Results of the search

See Figure 1.

Figure 1

---

---

Study flow diagram.

Included studies

See Characteristics of included studies.

All 25 included studies (ACT2545; APOLLO; AR3106116; ARTEMIS; Bern 2015; CALISTO; Cho 2013; DRI4090; DRI4757; EFFORT; EPHESUS; FONDACAST; Kolluri 2016; L‐8541; L‐8635; Li 2015; PEGASUS; PENTAMAKS; PENTATHLON; PENTATHLON 2000; PENTHIFRA; PENTHIFRA PLUS; Shen 2014; Yokote 2011; Fuji 2015) investigated fondaparinux for short‐term VTE prevention. No studies examined idraparinux or idrabiotaparinux for VTE prevention. Seventeen studies (ACT2545; APOLLO; AR3106116; ARTEMIS; CALISTO; DRI4090; DRI4757; EPHESUS; FONDACAST; L‐8541; L‐8635; PEGASUS; PENTAMAKS; PENTATHLON; PENTATHLON 2000; PENTHIFRA; PENTHIFRA PLUS) were run by a pharmaceutical company, two (EFFORT; Fuji 2015) received support from a pharmaceutical company and six (Bern 2015; Cho 2013; Kolluri 2016; Li 2015; Shen 2014; Yokote 2011) were not supported by a pharmaceutical company.

Fifteen studies (ACT2545; Bern 2015; Cho 2013; DRI4090; DRI4757;EPHESUS;Fuji 2015;L‐8541;L‐8635; PENTAMAKS; PENTATHLON; PENTATHLON 2000; PENTHIFRA; PENTHIFRA PLUS; Yokote 2011) included participants undergoing orthopaedic procedures, three (APOLLO; AR3106116; PEGASUS) included participants undergoing abdominal surgery, one (Shen 2014) included participants following surgical resection of oesophageal cancer, one (ARTEMIS) acutely ill hospitalised medical patients, one (FONDACAST) participants requiring rigid or semirigid immobilisation (e.g. with a plaster cast or brace), one (Li 2015) intensive care unit (ICU) patients with hypercoagulability secondary to traumatic infection, one (EFFORT) people undergoing bariatric surgery, one (Kolluri 2016) patients undergoing coronary artery bypass graft surgery and one (CALISTO) participants with superficial venous thrombosis. Fuji 2015 involved participants with renal impairment with creatinine clearance ≥ 20 to < 30 mL/min undergoing orthopaedic procedures in the fondaparinux group and its comparator edoxaban. The other studies did not involve participants with renal impairment.

In most studies, the follow‐up period was less than 30 days, except for CALISTO and FONDACAST, both of which had a follow‐up period up to 45 days. All included studies were RCTs with comparable baseline demographic characteristics of participants in different treatment groups and low withdrawal rates (< 15%, with most studies showing a withdrawal rate < 10%). Most studies were carried out in Western countries, except for Cho 2013, DRI4090, DRI4757, Fuji 2015, Yokote 2011, Li 2015, L‐8541, L‐8635 and Shen 2014. The first five studies were carried out in Japan, and the last four in China.

Shen 2014 used effective methods to warrant the randomisation sequence, effective allocation concealment and blinding of the outcome assessor but reported VTE rates for fondaparinux and LMWH groups instead of numbers of VTE events. We were unable to obtain these details when we contacted the study authors. Therefore, we did not include this study in the meta‐analyses.

Reports of L‐8541 and Yokote 2011 did not contain sufficient information to clearly stratify them into low risk or high risk of bias groups, as discussed above. The L‐8541 report did not mention whether personnel who evaluated outcomes were independent or were blind to group treatment details. Yokote 2011 was a single‐centre study that reported no reliable randomisation method. Therefore, we included L‐8541 and Yokote 2011 in the sensitivity analyses only. Investigators in the EFFORT study administered 5 mg/d fondaparinux in the study arm and 40 mg enoxaparin twice daily in the control arm after participants had undergone bariatric surgery. Doses of fondaparinux and enoxaparin used by EFFORT were double those used for prevention of VTE in other included studies for other types of patients. Therfore, we performed sensitivity analyses excluding this study to demonstrate whether this factor would affect the results.

