Heparin-induced thrombocytopenia: a historical perspective

John G. Kelton and Theodore E. Warkentin

Heparin was discovered 90 years ago,1 and within 2 decades it was being widely used as an anticoagulant.2 The numerous advantages of heparin lend to its broad use including its immediate onset of action, relatively short half-life (∼60 minutes), simple laboratory monitoring (aPTT), ability to be reversed (using protamine), and low cost. Despite a long experience of using heparin, the counterintuitive notion that an anticoagulant could be intensely prothrombotic took almost 4 decades to become recognized.

Our understanding of heparin-induced thrombocytopenia (HIT) has increased considerably and to a large measure exemplifies the best of translational research. Sometimes, basic studies investigating the pathophysiology of HIT lead to clinical insights. In other instances, clinical observations prompted basic studies. Together, these basic and applied investigations have dramatically enhanced our understanding of the pathogenesis and treatment of HIT.

Heparin-induced arterial embolism

On June 1, 1957, at the Fifth Scientific Meeting of the International Society of Angiology in New York, Rodger E. Weismann, an Assistant Professor of Clinical Surgery at the Dartmouth Medical School, and his Resident in Surgery, Dr Richard W. Tobin, described 10 patients who developed arterial embolism during systemic heparin therapy.3 The first reported embolic event was a femoral artery embolic occlusion, which occurred in a 62-year-old woman who was receiving heparin because of a deep-vein thrombosis (DVT). Three days after successful femoral embolectomy, and while continuing to receive heparin, she developed sudden occlusion of the distal aorta requiring distal aortic and bilateral iliac embolectomies.3

Multiple thromboemboli were observed in 9 of the 10 patients reported. Six patients died as a result of the thromboembolism, and 2 survivors underwent above-knee amputation.3 The clots were described as “pale, soft, salmon-colored” and were composed mostly of fibrin, platelets, and leukocytes. The arterial emboli began on average 10 days after commencing heparin treatment.

An additional 11 patients were described 5 years later by Roberts and colleagues.4 These vascular surgeons noted the paradox of “unexplained arterial embolization [occurring] for the first time while being treated with heparin for some condition that could not of itself reasonably be expected to cause arterial emboli. All patients had been receiving heparin for 10 days or more when the initial embolus occurred.” Thus, both groups3,4 noted the temporal hallmark of HIT, a delay of approximately 1 week from initiation of heparin to onset of its thrombotic manifestations.

Heparin-induced thrombocytopenia

Routine platelet count measurements were not routinely performed until the 1970s. This may explain why thrombocytopenia was not reported in the first 24 patients with heparin-induced arterial emboli.36 In 1969, the term “heparin-induced thrombocytopenia” was used by Natelson7 to describe a 78-year-old man with pulmonary embolism who developed severe thrombocytopenia after heparin. Over the ensuing days, 2 separate periods of heparin readministration were each characterized by abrupt platelet count declines and corresponding increases in fibrinogen. The addition of heparin to the patient's citrated platelet-rich plasma caused platelet aggregation. The striking dichotomy of a heparin-induced platelet count decrease, but rise in the level of fibrinogen, captures the essence of HIT: heparin simultaneously promotes and treats the thrombosis.

The syndrome of heparin-induced thrombocytopenia and thrombosis

The first to identify the central features of the HIT syndrome—thrombocytopenia, thrombosis, and its immune pathogenesis—were Drs Silver, Rhodes, and Dixon.8 In their 1973 paper, they described 2 patients with severe thrombocytopenia, myocardial infarction, and heparin resistance, with platelet count recovery on discontinuing heparin treatment.8 Both patients developed rapid recurrence of thrombocytopenia when heparin rechallenges were given.

An immune basis for this syndrome was suggested by increased numbers of bone marrow megakaryocytes and a rapid recurrence of the thrombocytopenia upon heparin reexposure. A circulating heparin-dependent, platelet-activating substance (subsequently identified as IgG) was found in the patients' blood, which caused aggregation of donor platelets in the presence of heparin.

A subsequent report by Rhodes and coinvestigators9 helped establish HIT as a distinct syndrome. Eight patients were described with thrombocytopenia (platelet count nadir, 25 × 109/L) that occurred during heparin administration. Thrombotic, rather than hemorrhagic, complications predominated: 7 patients had new or recurrent thromboembolic events, and one patient had a hemorrhagic stroke. Complement-fixing, heparin-dependent antibodies were found in blood from 5 patients.

