagulant prophylaxis-related bleeding.

Risk factors for bleeding following cardiac surgery varied across studies (Table 8). One study found that the risk of bleeding was similar for on-pump compared with off-pump CABG.¹³⁵ Two others reported that the risk of bleeding was approximately twice as high in patients treated with aspirin⁹⁰ or clopidogrel, at least when given within 3 days of surgery.⁹¹ In one series of 2,898 consecutive patients undergoing CABG, independent risk factors for bleeding requiring reexploration included BMI ≥ 25 kg/m², nonelective surgery, placement of five or more grafts, and older age.⁹² In an earlier study of 6,015 patients undergoing cardiopulmonary bypass between 1986 and 1993, independent risk factors for bleeding included older age, renal insufficiency, operation other than CABG, and longer bypass time.⁹³

4.3 Explanation of Evidence Profiles and Rationale for Recommendations in Cardiac Surgery

We classify most patients undergoing cardiac surgery as being at moderate risk for VTE and at high risk for major bleeding complications. In these patients, low-quality evidence (moderate-quality evidence downgraded for indirectness) suggests that the benefits of pharmacologic prophylaxis with either LDUH (Table 9) or LMWH (Table 10) are probably outweighed by the potential harms. In contrast, low-quality evidence suggests that the balance between desirable and undesirable outcomes is more favorable with mechanical prophylaxis (Tables 11, 12), which is expected to result in 15 to 20 fewer nonfatal VTE events at the expense of an uncertain number of skin complications.

When additional risk factors for VTE are present and the baseline risk of VTE is high, the trade-offs still appear to favor mechanical prophylaxis over both no prophylaxis (Tables 11, 12) and pharmacologic prophylaxis (Tables 13, 14). However, the relatively high risk of postoperative bleeding almost surely decreases over time in patients whose hospital course is prolonged by one or more nonhemorrhagic surgical complications. We classify such patients as being at high risk for VTE and low (or average) risk for bleeding, and these patients may benefit from the addition of pharmacologic prophylaxis to mechanical prophylaxis, although the trade-offs only slightly favor combined prophylaxis over mechanical prophylaxis alone (Table 24).

4.4 Recommendations for Cardiac Surgery

4.4.1. For cardiac surgery patients with an uncomplicated postoperative course, we suggest use of mechanical prophylaxis, preferably with optimally applied IPC, over either no prophylaxis (Grade 2C) or pharmacologic prophylaxis (Grade 2C).

4.4.2. For cardiac surgery patients whose hospital course is prolonged by one or more nonhemorrhagic surgical complications, we suggest adding pharmacologic prophylaxis with LDUH or LMWH to mechanical prophylaxis (Grade 2C).

5.1 Baseline Risk, Risk Factors, and Risk Stratification for VTE

Based on results of observational studies and our clinical judgment, we consider most thoracic surgery patients to be at least at moderate risk for VTE. Three relatively large retrospective studies reported the risk of symptomatic VTE events. In one study of 693 thoracotomies for lung cancer, symptomatic VTE was observed in 1.7% of patients, including PE in 1.3%, despite routine use of prophylaxis with LDUH or LMWH. In another analysis of 1,735 lung resections for malignancy, autopsy-confirmed fatal PE occurred in 1.2% of patients, despite ongoing heparin prophylaxis in most of them.¹³⁸ Another study of 706 thoracic surgery patients reported objectively confirmed PE in 20 of 344 (7%) patients who did not receive prophylaxis, but there were no episodes of PE among 362 patients who wore IPC.¹³⁹ Finally, the 91-day risk of clinically detected VTE for almost 13,000 patients undergoing major lung resection for malignant disease was 1.6% in the California Patient Discharge Data Set.⁷⁵

Thoracic surgery patients undergoing extended pulmonary resection, pneumonectomy, extrapleural pneumonectomy, or esophagectomy are probably at higher risk for VTE. In a prospective study of 336 patients undergoing pneumonectomy for malignancy, the risk of symptomatic VTE was 7.4%.¹⁴⁰ Similarly, in a study of 496 patients undergoing extrapleural pneumonectomy for malignant mesothelioma, DVT occurred in 6.4% of patients, and fatal PE was observed in 1.2%.¹⁴¹ Other risk factors for VTE in thoracic surgery have not been rigorously evaluated, although individual studies have implicated malignancy, larger tumors, and pack-years of smoking as possible risk factors.^{140,142‐144} In another study, age, sex, BMI, operation time, time to ambulation, operative method, and malignancy were not associated with VTE.¹³⁹

