Scope

The 2012 Comparative Effectiveness Review on Venous Thromboembolism Prophylaxis in Orthopedic Surgery¹³ (hereafter “the 2012 VTE report”) addressed many of the uncertainties in this area, including questions regarding the natural history of VTE, predictors of VTE, and the likelihood that DVTs result in PE in patients undergoing THR, TKR, or HFx surgery; the comparative efficacy of VTE prophylaxis strategies with no VTE prophylaxis, within and between classes of VTE prophylaxis modalities, and duration of VTE prophylaxis in patients undergoing these surgeries; and the efficacy of VTE prophylaxis in nonmajor orthopedic surgeries (knee arthroscopy, surgical repair of lower extremity injuries distal to the hip, and elective spine surgery). The 2012 VTE report included studies published from 1980 through May 2011. It found a general dearth of evidence regarding important clinical outcomes (nonfatal PE, fatal PE, major bleeding, reoperation), but high strength of evidence (SoE) that pharmacologic VTE prophylaxis reduces the risk of DVT compared to no VTE prophylaxis and increases the risk of minor bleeding. Comparisons of mechanical VTE prophylaxis versus no VTE prophylaxis did not provide strong evidence that mechanical prophylaxis reduced the risk of VTE, including, specifically, DVT. The comparisons of different classes of VTE prophylaxis modalities (e.g., different pharmacologic classes or pharmacologic versus mechanical devices) provided neither adequate evidence for important clinical outcomes nor strong evidence for other outcomes, including DVT. There were few studies evaluating the new FXaIs. In general, different interventions within classes were not statistically significantly different in their effects on DVT or bleeding. There was not strong evidence for other Key Questions.

We conducted a surveillance review of new studies potentially eligible to update all Key Questions from the 2012 VTE report. The surveillance review is summarized in the online protocol for this review.¹⁴ Upon discussion of the current state of the evidence with a panel of technical experts, we determined that a focused update of the 2012 Agency for Healthcare Research and Quality (AHRQ) report would be of greatest value. Based on their input and the findings of the surveillance review, we focused the update on comparisons between specific prophylaxis interventions; different classes of intervention; different doses, regimens, and treatment durations of interventions; different combinations of interventions; and different timing of starting prophylaxis (in relation to the time of surgery).

The objectives for the systematic review are to update the 2012 VTE report focused on the comparative effectiveness (for VTE outcomes and harms) of different thromboprophylaxis interventions for patients undergoing major orthopedic surgery (THR, TKR, and HFx surgery).

Key Questions

The following are the Key Questions (KQs) addressed by the review:

KQ 1 (update of original KQ 5)

In patients undergoing major orthopedic surgery (total hip or knee replacement, hip fracture surgery), what is the comparative efficacy between classes of thromboprophylaxis interventions on venous thromboembolism outcomes, treatment adherence, major bleeding, and other adverse events?

KQ 2 (update of original KQ 6)

In patients undergoing major orthopedic surgery (total hip or knee replacement, hip fracture surgery), what is the comparative efficacy of individual thromboprophylaxis interventions within classes (low molecular weight heparin, factor Xa inhibitors, direct thrombin inhibitors, and mechanical devices) on venous thromboembolism outcomes, treatment adherence, major bleeding, and other adverse events?

KQ 3 (new KQ based on original KQ 8)

In patients undergoing major orthopedic surgery (total hip or knee replacement, hip fracture surgery), what is the comparative efficacy of different doses, regimens, or treatment durations of the same thromboprophylaxis interventions (low molecular weight heparin, factor Xa inhibitors, direct thrombin inhibitors, and mechanical devices) on venous thromboembolism outcomes, treatment adherence, major bleeding, and other adverse events?

KQ 4 (update of original KQ 7 plus expansion)

In patients undergoing major orthopedic surgery (total hip or knee replacement, hip fracture surgery), what is the comparative efficacy of combined classes of thromboprophylaxis interventions versus single classes on venous thromboembolism outcomes, treatment adherence, major bleeding, and other adverse events?

KQ 5 (new KQ)

In patients undergoing major orthopedic surgery (total hip or knee replacement, hip fracture surgery), based on network meta-analysis, what are the comparative effects of thromboprophylaxis interventions on deep vein thrombosis and, separately, major bleeding?

5.1.

What are the comparative effects of different classes of thromboprophylaxis interventions?

5.2.

What are the comparative effects of different individual thromboprophylaxis interventions?

KQ 6 (new KQ)

In patients undergoing major orthopedic surgery (total hip or knee replacement, hip fracture surgery), what is the comparative efficacy of starting pharmacologic thromboprophylaxis at different times (i.e., preoperative, intraoperative, postoperative) on venous thromboembolism outcomes, treatment adherence, major bleeding, and other adverse events?

Search Strategy

A comprehensive search of the scientific literature was conducted to identify relevant studies addressing the KQs that have been published since the 2012 VTE report, which included studies published from 1980 through May 2011. We searched PubMed®, both the Cochrane Central Trials Registry® and Cochrane Database of Systematic Reviews®, and Embase® databases. Searches were limited to January 2010 through June 3, 2016. We included an overlap of more than 1 year with the search done for the 2012 VTE report. The updated literature searches replicated the searches from the 2012 VTE report and added additional terms for new treatments (e.g., factor Xa inhibitors [FXaI]). The search strategy was peer reviewed by an independent, experienced information specialist/librarian.

