Jan Beyer-Westendorf, Center for Vascular Medicine, University Hospital ‘Carl Gustav Carus’, Technical University Dresden, Fetscherstrasse 74, 01307 Dresden, Germany. Tel.: +49 351 4583659; fax: +49 531 4584359. E-mail: firstname.lastname@example.org
Summary. Background: Thromboprophylaxis with rivaroxaban (R) is superior to enoxaparin in patients undergoing major orthopedic surgery (MOS). However, rivaroxaban has never been directly compared with fondaparinux (F), which also shows superior efficacy over enoxaparin. The clinical impact of switching from fondaparinux to rivaroxaban thromboprophylaxis is unclear.
Objectives: To evaluate the efficacy and safety of rivaroxaban or fondaparinux thromboprophylaxis in unselected patients undergoing MOS.
Patients/Methods: This is a monocentric, retrospective cohort study in 5061 consecutive patients undergoing MOS at our centre, comparing rates of symptomatic VTE, bleeding and surgical complications, length of hospital stay and risk factors for VTE.
Results: Rates of symptomatic VTE were 5.6% (F) and 2.1% (R; P < 0.001), with rates for distal DVT being 3.9 vs. 1.1% (P < 0.001). Rates of major VTE were numerically higher with fondaparinux (1.8 vs. 1.1%), but not statistically significant. Rates of severe bleeding (bleeding leading to surgical revision or death, occurring in a critical site, or transfusion of at least two units of packed red blood cells) were statistically lower with rivaroxaban compared with fondaparinux (2.9 vs. 4.9%; P = 0.010). The mean length of hospital stay was significantly shorter in the rivaroxaban group (8.3 days, 95% CI 8.1–8.5 vs. 9.3 days, 9.1–9.5; P < 0.001).
Conclusion: Based on an indirect comparison of two consecutive cohorts, our data suggest that thromboprophylaxis with rivaroxaban is associated with less VTE and bleeding events than fondaparinux in unselected patients undergoing MOS. Prospective comparisons are warranted to confirm our findings.
Patients undergoing major orthopedic surgery (MOS) are at high risk of venous thromboembolism (VTE), manifesting as deep vein thrombosis (DVT) or pulmonary embolism (PE) [1–3]. Despite several limitations, low-molecular-weight heparin (LMWH) has been the standard for VTE prophylaxis in patients undergoing MOS for nearly two decades. The indirect selective factor (F) Xa inhibitor fondaparinux, which is a pentasaccharide of synthetic origin, has been developed to overcome some of the limitations of LMWH such as allergic reactions. In orthopedic surgery, a 50% relative risk reduction for VTE was found in patients receiving fondaparinux 2.5 mg compared with the LMWH enoxaparin [4,5]. However, higher costs and increased bleeding complications probably prevented fondaparinux from becoming the reference standard in MOS despite superior efficacy over LMWH .
A prolonged VTE prophylaxis is recommended following MOS as thromboembolic risk remains high for weeks after hospital discharge , but patients’ adherence to self-injections declines after hospital discharge . Recently, the oral direct FXa inhibitor rivaroxaban was developed and tested against the reference standard enoxaparin in four large phase-III trials (RECORD 1–4), where it was found to have superior efficacy with a relative risk reduction for thromboembolic events between 55% (at day 14) and 62% (at day 35) without increasing bleeding risks .
However, the translation of results from RCTs into clinical practice is limited by the fact that selected study populations in phase-III trials may not be representative of cohorts of unselected patients in daily care who do not need to fulfil strict inclusion and exclusion criteria and exhibit relevant co-morbidities [9–11]. This could impact the efficacy and safety of thromboprophylaxis regimens in daily care.
In addition, new oral anticoagulants have been compared with enoxaparin but not with the more efficacious fondaparinux. Therefore, the clinical impact of switching from fondaparinux to rivaroxaban thromboprophylaxis in daily care is unclear.
