Articles |

Heparin-Induced Thrombocytopenia: Recognition, Treatment, and Prevention : The Seventh ACCP Conference on Antithrombotic and Thrombolytic Therapy FREE TO VIEW

Theodore E. Warkentin, MD, Chair; Andreas Greinacher, MD
Author and Funding Information

Correspondence to: Theodore E. Warkentin, MD, Hamilton Regional Laboratory Medicine Program, Hamilton Health Sciences, General Site, 237 Barton St E, Hamilton, Ontario L8L 2X2, Canada; e-mail: twarken@mcmaster.ca

Chest. 2004;126(3_suppl):311S-337S. doi:10.1378/chest.126.3_suppl.311S
Text Size: A A A
Published online

This chapter about the recognition, treatment, and prevention of heparin-induced thrombocytopenia (HIT) is part of the Seventh ACCP Conference on Antithrombotic and Thrombolytic Therapy: Evidence Based Guidelines. Grade 1 recommendations are strong and indicate that the benefits do, or do not, outweigh risks, burden, and costs. Grade 2 suggests that individual patients’ values may lead to different choices (for a full understanding of the grading, see Guyatt et al, CHEST 2004; 126:179S–187S). Among the key recommendations in this chapter are the following: For patients in whom the risk of HIT is considered to be > 0.1%, we recommend platelet count monitoring (Grade 1C). For patients who are receiving therapeutic-dose unfractionated heparin (UFH), we suggest at least every-other-day platelet count monitoring until day 14, or until UFH is stopped, whichever occurs first (Grade 2C). For patients who are receiving postoperative antithrombotic prophylaxis with UFH (HIT risk > 1%), we suggest at least every-other-day platelet count monitoring between postoperative days 4 to 14 (or until UFH is stopped, whichever occurs first) [Grade 2C]. For medical/obstetric patients who are receiving prophylactic-dose UFH, postoperative patients receiving prophylactic-dose low molecular weight heparin (LMWH), postoperative patients receiving intravascular catheter UFH “flushes,” or medical/obstetrical patients receiving LMWH after first receiving UFH (risk, 0.1 to 1%), we suggest platelet count monitoring every 2 days or 3 days from day 4 to day 14, or until heparin is stopped, whichever occurs first (Grade 2C). For medical/obstetrical patients who are only receiving LMWH, or medical patients who are receiving only intravascular catheter UFH flushes (risk < 0.1%), we suggest clinicians do not use routine platelet count monitoring (Grade 2C). For patients with strongly suspected (or confirmed) HIT, whether or not complicated by thrombosis, we recommend use of an alternative anticoagulant, such as lepirudin (Grade 1C+), argatroban (Grade 1C), bivalirudin (Grade 2C), or danaparoid (Grade 1B). For patients with strongly suspected (or confirmed) HIT, we recommend routine ultrasonography of the lower-limb veins for investigation of deep venous thrombosis (Grade 1C); against the use of vitamin K antagonist (VKA) [coumarin] therapy until after the platelet count has substantially recovered; that the VKA antagonist be administered only during overlapping alternative anticoagulation (minimum 5-day overlap); and begun with low, maintenance doses (all Grade 1C). For patients receiving VKAs at the time of diagnosis of HIT, we recommend use of vitamin K (Grade 2C). For patients with a history of HIT who are HIT antibody negative and require cardiac surgery, we recommend use of UFH (Grade 1C). [Editor's note: These Grades have been changed as an erratum to the original printed version of this article.]

Heparin-induced thrombocytopenia (HIT) is an antibody-mediated, adverse effect of heparin that is important because of its strong association with venous and arterial thrombosis.14 Patients treated with heparin who acquire HIT constitute a cohort with substantially increased thrombotic risk, both in relative (odds ratio for thrombosis, 20 to 40)15 and absolute (thrombosis risk, 30 to 75%)110 terms, depending on the patient population affected.