5.2 Baseline Risk, Risk Factors, and Risk Stratification for Major Bleeding Complications

A review of complication rates from 14 studies of patients undergoing major lung resections for cancer distinguished between standard resection and pneumonectomy or extended resection.⁹⁴ Across nine studies of almost 17,000 patients who underwent standard resection, bleeding requiring reoperation was reported in 1%. However, in five studies of 1,223 patients who underwent pneumonectomy or extended resection, ∼5% of patients required reexploration for bleeding. In a more-recent retrospective analysis of 1,100 patients who underwent video-assisted thoroscopic lobectomy at a single center, intraoperative bleeding required conversion to thoracotomy in six (0.55%) patients, and 45 (4.5%) patients required postoperative red cell transfusion, but there were no episodes of fatal bleeding or bleeding requiring reoperation.¹⁴⁵

5.3 Explanation of Evidence Profiles and Rationale for Recommendations in Thoracic Surgery

Most thoracic surgery patients are at moderate risk for VTE. In these patients, moderate-quality evidence suggests that compared with no prophylaxis, pharmacologic prophylaxis with either LDUH or LMWH will result in more cases of VTE events prevented than bleeding episodes caused (Tables 9, 10). Low-quality evidence supports the use of mechanical prophylaxis over no prophylaxis, preferably with IPC (Tables 11, 12). The addition of mechanical prophylaxis with either ES (Table 16) or IPC (Table 17) to pharmacologic prophylaxis will prevent a few additional VTE events at the expense of skin complications, added cost, comfort, and convenience.

For thoracic surgery patients at high risk for VTE (including those undergoing extended pulmonary resection, pneumonectomy, extrapleural pneumonectomy, and esophagectomy), moderate-quality evidence suggests that when compared with no prophylaxis, the benefits of pharmacologic prophylaxis with LDUH (Table 9) or LMWH (Table 10) outweigh the harms. Because the risk of bleeding requiring reexploration appears to be elevated in patients who require pneumonectomy or extended-lung resection, prudence dictates that mechanical prophylaxis should be used until adequate hemostasis has been established and the risk of bleeding diminishes.

5.4 Recommendations for Thoracic Surgery

5.4.1. For thoracic surgery patients at moderate risk for VTE who are not at high risk for major bleeding, we suggest LDUH (Grade 2B), LMWH (Grade 2B), or mechanical prophylaxis with optimally applied IPC (Grade 2C) over no prophylaxis.

Remarks: Three of the seven authors favored a strong (Grade 1B) recommendation in favor of LMWH or LDUH over no prophylaxis in this group.

5.4.2. For thoracic surgery patients at high risk for VTE who are not at high risk for major bleeding, we suggest LDUH (Grade 1B) or LMWH (Grade 1B) over no prophylaxis. In addition, we suggest that mechanical prophylaxis with ES or IPC should be added to pharmacologic prophylaxis (Grade 2C).

5.4.3. For thoracic surgery patients who are at high risk for major bleeding, we suggest use of mechanical prophylaxis, preferably with optimally applied IPC, over no prophylaxis until the risk of bleeding diminishes and pharmacologic prophylaxis may be initiated (Grade 2C).