Platelet activation in HIT

During the latter half of the 1970s, at least 7 groups of investigators1016 reported patients who resembled those of Rhodes and colleagues. Together, these studies provided further evidence of an immunologic basis for HIT. A common feature of these reports was evidence of platelet activation induced by patient serum, plasma, or purified immunoglobulin.1117

HIT: an immune disorder or a consumptive coagulopathy?

Doubts persisted regarding the immune basis of HIT. After all, Rhodes and colleagues had also noted that when their patients were deliberately rechallenged with heparin several months after their episode of HIT, thrombocytopenia failed to recur.8,9 Moreover, HIT patients typically did not exhibit severe thrombocytopenia and purpura—features characteristic of established drug-induced immune-mediated thrombocytopenia reported with quinine, among other drugs. Also arguing against the model of typical immune-mediated drug-induced thrombocytopenia were our observations that thrombocytopenia could resolve while the patient continued on heparin therapy.18

But other evidence supported an immune pathogenesis: for example, Cines and associates19 documented heparin-dependent binding of complement to platelets. Another observation that suggested a complex antigen was provided by Green and coworkers.12 These investigators demonstrated that the antibodies did not bind to heparin per se, but required heparin plus a cofactor found in platelet lysates.

The uncertainty regarding etiopathogenesis was also reflected in the name of the condition: HIT was often termed heparin-associated thrombocytopenia to reflect the assumption that the thrombocytopenia might not be directly related to the heparin.20 Some investigators2123 raised the possibility that a procoagulant contaminant within the heparin preparation could be causing the reaction, perhaps by triggering disseminated intravascular coagulation. Further support for this hypothesis was provided by the demonstration that different heparin preparations carried different risks for thrombocytopenia.18,24 In addition, consistent with a “contaminant” possibility was the fact that heparin is unique among almost all medications: it consists of polydispersed and heterogeneous polysaccharide chains that have been extracted and partially purified from either beef lung or pork intestinal mucosa.

Heparin itself can directly interact with platelets and cause variable platelet aggregation when added to platelets in vitro. In addition, some investigators have reported that when heparin is injected into patients, there can be an early, transient drop in platelet count.25,26 At a platelet immunobiology workshop held in Milwaukee, it was proposed that the early, nonimmune disorder would be termed “HIT, Type 1” and the later onset immune disorder would be called “HIT, Type 2.27 Today these designations are not commonly used and most physicians designate the immune disorder simply as heparin-induced thrombocytopenia (HIT).

Our research group became interested in developing a laboratory test to diagnose HIT. Although platelet-associated IgG levels were elevated in HIT patients, we found that it was not a diagnostically useful assay. Subsequent investigations have documented that the various assays for PAIgG have minimal diagnostic usefulness in any thrombocytopenia.19,28,29 We next prospectively evaluated the widely used assay of heparin-dependent platelet aggregation.30 The results indicated that heparin-dependent platelet aggregation was only a moderately useful test for HIT and although it had a high specificity, its sensitivity was low. We also made an observation that subsequently proved informative, namely that platelet aggregation was maximal at pharmacologic concentrations of heparin (0.1 U/mL), but that higher heparin concentrations (above 5 U/mL final) often produced a negative result.

Our next aim was to improve the diagnostic usefulness of the functional (platelet activation) assay for HIT.31 The sensitivity was increased by optimizing the reactivity of the test platelets and by using the release of radiolabeled 14C-serotonin as the end point. The specificity was increased by observing that the platelet-activating effects of HIT antibodies were inhibited at high heparin concentrations (Figure 1). The serotonin release assay (SRA) was prospectively evaluated against a group of 600 patient and control samples, and proved both sensitive and specific. Subsequent modifications have further enhanced the assay's diagnostic utility.32

Figure 1

The relationship between the percentage release of C-serotonin from platelets obtained from healthy individuals (ordinate) and the final concentration of heparin (abscissa). Each series of points is the result obtained with serum from a different patient with definite heparin-induced thrombocytopenia, The lines numbered 1 to 4 represent 4 different serum samples not from patients with heparin-induced thrombocytopenia. These samples induced release at 0.1 and 100 U/mL of heparin. Numbers 1 and 2 are sera from 2 patients with immune thrombocytopenic purpura. Number 3 is the serum from a hospitalized control, and number 4 is heat-aggregated IgG. Adapted from Sheridan et al31 with permission.