HIT should be considered a clinicopathologic syndrome because the diagnosis is based on both clinical and serologic grounds.1114 Thus, HIT antibody seroconversion without thrombocytopenia or other clinical sequelae is not considered HIT, whereas a diagnosis of HIT is made when HIT antibody formation is accompanied by an otherwise unexplained platelet count fall (usually ≥ 50% fall, even if the platelet count nadir remains > 150 × 109/L),,2 or by skin lesions at heparin injection sites15 or acute systemic reactions (eg, chills, cardiorespiratory distress) after IV heparin bolus administration.,7 Diagnostic specificity can be further increased by use of a sensitive washed platelet activation assay, as a positive platelet activation assay is more specific for clinical HIT than a positive antigen assay.1617

The neoepitopes recognized by HIT antibodies are located on platelet factor 4 (PF4), and are formed when PF4 binds to heparin.18PF4 is a member of the C-X-C subfamily of chemokines, and is found in platelet α-granules. At least two distinct neoepitopes have been identified.1920 Only a subset of high-titer, IgG anti-PF4 antibodies activate platelets,16,21 however, which probably explains the greater diagnostic specificity of certain platelet activation assays (eg, platelet serotonin release assay [SRA]) for HIT compared with PF4-dependent enzyme immunoassay (EIA).,16,22

Our chapter is organized into recognition, treatment, and prevention of HIT. The scope of our recommendations include both platelet count monitoring for HIT, as well as management of HIT, both in patients detected by thrombocytopenia alone (“isolated HIT”) and patients who present with HIT-associated thrombosis. The interrelatedness of platelet count monitoring and treatment recommendations is clear, when one considers that isolated HIT (a patient population with substantial risk of thrombosis) by definition can only be detected by platelet count monitoring. Table 1 lists the inclusion and exclusion criteria for the studies used to formulate our recommendations.

1.1 Platelet count monitoring for HIT

HIT is a common adverse event in certain patient populations who receive standard, unfractionated heparin (UFH) for ≥ 1 week.6 The frequency of an adverse reaction can be described as “common” (or “frequent”) if its incidence is > 1%.23 As described later, there is evidence that isolated HIT has a substantial risk of symptomatic and fatal thrombosis. Further, prospective cohort studies (with historical controls) suggest that antithrombotic therapy reduces the risk of thrombosis in patients with isolated HIT. In other clinical settings, the risk of HIT can be described as “infrequent” (or “uncommon”; 0.1 to 1%) or even “rare” (< 0.1%).23 These considerations suggest that routine platelet count monitoring for HIT is appropriate in at least some clinical situations, and that it is reasonable to stratify the intensity of and/or need for platelet count monitoring in relation to the risk of HIT in a given patient population.

Another consideration that supports a role for platelet count monitoring in some clinical settings is that HIT antibody seroconversion and clinical HIT (thrombocytopenia) usually occur during specific time periods following initiation of heparin, namely days 5 to 10 (seroconversion and initial platelet count fall) and days 7 to 14 (reaching a threshold defining thrombocytopenia).12,67,2425 Further, “rapid-onset HIT” (in which the platelet count fall begins within 24 h of starting heparin) is strongly associated with recent heparin exposure (within the past 100 days).2425

The frequency of HIT among patients exposed to heparin is highly variable, and is influenced by the heparin preparation (bovine UFH > porcine UFH > low molecular weight heparin [LMWH])12,6,16,2630 and the exposed patient population (after surgery > medical > pregnancy).12,4,6,16,3134 Thus, whether to perform platelet count monitoring, and the intensity of such monitoring, depends on these considerations. Therefore, it is appropriate to perform platelet count monitoring in certain clinical situations, and to focus platelet count monitoring during those times when HIT usually occurs.

1.1. For patients receiving heparin in whom the risk of HIT is considered to be > 0.1%, we recommend platelet count monitoring over no platelet count monitoring (Grade 1C).

Underlying values and preferences. This recommendation places a high value on diagnosis and early treatment of HIT to prevent sequelae, and a lower value on the burden and cost of monitoring platelet counts.

