Bivalirudin is an alternative to unfractionated heparin (UFH) anticoagulation during percutaneous coronary intervention. Previously, we have reported clinical benefit on major bleeding in favor of bivalirudin compared with UFH monotherapy but inconclusive results on mortality. Controversial data have been reported in the last 2 years. We conducted an updated meta-analysis including randomized trials and observational studies, which evaluated ischemic and bleeding outcomes for bivalirudin compared with UFH-only during percutaneous coronary intervention. We included 18 observational studies and 12 randomized trials published from 2003 to 2015. Primary outcomes were major adverse cardiovascular events within 30 days including death, myocardial infarction, and urgent revascularization and stent thrombosis, major bleeding, and transfusion. Overall, we found a significant risk reduction with bivalirudin for major bleeding (odds ratio [OR] 0.59, 95% confidence interval [CI] 0.49 to 0.71, p <0.0001) and for transfusion (OR 0.79, 95% CI 0.66 to 0.95, p = 0.01) and similar risk for major adverse cardiovascular events (OR 0.98, 95% CI 0.86 to 1.12, p = 0.80). However, there was a substantial increased risk of stent thrombosis associated with bivalirudin (OR 1.52, 95% CI 1.11 to 2.08, p = 0.009). No impact on mortality was found. Meta-regression analyses on major bleeding suggested that bivalirudin was more effective than UFH at doses >60 IU/kg and independent of radial access. In conclusion, compared with UFH monotherapy, bivalirudin remains associated with less bleeding risk but higher stent thrombosis risk. Further study remains required to define its role in current antithrombotic armamentarium.
The optimal anticoagulant treatment approach for patients who undergo percutaneous coronary intervention (PCI) in contemporary practice remains unclear. For the last 10 years, the use of bivalirudin, a direct thrombin inhibitor, has been associated with reduction of major adverse cardiovascular event (MACE) rate compared with standard use of unfractionated heparin (UFH). Most large randomized controlled trials (RCTs) have compared anticoagulation with bivalirudin to UFH plus a planned glycoprotein IIb/IIIa inhibitor (GPI). Whether the net clinical benefit of bivalirudin compared with UFH in the absence of routine GPI is maintained has been questioned. In a previous meta-analysis comparing bivalirudin to UFH monotherapy in femoral-treated patients with PCI, we reported that the benefit of bivalirudin over UFH monotherapy was driven by a significant decrease in major bleeding (odds ratio [OR] 0.55, 95% confidence interval [CI] 0.43 to 0.72), with similar rates of MACE (OR 0.92, 95% CI 0.75 to 1.12). A decrease in mortality was seen in observational studies (OR 0.62, 95% CI 0.45 to 0.85) but remained inconclusive in randomized trials (OR 0.63, 95% CI 0.20 to 2.01). Recently, a few original reports and meta-analyses of RCTs have suggested that the beneficial effect of bivalirudin compared with UFH plus provisional GPI on bleeding events had been reduced or vanished and was counterbalanced by an increased rate of stent thrombosis. Since our previous report, 9 RCTs, including the large MATRIX (Minimizing Adverse Haemorrhagic Events by TRansradial Access Site and angioX) trial (n = 7,213) and 5 observational studies comparing bivalirudin with UFH monotherapy have been reported. We aimed to get insights into the optimal anticoagulation strategy in PCI on ischemic and bleeding outcomes by conducting an updated meta-analysis and meta-regression analyses including all RCTs and observational studies.
Methods
We carried out this review and meta-analysis according to the protocols recommended by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses group for randomized trials and Meta-analysis of Observational Studies in Epidemiology group for observational studies. We searched in PubMed database, EMBASE, and Cochrane Library, using the terms “bivalirudin,” “hirulog” and “antithrombin,” “stent,” and “percutaneous coronary intervention” (last update September 1, 2015). In addition, we performed a manual search in cardiology journals and abstracts from international cardiology meetings for relevant data. Inclusion criteria required to report clinical outcomes in bivalirudin and UFH groups separately. No language restriction was applied. Because our objective was to assess UFH monotherapy, studies involving >25% of GPIs (bailout) were excluded. Case reports, studies involving noncoronary interventions, and abstracts not reported as peer-reviewed manuscripts were also excluded. Two evaluators performed literature searches and extracted data independently. Discrepancies between data sets were resolved by consensus, if necessary, after contact with the investigators. We classified the studies by randomized or observational design.
Clinical outcomes evaluated were MACE including death, myocardial infarction (MI), or urgent revascularization and death from any cause, MI, major bleeding, transfusions, and stent thrombosis. Events were censored either during inhospital phase or at 30 days post-procedure. For bleeding, we used Thrombolysis In Myocardial Infarction major criteria when available or the study-defined major bleeding. For an easier study data management, we entered every relevant parameter, such as study design, sample size, population demographics, UFH and bivalirudin dose, procedural adjuvant pharmacotherapy, and ischemic and bleeding outcomes into a data sheet.