APOLLO, ARTEMIS, Bern 2015, CALISTO, Cho 2013, EFFORT, EPHESUS, Fuji 2015, Kolluri 2016, PENTAMAKS, PENTATHLON 2000, PENTHIFRA and PENTHIFRA PLUS reported using adjunctive anticoagulation methods. APOLLO used an intermittent pneumatic compression device (IPC) equally in different treatment groups. CALISTO, Cho 2013, EPHESUS and PENTAMAKS mentioned that they used compression stockings equally in different treatment groups as the adjunctive anticoagulation method. CALISTO also used a low dose of oral aspirin (≤ 325 mg per day) or other antiplatelet agents as an adjunctive anticoagulation method, and again this was done equally between treatment groups. ARTEMIS, Bern 2015, EFFORT, Kolluri 2016, PEGASUS, PENTATHLON 2000, PENTHIFRA, PENTHIFRA PLUS and Yokote 2011 mentioned that researchers used adjunctive anticoagulation methods (aspirin, thienopyridines and non‐steroidal anti‐inflammatory drugs were discouraged during the study; other antiplatelet agents, intermittent pneumatic leg compression, dextran and anticoagulant or thrombolytic agents were prohibited; graded‐pressure elastic stockings were permitted; and early mobilisation was strongly recommended) but did not report whether they were used equally in the different treatment groups. Because the adjunctive methods used by most participants in seven studies (ARTEMIS; EFFORT; Kolluri 2016; PEGASUS; PENTATHLON; PENTHIFRA; PENTHIFRA PLUS) were graduated compression stockings (GCS) and IPC, and no evidence suggests that these mechanical methods would substantially affect the symptomatic VTE rate, we decided to include these in the meta‐analysis. We used Yokote 2011 only for the sensitivity analysis, as described above. Another reason for including these studies was that they included large samples. If we excluded these from the meta‐analysis, the result would be biased. In Bern 2015, participants had early postoperative ambulation. All participants wore pneumatic compression stockings while in‐patients. Elastic compression stockings were prescribed to be used after discharge until follow‐up ultrasonography. Hydroxyethyl starch (HES) 6% was allowed intraoperatively for case‐specific reasons. Use of platelet function suppressive drugs, such as non‐steroidal anti‐inflammatory drugs (NSAIDs), was discouraged but not prohibited by the protocol. As Bern 2015 was the single study comparing fondaparinux and warfarin for VTE prevention, we included this study to present its results. Fuji 2015 allowed concomitant physiotherapy (intermittent pneumatic compression devices or elastic stockings) throughout the treatment period. AR3106116 compared the efficacy of fondaparinux versus IPC for VTE prevention in patients undergoing abdominal surgery who were at high risk of VTE, so the two groups in this study received fondaparinux injection and IPC treatment separately.

ACT2545, EFFORT, EPHESUS, FONDACAST, L‐8635, Li 2015, PEGASUS, PENTAMAKS, PENTATHLON, PENTATHLON 2000, PENTHIFRA and Shen 2014 compared the efficacy and safety of fondaparinux versus LMWH (approximately 100 anti‐factor Xa units per kilogram per day). ACT2545, EPHESUS and PENTATHLON 2000 compared the efficacy and safety of fondaparinux (2.5 mg subcutaneous (sc) once daily) versus LMWH after total hip replacement and hip fracture surgery. PENTATHLON was a dose‐ranging study involving five fondaparinux dose groups (0.75 mg, 1.5 mg, 3.0 mg, 6.0 mg, 8.0 mg) and a 60 mg enoxaparin control group of people undergoing elective primary hip replacement or revision of a primary procedure. We chose the 1.5 mg and 3.0 mg groups and combined their data for inclusion; we included the enoxaparin group as a comparator. FONDACAST compared the efficacy and safety of fondaparinux 2.5 mg sc once daily versus LMWH in participants requiring rigid or semirigid immobilisation for at least 21 days and up to 45 days because of isolated non‐surgical below‐knee injuries that were treated with a plaster cast. L‐8635 and PENTAMAKS compared the efficacy and safety of fondaparinux 2.5 mg sc once daily versus LMWH after major knee surgery, L‐8635 after elective knee replacement and PENTAMAKS for major knee surgery. PEGASUS compared the efficacy and safety of fondaparinux 2.5 mg sc once daily versus LMWH in participants after major abdominal surgery who were at high risk of VTE. PENTHIFRA compared the efficacy and safety of fondaparinux (2.5 mg sc once daily) versus LMWH in participants undergoing standard surgery for fracture of the upper third of the femur, including the femoral head and neck. Li 2015 compared the efficacy and safety of fondaparinux (2.5 mg sc once daily) versus LMWH in ICU patients with hypercoagulability accompanied by traumatic infection. EFFORT compared the efficacy and safety of fondaparinux (5 mg sc once daily) versus LMWH (40 mg twice daily) in bariatric surgical patients. Shen 2014 compared the efficacy and safety of fondaparinux (2.5 mg sc once daily) versus LMWH in participants undergoing surgical resection for oesophageal cancer.