A sensitive and specific test for a disorder can also be used to characterize the component reactions. For example, we were able to confirm the work of previous investigators that the reaction was mediated by IgG. Furthermore, the unimodal pattern of antigen-dependent progressive platelet activation followed by reduced platelet activation (at higher heparin concentrations) suggested to us that this was an immune complex type of reaction.31

In an attempt to identify the platelet target for the HIT antibodies, we tested platelets with congenital deficiencies of glycoprotein IIb/III (Glanzmann thrombasthenia) and platelets deficient in glycoprotein Ib/IX/V complex (Bernard-Soulier syndrome).33 The identical reaction to normal platelets indicated that these glycoproteins did not directly participate in the reaction. Next, IgG from patients with HIT was purified and digested. The requirement of F(ab′)2 indicated that HIT was an immunologic disorder recognizing a specific, but as yet unidentified, epitope. Inhibition by human or animal Fc indicated that HIT was an immune complex disorder. Finally, the complete inhibition of platelet reactivity via a monoclonal antibody against the platelet Fc receptor confirmed that platelet activation in HIT was mediated through the platelet Fc receptor.33

We now know that HIT-mediated platelet activation is a dynamic process34 in which the Fab from the HIT-IgG binds to platelet-associated PF4. Subsequently the Fc region on the IgG molecule binds to the Fc receptor on the same35 or adjacent platelets, which in turn triggers platelet activation.33,36

The platelet Fc receptor was identified by Karas et al37 with further characterization by Rosenfeld et al38 as well as by our group.39 Platelets carry the FcγIIa receptor, which is a low affinity receptor present at relatively low copy numbers (∼1000 to 2000 copies per platelet).3739 IgG immune complexes, which form in patients with HIT, produce receptor cross-linking, which strongly activates platelets. A common polymorphism of FcγIIa (arg131/his131) have proven to be of minor relevance in HIT.

Microparticles: the platelet procoagulant response

It is rare for a prothrombotic disorder to cause both arterial and venous thrombosis. HIT is an exception. In many HIT patients there was evidence of hypercoagulability, but the pathway that could initiate both arterial as well as venous thrombosis remained unexplained. Three explanations were proposed: (1) HIT antibodies could bind to and injure endothelial cells, thereby initiating coagulation40,42; (2) HIT antibodies could bind to monocytes and release tissue factor43,44; and (3) HIT antibodies induce a platelet procoagulant response.

Our group focused on the platelet procoagulant explanation. Using flow cytometry, we found that HIT sera caused platelets to form microparticles (Figure 2).45 In contrast, quinine/quinidine-induced thrombocytopenia sera did not cause microparticles, even though there was a dramatic increase in drug-dependent antibody binding to platelets. The procoagulant nature of the HIT microparticles was evidenced by shortening of the Russell viper venom time.45 We also showed that in at least some patients, there were detectable circulating platelet-derived microparticles during the acute episode of HIT.45 In subsequent comparative studies, we found that the magnitude of this platelet procoagulant response induced by HIT antibodies exceeded that of other physiologic agonists, including thrombin and collagen.46 Microparticle formation induced by HIT sera was sufficiently reliable to be used as a diagnostic test for HIT.47

Figure 2

Platelet-derived microparticles generated by heparin-induced thrombocytopenia serum. Platelets and microparticles were identified using fluorescence (FL1, FITC-antiGPIb3) and size (FSC, forward scatter) characteristics. The platelets had been incubated with heparin-induced thrombocytopenia serum in the presence of 0, 0.1, 0.3, or 100 U/mL heparin. The microparticles (inset box) were generated in the presence of 0.1 and 0.3 U/mL heparin, but not at 0 or 100 U/mL heparin. Results of the representative experiment are shown. Adapted from Warkentin et al45 with permission.

To better understand these microparticles, we examined their morphology using a variety of microscopic techniques (Figure 3).48 Using confocal microscopy, HIT antibodies plus heparin could be shown to produce platelet particles. Scanning and transmission electron microscopy showed that these microparticles ranged in size from 0.1 to 1.0 μm in diameter. These studies demonstrated that platelet activation resulted in the formation of localized points of swelling on the platelet body with the formation of well-defined buds. These platelet buds are released from the platelets to form the microparticles.48

Figure 3

Representative scanning electron micrographs of resting platelets and platelets activated with HIT serum in the presence of heparin. Resting platelets or platelets incubated with HIT serum in the presence of 0.1 U/mL heparin were fixed with 2% glutaraldehyde and processed for scanning electrom microscopy (A) The morphology of the representative normal resting platelet. Resting platelets were observed to generally maintain a discoid form. (B) The morphology of a representative platelet incubated with HIT serum. These platelets demonstrated several morphologic changes including absence of a discoid form, presence of pseudopodia and presence of microparticles near the end of the pseudopodia (indicated by arrows). Reprinted from Hughes et al48 with permission.

