Critical Analysis of Thrombocytopenia Associated With Glycoprotein IIb/IIIa Inhibitors and Potential Role of Zalunfiban, a Novel Small Molecule Glycoprotein Inhibitor, in Understanding the Mechanism(s)

A meta‐analysis of 29 randomized studies of 123 419 patients treated with the currently available intravenous agents and 4 oral agents that were studied but never approved for clinical use found unweighted incidence values for mild thrombocytopenia (<100 000 platelets/μL) of 3.3% in GPI‐treated patients and 2.2% in placebo‐treated patients (relative risk [RR], 1.63 [95% CI, 1.48–1.79]); for severe thrombocytopenia (<50 000 platelets/μL) the results were 0.8% and 0.2% (RR, 3.51 [95% CI, 2.68–4.58]).¹⁸ An analysis limited to the 3 approved GPIs, abciximab, eptifibatide, and tirofiban, found similar RR values of 1.61 (95% CI, 1.46–1.78) and 3.9 (95% CI, 3.08–4.95), respectively, for mild and severe thrombocytopenia. Subgroup analyses identified increased RR of mild thrombocytopenia for abciximab (2.93 [95% CI, 2.43–3.52]) and tirofiban (2.79 [95% CI, 1.17–6.63]), but not eptifibatide (1.05 [95% CI, 0.86–1.29]). One confounding factor in comparing different trials is the higher rate of thrombocytopenia in the placebo groups of studies of patients with acute coronary syndromes compared with studies of patients undergoing elective PCI. The difference has been postulated to be due to longer duration heparin therapy in the former studies and the known impact of heparin on developing mild thrombocytopenia.^{19, 20} If GPIs prevent some of the mild thrombocytopenia associated with heparin therapy, comparing the rates of thrombocytopenia in the GPI group to the placebo group may underestimate the frequency of GPI‐associated thrombocytopenia. Consistent with such a hypothesis is the finding that in the 9217‐patient PURSUIT (Platelet Glycoprotein IIb/IIIa in Unstable Angina: Receptor Suppression Using Integrilin Therapy) acute coronary artery study, although the overall frequency of thrombocytopenia was indistinguishable between eptifibatide‐treated and control patients, the onset of thrombocytopenia was more rapid with eptifibatide treatment.¹⁵ For context, all 3 GPIs appear on lists of the drugs most commonly associated with thrombocytopenia.²¹

In the 2099‐patient EPIC (Evaluation of 7E3 for the Prevention of Ischemic Complications) study, mild, severe, and profound acute thrombocytopenia were all more frequent with abciximab bolus+infusion than placebo (mild: 5.2% versus 3.3%; severe 1.6% versus 0.7%; profound 0.3% versus 0%).^{17, 22} The median time to any thrombocytopenia was also shorter with abciximab bolus+infusion therapy than placebo (0.50 versus 1.68 days). Thrombocytopenia, whether in abciximab‐treated or placebo‐treated patients, was associated with increased mortality (12.4% versus 1.1%), myocardial infarction (33.3% versus 5.6%), coronary artery bypass surgery (33.3% versus 3.7%), and balloon pump insertion (23.5% versus 2.2%). Of note, however, thrombocytopenic patients treated with abciximab bolus+infusion had fewer ischemic complications than thrombocytopenic placebo‐treated patients (mortality 10.8% versus 13.0%; myocardial infarction 21.6% versus 65.2%; coronary artery bypass surgery 24.3% versus 60.9%; balloon pump insertion 13.5% versus 47.8%). Thrombocytopenic patients had increased major bleeding (50.6% versus 9.0%) and nonsurgical hemorrhage (bleeding index 4.0 versus 2.0). However, abciximab bolus+infusion‐treated thrombocytopenic patients had less major bleeding than placebo‐treated thrombocytopenic patients (43.1% versus 69.6%), less nonsurgical bleeding (bleeding index 3.8 versus 8.0), and less overall bleeding (bleeding index 4.8 versus 8.7). The authors noted that the placebo‐treated patients who developed thrombocytopenia were sicker than the abciximab‐treated patients who developed thrombocytopenia, which may explain their greater tendency to bleed with thrombocytopenia. None of the patients with thrombocytopenia had intracranial hemorrhage. Predictors of thrombocytopenia included lower baseline platelet count, older age, and lighter weight. Similar data were observed in the 1265‐patient CAPTURE (c7E3 Fab Antiplatelet Therapy in Unstable Refractory Angina) study,²³ the 2702‐patient EPILOG (Evaluation in PTCA [Percutaneous Transluminal Coronary Angioplasty] to Improve Long‐Term Outcome With Abciximab GPIIb/IIIa Blockade) study,^{24, 25} the 2399‐patient EPISTENT (Evaluation of Platelet Inhibition in Stenting) study,²⁵ the abciximab arm of the 4623‐patient TARGET (Do Tirofiban and ReoPro Give Similar Efficacy Outcomes) study,²⁶ and the 7800‐patient GUSTO IV‐ACS (Global Use of Strategies to Open Occluded Coronary Arteries IV‐Acute Coronary Syndrome) study.²⁷