In our hospital, fondaparinux 2.5 mg OD was used as standard prophylaxis for patients undergoing MOS until 2009, followed by oral rivaroxaban 10 mg OD. We performed a retrospective analysis of the efficacy and safety of prophylaxis with oral rivaroxaban compared with subcutaneous fondaparinux in two large cohorts of consecutive patients undergoing MOS.
All patients undergoing MOS at the Orthopaedic Department of the University Hospital in Dresden, Germany, between January 2006 and June 2011, were included in the ORTHO-TEP registry. A detailed description of the registry design has previously been published . In brief, all patients undergoing major orthopedic surgery are prospectively entered in a quality management database. Information on surgical and anesthesiology procedures, VTE and bleeding complications was extracted from the patient charts, surgical and anesthesiology protocols, VTE database and the transfusion database. For a final cross-check, discharge letters were reviewed for consistency of complications and outcome for all patients (Fig. 1).
Surgical procedures classified as MOS included primary as well as revision total hip and knee replacement. Surgical techniques and selection of prosthetic devices did not change substantially during the observational period. Furthermore, the standards and procedures of the Department of Anesthesiology for patients undergoing MOS were not changed between 2008 and 2011.
From January 2008 to December 2009, hospital guidelines recommended VTE prophylaxis with fondaparinux 2.5 mg OD. In January 2010, the thromboprophylaxis regimen was changed to oral rivaroxaban 10 mg OD. Both types of VTE prophylaxis were started on the evening after surgery and at least 6 h after wound closure. Patients with contraindications to fondaparinux or rivaroxaban received alternative thromboprophylaxis. These patients were also identified but excluded from further analysis. At hospital discharge, continued thromboprophylaxis was recommended until day 35 post-surgery for all patients.
Only in-hospital symptomatic VTE events were used as efficacy outcomes. For all patients with suspected VTE events, reports and images of ultrasound and CT lung scans were reviewed. In our hospital, all patients with suspected DVT undergo bilateral standardized complete compression ultrasound (CCUS) [13,14] within 24 h. Rates of proximal DVT (defined as lower extremity DVT at the level of the popliteal vein or above), distal DVT (DVT below the popliteal vein) and PE (objectively confirmed by CT angiography) were evaluated.
The adjudication criteria for VTE and VTE-related deaths were:
1 symptoms leading to CCUS or CT scan and presence of a thrombus in objective testing; or
2 autopsy confirming VTE as the cause of death; or
3 any sudden unexplained death.
Causes of in-hospital death were evaluated by review of death certificates and autopsy reports.
The primary outcome was the rate of any symptomatic VTE at hospital discharge. Secondary outcomes were the rates of proximal or distal symptomatic DVT, symptomatic PE or VTE-related death.
The primary outcome for safety analysis was the rate of major bleeding using a modification of the ISTH definition . In the absence of overt bleeding, transfusions of more than two units of red blood cells (RBCs) or drop of hemoglobin after MOS were excluded to exclude potential bias. Consequently, the outcome of major bleeding was defined as overt bleeding with:
1 a documented transfusion of at least two units of RBCs; or
2 a surgical revision due to bleeding; or
3 a bleeding into a critical site (intracranial, intraocular, intra-articular, retroperitoneal or overt gastrointestinal bleeding); or
4 fatal bleeding.
Further safety outcomes were rates of transfusion of RBCs, plasma or platelet concentrates, rates of surgical revision, length of hospital stay and in-hospital mortality. All safety outcomes were evaluated until hospital discharge.
Differences in baseline variables were compared by Fisher’s exact test or Student’s t-test and ancova including interaction effects. Binary data were analyzed in the same way using logistic regression models; 95% confidence intervals for proportions are given according to Clopper-Pearson. Age and sex differences were considered to be potential confounders and were standardized by using the overall cohort’s mean as the standard. Length of hospital stay was analyzed by Kaplan–Meier estimation and assessed by log-rank testing. The risk for VTE was modelled using Cox proportional hazard models. The proportional hazard assumption was not violated. In a first step, the crude hazard ratio was estimated. In a second step, this hazard ratio was adjusted for confounding factors. In a third step, both potential risk factors and their interaction terms were included in the model. In the last step, model terms contributing no significant effect to the model were removed.