APOLLO, AR3106116, ARTEMIS, Bern 2015, CALISTO, Cho 2013, DRI4090, DRI4757, Kolluri 2016 and PENTHIFRA PLUS compared the efficacy and safety of fondaparinux versus placebo or the mechanical method IPC for VTE prevention. CALISTO investigated the efficacy and safety of fondaparinux 2.5 mg sc once daily for prevention of venous thromboembolic complications for acute symptomatic isolated superficial thrombophlebitis of the lower limbs. DRI4090 and DRI4757 were dose‐ranging studies conducted to determine the dose‐response effect of fondaparinux on the prophylaxis of VTE after total hip replacement (THR) and total knee replacement (TKR), respectively. These studies included four fondaparinux dose groups (0.75 mg, 1.5 mg, 2.5 mg, 3.0 mg) and a placebo control group. We combined the 1.5 mg, 2.5 mg and 3.0 mg groups into a single group and dropped the data for the 0.75 mg group because the 0.75 mg dose of fondaparinux was too small and was not effective for routine VTE prevention and treatment (PENTATHLON). In PENTHIFRA PLUS, participants received open‐label fondaparinux 2.5 mg sc once daily postoperatively for 7 ± 1 days (day 1 = day of surgery) after hip fracture surgery. They were then randomised to receive double‐blind fondaparinux 2.5 mg sc once daily for 21 ± 2 days or placebo for three weeks. AR3106116 compared fondaparinux (2.5 mg sc once daily) with IPC for VTE prevention. Cho 2013 investigated fondaparinux (2.5 mg sc once daily) in participants after unilateral total knee arthroplasty (TKA), and ARTEMIS focused on the efficacy and safety of the anticoagulant fondaparinux (2.5 mg sc once daily) in older acutely ill medical in‐patients at moderate to high risk of VTE. Kolluri 2016 compared fondaparinux 2.5 mg sc once daily versus placebo for prevention of VTE after coronary artery bypass graft surgery. APOLLO compared fondaparinux 2.5 mg sc once daily combined with IPC versus IPC alone in participants at high risk of VTE after major abdominal surgery.

Bern 2015 compared fondaparinux 2.5 mg sc once daily versus variable (target international normalised ratio (INR) 2.0 to 2.5) and fixed (1 mg) doses of warfarin after elective hip or knee replacement surgery.

Fuji 2015 compared fondaparinux (1.5 mg sc once daily) versus edoxaban (15 mg oral once daily) in participants with serious renal impairment undergoing orthopaedic procedures.

We also identified six ongoing studies (EUCTR2007‐003746‐15‐DE; EUCTR2008‐001779‐31‐IT; JPRN‐UMIN000002444; JPRN‐UMIN000007005;JPRN‐UMIN000008435;PROTECT), which were recruiting participants and had reported no results at the time of the search for this review.

Excluded studies

See Characteristics of excluded studies.