PF4/heparin complexes are the major antigen of HIT

In 1992, in a significant advance for HIT-related research, Amiral et al49 identified the elusive platelet component of the HIT antigen when he demonstrated that platelet factor 4 (PF4), a tetrameric member of the C-X-C subfamily of chemokines, formed complexes with heparin that bound HIT antibodies. This seminal observation was confirmed and extended by other groups.42,50,51 The antigen was shown to be present on multimolecular complexes of PF4 and heparin that required an optimal stoichiometric ratio of PF4:heparin of 1:1 to 2:1. In addition, it was shown that endothelial cells could bind PF4 and also produce the antigen. The identification of PF4 as the target antigen allowed the development of enzyme-immunoassay (EIA) techniques.52

The amino acid composition of PF4 had already been identified in 1987 by Poncz and colleagues.53 The small size of the PF4 monomer (70 amino acids) allowed Horsewood to synthesize a series of overlapping peptides that spanned the entire length of the PF4 molecule.54 We found that a minimal length of PF4 (19 amino acids encompassing the carboxy-terminal peptide including the lysine-rich moiety that binds heparin) was required for reactivity with certain HIT antibodies.54 But, it was not possible to identify a linear epitope on PF4 that served as a target antigen. These results suggested that heparin molecules bundle the PF4, resulting in conformational changes to the molecule, which in turn, become the binding sites for the HIT-IgG.54 Studies by the groups in Milwaukee55 and Philadelphia56,57 have identified several regions on PF4 that are binding sites for HIT antibodies. More recent studies suggest that the antigen sites are located at apposition points where PF4 tetramers are brought into proximity through charge neutralization by heparin.58 Figure 4 presents a schematic of the pathogenesis of HIT.

Figure 4

A schematic representation of the interaction HIT-IgG with platelets. In panel A platelet factor 4 (PF4) is released from platelet α-granules after platelet activation. The PF4 binds to the platelet surface through glycosaminoglycans (GAGs) that are associated with tetramers of PF4. Panel B shows heparin binding to the PF4. The size of the PF4 bundle in part is related to the chain length of the heparin. Panel C shows HIT-IgG binding to epitopes on the PF4/heparin complex. Panel D shows ultra-large complexes (ULC) of heparin/PF4 and IgG binding to platelets via the platelet FcRγIIa. The immune complexes are potent platelet activators leading to the release of platelet-derived procoagulant microparticles.

Amiral et al59 have also documented several other potential heparin-dependent antigens, including interleukin-8 and neutrophil-activating peptide-2, but antibodies against these chemokines probably explain only a tiny minority of cases of HIT.60

Characterization of the immune response in HIT

The demonstration that HIT was caused by PF4/heparin-containing immune complexes that activated platelets through their Fcγ receptors suggests a central role of the IgG immunoglobulin class in the pathophysiology of HIT. The HIT-IgG is polyclonal and IgG1 predominates with smaller amounts of IgG2.61 Although PF4/heparin-reactive IgM and IgA antibodies are also formed in heparin-treated patients, their role in HIT remains uncertain.60,62,63

Certain aspects about the immunobiology of HIT remain unexplained. The high frequency of heparin-induced autoantibodies against the self protein PF4 is perplexing. In addition, the reason for the remarkably rapid production of HIT-IgG antibodies, occurring as early as 5 days after commencing heparin, is uncertain. As well, the brief persistence of this antibody in patients with HIT (Figure 5)64 also remains unexplained, as does the lack of an anamnestic immune response upon reexposure.64,65

Figure 5

Kaplan-Meier analysis of the proportion of patients with heparin-dependent antibodies after an episode of heparin-induced thrombocytopenia. The time (in days) to a negative test by the activation assay (144 patients) or the antigen assay (93 of the 144 patients) is shown. All 144 patients initially had positive tests for heparin-dependent antibodies and underwent subsequent testing within 180 days. The date the test result became negative was estimated to be the date midway between the date of the last positive test and the date of the first negative test. The antigen assay tended to become negative more slowly than did the activation assay (P = .007 by the sign test). The bars indicate 95% confidence intervals. Reprinted from Warkentin and Kelton64 with permission.