Length of hospital stay was analyzed by Kaplan–Meier estimation and assessed by log-rank testing. Of note, subgroups with VTE or bleeding complications were also assessed by Kaplan–Meier estimation, but results are of descriptive value only, because interference of confounding factors such as early discharge policy and the potential for interaction between VTE and bleeding complications need to be considered. All statistical analyses were carried out using the IBM®, SPSS® Statistics Version 19 (IBM SPSS Statistics, Chicago, IL, USA) and R (Comprehensive R Archive Network) (R Foundation for Statistical Computing, Vienna, Austria).
The study protocol was approved by the local ethics committee. Due to the retrospective nature of our study and the anonymization of patient data, the patients’ informed consent was not requested.
Between January 2008 and June 2011, 3723 patients underwent MOS. Of these, 1994 received fondaparinux (53.6%) and 1043 received rivaroxaban (28.0%). Patients with contraindications for fondaparinux or rivaroxaban (mainly severe renal impairment) or patients with indications for therapeutic anticoagulation, such as atrial fibrillation, recent VTE or mechanical heart valves, received other types of prophylaxis according to guidelines and hospital standards (n = 686; 18.4%; Table 1).
Table 1. Patient characteristics and type of VTE prophylaxis in all patients undergoing major orthopedic surgery between January 2008 and June 2011
Fondaparinux 2.5 mg OD mean ± SD
Rivaroxaban 10 mg OD mean ± SD
UFH 5000 IU TID mean ± SD
Other prophylaxis (therapeutic LMWH, mechanical) mean ± SD
UFH, unfractionated heparin; LMWH, low-molecular-weight heparin; OD, once daily; TID, three times daily; values in bold indicate p>0.05.
History of VTE
66.8 ± 11.9
64.9 ± 12.5
72.6 ± 10.6
68.2 ± 12.8
167 ± 9.7
168 ± 10.0
167 ± 9.8
168 ± 10.0
80.9 ± 18.4
81.9 ± 17.6
83.0 ± 19.3
81.3 ± 17.9
BMI (kg m−2)
29 ± 5.8
28.9 ± 6
29.8 ± 6.5
28.6 ± 5.6
Platelet count (Gpt/l)
260.3 ± 71.8
255.4 ± 66.7
258.6 ± 119.0
247.9 ± 80.3
0.98 ± 0.1
1.0 ± 0.1
1.1 ± 0.2
1.2 ± 0.6
29.7 ± 2.1
28.2 ± 2.1
30.5 ± 8.2
30.9 ± 6.4
76.9 ± 20.2
76.9 ± 19.6
190 ± 135.8
81.6 ± 36.0
GFR (Cockroft-Gault) (mL min−1)
91 ± 34.9
94.9 ± 35.1
45.7 ± 24.6
89.1 ± 36.8
Of the 3037 patients receiving either fondaparinux or rivaroxaban, 1936 underwent hip arthroplasty (63.7%), 991 underwent knee arthroplasty (32.6%) and 110 underwent hip or knee revision surgery (3.6%).
Patients in the fondaparinux group were more often female, were older, smaller, and had a lower INR, higher aPTT values and a lower GFR. These differences were statistically significant (Table 1). Patients in the rivaroxaban group had significantly more often a positive history of VTE (4.03 vs. 0.95%).
The rates of all symptomatic VTE events were 5.6% in the fondaparinux group and 2.1% in patients receiving rivaroxaban (P < 0.001), mainly driven by higher rates of distal DVT (3.9 vs. 1.1%; P < 0.001). No significant differences in rates for proximal DVT or PE were found (Table 2).