We also searched the ClinicalTrials.gov registry and the Food and Drug Administration, Healthy Canadians, and the U.K. Medicines & Healthcare products Regulatory Agency Web sites for relevant documents from 2011 through July 18, 2016. In addition, the reference lists of published clinical practice guidelines, systematic reviews, and Scientific Information Packages from manufacturers were hand-searched, and the Technical Expert Panel (TEP) members were invited to provide references of new studies. Existing systematic reviews were used primarily as sources of new studies. With the exception of studies included in the 2012 VTE report, we extracted and incorporated any studies de novo and did not summarize or incorporate the existing systematic reviews. All articles identified through these sources were screened for eligibility using the same criteria as was used for articles identified through literature searches.

All studies cited and tabulated in the 2012 VTE report were screened for eligibility on a par with new studies. However, as noted below, we relied on the summary tables in the 2012 VTE report for data from these studies.

Study Eligibility Criteria

The eligibility criteria for this update are mostly similar to the criteria used in the 2012 VTE report, as pertain to updated KQs.

Study Selection

We assessed titles and abstracts of citations identified from literature searches for inclusion, using the above eligibility criteria. Abstract screening was done in the open-source, online software Abstrackr (http://abstrackr.cebm.brown.edu/). Full-text articles of potentially relevant abstracts were retrieved and a second review for inclusion was conducted by reapplying the eligibility criteria. Both abstract and full-text screening was conducted in duplicate with conflicts resolved by reconciliation among the whole research team. All rejected full-text articles were confirmed by the project lead.

Studies included in the 2012 VTE report were reassessed for inclusion based on the summarized data available in the 2012 VTE report. In general, we did not confirm eligibility criteria for these studies from the full-text articles.

Data Extraction

Each study was extracted by one methodologist and confirmed by at least one other experienced methodologist. Disagreements were resolved by open, free-flowing discussion among the team to achieve consensus. Data extraction was conducted into customized forms in the Systematic Review Data Repository (SRDR) online system designed to capture all elements relevant to the KQs (http://srdr.ahrq.gov); the completed extraction forms are available for public review at this site. These included population characteristics, including description of patients’ surgery, descriptions of the interventions analyzed, descriptions of relevant outcomes, sample sizes, study design features, funding sources, results (including adverse events), and risk of bias assessment. The forms were tested on several studies and revised as necessary.

New studies added to the 2012 VTE report were extracted from the full-text articles and any available supplemental material. With few exceptions, eligible studies from the 2012 VTE report extracted and entered into SRDR based only on the available data presented in the 2012 VTE report.

Risk of Bias Assessment

We based the methodological quality of each study on predefined criteria. For RCTs, we used the Cochrane risk of bias tool,¹⁶ which asks about risk of selection bias, performance bias, detection bias, attrition bias, reporting bias, and other potential biases. For observational studies, we used selected questions from the Newcastle Ottawa Scale about comparability of cohorts, representativeness of the population, and adjustment for different lengths of follow-up.¹⁷ The methodological quality of the eligible studies from the 2012 VTE report was based solely on what was reported in that report’s methodological quality tables. Risk of bias questions included in the current review that were not assessed in the 2012 VTE report were marked as “NR” (not reported).

Data Synthesis

Grading the Strength of Evidence

We graded the strength of the body of evidence as per the AHRQ methods guide on assessing the SoE.²⁰ We assessed the SoE for each health outcome, as determined with input from the panel of technical experts: total VTE, symptomatic VTE, PE, DVT, and adverse events. Following the standard AHRQ approach, for each intervention and comparison of intervention, and for each outcome, we assessed the number of studies, their study designs, the study limitations (i.e., risk of bias and overall methodological quality), the directness of the evidence to the KQs, the consistency of study results, the precision of any estimates of effect, the likelihood of reporting bias, and the overall findings across studies. Throughout the report, all estimates with 95 percent CI or CrI beyond 0.5 and 2.0 were considered to be highly imprecise. Based on these assessments, we assigned a SoE rating as being either high, moderate, low, or there being insufficient evidence to estimate an effect. Conclusions based on statistically nonsignificant findings could have at best a low SoE. Outcomes with highly imprecise estimates, highly inconsistent findings across studies, or with data from only one or two studies were deemed to have insufficient evidence to allow for a conclusion (with the exception that particularly large, generalizable single studies could provide at least low SoE). The data sources, basic study characteristics, and each strength-of-evidence dimensional rating are summarized in a “Strength of Evidence” table detailing our reasoning for arriving at the overall SoE rating.²⁴

Peer Review

A draft version of this report was reviewed (from July 27 to August 23, 2016) by invited and public reviewers, including representatives from orthopedic societies, industry, our TEP, and the general public. These experts were either directly invited by the EPC or offered comments through a public review process. Revisions of the draft were made, where appropriate, based on their comments. The draft and final reports were also reviewed by the Task Order Officer and an Associate Editor from another EPC. However, the findings and conclusions are those of the authors, who are responsible for the contents of the report.

Summary of Studies

The literature searches yielded 1738 citations. We rescreened 118 studies that had been included in the 2012 VTE report and 107 references found in relevant existing systematic reviews. In total, 455 articles were screened in full text, of which 313 were excluded for the reasons listed in Figure 2 and Appendix B of the full report. The 142 studies included 127 RCTs and 15 NRCSs; they provided 85 studies of THR, 60 of TKR, and 12 of HFx surgery. The publication status and sources of the studies are listed in Figure 2 of the full report. The grey literature searches added two studies, both unpublished studies with results in ClinicalTrials.gov.

Studies evaluated the following thromboprophylaxis classes (and combinations thereof): antiplatelet drugs, DTI, FEI, FXaI, FXIi, LMWH, mechanical devices, UFH, and VKA. The studies evaluated the following specific interventions (and combinations thereof): aspirin (antiplatelet drug); dabigatran and desirudin (DTIs); TB402 (FEI); apixaban, darexaban, edoxaban, eribaxaban, fondaparinux, rivaroxaban, and TAK422 (FXaIs); factor XI antisense oligonucleotide (FXIASO; FXIi); dalteparin, enoxaparin, semuloparin, and tinzaparin (LMWHs); flexion devices, graduated compression stockings (GCS), intermittent pneumatic compression (IPC), and venous foot pumps (VFP) (mechanical devices); UFH; and warfarin (VKA).