In total, we excluded 33 studies (ACT1840; Amadeus 2008; AR3106206; AR3106333; AR3106335; Argun 2013; Bonneux 2006; Buller 2014; Cassiopea; Cohen 2007; EQUINOX; extended van Gogh; FLEXTRA; Kawaji 2012; Li 2013; MATISSE‐DVT; MATISSE‐PE; NCT00521885; NCT00539942; PENTATAK; PERSIST; Rembrandt; SAFE‐AF; Sasaki 2009; Sasaki 2011; Savi 2005; Tsutsumi 2012; van Gogh‐DVT; van Gogh‐PE; Xin 2013; Yamaoka 2014; Zhao 2013; Zhao 2015). Cassiopea, extended van Gogh, PERSIST, van Gogh‐DVT and van Gogh‐PE investigated idraparinux or idrabiotaparinux for long‐term VTE treatment, not for primary VTE prevention. Amadeus 2008 focused on idraparinux for prevention of thromboembolism in participants with atrial fibrillation. Because most thrombotic events in this study among participants with atrial fibrillation consisted of arterial embolism (stoke or non‐central nervous system systemic embolism), not venous thrombosis, published results did not distinguish the arterial embolism and venous thrombosis events; therefore, we excluded this study from the review. Buller 2014 focused on idrabiotaparinux for prevention of thromboembolism in participants with atrial fibrillation. Published results did not distinguish systemic embolism from venous thrombosis events; therefore, we excluded this study from the review. We excluded AR3106206, Rembrandt, MATISSE‐DVT and MATISSE‐PE because researchers investigated initial treatment of VTE, but not prophylaxis of VTE. In AR3106333, AR3106335, Cohen 2007, EQUINOX, FLEXTRA and PENTATAK, all study groups received pentasaccharides for VTE prevention, so investigators included no comparator. Kawaji 2012, Tsutsumi 2012 and Yamaoka 2014 were not RCTs, and Sasaki 2009 and Sasaki 2011 were quasi‐RCTs, so we excluded them. In ACT1840, the fondaparinux dose (3 mg sc twice daily) markedly extended the normal VTE prevention dose; therefore we excluded this study (PENTATHLON). NCT00539942 and NCT00521885 were terminated owing to problems with accrual and reported no results. Savi 2005 reported no VTE rate; as this was our primary outcome, we excluded this study. Argun 2013 and Bonneux 2006 reported no reliable primary efficacy and safety results. Li 2013, SAFE‐AF, Xin 2013, Zhao 2013 and Zhao 2015 investigated the efficacy and safety of fondaparinux for treatment of acute coronary syndrome and atrial fibrillation, not for VTE prevention.

Allocation

Seventeen studies (ACT2545;APOLLO; AR3106116; ARTEMIS; CALISTO; DRI4090; DRI4757; EPHESUS; FONDACAST; L‐8541; L‐8635; PEGASUS; PENTAMAKS; PENTATHLON; PENTATHLON 2000; PENTHIFRA; PENTHIFRA PLUS) were multi‐centre RCTs organised by a pharmaceutical company. Five of them (ARTEMIS; CALISTO; PENTAMAKS; PENTATHLON 2000; PENTHIFRA PLUS) clearly described in their published articles the use of computerised or central randomisation at an independent centre and were assessed to be at low risk of selection bias. We assumed that the remaining 12 multi‐centre randomised studies organised by the same company (GlaxoSmithKline) (ACT2545; APOLLO; AR3106116; DRI4090; DRI4757; EPHESUS; FONDACAST; L‐8541; L‐8635; PEGASUS; PENTATHLON; PENTHIFRA) were also randomised centrally by reliable methods. In addition, we believed that the 17 studies had sufficient allocation concealment and assessed them as having low risk of selection bias. Another seven studies also used reliable randomisation methods (Bern 2015; Cho 2013; EFFORT; Fuji 2015; Kolluri 2016; Li 2015; Shen 2014); we therefore assessed them as having low risk of selection bias. Yokote 2011 was a single‐centre study without sufficient information to estimate whether its randomisation was reliable, and we assessed it as having unclear risk of selection bias. We did not include this study in the efficacy analysis but did include it in a sensitivity analysis.