The duration of abciximab therapy may have an impact on the development of thrombocytopenia. In the EPIC study, patients treated with a bolus of abciximab compared with a bolus and a 12‐hour infusion had a lower frequency of mild thrombocytopenia (placebo 3.3%, bolus 3.0%, bolus and infusion 5.2%; severe thrombocytopenia [0.7% versus 0.4% versus 1.6%, respectively]); and acute profound thrombocytopenia (0% versus 0% versus 0.3%, respectively). In a 1005‐patient study comparing bolus versus bolus and infusion abciximab for PCI, however, the frequency of severe thrombocytopenia was identical (0.6%) in each arm.²⁸

The ReoPro Re‐administration Registry analyzed data on 500 patients retreated with abciximab and found that thrombocytopenia (<100 000 platelets/μL) occurred in 23 patients (4.6%), of whom 12 (2.4%) had profound thrombocytopenia (<20 000 platelets/μL).^{29, 30} The onset was delayed until after discharge in 2 patients (0.4%). All patients with profound thrombocytopenia received platelet transfusions, with 4 requiring more than 1 platelet transfusion and 2 requiring more than a week to recover. Any type of bleeding occurred in 25.4% of thrombocytopenic patients compared with 9.6% of those without thrombocytopenia. Major bleeding occurred in 8 patients (1.6%), 6 of whom had normal platelet counts at the time, and 1 of whom had profound thrombocytopenia. Separately, Curtis et al reported on 9 patients who developed acute, profound thrombocytopenia after receiving a second dose of abciximab 2 to 15 weeks after the initial dose.³¹ Two patients had acute hypersensitivity reactions, and 1 of them developed intrapulmonary hemorrhage. Bleeding symptoms varied from petechial hemorrhage to ecchymoses to oozing from venipuncture sites. The response to platelet transfusion varied widely, with 3 patients obtaining good increments, 2 responding only transiently and requiring repeat transfusions, and 2 having no response. Platelet recovery occurred within 4 to 7 days in 7 patients, but 2 patients required more than a week.

A post hoc analysis of the PRISM (Platelet Receptor Inhibitor Ischemic Syndrome Management) trial found that the incidence of mild thrombocytopenia in the early phase of the study (n=1753) was higher in those treated with phosphate‐buffered tirofiban compared with those treated with unfractionated heparin (2.0% versus 0.7%, respectively; P=0.034).³² Treatment with citrate‐buffered tirofiban in the later phase of the study (n=1479), however, was not associated with an increased incidence of mild thrombocytopenia (0.3% versus 0.1%; P=0.62). Severe thrombocytopenia was observed in 4 patients treated with phosphate‐buffered tirofiban (0.5%) versus none treated with heparin in the early phase, and in 1 patient treated with citrate‐buffered tirofiban (0.1%) versus none with heparin in the later phase. Thrombocytopenia was not associated with an increase in ischemic events. A lower baseline platelet count and diabetes were associated with an increased risk of developing thrombocytopenia. In the TARGET study, 0.5% of patients treated with tirofiban developed thrombocytopenia, but none developed severe thrombocytopenia.²⁶ Additional cases of variably severe thrombocytopenia in association with tirofiban treatment have been reported, including 1 patient who developed hypotension, rigors, and chills soon after drug administration and 1 who developed fatal pulmonary hemorrhage.^{33, 34, 35, 36, 37, 38, 39, 40, 41}

Paradoxically, platelet activation and thrombotic complications have been associated with tirofiban‐induced thrombocytopenia. Dunkley et al reported on 11 cases of profound thrombocytopenia (9 acute and 2 delayed in onset) observed in 871 treated patients (1.3%).^{42, 43} All of these patients had evidence of circulating antibodies to platelets in the presence of tirofiban and in all 5 cases tested, the antibodies bound to αIIbβ3 in the presence of tirofiban. In 6 patients the antibodies induced platelet activation and all of these patients had ischemic coronary artery complications; in contrast, none of the 5 patients who did not manifest platelet activation had an ischemic coronary complication.⁴²