6.1 Baseline Risk, Risk Factors, and Risk Stratification for VTE

Although no VTE risk stratification scheme has been validated for patients undergoing craniotomy, data from published observational studies suggest that cancer, advanced age, longer duration of surgery, and paresis are associated with an increased risk of VTE.^148‐150 To estimate the baseline risks of nonfatal PE and symptomatic VTE in the absence of prophylaxis, we used data from an observational study of almost 2,400 neurosurgical admissions¹⁵¹ and a structured literature review that pooled results from numerous smaller studies of patients undergoing craniotomy.¹⁵² In the study by Chan et al,¹⁵¹ the risk of clinically diagnosed VTE within 30 days after discharge was 3.9% among all patients, and it was especially high for patients undergoing craniotomy for primary malignancy (7.5%) or metastasis (19%). In this study, pharmacologic and mechanical prophylaxis was used in 67% of patients with cancer and 84% of patients without cancer. Smaller studies of malignant glioma patients reported risks of symptomatic, postoperative DVT that ranged between 3% and 25%.¹⁵³ In the analysis by Danish et al,¹⁵² the pooled risk of symptomatic VTE was 2.1% across 13 studies that included almost 3,000 patients who received prophylaxis with IPC alone, whereas it was 2.2% in five studies that included > 3,500 patients who received combined prophylaxis with unfractionated heparin and IPC. Accordingly, we classify craniotomy patients as being at high risk for VTE, especially those who undergo craniotomy for malignancy.

6.2 Baseline Risk, Risk Factors, and Risk Stratification for ICH

In craniotomy patients, we focus on ICH rather than on the more generic major bleeding outcome because ICH is a potentially devastating complication of craniotomy and because pharmacologic VTE prophylaxis increases the risk of operative site bleeding. Although the risk of ICH probably varies depending on patient- and procedure-specific factors, we found no validated bleeding risk stratification system for craniotomy patients. Data from a structured literature review that included 20 different studies and > 31,000 patients who underwent craniotomy without pharmacologic prophylaxis indicate that the baseline risk of ICH is ∼1.1% (95% CI, 0.9%-1.4%).¹⁵² We favor this relatively precise estimate of baseline risk, although we recognize that it is higher than those reported in the meta-analysis by Collen et al,¹⁰ in which the pooled risks of intracranial hemorrhage were 0.04% (95% CI, 0%-3.7%) among those who did not receive pharmacologic prophylaxis, 0.35% (95% CI, 0%-7.4%) among those who received LDUH, and 1.5% (95% CI, 1.1%-1.9%) among those who received LMWH.

6.3 Explanation of Evidence Profiles

We classify patients undergoing craniotomy for nonmalignant disease as being at high risk for VTE (∼5%) and those with malignant disease as being at very high risk (≥ 10%). Although the baseline risk of bleeding (ICH) is probably ∼1%, the consequences of ICH are likely to be very severe.

For craniotomy patients at high risk for VTE (∼5%), such as those undergoing craniotomy for vascular disease, there is low-quality evidence that mechanical prophylaxis with IPC is beneficial. Compared with no prophylaxis, one can expect 11 to 40 fewer symptomatic VTE events per 1,000 patients treated with IPC (Table S10). As mentioned previously (Table S11), we favor IPC over ES primarily on the basis of indirect evidence from the CLOTS1 trial in stroke patients that ES increased the risk of skin complications without reducing the risk of VTE,²⁹ although differences between IPC and ES were neither demonstrated nor excluded in the meta-analysis of studies in neurosurgery.

In this group, there is moderate-quality evidence that compared with no prophylaxis, the benefits of pharmacologic prophylaxis with low-dose LMWH are probably outweighed by the harms (Table S12). First of all, LMWH was associated with a possible increase in the risk of death from any cause. In addition, although LMWH can be expected to prevent between eight and 36 VTE events, this comes at a cost of four to 22 additional intracranial bleeds. Based on the assumption that the disutility of intracranial hemorrhage is approximately two to three times greater than that associated with an average VTE event, the trade-offs favor no prophylaxis over LMWH. Although the trade-offs appear to be somewhat more favorable for LDUH compared with no prophylaxis (Table S13), the evidence is low in quality and sufficiently indirect to cast doubt on its relevance to craniotomy patients. Low-quality evidence for the comparison between LMWH and IPC suggests that the trade-offs favor IPC over pharmacologic prophylaxis in this group (Table S14).

For craniotomy patients at very high risk for symptomatic VTE (∼10%), such as those with cancer, low-quality evidence favors IPC, LDUH, and (possibly) LMWH over no prophylaxis (Tables S10, S12, S13). Trade-offs for the comparison between LMWH and IPC probably favor IPC, with six to 26 more nonfatal VTE events per 1,000 patients treated but four to 22 fewer episodes of nonfatal ICH (Table S14).