Blinding

ACT2545, AR3106116, FONDACAST, Fuji 2015 and L‐8635 were open‐label studies. ACT2545, FONDACAST and L‐8635 compared fondaparinux versus LMWH. AR3106116 compared fondaparinux versus IPC, and this was impossible to blind to both participants and personnel. Although the five studies were open‐label studies, their outcome evaluators were independent and were blinded to the randomisation. Therefore, we believed that failure to realise blinding to participants and healthcare personnel might not cause serious bias, and we included these trials in the data analyses. We therefore assessed these five studies as having unclear risk of performance bias but low risk of detection bias. Bern 2015, DRI4090, DRI4757, PENTATHLON, Shen 2014 and Yokote 2011 did not clearly describe the method used to ensure blinding of participants and personnel. However, they did describe effective methods that they used to ensure that outcome evaluators were blinded to the randomisation. According to our protocol, we included these studies in the data analyses. We assessed Bern 2015, DRI4090, DRI4757, PENTATHLON, Shen 2014 and Yokote 2011 as having unclear risk of performance bias but low risk of detection bias. Kolluri 2016 did not clearly describe whether outcome assessors were blinded; we therefore judged this study to have unclear risk of detection bias. Li 2015 reported reliable blinding of participants and outcome assessors but not of healthcare personnel; we therefore judged this study to have unclear risk of performance bias but low risk of detection bias. L‐8541, a single‐blind study, did not mention whether investigators used methods to ensure that outcome assessors were blinded to the treatment that participants received. We assessed L‐8541 to have unclear risk of performance and detection bias. We did not include this study in the data analysis but included it in a sensitivity analysis.

The remaining included studies were all triple‐blind and reported use of effective blinding strategies; we judged these studies to have low risk of performance and detection bias.

Incomplete outcome data

More than 90% of participants in every treatment group in all but two studies (PENTATHLON; PENTHIFRA PLUS) completed treatment as planned in the protocol, and we included them in the analyses of study results. In PENTATHLON, 10.4% of participants in the comparator enoxaparin group withdrew from the study, and in PENTHIFRA PLUS, 12.3% and 13.9% of participants from the fondaparinux group and placebo group, respectively, discontinued studies midway. We deemed all studies to have low risk of attrition bias.

Selective reporting

All included studies reported the primary efficacy outcomes listed in the Methods section of this review. However, Shen 2014 reported VTE rates rather than numbers of events, and we could not include this study in the meta‐analysis. All but one study (Shen 2014) reported the primary safety outcome. Shen 2014 reported drainage volume rather than major bleeding; we therefore judged this study to have unclear risk of reporting bias.

We used the funnel plot to detect publication bias.

Other potential sources of bias

All studies except Bern 2015, Cho 2013, Kolluri 2016, Li 2015, Shen 2014 and Yokote 2011 were organised by pharmaceutical companies, and EFFORT was supported by a pharmaceutical company, which provided the study medication. Therefore, we judged all studies except Bern 2015, Cho 2013, EFFORT, Kolluri 2016, Li 2015, Shen 2014 and Yokote 2011 as having unclear risk of other bias.

Fondaparinux versus placebo

Eight studies (APOLLO; ARTEMIS; CALISTO; Cho 2013; DRI4090; DRI4757; Kolluri 2016; PENTHIFRA PLUS) compared the efficacy and safety of fondaparinux versus placebo for VTE prevention. As the outcomes total VTE, total DVT and proximal DVT showed significant heterogeneity, we analysed the data by using the random‐effects model. Remaining outcomes (symptomatic VTE, total PE, fatal PE, non‐fatal PE, major bleeding, fatal bleeding, MI, all causes of death, other serious adverse effects) showed low heterogeneity and were analysed with the fixed‐effect model. Our results showed less total VTE (RR 0.24, 95% CI 0.15 to 0.38; P < 0.00001; 8 studies; 5717 participants), less symptomatic VTE (RR 0.15, 95% CI 0.06 to 0.36; P < 0.0001; 8 studies; 6503 participants), less total DVT (RR 0.25, 95% CI 0.15 to 0.40; P < 0.00001; 8 studies; 5715 participants), less proximal DVT (RR 0.12, 95% CI 0.04 to 0.39; P = 0.0004; 7 studies; 2746 participants) and less total PE (RR 0.16, 95% CI 0.04 to 0.62; P = 0.008; 8 studies; 6412 participants) for fondaparinux compared with placebo. All PE events were symptomatic. Fondaparinux also showed less fatal PE and non‐fatal PE, but results from eight studies and 6412 participants were not significant compared with placebo (RR 0.14, 95% CI 0.02 to 1.17; P = 0.07; and RR 0.22, 95% CI 0.05 to 1.03; P = 0.05).