Table 2. (a) Efficacy outcomes and (b) safety outcomes in all patients receiving VTE prophylaxis with fondaparinux or rivaroxaban after major orthopedic surgery (CI: 95% Clopper–Pearson confidence intervals)
Fondaparinux n = 1994
Rivaroxaban n = 1043
Values in bold indicate p>0.05.
Overt bleeding with transfusion > 2 RBC concentrates
Surgical revisions due to bleeding complications
Bleeding into critical sites
Any RBC transfusion
Any surgical revision
Transfusion of plasma concentrates
Transfusion of platelet concentrates
Length of hospital stay (days)
Length of hospital stay (days) Median (25th and 75th percentile)
9 (8; 9)
8 (7; 9)
Patients receiving rivaroxaban experienced lower rates of major bleeding (2.9 vs. 4.9%; P = 0.01). This finding was driven by a reduction of bleeding requiring transfusion of more than two units of RBCs (2.6 vs. 4.3%; P = 0.02) and a reduction of surgical revisions due to bleeding complications (0.4 vs. 1.1%; P = 0.041). No differences were found with regard to bleedings into a critical site or fatal bleedings (Table 2).
Five patients died during hospital stay (three treated with fondaparinux and two treated with rivaroxaban). All deaths were caused by septic complications leading to multi-organ failure. One of the three deaths in the fondaparinux group occurred in a 74-year-old woman with septic hip infection who developed wound hematoma as well as proximal DVT after surgery, both of which did not cause death. None of the other patients with a fatal outcome developed a VTE event.
In univariate analysis, the use of rivaroxaban was associated with a reduced VTE risk (hazard ratio 0.53; 95% CI 0.33–0.84; P = 0.007). On the other hand, a history of VTE (HR 5.9), female gender (HR 2.0) and BMI > 30 (HR 1.04 for every unit of BMI) increased VTE risk (Table 3). These factors were confirmed to be independent risk factors in multivariate analysis (Table 3).
Table 3. (a) Univariate and (b) multivariate analysis using Cox logistic regression model to evaluate potential risk factors for the occurrence of VTE in patients receiving fondaparinux or rivaroxaban for thromboprophylaxis after major orthopedic surgery (n = 3037)
Patients with/without VTE
HR, hazard ratio; values in bold indicate p>0.05.
Use of rivaroxaban
BMI (> 30 kg m−2; per unit)
Age (> 65 years; per year)
Use of rivaroxaban
BMI (> 30 kg m−2; per unit)
The mean length of hospital stay was significantly shorter in the rivaroxaban group (8.3 days, 95% CI 8.1–8.5 vs. 9.3 days, 9.1–9.5; P < 0.001; Table 2). This finding was also confirmed in the subgroup of patients without any complications (8.1 days, 95% CI 7.9–8.3 vs. 8.8 days, 8.7–8.9; P < 0.001 (Fig. 2A). As indicated by Fig. 2(B–D), the occurrence of VTE, bleeding or surgical complications contributed to prolonged hospitalization in both treatment groups and, at least numerically, this effect was more pronounced in patients receiving fondaparinux. Of note, due to the potential interference of complications (bleeding or surgical complications may increase VTE risk and VTE treatment might increase bleeding) and changes in discharge policy over time, the Kaplan–Meyer curves in Fig. 2(B–D) are of descriptive nature only and were not statistically tested for significance.
VTE prevention with fondaparinux or rivaroxaban
In two large cohorts of unselected patients undergoing MOS, VTE prophylaxis with rivaroxaban was found to be more effective than with fondaparinux, with a reduction of in-hospital rates of symptomatic VTE. This finding was mainly driven by an increase in distal DVT in the fondaparinux group. The clinical and prognostic relevance of isolated distal DVT, however, is still a matter of intense debate [16–18] and a therapeutic benefit of anticoagulant treatment in isolated distal DVT has yet to be proven.
The lower rate of symptomatic DVT in patients receiving rivaroxaban compared with fondaparinux is in contrast to the results of two large and very similar phase-III trial programs [4,8] evaluating efficacy and safety of fondaparinux or rivaroxaban against enoxaparin in MOS. In both programs, a similar 50% VTE risk reduction was found. Therefore, a comparable rate of VTE events would have been expected for both substances.