In the 9217 patients evaluated for thrombocytopenia in the PURSUIT trial the overall frequencies of mild and severe thrombocytopenia were similar among patients treated with eptifibatide or placebo (4.9% versus 4.9% and 0.5% versus 0.4%, respectively).¹⁵ In contrast, 5 of the 7 patients who developed profound thrombocytopenia were treated with eptifibatide, resulting in a frequency of 0.1% for eptifibatide‐associated profound thrombocytopenia.¹⁹ Patients who developed thrombocytopenia, whether treated with placebo or eptifibatide, had an increased risk of both bleeding (75.8% versus 27.8%) and ischemic events (mortality 7.4% versus 3.3%; myocardial infarction 27.2% versus 12.2%; repeat ischemia leading to a cardiac procedure 22.4% versus 12.8%).¹⁵ Thrombocytopenic patients treated with eptifibatide (n=314) compared with placebo (n=319) had increased frequency of severe bleeding (9.9% versus 5.6%) but decreased mortality (6.7 versus 8.2%), myocardial infarction (23.9% versus 30.4%), and stroke (2.9% versus 3.1%).¹⁵

Thrombocytopenia after eptifibatide treatment has been reported in other studies. Masood et al reported on 57 cases of eptifibatide‐induced thrombocytopenia identified in the US Food and Drug Administration adverse reporting system.⁴⁴ The mean (±SD) nadir in 35 patients was 9000±19 000 platelets/μL. Bleeding symptoms were present in 57% of cases and included bleeding from the catheterization site (14%), petechial hemorrhages (14%), groin hematomas (7%), bleeding from infusion site (7%), gastrointestinal bleeding (3.6%), hemoptysis (3.6%), epistaxis (3.6%), and hematuria (3.6%). Thrombosis was diagnosed in 3 of 28 patients (11%) and included deep vein thrombosis, pulmonary embolism, and stent thrombosis. Because the total number of patients treated with eptifibatide was not known, the frequency could not be ascertained. A study of 412 consecutive patients who received eptifibatide during PCI reported 4 patients (1%) experiencing acute, profound thrombocytopenia.⁴⁵ In 3 cases the drop in platelet count was precipitous (≤3 hours). Three of the cases had prior exposure to abciximab and 1 had prior exposure to eptifibatide. In another study, 3 cases of eptifibatide‐induced acute, profound thrombocytopenia were reported, 2 of which were temporally associated with stent thrombosis; 1 patient treated a month after receiving eptifibatide a first time without incident had an anaphylactic reaction in addition to acute, profound thrombocytopenia and another patient also previously received eptifibatide.⁴⁶ They also provided data on 9 other patients reported to have eptifibatide‐associated acute, profound thrombocytopenia, 7 of whom were detected in the first 4 hours; 3 of the latter had previously been treated with abciximab and 2 had been treated with eptifibatide. Thrombosis in association with eptifibatide‐associated thrombocytopenia has also been reported by multiple investigators.^{47, 48, 49} In 1 case the patient's serum contained antibodies that could induce platelet aggregation and release of platelet granule ATP in the presence of eptifibatide but not in its absence.⁴⁸ The Fab domain of the antibodies bound to eptifibatide‐occupied αIIbβ3 whereas the Fc domain simultaneously bound to the platelet‐activating receptor, FcγRIIA, on the same or other platelets, a mechanism demonstrated to also be responsible for platelet activation by a murine monoclonal antibody to αIIbβ3.⁵⁰ In another case, the patient's serum contained antibodies that bound to platelets only in the presence of eptifibatide; in addition, the eptifibatide‐dependent antibodies were able to activate the platelets as shown by release of serotonin and increased expression of platelet P‐selectin.⁴⁹ It is possible that activating antibodies are also present in patients with thrombocytopenia without thrombosis since systematic assessment for this activity has not been reported and thrombosis likely stimulated the search for activating antibodies.