A more difficult question is whether and when to add pharmacologic prophylaxis to mechanical prophylaxis in the very-high-risk craniotomy patient. Indirect evidence from studies in patients undergoing abdominal or elective orthopedic surgery suggests that the addition of fondaparinux or warfarin to mechanical prophylaxis further reduces DVT or PE by ∼60%. Assuming that the risk of symptomatic VTE is 4.1% in those who receive IPC alone, low-quality evidence suggests that adding pharmacologic prophylaxis to IPC will prevent 23 additional VTE events per 1,000 patients treated (95% CI, 34 fewer to three more) at the expense of 11 more intracranial bleeds (95% CI, four more to 22 more) (Table S15). Because most ICH occur in the first 12 to 24 h after craniotomy, whereas approximately one-half of VTE events occur after the first week,¹⁵⁰ it is advisable to delay adding LMWH or LDUH until adequate hemostasis is established and the risk of bleeding is judged not to be excessively high.

6.4 Recommendations for Craniotomy

6.4.1. For craniotomy patients, we suggest that mechanical prophylaxis, preferably with IPC, be used over no prophylaxis (Grade 2C) or pharmacologic prophylaxis (Grade 2C).

6.4.2. For craniotomy patients at very high risk for VTE (eg, those undergoing craniotomy for malignant disease), we suggest adding pharmacologic prophylaxis to mechanical prophylaxis once adequate hemostasis is established and the risk of bleeding decreases (Grade 2C).

7.1 Baseline Risk, Risk Factors, and Risk Stratification for VTE

Three systematic reviews described in Appendix S1 have examined the baseline risk of VTE in spinal surgery.^159‐161 Most of the studies were limited by small sample sizes and the measurement of asymptomatic DVT, although one large retrospective study reported a very low risk of symptomatic DVT (0.05%) among 1,919 patients who received heparin prophylaxis and did not undergo routine surveillance for DVT.¹⁶²

Risk factors for VTE in spinal surgery patients likely include a combined anterior-posterior approach; multiple operative levels; and patient-related factors, such as older age, prior VTE, and malignancy.^163,164 In a population-based retrospective analysis of discharges from California hospitals in 1992 to 1996, the risk of symptomatic VTE within 91 days of surgery was 0.5% (95% CI, 0.4%-0.5%) among 34,355 patients who underwent spinal surgery for nonmalignant disease, whereas the risk of VTE was 2.0% (95% CI, 1.4%-2.6%) among 1,545 who underwent spinal surgery for malignant disease.⁷⁵ Accordingly, we classify the baseline risk of VTE in spinal surgery as low for most patients with nonmalignant disease and moderate for those with malignancy.

7.2 Baseline Risk, Risk Factors, and Risk Stratification for Major Bleeding Complications

In a large retrospective study of spinal surgery patients treated with nadroparin,¹⁶² major bleeding (defined as hemorrhage associated with a mass effect on postoperative spinal MRI and neurologic deterioration or a large-wound hematoma with intractable pain) was observed in 13 of 1,954 (0.7%) patients. In another observational study, 720 noncranial neurosurgical patients who were not at high risk for bleeding received twice-daily prophylaxis with LDUH. Two patients (0.3%) developed epidural hematomas that required reoperation.¹⁶⁵ In a small randomized trial of LDUH vs placebo, deep hematomas were noted in two patients in the placebo group and no patients in the heparin group.¹⁵⁴ In another trial comparing LMWH plus dihydroergotamine vs LDUH plus dihydroergotamine, there were no hematomas in either group, although increased intraoperative bleeding was noted to be more common in the LDUH group.¹⁵⁵ Based on these data, we believe that the baseline risk of major bleeding in spinal surgery is probably < 0.5%, but the consequences are potentially very severe.