We chose the principal outcomes for this review (the various VTE outcomes, major bleeding, and serious adverse events) based on an a priori determination of their importance in regards to thromboprophylaxis choice decisionmaking and the high likelihood that these outcomes would be available to researches of almost all RCTs. However, only total DVT was reported by more than 80 percent of the studies (82%), an arbitrary threshold we chose to suggest high risk of reporting bias. In descending order, the remaining principal outcomes were proximal DVT (66% of studies reported), total PE (52%), major bleeding (52%), fatal PE (48%), symptomatic DVT (40%), symptomatic VTE (18%), symptomatic PE (17%), total VTE (15%), and study-defined serious adverse events (11%).

Of note, almost all studies that reported serious adverse events did not define the outcome. Presumably, it included major bleeding, but this is not clear. Two studies described them as treatment-related events that lead to death, are life-threatening, require or prolong hospitalization, cause disability or incapacity, jeopardize the subject, or require an intervention. One study referred to “standard regulatory definitions”, but did not further define.

Key Question 1. Comparison of Thromboprophylaxis Intervention Classes

Note that the results of comparisons with sufficient evidence are summarized here; other comparisons were deemed to have insufficient evidence.

Total Hip Replacement

Key Points

• There were 46 RCTs and 5 NRCSs that compared classes of interventions in patients undergoing THR.
• Pairwise comparisons between classes had sufficient data for only six pairs of classes.
  • LMWH vs. DTI: Across outcomes there is a tradeoff between the two drug classes. Moderate SoE favors DTI to prevent total DVT and, separately, proximal DVT, but low SoE favors LMWH to avoid major bleeding.
  • LMWH vs. FXaI: Across outcomes, the evidence is inconsistent. The studies found that FXaI better lowers the risk of total VTE (low SoE), total DVT (moderate SoE), and proximal DVT (moderate SoE), but LMWH better lowers the risk of symptomatic VTE (low SoE) and symptomatic DVT (low SoE). There was high SoE that LMWH is better to prevent major bleeding, but both classes were similar in rates of study-defined serious adverse events (moderate SoE). The inconsistencies in these finding suggest important reporting bias.
  • LMWH vs. UFH: Overall, favors LMWH, with lower risk of total PE (high SoE), proximal DVT (moderate SoE), and major bleeding (moderate SoE); risk of total DVT was similar between drug classes (moderate SoE).
  • LMWH vs. VKA: Overall unclear. There is insufficient evidence regarding the relative benefit of either drug class to lower the risk of any VTE outcome, but VKA results in lower risk of major bleeding (high SoE).
  • LMWH vs. aspirin: Based primarily on a very large propensity-score-adjusted NRCS, LMWH and aspirin result in similar rates of total PE, symptomatic DVT, and major bleeding (all low SoE).
  • Mechanical devices vs. VKA: Overall, unclear. VKA results in lower risk of proximal DVT (high SoE), but insufficient evidence all favors mechanical devices to lower the risk of total DVT, and adverse events data have not been reported.
  • For all other class comparisons and outcomes there was insufficient evidence.
  • Although studies reasonably should have had data for all VTE-related outcomes and for major bleeding and other serious adverse events, most outcomes were not reported by many studies, resulting in a high risk of reporting bias across the evidence base.
• A within-study subgroup analysis by chronic kidney disease category was inconclusive regarding differential risks of bleeding with LMWH and DTI.
• Industry-funded studies had similar finding as other studies. Asian studies had similar findings as non-Asian studies.

Summary Results for THR Studies

Pairwise comparisons between classes had sufficient data for at least one outcome for six pairs of classes (Table A). For the comparison of LMWH versus DTI, among four RCTs, three favored DTI to prevent total DVT and to prevent proximal DVT. Meta-analysis of the four trials found a nonsignificant difference between drug classes regarding major bleeding favoring LMWH.

Table A

Total hip replacement, intervention class versus class: Summary of “sufficient” evidence.

LMWH versus FXaI: For the comparison of LMWH versus FXaI, among 13 RCTs there is high risk of reporting bias. Most meta-analyses of VTE outcomes significantly favored FXaI (total VTE [6 RCTs, low SoE], total DVT [10 RCTs, moderate SoE], and proximal DVT [10 RCTs, moderate SoE]). The meta-analyses of symptomatic VTE (7 RCTs, low SoE) and symptomatic DVT (9 RCTs, low SoE) found no significant differences between LMWH and FXaI, but favored LMWH; however, these RCTs mostly did not report other VTE outcomes. Major bleeding was significantly less likely with LMWH (10 RCTs, high SoE), but there was no significant difference in study-defined serious adverse events (5 RCTs, moderate SoE). Given the inconsistent findings across VTE outcomes, the relative benefit of either drug class is unclear.

LMWH versus mechanical devices: Among 3 RCTs of LMWH versus mechanical devices, none found significant differences for multiple VTE outcomes (total VTE, total PE, symptomatic PE, fatal PE, total DVT, proximal DVT). A NRCS found no difference in total PE. A single RCT reported significantly more frequent major bleeding with LMWH. Overall, the evidence was deemed to be insufficient to make conclusions about relative effect or harms between the two intervention classes.

