Should Bivalirudin Replace Heparin During Percutaneous Coronary Interventions?

In 1994, the development of bivalirudin, a novel direct thrombin inhibitor, was precipitously suspended by the manufacturer based on an unfavorable economic analysis comparing bivalirudin with heparin in patients undergoing angioplasty.^{1 - 4} However, the convergence of several events over the last 8 years maintained interest in this agent: a new, less expensive manufacturing process was developed; the drug was licensed to a new sponsor in 1997; and the original angioplasty trial was reanalyzed with its results cast in a more favorable light.^{5 - 6} Now, in this issue of THE JOURNAL, the investigators of the Randomized Evaluation in PCI Linking Angiomax to Reduced Clinical Events (REPLACE)–2 trial⁷ suggest that clinicians should consider bivalirudin as the core anticoagulant in patients undergoing a percutaneous coronary intervention (PCI).

Why would a direct thrombin inhibitor that works only distally in the coagulation cascade possibly be effective in PCI, a procedure for which it is established that platelet aggregates play an important role in thrombus formation? Such platelet aggregates can occur in response to spontaneous disruption of a vulnerable plaque but can also develop in response to high-pressure balloon inflations and deployment of coronary stents. An especially potent stimulus for platelet activation is thrombin, which binds to the thrombin receptor on platelets ultimately leading to the expression of activated glycoprotein IIb/IIIa (Gp IIb/IIIa) receptors on the platelet surface. Ligands such as fibrinogen cross-link activated platelets via the Gp IIb/IIIa receptor forming platelet aggregates. In addition to contributing to thrombus formation, such platelet aggregates increase the surface area for the prothrombinase complex by providing a phospholipid membrane platform on which a complex of activated factors V and X and calcium ions can form. A potent direct thrombin inhibitor such as bivalirudin could theoretically reduce the thrombotic burden during PCIs by blocking the formation of fibrin and by suppressing platelet aggregation.⁸ However, direct thrombin inhibitors would not be expected to block platelet activation in response to agonists other than thrombin and would not disaggregate existing thrombi.

Additional features of bivalirudin that make it an attractive agent for clinical study include its ability to inhibit both fluid-phase and clot-bound thrombin, its lack of binding to plasma proteins, and its lack of neutralization by platelet factors. Because of its stable antithrombotic effect, bivalirudin can be administered as a constant infusion and adjustments in response to hematologic measurements are infrequently needed. Less than 20% of the dose is cleared by the renal route, thereby minimizing the impact of renal dysfunction on dosing except when renal failure is severe. A mechanistic explanation for the lower rates of bleeding observed in trials of bivalirudin is that its binding to thrombin is noncompetitive and reversible since thrombin can slowly cleave an Arg3-Pro4 bond in bivalirudin with resultant reexposure of the catalytic center of thrombin.⁹ Exposure of the catalytic center of thrombin restores hemostatic capacity by allowing the formation of thrombin.

In 2000, the US Food and Drug Administration approved bivalirudin for use as an anticoagulant in patients with unstable angina undergoing percutaneous transluminal angioplasty. The package insert⁹ reminded clinicians that the safety and effectiveness of bivalirudin were not established for its use in conjunction with platelet inhibitors other than aspirin because the angioplasty trial that formed the basis for approval was conducted between 1993 and 1994. Since then, 3 important tools were added to the interventionalist's armamentarium: intravenous Gp IIb/IIIa inhibitors, coronary stents, and thienopyridines. The stage was thus set for another trial of bivalirudin in the modern interventional era.

Such a trial evaluating bivalirudin might have included 4 treatment groups: (1) heparin plus provisional Gp IIb/IIIa inhibitor, (2) heparin plus planned Gp IIb/IIIa inhibitor, (3) bivalirudin plus provisional Gp IIb/IIIa inhibitor, and (4) bivalirudin plus planned Gp IIb/IIIa inhibitor. However, only groups 2 and 3 were studied in REPLACE-2. The investigators argued that because Gp IIb/IIIa inhibitors reduce ischemic events after a PCI, it would be unethical to withhold them and therefore an active control (group 2) was needed.¹⁰ However, an imputed comparison to heparin alone was constructed. Group 4, a potentially very interesting combination scientifically, was not included in REPLACE-2 for practical reasons associated with sample size and concerns that planned use of Gp IIb/IIIa inhibitors with bivalirudin could attenuate some of its bleeding advantage and detract from the economic appeal of minimizing the use of Gp IIb/IIIa inhibitors in favor of bivalirudin.