Structural requirements of heparin to form the antigen

The chemical and structural determinants allowing heparin to induce antigenic changes in the PF4 molecule have been studied by a number of investigators.50,66,67 The impetus was both scientific and commercial: one goal was the identification of a nonheparin polyanion that could be useful for in vitro testing. Visentin and coworkers identified such a substance (polyvinyl sulfonate) that is currently used in a commercial EIA.67 Another focus of research has been to identify a heparin species that would neither initiate nor propagate HIT. In vitro studies indicated that a variety of sulfated polysaccharides could substitute for heparin to inducing the antigenic changes in PF4. However, there were 2 key determinants that were required for antigenicity. First, a certain chain length (ie, approximately 1000 Da); and second, a minimal amount of sulfation per saccharide unit were required.50,66 These results suggested that progressively smaller heparin preparations carry a progressively lower risk of HIT. Clinical validation of this potential has been provided by the demonstration that low-molecular-weight heparin (enoxaparin) carries a lower risk of HIT of inducing antibodies and of causing clinical HIT, compared with UFH.68

Recently, there has been interest in whether fondaparinux, a pentasaccharide modeled after the antithrombin-binding region of heparin, will be even less likely to cause or propagate HIT. Interestingly, despite a negligible potential to create epitopes recognized by HIT antibodies in vitro, studies suggest that anti-PF4/heparin antibodies identical to those formed by unfractionated heparin can be generated during treatment with fondaparinux.69 However, these antibodies do not usually bind to PF4 in the presence of fondaparinux,69 suggesting a very low potential to cause or to potentiate HIT. To date, only 2 patient cases of HIT possibly caused by fondaparinux have been reported.70,71 It is possible that in the future, fondaparinux will be used as a treatment for HIT.

The relationship between HIT antibodies and clinical HIT—the “iceberg model”

Systematic serosurveillance studies of heparin use have shown that only a small proportion, at most 5% to 30% of patients who form HIT-IgG, will develop clinical HIT, depending upon the patient population in question.52,63,72 Among antibody-positive patients, only approximately half have detectable antibodies of IgG class; of these, approximately half again show biologic activity.63 Even among patients with platelet activating antibodies, the risk of HIT increases with the greater magnitude of HIT-IgG–induced platelet activation. However, even some patients with strong platelet-activating antibodies do not develop HIT, suggesting that patient-dependent risk factors, such as platelet Fcγ receptor numbers73 or increased platelet-associated PF4 levels7476 may also play a role. The relationship among the clinical presentation of HIT, the presence of HIT antibodies, and the type of heparin used can be conceptualized as an iceberg (Figure 6).77

Figure 6

The relationship among the clinical expression of HIT (thrombocytopenia with or without thrombosis), the type of heparin used, and the antibodies that cause HIT. This can be conceptualized as an “iceberg.” The visible component of the iceberg (the portion above the waterline) represents clinically evident features of HIT, such as thrombocytopenia and/or thrombosis. The mass of the iceberg corresponds to the entire spectrum of anti-PF4/heparin antibodies generated. Some of these antibodies will be biologically active (platelet-activating), and others will be nonplatelet-activating which may make them unlikely to cause clinical consequences. The type of heparin given to the patient determines the overall size of the iceberg, with unfractionated heparin (UFH) being the largest (most immunogenic), and low-molecular-weight heparin and fondaparinux have lesser immunogenicity. Also illustrated in this figure is the vivo cross-reactivity. UFH forms antigens which are readily recognized by HIT antibodies. In contrast, fondaparinux forms poorly recognized antigens. LMWH is intermediate. Illustration by A.Y. Chen.

Whether the presence of anti-PF4/heparin antibodies confers adverse prognosis in certain clinical situations—even in the absence of clinical HIT—is an emerging topic of debate.7881

Animal models of HIT

The development of an animal (murine) model of HIT has been difficult. The problem reflects the lack of FcγIIA receptors on murine platelets. Furthermore, murine platelets do not express a PF4 immunologically similar to human PF4. Nonetheless progress is being made in this area. Using a double-transgenic mouse expressing both FcγRIIA and human PF4, Reilly and coinvestigators82 injected the mice with KKO,83 a mouse monoclonal antibody raised against human PF4/heparin. The mice were then injected with heparin, developing thrombocytopenia, including some with thrombosis. This model was used to demonstrate the importance of platelet-bound PF4 in the development of thrombocytopenia. The investigators further evolved the murine model using the same double transgenic mice, but with a murine PF4 knockout. This model demonstrated the importance of host factors such as hypercholesterolemia in thrombocytopenia and thrombosis.84 A murine model was used to investigate the immunogenic potential of various ratios of PF4 (murine) to unfractionated heparin.85 These investigators determined that high ratios of murine PF4 to heparin were more immunogenic than the more equimolor ratio that serves as the optimum target for antibody binding.