The development of drug‐dependent antibodies after GPI treatment is the presumed mechanism for both delayed onset thrombocytopenia and rapid onset of thrombocytopenia in patients previously treated with the same drug.^{39, 51} Acute profound thrombocytopenia in previously untreated patients, however, is presumed to result from the presence of preformed antibodies that react with the drug‐αIIbβ3 complex. Bednar et al. performed studies in 201 nonhuman primates and found that 1 of them (0.5%) developed acute thrombocytopenia in response to 2 of 3 experimental small molecule GPIs.⁵² Antibodies in the plasma of that animal bound to both chimpanzee and human platelets as well as to purified αIIbβ3 in the presence of the 2 drugs but not in their absence or in the presence of the third drug that did not induce thrombocytopenia. They then tested 1032 human plasma samples and found that 0.7% were positive for antibodies that could bind to human platelets only in the presence of 1 of the GPIs, whereas 0.4% were positive only in the presence of another GPI, and 0.4% were positive with both GPIs. This provided support for the role of preformed antibodies in producing acute thrombocytopenia with the approved GPIs. Billheimer et al performed similar studies in humans in response to the oral GPI roxifiban and found that 1% of 1000 blood donors had a high level of antibodies to platelets in the presence of the active metabolite of roxifiban.⁵³

Bougie et al studied 9 patients treated with tirofiban (n=4) or eptifibatide (n=5), 5 after first exposure and 4 after second or third exposure, who developed platelet counts between 1000 and 25 000/μL within 1 to 24 hours after treatment.⁴¹ Antibodies from all of the patients bound to platelets and purified αIIbβ3 in the presence of the administered drug but not its absence, including antibodies in serum obtained before the drug was administered to 2 patients. Eight of the antibodies failed to react with platelets in the presence of the other small molecule GPI, whereas antibodies from 1 of the eptifibatide‐treated patients reacted weakly with platelets in the presence of tirofiban. None of the 100 serum samples from untreated control individuals reacted with platelets in the presence of either drug and 21 of 23 sera from patients treated with the drugs who did not develop thrombocytopenia were also negative, with the remaining 2 being only weakly positive. The antibodies in the sera from the patients who developed thrombocytopenia were heterogeneous in both their binding sites and sensitivity to differences in divalent cations. Thus, these antibodies have many of the same properties as monoclonal antibodies that are specific for ligand‐induced binding sites.^{54, 55, 56, 57}

Bougie et al later studied 43 additional cases of acute thrombocytopenia (range 1000–102 000/μL; median 10 000/μL) associated with tirofiban or eptifibatide treatment. Of particular note, 47% of the patients had no bleeding symptoms.³⁹ Antibodies from all 38 patients treated with eptifibatide bound to platelets in the presence of eptifibatide, whereas 7 bound to platelets in the presence of the pseudo‐ligand peptide RGDW, which both blocks ligand binding and induces the receptor to undergo extension and swing‐out, and 5 bound to platelets in the presence of tirofiban. Antibodies from all 5 patients treated with tirofiban bound to platelets in the presence of tirofiban, but only 1 bound to platelets in the presence of eptifibatide and none bound in the presence of RGDW. They confirmed the ability of both tirofiban and eptifibatide to induce conformational changes in αIIbβ3 as judged by their ability to enhance the binding of nearly all of the 18 different ligand‐induced binding sites antibodies they studied. Thus, while some of the drug‐associated antibodies could bind αIIbβ3 in response to the conformational changes induced by pseudo‐ligand or drug binding, most showed greater specificity than the ligand‐induced binding sites antibodies for the inciting drug. They went on to perform studies in which monoclonal antibodies against known parts of αIIbβ3 were tested for their ability to inhibit the binding of the patients' antibodies. Strikingly, monoclonal antibody 7E3 and abciximab effectively blocked the binding of all 6 antibodies tested (3 associated with eptifibatide treatment and 3 with tirofiban treatment), whereas there was heterogeneity in the ability of other well‐characterized antibodies to αIIbβ3 to block the binding, indicating that all of the antibodies make a major contact near the β3 MIDAS domain³ but with varying additional sites of interaction. Thus, the drugs may contribute to the epitope by (1) binding to αIIbβ3 to create a shared epitope with the receptor independent of inducing the conformational change (Figure 1, model 1); (2) both binding to αIIbβ3 and inducing the conformational change to create a shared epitope with the receptor (Figure 1, model 2); and (3) binding to αIIbβ3 and inducing the conformational change without directly contributing to the epitope (Figure 1, model 3). Alternatively, the drug may enhance antibody binding to αIIbβ3 by binding to the antibody and enhancing its ability to bind to αIIbβ3 rather than binding to αIIbβ3.

Billheimer et al studied the antibodies of 2 patients who developed thrombocytopenia in response to the oral GPI roxifiban, which binds to αIIbβ3 in a manner expected to induce the conformational changes produced by eptifibatide and tirofiban.⁵³ They concluded that the drug resulted in antibody binding by inducing a conformational change in αIIbβ3 and that the drug itself was not part of the antibody epitope based on (1) the failure of increasing concentrations of free drug to inhibit the binding; (2) the ability of the drug to alter the electrophoretic mobility of αIIbβ3 in sodium dodecyl sulfate‐polyacrylamide gel electrophoresis; and (3) the ability of the antibody to immunoblot the αIIbβ3 band even though the drug was no longer bound. This proposed mechanism fits model 3 in Figure 1.

