Andrew F. Shorr, Pulmonary and Critical Care Medicine, Washington Hospital Center, 110 Irving St, NW 20010, Washington, DC, USA. Tel.: +1 2028777856; fax: +1 2022910386. E-mail: firstname.lastname@example.org
Summary. Background: Venous thromboembolism (VTE) remains a significant complication of major orthopedic surgery, and chronic kidney disease (CKD) is common among elderly patients undergoing total hip replacement (THR).
Objectives: The purpose of this study was to evaluate thrombosis and bleeding outcomes in patients with stage 3B CKD treated with either desirudin or enoxaparin after elective THR.
Patients/Methods: This was a post hoc subgroup analysis of a randomized, multicenter, double-blind study of desirudin vs. enoxaparin in patients undergoing elective THR.
Results: Patients received either subcutaneous desirudin 15 mg twice daily or subcutaneous enoxaparin 40 mg once daily. Of the 2078 randomized patients who received study medication, 577 had stage 3B CKD or worse (27.8%), and the proportion of these patients who experienced a major VTE in the enoxaparin treatment group was found to be much higher than in the desirudin treatment group (11.1% vs. 3.4%, model-adjusted odds ratio 3.52, 95% confidence interval 1.48–8.40, P = 0.004). There was no statistically significant difference between treatment groups in terms of rates of major bleeding, regardless of stage of renal function.
Conclusions: CKD has been reported previously to increase the risk of bleeding with anticoagulants, and these findings suggest that CKD may also increase the risk of major VTE for patients treated with enoxaparin, but not for patients treated with desirudin. Clinicians should consider the impact of CKD on the risk of VTE when choosing a prophylaxis agent.
Major orthopedic surgery is an important risk factor for venous thromboembolism (VTE). In addition, chronic kidney disease (CKD) is common in elderly patients undergoing total hip replacement (THR). Patients with stage 3B CKD, defined as an estimated glomerular filtration rate (eGFR) of 30–44 mL min−1, have a four-fold increase in risk for VTE as well as an increased risk for bleeding owing to reduced clearance of VTE prophylaxis agents [1,2]. However, there are few data on outcomes in patients with stage 3B CKD undergoing major orthopedic surgery with respect to VTE prophylaxis.
Desirudin is a bivalent direct thrombin inhibitor indicated for the prevention of deep vein thrombosis (DVT) in patients undergoing elective hip or knee replacement surgery. It is the only subcutaneous-injectable direct thrombin inhibitor, and is eliminated exclusively through renal mechanisms . It has been studied in clinical trials involving > 13 000 patients in the settings of hip replacement surgery, acute coronary syndromes, and coronary angioplasty. In one pivotal, randomized, double-blind clinical trial comparing desirudin 15 mg every 12 h with unfractionated heparin 5000 IU every 8 h, and in a second pivotal, randomized, double-blind clinical trial comparing desirudin 15 mg every 12 h with enoxaparin 40 mg every 24 h, for DVT prophylaxis in patients undergoing elective total hip replacement, desirudin was superior to these two agents in reducing the incidence of overall DVT (P <0.0001 and P =0.001, respectively) after 8–12 days of treatment. The overall safety of desirudin was similar to that of heparin comparators [4,5]. In light of prior reports suggesting increased VTE and bleeding risk in patients with CKD , we hypothesized that, owing to its characteristic direct thrombin inhibition and high binding affinity, desirudin would provide a greater safety and efficacy benefit than heparins in this population. The purpose of the current supplementary analysis on the second pivotal study  was to evaluate thrombosis and bleeding outcomes in patients with stage 3B CKD treated with either desirudin or enoxaparin after elective THR surgery.
This was a post hoc subgroup analysis of a randomized, multicenter, double-blind study of desirudin vs. enoxaparin in patients undergoing elective THR surgery. Details of the methodology of the original clinical trial have been previously published . In short, 2079 patients were randomized to receive either subcutaneous desirudin 15 mg twice daily, initiated 30 min prior to surgery, or subcutaneous enoxaparin 40 mg once daily, with the first dose being given on the evening prior to surgery.
Determination of CKD stage
The eGRF was assessed at baseline with the Cockcroft–Gault equation .