Evolving concepts in HIT

The thrombocytopenia of HIT is typically moderate (median platelet count nadir, ∼60 × 109/L), which is different from classic drug-induced immune thrombocytopenic purpura (median nadir, 10 × 109/L).86 Some patients with HIT will have a drop in the platelet count that does not necessarily reach the conventional threshold that defines thrombocytopenia (platelet count < 150 × 109/L).87 This can pose diagnostic challenges, most often in the setting of postoperative thromboprophylaxis, where elevated platelet counts can be observed.

HIT was first described as an arterial prothrombotic disorder, perhaps because of its dramatic onset and its recognition by vascular surgeons.36 For a number of years physicians did not recognize venous thrombi as part of the syndrome. The problem with attributing venous thromboembolism (VTE) to the HIT syndrome was that patients requiring heparin were sometimes at the highest risk for developing VTE. Consequently, any association could be collateral rather than causal. Clarification of the importance of VTE as a manifestation of HIT was provided by Boshkov et al88 This retrospective study related the site of thrombosis to potential risk factors among patients with serologically confirmed HIT. These researchers found that the postoperative state was strongly associated with development of HIT-associated VTE. It is now appreciated that VTE predominates over arterial thrombosis in HIT in a ratio of approximately 4:1.89,90 It is also apparent that patient-dependent risk factors play a role in the site of the thrombosis. For example, recent arterial surgery or severe atherosclerosis is associated with arterial thrombosis in HIT, while a central venous catheter predisposes to development of upper limb DVT in patients with HIT.88,91 Inherited prothrombotic risk disorders such as factor V Leiden have not been shown to be play a major role in explaining HIT-associated thrombosis, although a contributory role in specific circumstances remains plausible.92

Venous limb gangrene is another manifestation of HIT, which usually occurs as a result of warfarin-induced microthrombosis due to an altered procoagulant-anticoagulant balance. In this disorder, HIT-associated hypercoagulability interacts with warfarin induced protein C depletion.9395 Other venous thrombotic events in HIT include unilateral or bilateral adrenal hemorrhagic necrosis (resulting from adrenal vein thrombosis) and cerebral venous (dural sinus) thrombosis.70,96

These concepts help explain the oftentimes severe clinical course of patients with delayed-onset HIT.9698 In this disorder, thrombocytopenia and thrombosis begin days to weeks after heparin has been stopped. Antibodies from these patients can cause platelet activation, in vitro, in the absence of heparin.97 It is possible that delayed-onset HIT is caused by binding of HIT-IgG to a complex of platelet-bound. PF4 and chondroitin sulfate.75,99

By 1998, there was sufficient agreement regarding the pathophysiologic, clinical, and laboratory diagnostic features of HIT that a consensus statement was prepared by representatives of 3 research groups.100 HIT was conceptualized as a clinicopathologic syndrome with one or more clinical events (thrombocytopenia with or without thrombosis) temporally related to heparin administration and caused by HIT antibodies.

Treatment strategies for HIT

The treatment of HIT has paralleled an understanding of its pathophysiology. Initial strategies did not treat the hyperco-agulability of HIT, and consisted of heparin cessation alone,89,101 warfarin,89,102 ancrod,102,103 and antiplatelet agents such as aspirin.104 More recently, agents that inhibit the generation of thrombin, or that inhibit thrombin itself, such as danaparoid,102,105107 lepirudin,108111 and argatroban have been used.112114

An important first step after diagnosing HIT is to stop the heparin. Ironically, for some patients, continuation of heparin does not result in clinical worsening, and even resolution of thrombocytopenia has been observed. In other patients, the thrombotic events would begin or worsen when the heparin was discontinued. In 1996, we described a group of patients with HIT and isolated thrombocytopenia.89 When these patients were managed conservatively by cessation of heparin (with or without initiation or continuation of warfarin) approximately 50% developed a thrombosis within the next days to weeks (Figure 7).89 Similar observations were made by Wallis and colleagues who also reported that heparin cessation often failed to prevent thrombotic events.101 The precise risk of thrombosis after discontinuation of heparin in a patient with HIT and isolated thrombocytopenia remains uncertain with risk estimates range from 20% to 50%.115 The duration of anticoagulation for a patient with HIT and isolated thrombocytopenia is not known, but we typically continue anticoagulation for 6 to 8 weeks. This is anecdotal and in part reflecting the ongoing prothrombotic risk we observed in our retrospective study (Figure 7).