7.3 Explanation of Evidence Profiles

Among spinal surgery patients at low risk for VTE, including those with nonmalignant disease, we estimate that compared with no prophylaxis, there will be similar reductions in the numbers of symptomatic VTE events when prophylaxis is given with IPC (five per 1,000), ES (six per 1,000), LDUH (five per 1,000), and LMWH (six per 1,000) (Tables S16-S19). These modest reductions are offset by similar increases in the absolute numbers of major bleeding complications with LDUH (three per 1,000) and LMWH (five per 1,000). Likewise, the benefits of IPC and ES are offset by an uncertain number of skin complications. Comparisons of IPC vs ES (Table S20), IPC vs LDUH (Table S21), and IPC vs LMWH (Table S22) suggest that the balance of desirable and undesirable outcomes favors IPC in these patients.

Among spinal surgery patients at moderate risk for VTE, including those with malignant disease and those undergoing surgery with a combined anterior-posterior approach, even greater reductions in symptomatic VTE events are anticipated with IPC (29 per 1,000), ES (31 per 1,000), LDUH (27 per 1,000), and LMWH (33 per 1,000), all compared with no prophylaxis (Tables S16-S19). Although the balance between benefits and harms favors either pharmacologic or mechanical methods over no prophylaxis, the trade-offs involved in the comparison between pharmacologic prophylaxis and mechanical prophylaxis with IPC are not as clear cut (Tables S21, S22). IPC may still be preferred over LMWH if the consequences of a nonfatal major bleeding event are believed to be at least two times more severe than those of nonfatal VTE.

7.4 Recommendations for Spinal Surgery

7.4.1. For patients undergoing spinal surgery, we suggest mechanical prophylaxis, preferably with IPC, over no prophylaxis (Grade 2C), unfractionated heparin (Grade 2C), or LMWH (Grade 2C).

7.4.2. For patients undergoing spinal surgery at high risk for VTE (including those with malignant disease and those undergoing surgery with a combined anterior-posterior approach), we suggest adding pharmacologic prophylaxis to mechanical prophylaxis once adequate hemostasis is established and the risk of bleeding decreases (Grade 2C).

8.1 Baseline Risk, Risk Factors, and Risk Stratification for VTE

Numerous studies have examined the risk of VTE in trauma (Appendix S1). Across four studies of patients with mixed trauma, the risk of symptomatic VTE ranged from > 1% to 7.6%.^61,177‐179 The risk is probably highest among patients with spinal trauma (2.2% despite near-universal prophylaxis), acute spinal cord injury (5%-6%), or traumatic brain injury (3%-5% among those who received pharmacologic prophylaxis within 24 to 48 h; up to 15% when initiation of pharmacologic prophylaxis was delayed beyond 48 h).^180‐185 A systematic review identified patients with spinal fractures (OR, 2.3; 95% CI, 1.4-3.6) or spinal cord injury (OR, 3.0; 95% CI, 1.8-5.4) as having a higher risk of VTE than other patients with trauma.¹⁸⁶ Older age has also been implicated as a risk factor for VTE in a number of studies.^177,178,187

Other independent risk factors for VTE, inconsistent across studies, included blood transfusion, surgery, femoral or tibial fracture, and spinal cord injury¹⁷⁷; head injury, major operation, lower-extremity fracture, venous injury, and (especially) > 3 days of mechanical ventilation¹⁷⁸; and male sex, black race, complete paraplegia (vs tetraplegia), and multiple comorbidities.¹⁸¹ Based on results of these studies, we believe that the baseline risk of VTE in most patients with major trauma is at least 3% to 5% and that the risk is even higher (8%-10%) among patients with traumatic brain or spinal cord injury and among those who require spinal surgery.

8.2 Baseline Risk, Risk Factors, and Risk Stratification for Major Bleeding Complications

Few studies have examined bleeding complications associated with thromboprophylaxis in trauma. In a prospective study of 525 patients with traumatic brain injury who were judged to be eligible to receive LMWH prophylaxis within 48 h of admission, progressive hemorrhagic changes were seen on head CT scan in 18 patients (3.4%), including in six (1.1%) in whom there was a change in management or outcome.¹⁸⁸ In a retrospective study of nosocomial complications in 525 adult patients with trauma, the reported risk of bleeding requiring red cell transfusion of > 4 units was 4.7%.¹⁷⁹