Delayed onset thrombocytopenia with the small molecules is not easily explained because in these cases the serum drug levels have declined to very low or unmeasurable levels. The likelihood and speed of developing thrombocytopenia probably reflect a complex combination of factors, including any residual platelet‐associated drug, the drug‐dependent antibody level and affinity (both of which may be changing rapidly during an immune response), the efficiency of the Fc region of the patient's antibodies in binding to Fc receptors on the cells that remove the platelets, and the number and competency of those cells. We would raise the theoretical possibility that although serum levels of the drugs may decline rapidly, platelet‐associated drug may persist because of the high affinity of the drugs for the receptor and the possibility that drug may redistribute to newly developed platelets or may bind to αIIbβ3 on megakaryocytes and thus continue to enter the circulation for periods beyond the serum level dropping to undetectable levels. Because estimates are that only a few hundred to a few thousand antibody molecules are required for platelet clearance,⁵⁸ and with ~100 000 αIIbβ3 receptors per platelet, only a very small fraction of the receptors potentially need to be occupied to support drug‐dependent antibody clearance. Consistent with this hypothesis, Brassard et al reported that a concentration of xemilofiban one‐twentieth of its peak plasma concentration after oral dosing was sufficient to recruit antibody to platelets.⁵⁹ Of course, the delay still needs an explanation related to the antibody's ability to induce clearance and so even under the scenario postulated one would need to additionally postulate a change in the antibody's ability to lead to clearance. This could theoretically be due to increased affinity or a change in Fc effectiveness because the in vitro assays may pick up even low affinity antibodies, as was found for the preformed hinge cleavage region‐specific antibodies identified in patients treated with abciximab that did not lead to thrombocytopenia or clearance of abciximab. Another possibility is that antibodies are originally produced against the drug‐receptor complex, but then recognize the receptor in the absence of the drug due to epitope spreading. In this case, however, the in vitro tests should show antibody binding in the absence of drug, and that has not been reported to our knowledge.

The analysis of antibodies to abciximab is more complicated than the analysis of antibodies to the small molecule GPIs because it is a chimeric Fab fragment of the parent 7E3 murine monoclonal.⁶⁰ Studies by Knight et al demonstrated that 6 out of 88 patients treated with abciximab developed antibodies that reacted with immobilized abciximab, but only 1 of the antibodies was inhibitable by the murine 7E3 Fab, indicating that it was the only one actually directed at the murine component of aciximab.⁶¹ The binding of 4 of the 6 antibodies was localized to the hinge cleavage site region of the human heavy chain in abciximab, a region to which many humans have low‐affinity antibodies.⁶² The presence of low levels of antibodies to the cleavage site before abciximab treatment did not affect the function of the drug in inhibiting αIIbβ3.

In the EPIC study, 5.8% of patients developed antibodies to the murine portion of abciximab.²² Additional data on development of antibodies to abciximab and thrombocytopenia were reported from the ReoPro Re‐administration Registry.^{29, 30} Antibodies to the murine component of abciximab were detected in 22 of 454 patients (4.8%) after the first administration and this increased by 19% (82/432) after the second administration. Of the 23/499 patients (4.6%) who developed platelet counts <100 000/μL upon readministration, only 2 of the 20 who were tested and had an interpretable result had antibodies to abciximab before receiving the repeat dose, and among those who developed platelet counts <20 000/μL, only 1 of 9 had detectable antibodies. Because the authors did not test for antibodies against the abciximab‐αIIbβ3 complex on platelets, it is possible that antibodies specific for the complex that could not bind to abciximab itself were responsible for the observed thrombocytopenia. In the 115 patients who were retreated within 1 month of their original treatment, mild and profound thrombocytopenia occurred in 16.5% and 12.2% of patients, respectively.³⁰ Having a preexisting antibody before administration increased the risk of mild and profound thrombocytopenia, but still only a small percentage of patients with such antibodies developed thrombocytopenia (14.1% versus 4.4% and 5.6% versus 1.6%, respectively). Seven of 14 patients (50%) who received a second course of abciximab despite having had thrombocytopenia with the first treatment developed recurrent thrombocytopenia, indicating the high risk of thrombocytopenia in this group. Thrombocytopenia was associated with an increase in bleeding (25.4% versus 9.6%) and clinical adverse events (18.4% in those with severe thrombocytopenia, 11.9% with any thrombocytopenia, and 4.8% in those without thrombocytopenia).