The absolute number of events for each outcome in every study was entered into Review Manager 5.1.6 (Cochrane Collaboration, Copenhagen, Denmark). The outcomes were reported as OR and 95% CI for randomized trials and observational studies separately and combined regardless of design. Taking into account the differences in study characteristics that may introduce a level of heterogeneity in the estimated treatment effect among studies, we used a random-effects model. Heterogeneity was evaluated across studies with the Cochrane Q statistic and I 2 test. An I 2 value <25% was considered to indicate low heterogeneity, a value between 25% and 50% was considered moderate heterogeneity, and a value >50% was considered substantial heterogeneity. The weight of each study on the overall result of meta-analysis outcome was calculated as the percentage of the number of patients in that given study over the total number of patients included in each outcome analysis. Publications bias was evaluated using the Egger regression test. Meta-regressions were performed where possible to investigate sources of heterogeneity in common clinical and demographic variables, including age, gender, diabetes, impaired renal function, use of GPI, and UFH dose. Meta-regressions were performed using Comprehensive Meta Analysis, version 2.0 (Biostat, Englewood, New Jersey). Statistical significance for hypothesis testing was set at the 0.05 2-tailed level.
Results
The study inclusion process in this updated meta-analysis is described in Figure 1 . Fourteen new studies were included: 9 RCTs (13,840 patients) and 5 observational studies (5,022 patients), for an overall total of 30 studies (51,593 patients). All studies were reported from 2003 to 2015. Among all included RCTs (total of 12 studies), 8 were performed in Europe, 2 in China, and 1 in the United States. Most observational studies (total of 18 studies) were done in the United States with the exception of 2 studies in Europe and 2 in Canada. Assessment of publication bias using the ORs of clinical outcomes of the included studies demonstrated a symmetrical funnel plot with no evidence of publication bias for the primary analysis by the Egger regression–based test (all p values not significant, data not shown). Study characteristics are summarized in Table 1 . In the UFH group, initial bolus dose was between 40 and 140 IU/kg. In 17 studies, the operator used an additional bolus according to activated clotting time, and in 9 studies UFH dose was not available. In the bivalirudin group, the dose was a bolus of 0.75 mg/kg/h and a 1.75 μg/kg/h infusion. Eleven studies did not use GPI in either anticoagulation group. Mean age of patients was 67 years (range 58 to 93 years), proportion of females was 32% (range 16% to 51%), and 35% (range 13% to 87%) were diabetics. Eleven studies enrolled only patients with acute coronary syndrome, otherwise proportion varied between 0% and 84%. Prevalence of chronic renal failure ranged from 2.5% to 100%.
| Study | Year | N patients | % ACS | H IU/kg | % Thieno | % GPI | Age | % Female | % DM | % CRF | |||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| B | H | B | H | B | H | ||||||||
| Observational studies | |||||||||||||
| Wolfram | 2003 | 1878 | 82 | 40 | 91 | 0 | 0 | 76 | 74 | 37 | 37 | 36 | 2.6 |
| Gurm | 2005 | 1725 | 59 | – | – | 0 | 0 | 66 | 66 | 30 | 27 | 34 | 5.9 |
| Rha | 2005 | 114 | – | 40 | 93 | 0 | 0 | 72 | 69 | 28 | 23 | 40 | – |
| Chu | 2006 | 672 | 100 | 40 | 100 | 15 | 20 | 66 | 65 | 41 | 37 | 35 | 18 |
| Gurm | 2007 | 5973 | 0 | – | – | 7.6 | 0 | 66 | 65 | 36 | 36 | 34 | 12 |
| Bonello | 2009 | 899 | 100 | 50 | 97 | 0 | 0 | 63 | 62 | 32 | 38 | 30 | 16 |
| Lemesle | 2009 | 171 | 82 | – | 100 | 1.3 | 17 | 93 | 92 | 49 | 53 | 28 | 26 |