Figure 7

Cumulative frequency of thrombosis in HIT patients presenting with isolated thrombocytopenia. Approximately 50% of HIT patients initially recognized with isolated thrombocytopenia developed objective evidence for thrombosis during the subsequent 30-day period. Reprinted from Warkentin and Kelton89 with permission.

In the 1990s, many physicians would diagnose HIT and immediately discontinue the heparin and anticoagulate with warfarin. But one patient who was managed this way, but dramatically worsened, led us to question this approach. In a case control study,93 we found that warfarin therapy in patients with HIT was often complicated by a dramatic progression of the DVT to distal extremity necrosis (venous gangrene), even when pulses were palpable. Some of these patients ultimately required limb amputation. Plasma samples from several patients demonstrated that warfarin therapy was associated with a severe reduction in the natural anticoagulant protein C, while at the same time the warfarin failed to inhibit the increased thrombin generation.93 One of the hallmarks of this syndrome was a dramatic increase in the INR to supratherapeutic levels (typically, greater than 4.0), which was explained by parallel reductions in factor VII. This syndrome, which has also been observed by other investigators,94,95 has led to the treatment principle that warfarin should be avoided during the acute (thrombocytopenic) phase of HIT.

Our current approach is to postpone warfarin therapy in a patient with acute HIT until there has been substantial resolution of the thrombocytopenia (platelet count rise to at least 150 × 109/L).65 Furthermore, careful overlap of the direct thrombin inhibitor (DTI) and warfarin, or of danaparoid/warfarin therapy, should occur, that is, starting with low (maintenance) doses of warfarin, ensuring at least a 5-day DTI/warfarin or danaparoid/warfarin overlap, and maintaining the DTI or danaparoid until the platelet count has reached a stable plateau within the normal range.65

Two distinctly different direct thrombin inhibitors (DTIs) have been evaluated in prospective studies using historical controls.108114 Lepirudin is a recombinant hirudin derivative and argatroban is a small, synthetic molecule. The strengths and weaknesses of these trials have been recently reviewed.115 The lepirudin studies108111 had the diagnosis confirmed serologically in treated patients, whereas the argatroban studies112114 relied on a clinical diagnosis of HIT (with approximately 65% showing positive antibodies). Both agents showed reductions in new thromboembolic events among patients with HIT-associated thrombosis (pooled analyses vs historical controls, relative risk [RR] of 0.28 and 0.45 for lepirudin and argatroban, respectively). Advantages of DTIs include their ability to inhibit thrombin rapidly, simple monitoring with the aPTT, and short half-life (in situations of bleeding or need for surgery). Potential drawbacks include difficulty judging adequacy of anticoagulation by aPTT in some circumstances, potential for rebound hypercoagulability if stopped prematurely, and drug accumulation in renal (lepirudin) or hepatobiliary (argatroban) dysfunction. Moreover, it is now recognized that for many patients, the manufacturer's recommended starting doses are too high.65 In addition, antihirudin antibodies have also been associated with acute anaphylaxis after lepirudin bolus.116

In many countries (although not in the United States), danaparoid, a mixture of anticoagulant glycosaminoglycans, predominantly heparan, dermatan, and chondroitin sulfate, is available.65 Danaparoid has been shown to be more effect than dextran-70, ancrod, or coumadin102,106 Advantages of danaparoid include its long half-life, ability to monitor drug levels directly (via anti-Xa levels), and the theoretical disruption of PF4-containing immune complexes on platelet surfaces.105 Disadvantages include potential for cross-reactivity of HIT antibodies; however, the success rate is high even when the drug is given without prior testing for cross-reactivity.102,106 An additional challenge is the requirement for measuring anti-Xa levels to achieve optimal anticoagulent monitoring.

Anecdotal reports describe success with bivalirudin and fondaparinux as treatments for patients with HIT.117,118

Heparin-induced thrombocytopenia: future prospects

HIT will continue to be a significant problem for years to come. Although novel inhibitors to the coagulation cascade are currently being evaluated both in the laboratory and in clinical trials, it is likely that heparin will continue to be widely used because of its ease of use, short half-life, low cost, and ability to be neutralized, among other attributes. For certain surgical procedure such as cardiopulmonary bypass, there are no obvious alternative agents. But the understanding of HIT has led to improved treatment strategies, as well as the ability to prevent many cases. The increased use of progressively smaller heparin moieties will continue to reduce the risk of HIT. The recent introduction of the pentasaccharide fondaparinux offers the potential of an exceptionally low risk of HIT.