Curtis et al studied 9 patients who developed profound thrombocytopenia after a second dose of abciximab.³¹ In contrast to the previous studies, they tested patients' sera for antibodies against platelets in the presence of abciximab rather than antibodies to abciximab itself. They found that all 9 patients' sera contained strong IgG antibodies to the abciximab‐treated platelets, and although 74% of sera from healthy donors also had antibodies that bound to abciximab‐treated platelets, the titers were lower and all of the control sera's reactivity was greatly diminished by incubation with Fab fragments of nonspecific IgG, suggesting that the reactions were with the cleavage site regions. In contrast, titers from only 2 of the 9 patients who became thrombocytopenic were greatly diminished by the Fab fragments.

In summary, acute thrombocytopenia associated with first administration of abciximab is primarily due to the presence of preformed antibodies that recognize abciximab or the abciximab‐αIIbβ3 complex on platelets. In contrast, delayed thrombocytopenia in association with the first dose of abciximab and thrombocytopenia associated with readministration of abciximab is due to antibodies generated to abciximab or the abciximab‐αIIbβ3 complex. Neither baseline anti‐abciximab antibody levels or the ones induced by the readministration of abciximab correlate well with the development of thrombocytopenia, suggesting that most of the antibodies causing thrombocytopenia are specific for the abciximab‐αIIbβ3 complex on platelets rather than abciximab itself.

Further support for an immune mechanism for GPI‐associated thrombocytopenia derives from studies of the oral GPI roxifiban. Among 386 treated patients, 8 developed thrombocytopenia (2.1%), with 3 mild, 2 severe, and 3 profound.⁶³ Pretreatment samples were available in 2 patients with early onset thrombocytopenia, and 1 of them had preformed antibodies to platelets in the presence of the drug. All 4 patients with delayed thrombocytopenia had drug‐dependent antibodies at the time of the thrombocytopenia, and 3 of 4 (75%) had preformed antibodies before therapy.⁶³ Based on these data, the investigators evaluated 1332 patients prospectively and excluded the 57 patients (4.3%) who tested positive for drug‐dependent antibodies from receiving roxifiban treatment. They then treated those who did not have preformed antibodies with roxifiban and 2.7% became antibody positive by day 14. Even after excluding all of these patients, 1 patient who entered the study developed profound thrombocytopenia in association with the development of drug‐dependent antibodies and 7 additional patients were removed from the study because they developed drug‐dependent antibodies even though they did not have thrombocytopenia. Thus, although prescreening for drug‐dependent antibodies reduced the risk of thrombocytopenia, it did not eliminate the risk. Moreover, it is not feasible to test patients for the presence of antibodies who need an emergency PCI.

The diagnosis of GPI‐associated thrombocytopenia usually begins with the report of a low platelet count from of an automated cell counter, but sometimes it begins with the onset of hemorrhagic symptoms resulting in the ordering of an unscheduled platelet count. Because thrombocytopenia may occur rapidly and be asymptomatic, a platelet count should be performed 2 to 4 hours after the bolus dose of the GPI and 24 hours later. Low platelet counts should be confirmed with inspection of the blood smear. The presence of red blood cell schistocytes would suggest the possibility of thrombotic thrombocytopenic purpura or disseminated intravascular coagulation, whereas platelet clumps would indicate that the automated count is likely an underestimate, most likely due to platelet clumping induced by the standard blood cell count anticoagulant EDTA. Because the platelet aggregates may be mistaken for leukocytes by the automated counter, a reciprocal increase in white blood cells may be observed. This anticoagulant‐dependent platelet clumping phenomenon is termed pseudothrombocytopenia and the presumptive diagnosis is usually confirmed by finding a higher platelet count and loss of the platelet clumps upon repeating the platelet count with blood anticoagulated with citrate.^{64, 65} Platelet clumping may also be present in citrated blood, and this may be more common with GPI‐associated pseudothrombocytopenia, and under these conditions, heparin or magnesium sulfate may be better anticoagulants.^{65, 66, 67}