LMWH versus UFH: From 10 RCTs, meta-analyses of LMWH versus UFH significantly favored LMWH to prevent total PE (8 RCTs, high SoE) and proximal DVT (6 RCTs, moderate SoE) and to avoid major bleeding (6 RCTs, moderate SoE), but showed no statistically significant difference in total DVT (10 RCTs, moderate SoE). Overall, the evidence favors LMWH.

LMWH versus VKA: Meta-analysis of the 4 RCTs of LMWH versus VKA found significantly lower rates of major bleeding with VKA (high SoE); however, the evidence regarding VTE is insufficient.

LMWH versus antiplatelet drug (aspirin): One very large NRCS (N=108,584) and another smaller NRCS (N=1533) compared LMWH versus antiplatelet drug (aspirin). The evidence suggests both drug classes have similar effects and harms. In both adjusted and propensity-score matched analyses, the very large NRCS found no differences in rates of total PE, symptomatic DVT, and major bleeding (all low SoE).

Mechanical devices versus VKA: Three RCTs evaluated mechanical devices versus VKA, overall yielding unclear findings regarding relative benefits and harms. The studies favored VKA to prevent proximal DVTs (high SoE), but insufficient evidence for total DVT favored mechanical devices, and there was no evidence regarding adverse events.

Other intervention classes compared by fewer studies (with insufficient evidence) included antiplatelet drug (aspirin) versus VKA (2 RCTs, one NRCS), LMWH versus antiplatelet drug (2 NRCSs), antiplatelet drug versus mechanical device (1 NRCS), mechanical device versus UFH (1 RCT), DTI versus FXaI (1 RCT), DTI versus UFH (2 RCTs), and FEI versus FXaI (1 RCT).

Subgroup Analysis in THR Studies

One RCT reported results for serious bleeding by level of chronic kidney disease in a comparison of LMWH and DTI. Event rates were low for all participants (2% in both the desirudin and the enoxaparin arms). They reported that for chronic kidney disease category 3B (n=569), more patients experienced a major bleed in the desirudin arm than in the enoxaparin arm, although the difference was not statistically significant (1.8% vs. 0.3%; P = 0.112). For chronic kidney disease category 3A (n=758), the rates were the same (0.3% in both arms). For chronic kidney disease categories 1–2 (n=700), DVT rates were lower in the enoxaparin arm (0.6% vs. 0%).

Studies were generally homogeneous in terms of patient eligibility criteria, such that most studies included all-comers without eligibility restrictions based on demographics, or other major patient or surgery subtypes. While some studies were restricted based on past bleeding history or chronic antiplatelet or VKA use, no RCTs were restricted to the converse populations (only patients with bleeding history or on antithrombotic medication). Thus, across-study comparisons of subgroup factors are limited.

Among THR RCTs, differences between studies based on industry funding was analyzable for only the comparison of LMWH versus UFH. For total DVT, by random effects model metaregression no significant difference (P=0.51) was found between the eight industry-funded studies (summary OR 0.91, 95% CI 0.59 to 1.41) and the two studies without reported industry support (summary OR 0.71, 95% CI 0.38 to 1.32). Similarly, for major bleeding, no significant difference (P=0.95) was found between the four industry-funded studies (summary OR 0.62, 95% CI 0.13 to 2.93) and the two studies without industry support (summary OR 0.56, 95% CI 0.26 to 1.20).

For the comparison of Asian versus non-Asian RCTs, only the comparison of LMWH versus FXaI was analyzable. For total DVT, no significant difference (P=0.56) was found between the five Asian studies (summary OR 1.63, 95% CI 0.81 to 3.31) and the four non-Asian studies (summary OR 2.08, 95% CI 1.40 to 3.09) by random effects model metaregression. The non-Asian studies included more patients, largely explaining the difference in statistical significance between the two sets of studies. Overall, the same percentage of Asian and non-Asian study participants had a DVT among these RCTs (4.7%). Similarly, for major bleeding, no significant difference (P=0.16) was found between the four Asian RCTs with major bleeding events (summary OR 1.95, 95% CI 0.46 to 8.22) and the five non-Asian studies (OR 0.68, 95% CI 0.49 to 0.94). Again, the non-Asian studies included more patients, largely explaining the difference in statistical significance between the two sets of studies. The Asian RCTs had relatively few events, with an overall major bleeding rate of 0.7 percent compared to 1.5 percent among all non-Asian RCTs (P=0.041); however, if the European study with an atypically high reported major bleeding rate (3.5%) is excluded, the non-Asian RCTs have a major bleeding rate of 0.9 percent, similar to the reported Asian rate (P=0.59).

Total Knee Replacement

Key Points

• There were 29 RCTs and 6 NRCSs that compared classes of interventions in patients undergoing TKR.
• Pairwise comparisons between classes had sufficient data for meta-analyses for only two pairs of classes.
  • LMWH vs. FXaI: Overall, the evidence is unclear. FXaI results in a lower risk of total VTE (low SoE), total DVT (low SoE), and proximal DVT (moderate SoE), but similar risks for total VTE (moderate SoE) and symptomatic DVT (low SoE); risk of major bleeding is lower with LMWH (low SoE) but risk of study-defined serious adverse events is lower with FXaI (low SoE).
  • LMWH vs. VKA: There is a tradeoff in risks between the two drug classes, such that LMWH better lowers risk of total DVT (high SoE) and proximal DVT (low SoE), but VKA has a lower risk of major bleeding (low SoE).
  • For all other class comparisons and outcomes there was insufficient direct comparative evidence.
  • Although studies reasonably should have had data for all VTE-related outcomes and for major bleeding and other serious adverse events, most outcomes were not reported by many studies, resulting in a high risk of reporting bias across the evidence base.
• A within-study subgroup analysis did not find a substantial difference in relative effect of antiplatelet drug vs. mechanical device between unilateral or bilateral TKR surgery.
• Industry-funded studies had similar finding as other studies. Asian studies had similar findings as non-Asian studies.