On the other hand, screening of asymptomatic patients was routinely performed in phase-III trials and the results of the cited meta-analyses are mainly based on high VTE event rates found on screening ultrasound, whereas our study only evaluated symptomatic VTE events, which may partly explain our findings.
In our study, the rate of all symptomatic VTE events was found to be surprisingly high, at 5.6% (fondaparinux) and 2.1% (rivaroxaban). The rates of symptomatic PE in our study were 0.5% and 0.2%, respectively. This translates into a significant increase of VTE in large unselected cohorts of routine patients compared with the highly selected cohort of patients evaluated in phase-III trials, in which symptomatic VTE events were reported to occur in about 0.5–1% of patients, with a PE rate of 0.2% [4,5,8].
The high rate of symptomatic VTE in our study was not due to insufficient prophylaxis, because all patients received approved thromboprophylaxis according to national and hospital guidelines. Furthermore, all patients received graduated compression stockings (if not contraindicated) and early mobilization.
Higher VTE rates were also not attributable to a prolonged observational period, because our study documented symptomatic VTE events until hospital discharge (mean 9.3 days), while phase-III trials evaluated symptomatic VTE events until day 11.
Differences in trial design and reporting of events influence VTE event rates in prospective randomized controlled trials or retrospective analyses. However, similar effects would be expected in both treatment groups and, usually, under-reporting would be expected outside of prospective trials. Therefore, rates of symptomatic VTE would be expected to be lower, not higher, in a retrospective trial. It seems reasonable to conclude that the high rates of VTE events were mainly caused by the unselected nature of our study cohort, which also included old patients with relevant co-morbidities and revision arthroplasty.
Risk factors for VTE
In multivariate analysis, rivaroxaban was found to reduce the risk of VTE compared with fondaparinux prophylaxis, with a hazard ratio of 0.5, which is comparable to the risk reduction found in phase-III trials testing rivaroxaban against enoxaparin . However, because no direct comparisons are available for rivaroxaban and fondaparinux, it is impossible to compare our results with previous findings. Therefore, our data close a significant gap in the field. Previous VTE (HR 5.0), female gender (HR 1.9) and BMI > 30 (HR 1.04 per unit) were further independent risk factors for VTE, which is in agreement with the findings of other studies .
Safety of fondaparinux or rivaroxaban thromboprophylaxis
The rate of major bleeding complications was significantly higher in patients receiving fondaparinux (4.9%) compared with rivaroxaban (2.8%). This finding was mainly driven by higher rates of overt bleeding with more than two RBC transfusions and surgical revisions due to bleeding complications and is in line with our findings for secondary safety outcomes: significantly more RBC transfusions, numerically higher rates of plasma or platelet transfusions and higher rates of non-bleeding complications leading to secondary surgical interventions in patients receiving fondaparinux. Our data further support published data indicating increased bleeding complications with fondaparinux .
The length of hospitalization was significantly shorter in the rivaroxaban group (8.3 vs. 9.3 days). However, length of hospital stay in general has been decreasing over the years, mostly due to economic reasons. This effect contributed to our finding, because a shorter hospitalization was also found in the subgroup of patients without complications (8.1 vs. 8.8 days). However, despite the potential for interference of complications, which prevented testing for statistical significance, our subgroup analyses suggest that complications are less severe with rivaroxaban, because, at least numerically, prolongation of hospitalization was more pronounced in fondaparinux patients with complications. Thus, both the reduction of complication rates and the probably reduced severity of complications may have contributed to the reduction in length of hospital stay.
There are several limitations to our study. The retrospective design of our study did not allow for randomization. Therefore, significant differences in demographic characteristics exist between groups. Patients in the fondaparinux group were slightly older and more often female, while patients in the rivaroxaban group had significantly more often a history of VTE, which was found to be the strongest risk factor for VTE in both cohorts. However, statistical analysis of outcome events was adjusted for the imbalances between both groups.