The REPLACE-2 investigators departed from an approach traditionally taken in registration pathway trials in which the primary end point consists exclusively of efficacy components while safety observations are reported separately and/or included as a secondary end point such as net clinical benefit.¹¹ In REPLACE-2 the primary end point was a quadruple composite of death, myocardial infarction, urgent revascularization, or in-hospital major bleeding. The triple composite end point of just the efficacy elements was the secondary end point in the trial. Combining efficacy and safety into one composite end point can be problematic from a regulatory perspective because drugs that are ineffective but have a safety advantage can appear to be better than drugs with proven efficacy.

Since it was believed that bivalirudin would not be superior to heparin plus planned Gp IIb/IIIa inhibitor, a noninferiority design was selected. What does that mean and how was it done? Noninferiority trials involve a boundary or margin that places a statistical limit on how much inferiority a new therapy may exhibit compared with standard therapy while operationally being considered to provide similar effectiveness to the standard therapy.^{12 - 13} Investigators may prespecify a margin that is a fixed amount on the basis of clinical judgment (eg, the odds ratio [OR] comparing the new and standard therapies cannot have an upper bound of the 95% confidence interval [CI] that exceeds 1.10).^{12 - 13} Investigators may also prespecify a margin that statistically ensures some proportion of the previously established treatment effect of the standard therapy (compared with placebo) is preserved.^{14 - 15} The larger the treatment effect of the standard therapy vs placebo, the easier it is to show preservation of a proportion of that effect. This can result in a situation that satisfies a statistical definition of noninferiority but is not that conservative from a clinical perspective because a considerable loss of the effectiveness of the standard therapy would still fall within the prespecified definition of noninferiority.

The REPLACE-2 investigators estimated the treatment effect of combining Gp IIb/IIIa inhibitors with heparin from a meta-analysis of 2 placebo-controlled trials in the modern interventional era—Evaluation of Platelet Inhibition in Stenting (EPISTENT) and Enhanced Suppression of the Platelet IIb/IIIa Receptor with Integrilin Therapy (ESPRIT)—and then set up 2 hypotheses.^{16 - 17} They first hypothesized that bivalirudin plus provisional Gp IIb/IIIa inhibition would be superior to heparin alone.⁷ This was accomplished by imputing an OR for bivalirudin vs heparin as the mathematical product of the OR from the prior meta-analysis times the OR comparing the 2 treatments actually tested in REPLACE-2 (groups 2 and 3). To establish superiority, the upper bound of the 95% CI for the imputed OR against heparin alone had to exclude unity. Using the definitions in the trial, REPLACE-2 demonstrated the superiority of bivalirudin compared with heparin alone for both the quadruple and triple end points. The second hypothesis was that provisional use of Gp IIb/IIIa inhibitors with bivalirudin would preserve at least 50% of the benefit of adding a Gp IIb/IIIa inhibitor to heparin compared with heparin alone. The results of REPLACE-2 show that bivalirudin plus provisional Gp IIb/IIIa inhibition preserves at least 50% of the treatment effect of combining Gp IIb/IIIa inhibition with heparin for the triple end point and 50% to 75% of the treatment effect for the quadruple end point.^{14 - 15}

The usual concerns about composite end points become more complicated in REPLACE-2 because the triple composite of efficacy elements favored the heparin plus planned Gp IIb/IIIa inhibitor group (OR, 1.09; 95% CI, 0.90-1.32), whereas the safety outcome (major bleeding) favored the bivalirudin plus provisional Gp IIb/IIIa inhibitor group. As noted in Table 3 and Table 4 of the article, there was a net excess of 23 non–Q-wave myocardial infarctions (predominantly in the intermediate and large categories as ascertained by creatine kinase-MB elevations) but 52 fewer major bleeding events in the bivalirudin group. Because the difference in major bleeding events was greater than the difference in efficacy events, the quadruple composite end point favored the bivalirudin group (OR, 0.92; 95% CI, 0.77-1.09). Thus, with 95% confidence, bivalirudin with provisional Gp IIb/IIIa inhibitor compared with heparin plus planned Gp IIb/IIIa inhibitor could be no more than 9% worse with respect to the quadruple end point (ie, would have satisfied a rather conservative noninferiority OR margin of 1.10 in a strict comparison of the 2 groups actually studied) but could be up to 32% worse with respect to the triple efficacy end point (ie, would not have satisfied a noninferiority OR margin of 1.30).