Summary Results for TKR Studies

Pairwise comparisons between classes had sufficient data for meta-analysis for only two pairs of classes (Table B).

Table B

Total knee replacement, intervention class versus class: Summary of “sufficient” evidence.

LMWH versus FXaI: For the comparison of LMWH versus FXaI, across 10 RCTs, meta-analysis significantly favored FXaI to prevent total DVT (7 RCTs) and proximal DVT (6 RCTs). While not statistically significant, the evidence favored FXaI to reduce the risk of total VTE (4 RCTs) with lower rates of study-defined serious adverse events (4 RCTs). Major bleeding occurred (nonsignificantly) less frequently with LMWH (7 RCTs). Rates of symptomatic DVT were the same with both drug classes (8 RCTs).

LMWH versus VKA: Among 4 RCTs that compared LMWH versus VKA, LMWH treatment resulted in less frequent total DVT (nonsignificantly) in 3 RCTs and proximal DVT across 4 RCTs (also not statistically significant); 4 RCTs found (nonsignificantly) lower risk of major bleeding with VKA.

Other intervention classes compared by fewer studies (with insufficient evidence) included antiplatelet drug versus FXaI (1 RCT), antiplatelet drug versus mechanical devices (1 RCT, 1 NRCS), antiplatelet drug (aspirin) versus VKA (1 RCT), DTI versus FXaI (1 RCT), LMWH versus antiplatelet drug (1 RCT), LMWH versus FXIi (1 RCT), LMWH versus mechanical devices (1 RCT and 1 NRCS), LMWH versus UFH (2 RCTs), and VKA versus mechanical devices (1 NRCS). Five RCTs evaluated LMWH vs. DTI but had highly inconsistent findings related to symptomatic DVT (3 RCTs) and rare episodes of major bleeding resulting in a highly imprecise effect estimate (5 RCTs).

Subgroup Analysis in TKR Studies

One RCT compared subgroups of patients who received unilateral or bilateral TKR surgery in a comparison of antiplatelet drug (aspirin) versus mechanical device; the trial was conducted in the 1980s and included an unrestricted sample of adult patients undergoing TKR. They found that in the unilateral surgery group (n=72) the percent of patients with a DVT was lower for those receiving mechanical prophylaxis through a compression boot (22%) compared to those receiving aspirin (47%, P<0.03). In the bilateral surgery group (n=47), DVT incidence was also lower in patients who used compression boots (48%) compared with those who received aspirin (68%), but this difference was not significant (P<0.20). Whether the treatment effect differed between unilateral and bilateral surgery subgroups was not analyzed.

Studies were generally homogeneous in terms of patient eligibility criteria, such that most across-study comparisons of subgroup factors are limited.

Among TKR RCTs, differences between studies based on industry funding was analyzable for only the comparison of LMWH versus FXaI. For total DVT, by random effects model metaregression no significant difference (P=0.21) was found between the six industry-funded studies (summary OR 2.04, 95% CI 1.68 to 2.49) and the single study without industry support (OR 4.71, 95% CI 1.31 to 16.9).

For the comparison of Asian versus non-Asian RCTs, only the comparison of LMWH versus FXaI was analyzable. For total DVT, no significant difference (P=0.97) was found between the four Asian studies (summary OR 2.15, 95% CI 1.35 to 3.41) and three non-Asian studies (summary OR 2.12, 95% CI 1.59 to 2.82) by random effects model metaregression. However, the total DVT rate was lower in the Asian RCTs (9.6%) than the non-Asian studies (16.0%, P<0.01). Similarly, for major bleeding, no significant difference (P=0.34) was found between the two Asian studies (summary OR 0.27, 95% CI 0.03 to 2.32) and the five non-Asian studies (OR 0.89, 95% CI 0.29 to 2.72). Major bleeding rates were similar between Asian studies (0.7%) and non-Asian studies (0.9%, P=0.57).

Hip Fracture Surgery

Key Points

• There were 6 RCTs that compared classes of interventions in patients undergoing HFx surgery.
• No drug class comparison had sufficient data for meta-analysis. One comparison had sufficient data for an effect conclusion.
  • LMWH vs. FXaI: Overall, the evidence is unclear. There is moderate SoE that LMWH results in a lower risk of total DVT. There is insufficient evidence for all other outcomes, including adverse events.
  • For all other class comparisons and outcomes there was insufficient direct comparative evidence.

Summary Results for HFx Studies

Only 6 RCTs of thromboprophylaxis have been conducted comparing intervention classes in patients undergoing HFx surgery. Pairwise comparisons between classes had sufficient data only for the comparison of LMWH versus FXaI (Table C). The 3 RCTs that compared LMWH versus FXaI found lower risk of total DVT with LMWH, but there was insufficient evidence regarding other outcomes. Other interventions classes compared included antiplatelet drug (aspirin) versus mechanical devices (1 RCT), antiplatelet drug (aspirin) versus VKA (1 RCT), and LMWH versus UFH (1 RCT); there was insufficient evidence regarding these comparisons.

Table C

Hip fracture surgery, intervention class versus class: Summary of “sufficient evidence”.

Key Question 2. Comparison of Within-Class Thromboprophylaxis Interventions

Relatively few RCTs of thromboprophylaxis compared specific interventions within any given class (3 for THR, 2 for TKR, and 2 for HFx surgery). No comparison was evaluated by more than two studies.