Lean body weight (LBW) was used and defined as 49.9 kg + 2.3 kg for every 0.025 M over 1.5 M for men, and 45.4 kg + 2.3 kg for every 0.025 M over 1.5 M for women. For individuals < 1.5 M, an LBW of 49.9 kg for men and 45.4 kg for women was used. For individuals with a missing value for height (n = 4), actual weight was be used in place of LBW. In patients with a baseline serum creatinine of < 1.0 mg dL−1, a value of 1.0 mg dL−1 was used.
Patients were classified according to the CKD staging system  as stage 1 or stage 2 (eGFR ≥ 60 mL min−1), stage 3A (eGFR = 45–59 mL min−1) or stage 3B (eGFR = 30–44 mL min−1). Most patients with stage 4/5 CKD (eGFR < 30 mL min−1) were excluded from this trial, but those few who were randomized were included in the stage 3B group for purposes of analysis.
The primary efficacy variable of major VTE was defined in the protocol as the combination of proximal DVT, fatal or non-fatal pulmonary embolism, death because of a thromboembolic event, or unexplained death.
Bleeding complications were defined in the protocol as major if the hemorrhage was either: (i) overt and produced a fall in hemoglobin of 2 g dL−1 or led to a transfusion of ≥ 2 units of whole or packed cells in the postoperative period (beginning 12 h after surgery and continuing to the end of drug therapy); and (ii) retroperitoneal, intracranial, intraocular, intraspinal, or occurred in a major prosthetic joint. In this new analysis, an additional bleeding analysis was included. The percentage of patients with a bleeding index (BI) of ≥ 2 in the postoperative period was calculated, in order to account for differences in overall major bleeding that could be attributed to differences in timing of study drug administration. The BI was based on the fall in hemoglobin (g dL−1) adjusted for whole blood or red blood transfusions occurring from 12 h postoperatively to day 6 inclusive (where day 0 is the day of operation) . To be included in the analysis of BI, patients needed to satisfy all of the following: (i) underwent the operation; (ii) had transfusion information for postoperative period recorded; (iii) had a hemoglobin value recorded on day 1 at the start of the postoperative period; and (iv) had a further hemoglobin value recorded on the day of (or on the day after) the end-date of the period covered by the postoperative transfusion information. The investigator responsible for calculating the BI was blinded to drug assignment (J.S.).
A prespecified analysis plan outlining the modeling and testing strategies was developed prior to the performance of any statistical testing. Baseline eGFR was the only subgroup variable chosen for analysis. All statistical tests were two-tailed, and a P-value of ≤ 0.050 was taken to represent statistical significance, except for interactions where, in accordance with usual practice  for FDA-regulated clinical trials, 0.100 was used instead.
The safety population, which was used for the analysis of all safety or demographic data, was defined as the set of randomized patients who received at least one dose of study medication. The evaluable primary outcome population, which was used for the analysis of major VTEs as well as for an additional analysis of postoperative BI, was defined in accordance with the protocol as described previously . Patients were excluded from analyses by CKD stage if there was insufficient information to calculate the Cockroft–Gault measure of eGFR.
To assess the differential effect of treatment on major VTEs by levels of baseline eGFR, we utilized logistic regression. The interaction of treatment with baseline eGFR was assessed from a logistic regression model with terms for baseline eGFR (as a continuous variable), age, gender, treatment, and the interaction of treatment with the baseline eGFR value. The set of terms included in this model was identified prospectively, with the age and gender covariates included because of their well-recognized prognostic importance. Two further supportive analyses of this interaction were also conducted in which baseline eGFR was represented by a binary variable (first with a cut-point of 45 mL min−1, and second with a cut-point of 60 mL min−1). We further prospectively planned to assess the effect of treatment on major VTEs within each strata of eGFR (as outlined above) separately, using a logistic regression model with (prespecified) terms for age, gender, and treatment. Adjusted odds ratios (ORs) for treatment were calculated from this model, together with corresponding 95% confidence intervals (CIs). Hosmer and Lemeshow’s c-statistic goodness-of-fit test was also used to assess model adequacy separately within each eGFR stratum. A similar modeling approach was employed to analyze the binary variable (≥ 2, < 2) for postoperative BI, except that, for the within-strata analyses, the covariate for gender needed to be omitted because of zero events in certain gender–subgroup combinations. For major bleeding, because the events were rare, we could not test the potential interaction term for eGFR with treatment. Also for this variable, owing to the low counts, treatment groups were compared by use of Fisher’s exact test.