Another source of data to estimate the baseline risk of major bleeding complications comes from patients who were assigned to receive nonpharmacologic management in randomized trials of thromboprophylaxis. Unfortunately, only three trials in patients with trauma reported major bleeding complications in four groups that did not receive pharmacologic prophylaxis.^189‐191 In these groups, the pooled (random-effects) risk of major bleeding was 0.7% (95% CI, 0.2%-1.7%). This is likely to represent a lower boundary for the baseline risk of bleeding because patients judged to be at increased risk for bleeding were excluded from most trials of thromboprophylaxis. Relative contraindications to pharmacologic prophylaxis in trauma include severe head injuries, nonoperatively managed liver or spleen injuries, renal failure, spinal column fracture with epidural hematoma, severe thrombocytopenia, and coagulopathy.¹⁹²

8.3 Explanation of Evidence Profiles

For patients with major trauma who are at average risk for VTE and average risk for major bleeding, low-quality evidence suggests that pharmacologic prophylaxis with LDUH or LMWH can be expected to prevent approximately four times as many nonfatal VTE events as nonfatal bleeding complications caused (Tables S23, S24). Low-quality evidence suggests that mechanical prophylaxis with ES or IPC can be expected to prevent a similar number of nonfatal VTE events (Tables S25, S26) at a cost of an uncertain number of skin complications.

For patients with major trauma who are at especially high risk for VTE and average risk for bleeding complications (eg, acute spinal cord injury, spinal surgery for trauma), low-quality evidence suggests that pharmacologic prophylaxis with LDUH or LMWH can be expected to prevent almost 10 times as many nonfatal VTE events as nonfatal bleeding complications caused (Tables S23,S24). Moderate-quality evidence suggests that both LDUH and LMWH can also be expected to prevent four deaths from PE per 1,000 patients treated. The addition of mechanical prophylaxis can be expected to prevent 15 additional nonfatal VTE events per 1,000 patients treated (Table S27) at a cost of an uncertain number of skin complications.

For patients with major trauma at high risk for major bleeding (including those with traumatic brain injury), low-quality evidence suggests that the numbers of nonfatal VTE events prevented by pharmacologic methods are only slightly larger than the numbers of nonfatal major bleeding complications caused (Tables S23, S24). In these patients, mechanical prophylaxis with ES or IPC prevents sizable numbers of nonfatal VTE events at the expense of skin complications but without increasing the risk of bleeding (Tables S25,S26).

Few studies address the optimal duration of prophylaxis for patients with acute spinal cord injury. However, in a retrospective study of > 16,000 patients discharged from California hospitals between 1991 and 2001, > 90% of all thromboembolic events reported within 1 year after injury occurred in the first 91 days.¹⁷⁸ Pending further evidence, we agree with others¹⁹³ that 3 months is a reasonable time for VTE prophylaxis in most patients with acute spinal cord injury. Shorter durations may be appropriate for patients who regain purposeful movement of the lower extremities before 3 months, but further study is needed. For information about the use of IVC filters and DVT surveillance with VCU in trauma patients, please see sections 2.12, 2.13 and 3.5, and Table 22.

8.4 Recommendations for Patients With Trauma

Recommendations for patients with isolated lower-extremity injuries are provided by Falck-Ytter et al³⁵ in this supplement.

8.4.1. For major trauma patients, we suggest use of LDUH (Grade 2C), LMWH (Grade 2C), or mechanical prophylaxis, preferably with IPC (Grade 2C), over no prophylaxis.

8.4.2. For major trauma patients at high risk for VTE (including those with acute spinal cord injury, traumatic brain injury, and spinal surgery for trauma), we suggest adding mechanical prophylaxis to pharmacologic prophylaxis (Grade 2C) when not contraindicated by lower-extremity injury.

8.4.3. For major trauma patients in whom LMWH and LDUH are contraindicated, we suggest mechanical prophylaxis, preferably with IPC, over no prophylaxis (Grade 2C) when not contraindicated by lower-extremity injury. We suggest adding pharmacologic prophylaxis with either LMWH or LDUH when the risk of bleeding diminishes or the contraindication to heparin resolves (Grade 2C).

8.4.4. For major trauma patients, we suggest that an IVC filter should not be used for primary VTE prevention (Grade 2C).

8.4.5. For major trauma patients, we suggest that periodic surveillance with VCU should not be performed (Grade 2C).