In patients undergoing THR or TKR (in separate analyses), one or two RCTs each evaluated enoxaparin versus semuloparin (LMWHs), enoxaparin versus tinzaparin (LMWHs), and graduated compression stockings versus intermittent pressure devices (mechanical devices). In patients with HFx surgery, one RCT each compared enoxaparin versus dalteparin (LMWHs) and enoxaparin versus semuloparin (LMWHs). Evidence was insufficient to evaluate within-class intervention comparisons.

Key Question 2. Comparison of Dosages and Treatment Durations of Thromboprophylaxis Interventions

Summary Results for Key Question 3

More than 300 specific comparisons of different drug doses or device regimens have been reported; the large majority of specific comparisons were made by a single study only. Comparisons with sufficient evidence are summarized here. These all pertain to class-level analyses; comparisons of individual thromboprophylaxis interventions within classes were not evaluated with sufficient frequency to allow a conclusion of sufficient evidence.

Total Hip Replacement

For three pairwise comparisons of dose or treatment duration, there was sufficient data (Table D). Among four RCTs comparing FXaI low versus high doses, meta-analysis yielded a nonsignificant effect favoring high dose FXaI to prevent total VTE. Data were insufficient for other outcomes.

Table D

Total hip replacement, comparison of different doses or treatment durations: Summary of “sufficient” evidence.

Five RCTs compared LMWH low versus high doses. Meta-analysis of the five RCTs found a nonsignificant effect on total DVT favoring higher dose LMWH. Meta-analysis found no difference in effect on proximal DVTs (4 RCTs). By meta-analysis, there was significantly less risk of major bleeding with lower dose LMWH (4 RCTs).

Among six RCTs of LMWH short versus long duration treatment, long duration LMWH resulted in fewer total PE (5 RCTs), but the summary OR was not statistically significant. Long duration LMWH resulted in statistically significantly lower risk of total DVT (6 RCTs) and proximal DVTs (5 RCTs). Data were insufficient for adverse events.

Total Knee Replacement

For only two pairwise comparisons of dose or treatment duration were there sufficient data (Table E). Among five RCTs of low versus high dose DTI, studies favored higher dose DTI (e.g., dabigatran 220 mg/day) over lower dose DTI (e.g., dabigatran 150 mg/day) to prevent total DVT (3 RCTs) and proximal DVT (4 RCTs). By meta-analysis the five RCTs nonsignificantly favored lower dose DTI to avoid major bleeding.

Table E

Total knee r eplacement, comparison of different doses or treatment durations: Summary of “sufficient” evidence.

Among four RCTs of low versus high dose FXaI, studies favored higher dose FXaI (multiple drugs, mostly twice the lower dose) over lower dose FXaI to prevent total VTE (4 RCTs), symptomatic DVT (4 RCTs), and proximal DVT (4 RCTs). Four RCTs were highly imprecise and inconsistent regarding difference in major bleeding risk, thus providing insufficient evidence.

Hip Fracture Surgery

One RCT each compared different duration FXaI and LMWH, providing insufficient evidence.

Key Question 4. Comparison of Single Versus Combination Thromboprophylaxis Intervention Classes

Key Question 5. Network Meta-Analyses Across Classes of Thromboprophylaxis Interventions

For all three major orthopedic surgeries, network meta-analyses that included more than sparse connections could be constructed only for total DVT and major bleeding. Due to incomplete and selective outcome reporting by most articles, other outcomes were too sparsely populated to allow interpretable networks. Overall, network meta-analysis findings were consistent with direct, pairwise comparisons, with the caveat that they pertain only to total DVT and major bleeding.

When interpreting the findings of the network meta-analyses, it is important to recognize that the exact ranking of interventions is susceptible to change with the addition of more studies. Interventions with relatively sparse data are likely to have imprecise rankings (i.e., to have flat rank graphs with similar likelihood across a range of ranks); see rank graphs for each network. Furthermore, while the pairwise comparisons with a network yield summary estimates and confidence intervals, the rankings of interventions are not supported by evaluations of statistical significance. Conclusions on total DVT may not translate to other, clinically significant, VTE outcomes, as suggested by the lack of correlation across studies between rates of total DVT and total PE.

Key Question 6. Comparison of Different Start Times of Thromboprophylaxis Interventions

Only two RCTs compared LMWH started at different times relative to THR surgery. No eligible studies evaluated patients with TKR or HFx surgery. There was insufficient evidence to yield conclusions.

Evidence Summary

Evidence and Analysis Limitations

As noted in the evidence summary, despite the large number of trials addressing thromboprophylaxis in patients undergoing major orthopedic surgery, there is inadequate evidence to confidently compare the effectiveness and the major adverse events of the myriad treatment options. As noted, the large majority of evidence pertains to LMWH (specifically enoxaparin), limiting the ability to compare all interventions. In particular, there are sparse RCTs or NRCSs that evaluated antiplatelet drugs (e.g., aspirin), VKA (e.g., warfarin), or mechanical devices.

The network meta-analyses provided greater power to compare all intervention classes and all interventions, but the sparseness of direct (within-study) comparisons for many of the interventions meant that meaningful conclusions could be derived for only a small subset of the interventions. However, the network meta-analyses are subject to important caveats. The sparseness of direct comparisons between most interventions within each network weakened the structure and the conclusions from the network meta-analyses. The only VTE outcome with sufficient evidence to allow network meta-analysis was total DVT, which is of questionable clinical significance since it includes asymptomatic and distal DVTs which have not been demonstrated to be associated with increased risk of PE. It is also important to recognize that the ranking of interventions by network meta-analysis may not be stable and may be susceptible to change with the addition of more studies; the ranking orders are also not supported by evaluations of statistical significance. However, network meta-analysis findings were consistent with direct, pairwise comparisons of interventions to lower the risk of total DVT and major bleeding.