On the other hand, results showed that major bleeding was increased in the fondaparinux arm (RR 2.56, 95% CI 1.48 to 4.44; P = 0.0008; 8 studies; 6659 participants) versus the placebo arm. Only two out of six studies reporting on fatal bleeding reported participants suffered from fatal bleeding, and the rate was not different between study arms (RR 4.87, 95% CI 0.58 to 40.84; P = 0.14; 6 studies; 5993 participants).

Four out of eight studies (APOLLO; ARTEMIS; CALISTO; PENTHIFRA PLUS) reporting on all causes of death reported deaths. We combined results and showed that the all causes of death outcome was not different between fondaparinux and placebo arms (RR 0.76, 95% CI 0.48 to 1.22; P = 0.26; 8 studies; 6674 participants).

Our results did not show differences in MI (RR 0.25, 95% CI 0.03 to 2.19; P = 0.21; 5 studies; 5777 participants) or other serious adverse effects (RR 0.98, 95% CI 0.77 to 1.24; P = 0.85; 7 studies; 6581 participants) between fondaparinux and placebo. Most of the included studies reporting MI did not classify MI into fatal and non‐fatal events.

The quality of evidence for total VTE, total DVT and proximal DVT was moderate, and the quality of evidence for symptomatic VTE and total PE was high. The quality of evidence for major bleeding and all causes of death was moderate. See summary of findings Table for the main comparison.

Fondaparinux versus LMWH

Twelve studies (ACT2545; EFFORT; EPHESUS; FONDACAST; L‐8635; Li 2015; PEGASUS; PENTAMAKS; PENTATHLON; PENTATHLON 2000; PENTHIFRA; Shen 2014) compared the efficacy and safety of fondaparinux versus LMWH for the prevention of VTE. We combined the first 11 studies into meta‐analyses. Total VTE, total DVT and proximal DVT outcomes showed significant heterogeneity; therefore, we used the random‐effects model. Outcomes of symptomatic VTE, total PE, fatal PE, non‐fatal PE, major bleeding, fatal bleeding, all causes of death, MI and other serious adverse effects showed low heterogeneity; therefore, we used the fixed‐effect model for analyses of these variables. The heterogeneity assessment of symptomatic VTE showed a P value for the Chi‐squared test of 0.16 and an I² statistic of 35%; after discussion, we decided to analyse this outcome by using the fixed‐effect model. Results showed that fondaparinux reduced VTE and DVT (RR 0.55, 95% CI 0.42 to 0.73; P < 0.0001; 11 studies; 9339 participants; and RR 0.54, 95% CI 0.40 to 0.71; P < 0.0001; 10 studies; 9356 participants, respectively) and showed a trend towards reduced proximal DVT (RR 0.58, 95% CI 0.33 to 1.02; P = 0.06; 9 studies; 8361 participants). EPHESUS, PEGASUS, PENTAMAKS, PENTATHLON, PENTATHLON 2000, PENTHIFRA, Li 2015, EFFORT and FONDACAST reported symptomatic VTE. The combined result showed no differences between fondaparinux and LMWH in symptomatic VTE (RR 1.03, 95% CI 0.65 to 1.63; P = 0.90; 9 studies; 12,240 participants). Fondaparinux increased major bleeding (RR 1.38, 95% CI 1.09 to 1.75; P = 0.008; 11 studies; 12,501 participants) but did not increase fatal bleeding (RR 0.71, 95% CI 0.14 to 3.62; P = 0.68; 6 studies; 10,293 participants). Total PE, fatal PE and non‐fatal PE were not different (RR 1.24, 95% CI 0.65 to 2.34; P = 0.51; 10 studies; 12,350 participants; and RR 0.72, 95% CI 0.25 to 2.05; P = 0.54; 9 studies; 11,107 participants; and RR 1.40, 95% CI 0.63 to 3.11; P = 0.41; 9 studies; 11,107 participants, respectively) between fondaparinux and LMWH groups. All PE events were symptomatic events. Shen 2014 reported VTE rates of fondaparinux and LMMH groups as 7.02% and 8.47%, respectively (P = 0.957). However, because Shen 2014 did not report the number of events for each group, we did not include this study in the meta‐analysis.