Contribution: J.G.K. and T.E.W. cowrote the review.

Conflict-of-interest disclosure: T.E.W. is a consultant for GlaxoSmithKline, GTI Inc, and Organon Inc. (part of Schering-Plough), and has received support for research studies and speaker honoraria from GlaxoSmithKline and Organon, and speaker honoraria from Sanofi-Aventis. J.G.K. is a scientific advisor for GlaxoSmithKline.

Correspondence: John G. Kelton, MD, McMaster University, 1200 Main St West, Room 2E1, Hamilton, ON, L8N 3Z5, Canada; e-mail: keltonj{at}


Many of the studies described in this report were supported by grants from the Heart and Stroke Foundation of Ontario.

  • Submitted February 14, 2008.
  • Accepted March 31, 2008.

John G. Kelton: My current responsibilities include Dean for the Faculty of Health Sciences, including the Michael G. DeGroote School of Medicine, and Vice President at McMaster University. I completed my MD training at the University of Western Ontario and then did internal medicine and hematology training at Duke University. The intense training experience, perhaps best described as a Stockholm Syndrome, led to lifelong friendships that continue today, including friendships with Drs Russel Kaufman, Farrell Collins, Wendell Rosse, Wayne Rundels (now deceased), and E. Holmes, among many others. Returning to McMaster University, I learned the true joy of translational research and developed friendships and collaborations (Drs Ted Warkentin, Cedric Carter, Jack Hirsh, D. Sheridan, L. Boshkov, D. Arnold, P. Horsewood, I. Nazi, C. Hayward, I. Walker, P. Wasi, N. Heddle, and M. Blajchman, among many others). Much of my laboratory's success reflects the extraordinary dedication of my research technologists. Recently, an inordinate amount of time has been spent worrying about whether long-term friends who haven't seen me for years will look at my picture and feel that I've gained weight and aged poorly.

Ted Warkentin: How did a teenager growing up in the Canadian prairies end up in the Great Lakes region (McMaster University) so focused on a drug reaction, HIT? In medical school (University of Manitoba), I was amazed how hematologic diagnoses could be made simply by viewing a blood film. In Toronto, a 1-month rotation with the late Dr Michael F. X. Glynn convinced me that hemostasis was really interesting; this endearing hematologist also pointed me westward alongside Lake Ontario to Hamilton.

But why HIT? During a hematology elective and subsequent research fellowship, I was introduced to this topic by my supervisor, Dr Kelton. At that time, the very existence of HIT was debated, and my own first forays into detecting HIT antibodies gave mixed results. I became discouraged, but then everything changed (“Eureka!” moment): a single experiment showed that the inconsistency of experimental data had resulted from widely variable—but hierarchically ordered—reactivities of HIT sera and target platelets. This insight improved detection of HIT antibodies and made me an HIT “believer.” Dr Kelton also engineered my subsequent appointment at the Hamilton General Hospital, with both clinical and laboratory responsibilities. I realized there that HIT was perplexingly counterintuitive, and something worth spending time on. Soon after, I met an HIT aficionado, Andreas Greinacher, who was roughly at the same stage of his research and clinical career (though in Germany), and who shared similar perspectives on HIT. And so I had a mentor near, and a soulmate far, with whom to pursue my fascination with HIT.

Looking back, there are 5 reasons why McMaster University provided an ideal environment to study HIT. First, my research fellowship supervisor—and ongoing career mentor—Dr John Kelton developed a superior test for identifying HIT antibodies; this tool distinguishes the (few) patients with HIT from those (many) without. Second, the “British” system of hematology practiced at McMaster, which emphasizes dual clinical-laboratory appointments, facilitates study of a clinical-pathological disorder such as HIT. Third, the tradition of methodologic rigor exemplified by McMaster's research leaders, particularly Dr Jack Hirsh, demands that observational research be performed as meticulously as possible. Fourth, McMaster fosters an entrepreneurial spirit, and collaborations with industry have led to remarkable research opportunities. And fifth, my primary research assistant, Jo-Ann Sheppard, demonstrates an affability and efficiency of output that never ceases to amaze me.