Further hampering evaluation of the trials, studies were not consistent in which specific outcomes were reported. Notably only total DVT was reported by more than 80 percent of the studies. However, as discussed, this outcome is of unclear clinical importance. Only about half of studies reported major bleeding, the adverse event of greatest concern for most interventions. Most of the VTE outcomes were reported by 50 percent or fewer of the studies. Only one study reported all VTE and adverse event outcomes of primary interest to our panel of stakeholders and only two studies reported all VTE outcomes. Full reporting of VTE outcomes and adverse events by trials would have allowed greater SoE for almost all intervention classes and several specific interventions. However, studies arbitrarily or selectively reported specific outcomes.

Our analyses did not find significant evidence of bias due to industry funding, based on subgroup meta-analysis comparisons of industry-funded vs. other studies. However, 54 percent of the trials were industry-supported and only 13 percent of RCTs explicitly reported no industry support, which might partially explain the selective outcome reporting (although, we did not find evidence of such an association).^{25, 26} The relatively small number of RCTs available for meta-analysis for any given comparison and the small percentage of studies explicitly with no industry support meant that our analyses of industry funded required us to combine RCTs with no industry support and those that did not report funding source. If many of the studies that did not report funding were in fact industry-funded, then any real funding-source bias would have been diluted by the misclassification of funding source. Under the assumption that industry is most likely to fund and publish studies designed to be favorable to their products, the fact that the majority of evidence is industry-supported may explain the selective outcome reporting across studies (if favorable outcomes were more likely to be reported and nonfavorable outcomes omitted), the preponderance of evidence regarding enoxaparin, the sparseness of evidence on aspirin and mechanical devices, and relative sparseness of head-to-head trials of newer drugs (as opposed to comparisons with UFH or placebo).

The RCTs were generally consistent in regard to their eligibility criteria, mostly including all-comers without contraindications. This approach improves the applicability of the individual trials (and thus of the systematic review). Nonetheless, effect sizes in subgroups were rarely reported in these RCTs, and it greatly hampered our ability to evaluate potential explanations for heterogeneity or to hypothesize about possible subgroup differences based on patient history or surgery or anesthesia characteristics. Other than funding source, we were able only to evaluate potential differences between Asian and non-Asian studies. Overall, we found no significant difference between studies conducted in different regions (among analyzable studies), except major bleeding for the comparison of LMWH and FXaI in patients undergoing THR (summary OR in Asian RCTs 1.95, 95% CI 0.46 to 8.22; summary OR in non-Asian studies 0.68, 95% CI 0.49 to 0.94). Nevertheless, the event rates in the Asian studies were generally lower than the non-Asian studies. It suggests incomparability in the two populations besides ethnicity, which might explain the potential difference in the treatment effects. Only two RCTs reported on within-study subgroup analyses based on chronic kidney disease category (major bleeding, enoxaparin vs. desirudin) and by unilateral versus bilateral TKR surgery (DVT, aspirin vs. compression boots). Neither study found a significant difference in treatment effect in the different subgroups. Differences in effectiveness and safety between numerous different subgroups could not be evaluated due to lack of reporting of such analyses, including by age, sex, race, thrombosis risk factors, bleeding risk factors, comorbidities, medication use, or surgery types or techniques.

Of note, this review evaluated the evidence as per the a priori protocol, which was built off of, and relied on, the 2012 VTE report.¹³ Acknowledging that evidence for some interventions (e.g., mechanical devices) was likely to be sparse, we included larger NRCS. However, the smaller NRCSs that were excluded may have provided additional evidence, particularly for mechanical devices. While we did not reevaluate (mostly old) placebo-controlled RCTs among the direct comparisons between interventions, these studies were included in the NMAs. This review also did not cover numerous pertinent clinically important questions including comparisons of different strategies (e.g., aspirin and mechanical devices for low-risk patients and LMWH for high-risk patients). There are multiple standard methods for accounting for evidence in three (or more) arm studies in meta-analyses, when two (or more) of the arms are the same intervention (e.g., at different doses). In these instances, we chose the simplest method, which may be most clinically relevant in that we chose to analyze only the FDA-approved dose. When this was not possible, we selected the arm with the largest sample size (among FDA-approved or commonly used doses).

Future Research Recommendations

Much of the evidence base is insufficient to allow confident conclusions. Much of this lack is due to a relative sparseness of evidence evaluating interventions other than LMWH, and enoxaparin in particular. A more complete evidence base for the other treatments would allow for a stronger ranking of intervention classes, and of specific interventions, in term of risk of VTE and risk of major bleeding (and other adverse events). In particular, there is only sparse or low SoE data on the comparative effectiveness of aspirin or mechanical devices with LMWH or other anticoagulants. Given the likely low risk of major bleeding and other adverse events with aspirin and mechanical devices, it would be clinically important to determine whether patients at low risk of VTE events, in particular, could get adequate VTE prophylaxis with these low-risk interventions. Currently, there has been substantially more research conducted in patients undergoing THR than TKR; further studies regarding TKR may be warranted. In particular, few RCTs have been conducted in HFx surgery.

To avoid real and perceived bias (including, in particular concerns about reporting bias), ideally, a greater number of studies should be funded independently of industry. Furthermore, to minimize bias, all studies should report the full range of outcomes of interest, regardless of study results. Trial registration in priori and standard reporting compliant with Consolidated Standards of Reporting Trials (CONSORT) statement also help reduce potential reporting bias. For VTE prophylaxis studies, there is a fairly standard list of VTE and adverse event outcomes that are generally accepted as being of interest. This systematic review covers a complete list of outcomes that should be reported by all studies. To reduce the risk of bias in systematic reviews, all outcomes, particularly symptomatic DVT and PE and including those with no events, should be reported. However, to improve applicability of future studies to real-world clinical practice (where radiographic searches for asymptomatic DVTs are not performed), we would recommend that RCT protocols not mandate postsurgical diagnostic testing for asymptomatic DVTs.