The benefit of bivalirudin pivots around the difference in the frequency of major bleeding. The criteria for major bleeding in REPLACE-2 are roughly comparable with the combination of major plus minor bleeding by the Thrombolysis in Myocardial Infarction (TIMI) criteria. As noted in Table 4 of the REPLACE-2 report, the drivers for the difference in major bleeding using the REPLACE-2 criteria were vascular access puncture and gastrointestinal tract bleeding episodes. There was little difference in the 2 groups by the TIMI major criteria and the quadruple end point rates would have been quite similar if that bleeding definition had been used. The TIMI major bleeding rate in the heparin plus Gp IIb/IIIa inhibitor group of REPLACE-2 (0.9%) is in the same range as the comparable groups of the EPISTENT (1.5%) and ESPRIT (1%) trials.^{16 - 17}

There was a significant excess of TIMI minor bleeding events in the heparin plus Gp IIb/IIIa inhibitor group, which necessitates a search for potential explanations. The bolus dose of heparin in REPLACE-2 was midway between that used in the EPISTENT (70 U/kg, maximum 7000 U) and ESPRIT (60 U/kg, maximum 6000 U) trials, yet the activated clotting time (ACT) values in REPLACE-2 tended to be higher than those observed in both EPISTENT and ESPRIT and exceeded the recommended ACT range of 200 to 300 seconds in the package inserts for abciximab and eptifibatide.^{16 - 17} It is difficult to know if the patients in the control group in REPLACE-2 were anticoagulated to a higher degree than comparable patients in the other trials because of limitations of comparisons of ACT values across trials where different generations of instrumentation may have been used. However, the ESPRIT investigators observed a trend toward higher rates of major bleeding events (TIMI criteria) comparing the lowest ACT tertile with the middle and highest tertiles, but they observed no loss of benefit of eptifibatide in patients with lower ACT values.^{17 - 18}

How should the available data on bivalirudin be translated into current practice? Bivalirudin is an attractive anticoagulant to support PCI in patients with renal failure and in those with heparin induced thrombocytopenia. Beyond that, the situation becomes more complicated. Because REPLACE-2 excluded patients with acute ST-elevation myocardial infarction, the trial results cannot be extrapolated to guide dosing for primary PCI. Interventionalists should plan to use a Gp IIb/IIIa inhibitor rather than beginning the PCI with an anticoagulant alone in patients with unstable angina or non–ST-elevation myocardial infarction when a thrombus is seen angiographically and in patients receiving treatment for diabetes.^{19 - 20} Much more experience with bivalirudin in combination with Gp IIb/IIIa inhibitors is needed to understand fully the safety of that combination before switching from heparin. It will also be important to see adjusted analyses in the future for clinicians to determine if the results with bivalirudin in REPLACE-2 are applicable to both Gp IIb/IIIa inhibitors studied. The findings with bivalirudin will need to be judged in the context of the rapidly changing menu of new anticoagulants available to support PCI. Clinicians may now use agents that act proximately in the coagulation cascade, such as low-molecular-weight heparins and possibly in the future inhibitors of tissue factor and factor Xa.^{21 - 22}

Whether bivalirudin should replace heparin in more elective PCI cases (about 56% of the REPLACE-2 trial population) will depend on whether interventionalists (and the US Food and Drug Administration) accept the quadruple primary end point, the definition of major bleeding, the definition of the noninferiority margin, and the dose-ACT range in the heparin group of REPLACE-2. Economic considerations are also likely to be important. The tradeoff appears to be fewer non–life-threatening bleeding events with bivalirudin compared with heparin at the cost of a few more moderately sized myocardial infarctions when Gp IIb/IIIa inhibitors are used provisionally rather than routinely.