Of the 2079 patients randomized, one patient did not receive any study medication, so the safety population consisted of the remaining 2078 patients. The evaluable primary outcome population consisted of 1587 patients, and, as shown previously , the most common reasons for exclusion were venogram not performed (11%), inadequate central venogram or venogram only performed locally (10%), major protocol violation (2%), or operation not performed (1%).
Of the 2078 patients in the safety population, 706 had stage 1 or 2 CKD (34.0%), 764 had stage 3A CKD (36.8%), 577 had stage 3B CKD or worse (27.8%), and 31 (1.5%) did not have sufficient information for calculation of eGFR. Of the 577 patients with stage 3B CKD or worse, 560 had stage 3B CKD and 17 had stage 4 CKD (eGFR = 15–29 mL min−1), but, for simplicity of presentation, all 577 such patients are summarized under ‘stage 3B’ from here onwards. No patients with stage 5 CKD (eGFR < 15 mL min−1) were included in the study. There were no noteworthy differences between treatment groups in terms of age, weight, gender and other key demographic variables for any CKD stage (1/2, 3A, or 3B) (Table 1). Patients with stage 3B CKD were older (median ages of 74 years vs. 59.5 years), more likely to be female (94% vs. 11%) and had lower weight (median weights of 65 kg vs. 82 kg) than patients with stage 1 or 2 CKD, whereas patients with stage 3A CKD were intermediate in terms of these three variables (median age of 66 years; 74% female; median weight of 72 kg).
Table 1. Demographics – safety population
Stage 1 and 2 CKD (CLcr ≥ 60 mL min−1)
Stage 3A CKD (CLcr = 45–59 mL min−1)
Stage B CKD* (CLcr = 30–44 mL min−1)
BMI, body mass index; CKD, chronic kidney disease; CLcr, creatinine clearance; VTE, venous thromboembolism. *Of the 577 patients included under CKD stage 3B above, 560 were stage 3B (enoxaparin, 288; desirudin, 272) and 17 were stage 4 (enoxaparin, 10; desirudin, 7).
Randomized and treated, n
Evaluable: primary outcome, n
Age (years), median (min.–max.)
Weight (kg), median (min.–max.)
Female, n/N (%)
Obese (BMI ≥ 30), n/N (%)
History of VTE, n/N (%)
Overall, as shown previously , the proportion of patients who experienced a major VTE was significantly lower (P = 0.019) in the desirudin treatment group (4.9%) than in the enoxaparin treatment group (7.6%). This proportion was lower in the desirudin treatment group than in the enoxaparin treatment group for each stage of CKD (1/2, 3A, and 3B) (Fig. 1). The interaction between treatment and eGFR was found to be statistically significant (P = 0.093), indicating a differential treatment effect. The supportive analysis of this interaction with a cut-point of 45 mL min−1 (for the presence/absence of stage 3B CKD) also showed statistical significance (P = 0.038). Examination of event rates for major VTEs by CKD stage showed that, for the enoxaparin treatment group, the event rate was much higher for stage 3B CKD than for stages 1/2 or 3A CKD, but for the desirudin treatment group this event rate was relatively constant. Comparison of treatment groups for stage 3B CKD patients showed that the proportion in the enoxaparin treatment group who experienced a major VTE was significantly higher than the proportion in the desirudin treatment group (11.1% vs. 3.4%, model-adjusted OR 3.52, 95% CI 1.48–8.40, P = 0.004). For stage 3B CKD, the Hosmer and Lemeshow goodness-of-fit test indicated that the prespecified model used here (with terms for treatment group, age, and gender) fitted the data well (c-statistic = 3.37 on 8 degrees of freedom; P = 0.909).
In the overall safety population, more patients in the desirudin treatment group (0.8%) than in the enoxaparin treatment group (0.2%) experienced a major bleed, although this difference was not statistically significant (P = 0.109) (Table 2). The results presented previously showed that the treatment groups had similar rates of serious bleeding (2.0% enoxaparin; 1.9% desirudin). In addition, postoperative BI ≥ 2 occurred significantly less often in the desirudin treatment group than in the enoxaparin treatment group (11.1% vs. 15.1%; P = 0.042). For stage 3B CKD, more patients experienced a major bleed in the desirudin treatment group than in the enoxaparin treatment group, although this difference was not statistically significant (1.8% vs. 0.3%; P = 0.112). Conversely, there were fewer patients with postoperative BI ≥ 2 in the desirudin treatment group (8.8% vs. 15.4%; P = 0.054), and this difference was more pronounced (5.6% vs. 15.8%; P = 0.018) when analyzed for the evaluable primary outcome population.