Due to the study design, changes in patient management over time need to be considered. However, during the observational period, surgical or anesthesiology management did not change and rates of performed ultrasounds (around 13–15% of all cases) remained constant, indicating that no over- or under-diagnosis influenced our VTE rates. Therefore, differences of event rates between treatment arms cannot be explained by such bias.
Furthermore, in a retrospective analysis data on complications might be documented incompletely or affected by selection or reporting bias. However, the information used for our analysis was directly extracted from the quality management database, in which all treatment-associated complications are prospectively entered during hospitalization and reviewed at discharge, providing a complete and robust set of data. Furthermore, careful review of surgical procedures and standards at our hospital did not identify changes in the management of surgery or surgical complications during the study period. Therefore, the influence of confounding factors was limited.
Another limitation derives from the existence of different definitions of ‘major bleeding complications’, some of them excluding bleeding at the surgical site. We used the ISTH definition for ‘major bleeding’ but excluded postoperative drop in hemoglobin levels (present in most patients after MOS), because indication and timing of blood sampling is not predefined in a retrospective analysis and will introduce a risk of bias. Furthermore, in a retrospective analysis it is difficult to establish the reason for blood transfusion: was it necessary to treat overt bleeding or was this mainly carried out due to a subjective assessment by the attending physician. Therefore, in our cohort ‘major bleeding’ was only assessed for patients with reported overt bleeding and a documented indication for transfusions of more than two units of blood in accordance with our hospital standards. However, we need to point out that some major bleedings occurred before the first dosage of thromboprophylaxis and that our retrospective data do not allow us to establish the exact time of bleeding on the day of surgery in all cases. The standard criteria of ‘bleeding requiring intervention or surgical revision’ and ‘bleeding into a critical organ site’ from ISTH criteria were included in our endpoint definition, as were bleeding events at the site of surgery. Therefore, in our analysis the ISTH definition of major bleeding complications was designed to limit the reported events to those of established clinical relevance to the patient, thus reducing the risk of bias by a potentially low transfusion threshold of some physicians.
Finally, no information on the rates of VTE, bleeding complication and death was available for the post-discharge period. Therefore, total event rates might be even higher than found in our in-hospital analysis.
On the other hand, the size of our cohorts, comparing nearly 2000 patients treated with fondaparinux with more than 1000 patients receiving rivaroxaban, is comparable to large phase-III RCTs and this is a significant strength of our study. Furthermore, the use of clinically relevant endpoints (objectively confirmed symptomatic VTE, severe bleeding complications, surgical revisions and length of hospital stay) contributes to the strength and clinical impact of our data.
Compared with the results of phase-III trials evaluating rivaroxaban in major orthopedic surgery, we found higher rates of symptomatic VTE events and major bleeding complications in our cohort of unselected patients, including high-risk patients, who are usually excluded in phase-III trials. Therefore, results of clinical trials evaluating efficacy of VTE prophylaxis are not entirely transferable to daily practice and need to be confirmed in large registries or phase-IV trials of unselected patients.
Our study is the first to provide efficacy and safety data for an unselected cohort of rivaroxaban-treated patients undergoing MOS, including a comparison with fondaparinux. Based on an indirect comparison of two consecutive cohorts, our data suggest that thromboprophylaxis with rivaroxaban is associated with fewer VTE and bleeding events than fondaparinux in unselected patients undergoing MOS. The reduction of VTE, bleeding and surgical complications seen with rivaroxaban seems to contribute to a significantly shorter hospitalization.
Disclosure of Conflict of Interests
The Ortho-TEP registry was (in part) supported by a grant from Bayer Healthcare, Germany, providing funding for a documentation assistant and statistical analysis. All the authors declare that Bayer Healthcare had no influence on the study design, conduct of the study, data collection, statistical analysis or preparation of the manuscript. None of the authors declared a conflict of interest with regard to the ORTHO-TEP registry or this manuscript.