EDTA is known to affect the conformation of αIIbβ3 and this is reflected in its effects on the binding of monoclonal antibodies specific for different conformations of αIIbβ3.^{57, 68} The binding site for fibrinogen contains a Mg²⁺ and 2 Ca²⁺ ions⁶⁹ and EDTA chelates both of these divalent cations. In addition, αIIb connects to β3 via a Ca²⁺‐dependent mechanism and so under some conditions, EDTA dissociates the subunits.⁷⁰ Platelet clumping in most patients with pseudothrombocytopenia is due to antibodies in patients' plasma, mostly IgM, but also IgG and rarely IgA, that appear to react primarily with αIIbβ3, in particular, the dissociated αIIb subunit.^{65, 71} αIIbβ3 is also the target for many autoantibodies in patients with immune thrombocytopenia who have true thrombocytopenia.⁷²

A pooled analysis of 4 randomized studies with abciximab (n=8555), demonstrated pseudothrombocytopenia in 2.1% of abciximab‐treated patients and 0.6% of placebo‐treated patients (P<0.001), suggesting that abciximab increases the likelihood of developing pseudothrombocytopenia.^{64, 73} For comparison, true thrombocytopenia occurred in 3.7% of abciximab‐treated patients and 1.8% of placebo‐treated patients (P<0.001). Thus, a sizable fraction of thrombocytopenia with abciximab is due to pseudothrombocytopenia. In contrast, pseudothrombocytopenia has not been specifically linked to treatment with tirofiban or eptifibatide.⁷⁴ Pseudothrombocytopenia does not confer an increased risk of hemorrhagic or ischemic complications.⁶⁴ It does, however, pose risks if it is not recognized because it may lead to inappropriate discontinuation of anticoagulant or antiplatelet therapy and inappropriate administration of platelet transfusions, all of which may enhance the likelihood of ischemic complications.

Because virtually all patients receiving a GPI are also receiving heparin, it is also important to exclude the diagnosis of heparin‐induced thrombocytopenia.^{75, 76, 77} Transient, mild thrombocytopenia due to a direct effect of heparin on platelets^{20, 75} is common with heparin therapy and so may contribute to mild thrombocytopenia in patients treated with GPIs. This phenomenon may partially explain why low baseline platelet counts are associated with the development of GPI‐associated thrombocytopenia. Because the thrombocytopenia in these cases is not due to an immune mechanism,⁷⁵ the screening tests for heparin‐induced thrombocytopenia based on the presence of antibodies to the PF‐4 (heparin‐platelet factor 4) complex will be negative,⁷⁸ and thus the thrombocytopenia may be inappropriately ascribed to the GPI.

More severe heparin‐induced thrombocytopenia is commonly associated with platelet activation and a high risk of thrombosis,^{76, 78} and so it is important to diagnose it as rapidly as possible. In patients who have never been previously exposed to heparin, it usually takes 4 to 5 days to develop heparin‐PF4 complex antibodies and so acute thrombocytopenia in such patients is unlikely to be due to heparin‐induced thrombocytopenia. Patients who have been previously exposed to heparin, especially in the recent past, however, may experience very rapid onset of thrombocytopenia,^{76, 78} making it more difficult to differentiate from acute severe thrombocytopenia produced by a GPI. Even in patients who do have rapid onset heparin‐induced thrombocytopenia, however, it is uncommon for the platelet count to drop below 20 000/μL, which may help in the differential diagnosis. Current guidelines recommend only performing 1 of the screening tests when the patient has a high probability of having heparin‐induced thrombocytopenia based on the 4Ts (Thrombocytopenia, Timing of thrombocytopenia, Thrombosis, and oTher reasons) score.^{76, 79} These tests are relatively sensitive but are less specific than functional assays, such as the serotonin release assay, but the latter require specialized equipment.^{77, 78} A new assay using cryopreserved platelets and measuring release of thrombospondin 1 from platelets in the presence of PF4 appears promising as a more specific assay that can be performed without the need for specialized equipment.⁸⁰

There are no controlled studies comparing different therapeutic approaches to GPI‐associated thrombocytopenia and so recommendations regarding therapy are tentative. Remarkably, as noted previously, many patients with profound thrombocytopenia have recovered without having clinically significant bleeding symptoms, whereas others have suffered severe bleeding despite aggressive therapy and so there is a wide range of variability, suggesting that other factors play an important role in both the clinical manifestations and the response to therapy.^{19, 31, 42, 46, 81, 82} Paradoxically, as also noted previously, GPI‐associated thrombocytopenia has also been associated with an increased risk of thrombotic complications and Table 2 contains a list of potential contributing factors. In addition, technical aspects related to the PCI, including the extent of disease; the size of the target coronary artery; and the size, number, and length of the stents may all contribute to the variable risks of ischemic complications. The possibility that thrombotic complications are due to platelet activation by GPI‐associated antibodies is mechanistically plausible, but unfortunately, there are no commercially available assays that can be performed rapidly enough to guide initial therapy. Perhaps the most important technical association with the risk of bleeding is the procedural access site, with radial artery access having a much lower risk of access site hemorrhage.^{5, 6}