Table 2. Major bleeding and postoperative bleeding index (BI) by renal function
Stage 1 and 2 CKD (CLcr ≥ 60 mL min−1)
Stage 3A CKD (CLcr = 45–59 mL min−1)
Stage 3B CKD (CLcr = 30–44 mL min−1
CKD, chronic kidney disease; CLcr, creatinine clearance. *Frequencies and percentages based on the safety population patients for whom either the presence or the absence of major bleed was recorded. †Frequencies and percentages based on operated patients with a valid value for the postoperative BI.
Patients with a major bleed
Patients with postoperative BI ≥ 2
This retrospective subgroup analysis of a large randomized controlled trial (RCT) of different VTE prophylactic choices following elective THR surgery shows that patients with CKD have higher rates of major VTEs, with this tendency being most pronounced in patients with stage 3B CKD. The relationship between lower eGFRs and major bleeding was less clear, as these events were infrequent. Patients with stage 3B CKD randomized to receive desirudin had significantly fewer major VTEs and less bleeding (as assessed by the postoperative BI) than patients randomized to enoxaparin. This relationship was significant after adjustment for several important potential confounding factors.
Our observation of the link between stage 3B CKD and major VTEs confirms the findings of others. For example, the Longitudinal Investigation of Thromboembolism Etiology (LITE) study followed > 19 000 middle-aged and elderly adults, for a median of 11.8 years, to determine risk factors for VTE. These investigators noted that subjects with stage 3/4 CKD (eGFR = 15–59 mL min−1) had a relative risk for VTE of 2.09 (95% CI 1.47–2.96) as compared with subjects with normal renal function . After additional adjustment for baseline cardiovascular disease risk factors, stage 3/4 CKD remained significantly associated with the diagnosis of a VTE (adjusted relative risk of 1.71, 95% CI 1.18–2.49). Our data confirm that, among patients treated with enoxaparin, those with stage 3B CKD have a higher rate of major VTEs than those with less severe renal impairment. However, for patients treated with desirudin, our data show that the rate of major VTEs for those with stage 3B CKD was similar to the rate for those with less severe renal impairment. Moreover, our findings are novel in that they are derived from a randomized trial in which the diagnosis was based on aggressive surveillance for this event. Other RCTs in the field of VTE prevention have not investigated the relationship between VTE and CKD.
The mechanism(s) explaining the greater risk for VTE in CKD remains unclear. Potentially, the increased inflammatory state associated with CKD may contribute to hypercoaguability. For example, the levels of various cytokines, such interleukin-6, are elevated in persons with CKD . Similarly, the levels of C-reactive protein, fibrinogen, factor VII, FVIII and D-dimer are all elevated in those with impaired renal function . Furthermore, as renal function declines, levels of many of these inflammatory and procoagulant proteins climb. Other hemostatic abnormalities have been reported in CKD, and include higher levels of thrombin–antithrombin complexes and reduced antithrombin activity . In addition to these baseline abnormalities in CKD, surgery itself represents a major inflammatory insult that propagates thrombosis in a multitude of ways. Hence, even if the risk for VTE in CKD alone may be mildly elevated, the interaction between surgery and CKD may lead to excessively robust disruption of the inflammatory and procoagulant cascades to foster VTE. Formal studies utilizing serial measurement of multiple biomarkers will be necessary to confirm this hypothesis. Our present study is limited, in that it can only note the association between surgery, CKD, and VTE, and cannot provide greater insights into the pathophysiology of this process.
Irrespective of the possible mechanism of the impact of CKD on VTE risk, our data indicate that patients with stage 3B CKD merit special focus when undergoing risk assessments for VTE. An estimated 20% of the US population > 60 years of age suffer from stage 3 CKD, with as many as 40% of these having stage 3B CKD . The number of patients with CKD is also expected to grow substantially in the future, given the prevalence of both hypertension and diabetes mellitus and the overall aging of the population. Taken together, these trends indicate that future RCTs exploring VTE prevention must specifically focus on these cohorts. More importantly, it would seem imprudent to conclude that interventions for VTE prophylaxis that appear effective in groups of patients with normal renal function will necessarily be similarly efficacious in those patients with CKD.