This review made no assumptions about unreported event rates. Therefore, since mechanical device studies rarely reported bleeding (or other adverse event) outcomes, our pairwise and network meta-analysis review of mechanical devices had insufficient evidence about risk of bleeding. Ideally, all existing RCTs should report their full set of outcome results. This can relatively easily be done by submitting trial results to a publicly-accessible registry such as ClinicalTrials.gov.

Larger RCTs should evaluate differences in treatment and adverse event effects in relevant subgroups of patients. Ideally, these analyses should be adequately powered. Based on our discussions with a panel of clinical experts and other key informants, the following subgroup analyses are of interest: sex, race/ethnicity, age, body weight, tobacco use, chronic disease, varicosities, history of bleeding disorders or surgical bleeding, prior VTE, presurgical use of antiplatelet drugs or warfarin, or hormones, unilateral versus bilateral surgery, use of cemented fixation, tourniquet use, tranexamic acid use, and anesthesia type. A small number of trials were explicitly limited to some of these subgroups (including no presurgical use of antithrombotics and unilateral surgery), the counterfactuals (e.g., only presurgical antithrombotics or bilateral surgery) have not been studied. Since it is unlikely that RCTs will focus on these rarer and higher-risk factors, it is more important for researchers to evaluate the subgroups within their studies, when available.

Conclusions and Clinical Implications

While a large body of RCT evidence exists on comparative effectiveness and harms of thromboprophylaxis interventions after major orthopedic surgery, none of the Key Questions are fully and adequately addressed. For most Key Questions, the evidence base was too sparse to allow conclusions with sufficient SoE. For the comparisons of different interventions classes, only selective pairs of intervention classes had sufficient evidence, but often only for selective outcomes. The largest body of evidence exists for THR, with fewer studies of TKR, and very few studies of HFx surgery. The large majority of head-to-head studies evaluated LMWH (enoxaparin, in particular) with relatively few studies evaluating other intervention classes. Only a small minority of studies reported no industry support. Studies did not regularly report on all VTE-related and adverse effect outcomes, resulting in important possible reporting bias. Studies mostly reported total DVT, an outcome with unclear clinical significance. Almost no studies reported subgroup analyses. These limitations restrict the conclusions that can be drawn from the body of evidence.

Based on head-to-head comparisons for which there is sufficient evidence to make conclusions, LMWH is more effective to prevent VTE outcomes (with moderate to high SoE) and safer to prevent major bleeding (moderate SoE) than UFH (in patients undergoing THR). There are tradeoffs between LMWH and DTI (for THR) such that DTI is more effective to prevent total and proximal DVTs (moderate SoE), but LMWH results in less major bleeding (low SoE). Similarly there are tradeoffs between LMWH and VKA (for TKR) such that LMWH is more effective to prevent proximal and total DVTs (low and high SoE, respectively), but VKA results in less major bleeding (low SoE). Based primarily on a very large, well conducted observational study (with propensity score analyses), there is low SoE that LMWH and aspirin result in similar rates of total PE, symptomatic DVT, and major bleeding after THR. Comparisons between LMWH and FXaI, and between other pairs of treatment classes, are inconclusive due to either conflicting evidence across specific types of VTE or different adverse events or because of insufficient direct comparative evidence.

Two other findings of note are that for both LMWH (in THR) and DTI (in TKR) there is variable SoE that higher dose LMWH or DTI is more effective to prevent DVT but lower doses result in less major bleeding. Evidence is insufficient regarding different doses of other drug classes, different durations of treatment, comparisons of specific interventions, evaluations of combinations of interventions, and comparisons of timing of when to start thromboprophylaxis.

Of particular note, the inconsistent evidence LMWH versus FXaI was very likely due to selective outcome reporting. As an example, for THR, among 11 RCTs, only 6 reported on total VTE (favoring FXaI) and only 7 reported on symptomatic VTE (favoring LMWH), of which only 3 trials reported both outcomes. Selective outcome reporting was a major concern across all the analyses and in this case may have resulted in inconsistent conclusions across outcomes.

Due to a lack of sufficient direct comparisons between interventions for most outcomes of interest, we were able to construct network meta-analyses (to simultaneously evaluate both direct and indirect comparisons among all interventions) only for total DVT and major bleeding. For these outcomes network meta-analysis found that, for THR there is moderate SoE that FXaI is most effective to prevent total DVT; LMWH has lower risk of major bleeding that FXaI (low SoE). For TKR, by network meta-analysis we can conclude only that there is low SoE that FXaI is more effective to prevent total DVT than LMWH; there is insufficient evidence regarding major bleeding. Data are too sparse for HFx surgery to make conclusions from network meta-analysis. These analyses pertain to total DVT and major bleeding only.

In the face of incomplete and unclear evidence, patient and clinician preferences and values regarding the relative importance of avoiding VTE (primarily DVT) and major bleeding (and subsequent sequelae). While clinicians, policymakers, and clinical practice guideline developers should consider this evidence regarding relative effectiveness and safety of different thromboprophylaxis regimens (and its deficiencies), it is reasonable to also consider other sources of evidence not covered here (e.g., other observational research and assumptions related to mechanisms of action) to aid with decisionmaking in the face of incomplete evidence.

Future studies, particularly of interventions other than enoxaparin, are needed to address most Key Questions. These studies, and if feasible existing studies, should report all VTE-related and adverse event outcomes. Larger trials should conduct and report subgroup analyses of interest. Ideally, more future studies should be funded independently of industry to avoid real and perceived bias.