EPHESUS, FONDACAST, PEGASUS, PENTAMAKS, PENTATHLON, PENTATHLON 2000 and PENTHIFRA reported deaths, and ACT2545, Li 2015, EFFORT and L‐8635 reported no cases of death. The pooled result showed no differences between fondaparinux and LMWH arms (RR 0.88, 95% CI 0.63 to 1.22; P = 0.44; 11 studies; 12,400 participants). EPHESUS, PENTATHLON and PENTHIFRA also reported deaths associated with VTE or bleeding. Pooled results did not show a difference (RR 0.89, 95% CI 0.38 to 2.07; P = 0.79; 5 studies; 4774 participants). Results showed no differences in MI (RR 1.28, 95% CI 0.69 to 2.37; P = 0.43; 6 studies; 10,720 participants) nor in other serious adverse effects (RR 1.06, 95% CI 0.94 to 1.19; P = 0.31; 10 studies; 12,465 participants) between the two types of medicine. Most included studies did not classify MI data into fatal and non‐fatal events.

The quality of evidence was moderate for total VTE, symptomatic VTE, total DVT, total PE and all causes of death, and was low for proximal DVT. The quality of evidence was high for major bleeding. See summary of findings Table 2.

Fondaparinux versus warfarin

Only one study (Bern 2015) compared fondaparinux versus variable dose warfarin (target INR 2.0 to 2.5) for VTE prevention after hip or knee replacement until day 28 ± 2 after operation. Investigators included 118 participants in each of the two treatment groups and reported no VTE events (VTE, DVT or PE) in either group. They reported three cases of major bleeding in the fondaparinux group and none in the variable dose warfarin groups, showing no differences between groups (RR 7.00, 95% CI 0.37 to 134.5; 236 participants). Study authors reported no fatal bleeding. The quality of the evidence for all outcomes was very low. For details, see the Data and analyses section, Comparison 9.

Bern 2015 also compared fondaparinux with the fixed dose 1 mg once daily warfarin for VTE prevention. They reported no VTE events in the fondaparinux group and two distal DVTs in the 1 mg warfarin group, showing no statistically significant differences (RR 0.2, 95% CI 0.01 to 4.12; 236 participants for VTE and total DVT analyses). Results showed three and one major bleeding events, respectively, in the fondaparinux and 1 mg warfarin groups (RR 3.00, 95% CI 0.32 to 28.43; 236 participants). The quality of the evidence for all outcomes was very low. For details, see the Data and analyses section, Comparison 10.

Bern 2015 reported no deaths and no other serious adverse events.

Fondaparinux versus edoxaban

Only one study (Fuji 2015) involving 43 participants compared fondaparinux (1.5 mg sc once daily) versus edoxaban (15 mg oral once daily) for patients with renal impairment undergoing lower‐limb orthopaedic surgery. Researchers reported no cases of thromboembolic and major bleeding events in the fondaparinux and edoxaban groups. In addition, they reported no deaths. The quality of the evidence for all outcomes was very low. For details, see the Data and analyses section, Comparison 11.

Fondaparinux versus mechanical thromboprophylaxis

Only one study (AR3106116) compared fondaparinux with the mechanical thromboprophylaxis method and intermittent pneumatic compression (IPC). This study showed no differences in total VTE (RR 0.61, 95% CI 0.22 to 1.67; 99 participants), DVT (RR 0.63, 95% CI 0.23 to 1.72; 100 participants) or other serious adverse effects (RR 0.18, 95% CI 0.02 to 1.67; 120 participants) between fondaparinux and IPC. Results included no cases of symptomatic VTE, proximal DVT, PE, major bleeding or death,although this is likely a result of the fact that the study lasted only up to eight days. The quality of the evidence for all outcomes was low. For details, see the Data and analyses section, Comparison 12.