In this respect, we noted, among subjects with stage 3B CKD, a differential impact of desirudin and enoxaparin on both bleeding (in terms of postoperative BI) and major VTEs. Both desirudin and enoxaparin are eliminated through renal mechanisms, and have been shown to accumulate in patients with CKD [14,15]. Hence, dose reductions are recommended for both agents to avoid bleeding complications. The situation may, however, be more complex. Although both drugs do accumulate in CKD, there may be more bioaccumulation of enoxaparin than of desirudin. However, if this fact accounted for more frequent bleeding with enoxaparin, one would expect fewer VTEs as well, because of an enhanced antithrombotic effect. Conversely, differences in the mechanism of anticoagulation may explain our observations. Desirudin differs from low molecular weight heparins (LMWHs) in that it directly inhibits both clot-bound and circulating thrombin. LMWHs predominately inhibit circulating FXa indirectly through catalyzing the effect of antithrombin. LMWHs, such as enoxaparin, have varying degrees of activity against thrombin, but cannot inhibit clot-bound thrombin or platelet-bound FXa. Direct thrombin inhibition, therefore, might theoretically provide better thrombosis protection in patients with decreased circulating antithrombin concentrations or increased concentrations of thrombin. As noted above, persons with CKD often have higher concentrations of thrombin and lower levels of antithrombin .
Despite our data deriving from a large prospective RCT, the present analysis has several important limitations. First, although the data were collected prospectively, this analysis is retrospective, and thus exposed to various forms of bias, although such possible bias would be expected to be reduced by the use of a prespecified analysis plan and the fact that only a single subgroup (i.e. based on eGFR) was considered and identified on a clinical basis prior to the performance of any supplementary analysis. Second, RCTs in VTE prophylaxis, such as the one explored in our analysis, rely on prospective, mandatory screening protocols (e.g. venography) of all patients for the diagnosis of VTE. This necessarily leads to the diagnosis of clinically silent events whose significance is often controversial. In turn, this affects the generalizability of our results. Likewise, as our data focus only on patients undergoing THR, the relevance of our conclusions to other settings is unclear, and confirmatory studies are therefore necessary. Third, we defined CKD on the basis of a baseline serum creatinine (SCr), so we did not capture the effect of changes in renal function during the course of the study and their impact on outcomes. Fourth, renal function was estimated from the Cockcroft–Gault equation. This equation may not be accurate in obese patients or those with very low weight. We attempted to adjust for this by using LBW in the calculation and a minimum SCr value of 1.0 mg dL−1. Nevertheless, eGFR is not as reliable as measured glomerular filtration rate, and may have led to overestimation or underestimation of the degree of renal function in this cohort. Finally, key variables used in the Cockcroft–Gault equation are gender and age, and we cannot rule out the impact of these variables on the observed outcomes, although they are both included as covariates in all efficacy analyses. It should also be noted that the treatment–eGFR interaction is statistically significant (P = 0.093 from the analysis that based eGFR on a continuous variable; P = 0.038 for the analysis based on a binary variable of the presence or absence of stage 3B CKD), whereas subsequent analysis has shown that none of the interactions of treatment with age, gender, body mass index, weight or history of VTE are statistically significant (data not shown).
In summary, CKD is an important risk factor for both VTE and bleeding complications of anticoagulation, and should be taken into consideration when considering VTE prophylaxis therapies. As highlighted by the present analysis of one large randomized clinical trial of patients undergoing elective THR surgery in which stage 3B CKD occurred in nearly 30% of patients and disproportionately affected women, the prevalence of CKD is an important and growing public health problem. Our analysis suggests that the presence of CKD may variably influence the efficacy of VTE prophylaxis through an as yet undefined mechanism. However, given the post hoc nature of our analysis, this hypothesis requires evaluation in a prospective randomized trial.
Disclosure of Conflict of Interests
Funding for this analysis was provided by Canyon Pharmaceuticals, Inc. A. Jaffer, A. F. Shorr and J. Smith have served as consultants to Canyon Pharmaceuticals.