| Lemesle | 2009 | 2766 | 61 | – | 100 | 2 | 16 | 84 | 84 | 49 | 52 | 28 | 22 |
| Delhaye | 2010 | 396 | 79 | 50- 70 | 100 | 0 | 0 | 66 | 65 | 43 | 54 | 68 | 100 |
| Lindsey | 2010 | 1364 | 65 | 70 | 100 | 2 | 20 | 65 | 64 | 39 | 32 | 36 | 5.4 |
| Lopes | 2010 | 3136 | 100 | – | 76 | 0 | 0 | 66 | 66 | 37 | 39 | 34 | 2.5 |
| Schulz | 2010 | 4794 | 20 | 100 | 100 | 0.2 | 0 | 68 | 68 | 24 | 22 | 29 | – |
| Bangalore | 2011 | 3022 | 34 | – | 48 | 0 | 0 | 65 | 66 | 35 | 33 | 34 | – |
| Mac Haalany | 2012 | 250 | 82 | 70 | 100 | 3 | 20 | 72 | 66 | 34 | 35 | 36 | 40 |
| Hibbert | 2012 | 1424 | 100 | 60 | 100 | 7.5 | 0 | 62 | 63 | 28 | 30 | 16 | – |
| Bangalore ACS | 2014 | 1036 | 100 | – | 52 | 0 | 0 | 65 | 65 | 33 | 35 | 36 | 5 |
| Bangalore SIHD | 2014 | 2062 | 0 | – | 50 | 0 | 0 | 66 | 66 | 32 | 32 | 35 | 5 |
| Akin | 2014 | 250 | 14 | 70-100 | 100 | 1 | 4 | 70 | 73 | 22 | 26 | 27 | – |
| Randomized trials | |||||||||||||
| Kastrati | 2008 | 4570 | 18 | 140 | 100 | 0.2 | 0.2 | 67 | 67 | 24 | 23 | 27 | – |
| Parodi | 2010 | 850 | 27 | 100 | 100 | 0 | 0 | 69 | 69 | 23 | 25 | 22 | – |
| Patti | 2012 | 401 | 29 | 75 | 100 | 12 | 14 | 70 | 70 | 29 | 27 | 63 | 21 |
| Steg | 2013 | 1549 | 100 | 60-100 | 98 | 11 | 25 | 61 | 62 | 25 | 23 | 13 | 17 |
| Waksman | 2013 | 100 | 100 | 60 | – | 5.9 | 20 | 63 | 62 | 27 | 37 | 17 | 5 |
| Xiang | 2013 | 217 | 75 | 130 | 100 | 0.9 | 3.7 | 57 | 59 | 16 | 18 | – | – |
| Schulz | 2014 | 548 | 100 | 70- 100 | 100 | 3.0 | 6.1 | 61 | 61 | 24 | 21 | 16 | – |
| Shahzad | 2014 | 1812 | 100 | 70 | 99 | 13 | 15 | 63 | 64 | 29 | 27 | 14 | – |
| Briguori | 2014 | 837 | 23 | 70 | – | 10 | 1.3 | 78 | 78 | 50 | 44 | 44 | 46 |
| Han | 2014 | 1464 | 100 | 100 | 100 | 4 | 6 | 57 | 58 | 17 | 18 | 21 | 10 |
| Feldman | 2014 | 100 | 32 | 60 | 100 | 0 | 0 | 68 | 65 | 22 | 40 | 87 | 22 |
| Valgimigli | 2015 | 7213 | 100 | 50-100 | 82 | 4.6 | 25 | 65 | 65 | 24 | 23 | 22 | 1.3 |
Overall, bivalirudin was associated with similar risks of MACE compared with UFH, with an incidence of 6.0% and 6.4%, respectively (OR 0.98, 95% CI 0.86 to 1.12, p = 0.80), with no difference regarding effect size between observational studies (4.9% vs 5.7%, OR 0.95, 95% CI 0.80 to 1.14, p = 0.60) and RCTs (7.4% vs 7.5%, OR 1.02, 95% CI 0.82 to 1.28, p = 0.85; Figure 2 ). In contrast, all-cause mortality was significantly different in the bivalirudin group compared with the UFH group (1.2% vs 1.5%, respectively, OR 0.75, 95% CI 0.63 to 0.89, p = 0.001), but this was entirely driven by the observational studies (0.9% vs 1.3%, OR 0.61, 95% CI 0.48 to 0.76, p <0.0001) with no impact in RCTs (OR 0.86, 95% CI 0.69 to 1.07, p = 0.17; Figure 3 ).
Bivalirudin was associated with a ∼35% relative reduction in major bleeding ( Figure 4 ; 1.9% vs 2.9%, OR 0.59, 95% CI 0.49 to 0.71, p <0.0001), with larger effect size observed in observational studies (2.5 vs 3.9%, OR 0.57, 95% CI 0.45 to 0.72, p <0.0001) compared with RCTs (1.2% vs 1.7%, OR 0.63, 95% CI 0.45 to 0.90, p = 0.01). There was no significant difference in the incidence of MI regardless of the study design (OR 1.05, 95% CI 0.89 to 1.23, p = 0.57; Figure 5 ). Almost 50% higher relative risk in the overall incidence of acute and subacute stent thrombosis was observed with bivalirudin (0.9% vs 0.6%, OR 1.52, 95% CI 1.11 to 2.08, p = 0.009). Surprisingly, this effect of stent thrombosis was particularly striking in the bivalirudin group in RCTs (1.1% vs 0.7%, OR 1.53, 95% CI 0.98 to 2.38, p = 0.06) but remained inconclusive in observational studies (0.5% vs 0.4%, OR 1.56, 95% CI 0.89 to 2.75, p = 0.12; Figure 6 ). Sixteen studies out of 30 reported blood transfusion rates ( Figure 7 ) with substantial clinical benefit for bivalirudin (OR 0.79, 95% CI 0.66 to 0.95, p = 0.01).