Masood et al reported data on therapy in 37 of the 57 eptifibatide‐associated reports from the US Food and Drug Administration adverse reporting system.⁴⁴ Eptifibatide was withheld in all cases, platelet transfusion were given in 54%, steroids in 5.4%, and IVIgG in 1 patient (2.7%), with steroids reserved for patients who did not respond to transfusions, and IVIgG given when a patient had a contraindication to steroids. As noted previously, 43% of the patients did not have bleeding symptoms and the others had a wide range of symptoms, primarily access site bleeding and petechiae. Kereiakes et al reported on 4 patients out of 575 (0.7%) who developed acute profound thrombocytopenia after abciximab therapy; 3 had groin hematomas and 1 a forearm hematoma. One patient was treated with IVIgG alone and had a slow recovery, 2 were treated with platelet transfusion (8–16 units) alone and had a brisk response, and 1 was treated with both and had a brisk response.⁸³

With the limited data available, Figure 3 provides a graded approach to therapy based upon whether the patient has hemorrhagic symptoms and the severity of the thrombocytopenia. It is not clear whether there is a reason to modify treatment based on whether the onset of the thrombocytopenia is relatively soon after treatment or delayed. In asymptomatic patients with mild thrombocytopenia, careful observation with at least daily platelet counts may be all that is required because hemorrhagic and ischemic complications appear to be rare in such patients, with spontaneous recovery occurring within days. If thrombocytopenia is detected while the GPI infusion is still ongoing, one could consider stopping the infusion, but this may compromise some of the antiplatelet benefit.

For asymptomatic severe thrombocytopenia, the risk of further progression to profound thrombocytopenia would seem to justify at least stopping the GPI and performing platelet counts every 12 to 24 hours until the thrombocytopenia resolves. Asymptomatic profound thrombocytopenia justifies both stopping the GPI and probably discontinuing the anticoagulant if one is still being administered or other antiplatelet agents (aspirin, P2Y12 inhibitor). If the bleeding risk is especially high, consideration should be given to platelet transfusion. As noted previously, bleeding risk should take into consideration the vascular access site because hemorrhagic complications are more common with femoral than radial artery access.^{5, 6} The patient's prior history of bleeding and current risk factors for PCI‐associated bleeding may also be relevant, and these can be judged with a standardized bleeding assessment tool, such as the one developed by the International Society on Thrombosis and Haemostasis, and analysis of patient risk factors as per the criteria established by the Academic Research Consortium, respectively.^{84, 85, 86}

The GPI should be discontinued in all symptomatic patients to prevent further exacerbation of the bleeding. For patients with mild thrombocytopenia one needs to weigh the potential benefit of also discontinuing anticoagulant and antiplatelet agents versus the increased risk of thrombotic complications if these drugs are discontinued. With severe thrombocytopenia in a symptomatic patient, the hemorrhagic risk increases and so in most cases it would be prudent to stop the anticoagulant and antiplatelet agents. Depending on the severity of symptoms and the temporal trend in the platelet count, it may also be appropriate to treat the low platelet count directly with steroids, IVIgG, or platelet transfusions.^{33, 40, 42, 83, 87} Each has a rationale based on treatment of other forms of immune thrombocytopenia,^{72, 76} but none has been demonstrated to provide a clear benefit in GPI‐associated thrombocytopenia.^{16, 72} In the case of symptomatic profound thrombocytopenia, in addition to stopping the GPI, anticoagulant, and other antiplatelet medications, platelet transfusions and steroids should probably be administered to prevent further symptoms, with additional consideration of administering IVIgG.

In patients who previously developed GPI‐related thrombocytopenia, the safety of switching to another GPI is not established, but there is a case report of a patient who developed abciximab‐associated thrombocytopenia who did not develop thrombocytopenia when subsequently treated with eptifibatide.⁸⁸ In vitro studies of antibody recruitment to platelets, however, have shown that in a small percentage of patients with thrombocytopenia in association with one of the small molecule GPIs, the other small molecule GPI could also increase antibody binding to platelets,³⁹ and as noted previously, some patients who developed thrombocytopenia to a GPI had a history of prior treatment with another GPI. As a result, retreatment with any GPI may confer an increased risk of repeat thrombocytopenia.