1.1.1 Platelet count monitoring of patients recently treated with heparin

Rapid-onset HIT refers to patients who have a large platelet count fall attributable to HIT antibodies within 24 h of starting heparin.2425 Contrary to popular assumption, this phenomenon is not caused by an anamnestic immune response, but rather results from the administration of heparin to a patient who has already-circulating HIT antibodies that resulted from a recent heparin exposure.2425 As a general rule, exposure within the past 100 days (and especially within the last month) is associated with the phenomenon of rapid-onset HIT.

1.1.1. For patients who are starting UFH or LMWH treatment and who have received UFH within the past 100 days, or those patients in whom exposure history is uncertain, we suggest obtaining a baseline platelet count and then a repeat platelet count within 24 h of starting heparin (Grade 2C).

1.1.2 Acute systemic reactions after IV UFH bolus

Rarely, patients acquire acute inflammatory (eg, fever, chills) or cardiorespiratory (eg, hypertension, tachycardia, dyspnea, chest pain, cardiorespiratory arrest) symptoms and signs within 30 min following an IV heparin bolus.7,35These reactions can mimic acute pulmonary embolism (“pseudo-pulmonary embolism”36) and strongly suggest acute in vivo platelet activation secondary to HIT. The platelet count should be promptly measured, as an abrupt platelet count fall in this clinical context supports the diagnosis of HIT. Further, the platelet count drop is frequently transient,,2 and thus a delay in determining the platelet count, especially if heparin is stopped, may lead to missing the diagnosis.

1.1.2. For patients who acquire acute inflammatory, cardiorespiratory, neurologic, or other unusual symptoms and signs within 30 min following an IV UFH bolus, we recommend performing an immediate platelet count measurement, and comparing this value to recent prior platelet counts, in comparison with not performing a platelet count measure (Grade 1C).

1.1.3 Platelet count monitoring in patients receiving therapeutic-dose UFH

For patients receiving porcine UFH in therapeutic doses, either by IV or subcutaneous (SC), for the treatment of venous or arterial thrombosis, the risk of HIT has been estimated at approximately 1%,6 based on a review of several studies4,30,3750 of the frequency of HIT in patients receiving porcine UFH for venous thromboembolism.

1.1.3. For patients who are receiving therapeutic-dose UFH, we suggest at least every-other-day platelet count monitoring until day 14, or until UFH is stopped, whichever occurs first (Grade 2C).

Underlying values and preferences. This recommendation places a high value on diagnosis and early treatment of HIT to prevent sequelae, and a lower value on the burden and cost of monitoring platelet counts.

1.1.4 Platelet count monitoring in postoperative patients receiving UFH antithrombotic prophylaxis

Patient groups at the highest risk of HIT (1 to 5%) include postoperative orthopedic, cardiac, and vascular surgery patients who are receiving UFH for 1 to 2 weeks.12,6,16,2627,5155 Data are not available for general surgery patients. However, we have included this patient population in this section, because patients undergoing major abdominal surgery might be at similar risk as the other major surgical procedures discussed. Thus, this section includes all “Postoperative Patients Receiving UFH Antithrombotic Prophylaxis.”

Our recommendation for platelet count monitoring in this and other patient populations (see also recommendations 1.1.4 to 1.1.6, inclusive) have been given a weak (Grade 2) recommendation because no study exists comparing outcomes using any particular platelet count monitoring strategy. Our suggestion to perform every-other-day monitoring takes into account the observation that platelet count declines in HIT, when they occur, are relatively rapid (median of 3 days from baseline [postoperative peak] to ≥ 50% platelet count decline).12

1.1.4. For patients who are receiving postoperative antithrombotic prophylaxis with UFH (HIT risk > 1%), we suggest at least every-other-day platelet count monitoring between postoperative days 4 to 14, or until UFH is stopped, whichever occurs first (Grade 2C).

Underlying values and preferences. This recommendation places a high value on diagnosis and early treatment of HIT to prevent sequelae, and a lower value on the burden and cost of monitoring platelet counts.

1.1.5 Platelet count monitoring in patients in whom HIT is infrequent (0.1 to 1%)

There are several patient groups in which the risk of HIT can be classified as “infrequent,” ie, 0.1 to 1%. These include medical or obstetric patients receiving prophylactic-dose UFH4,6,34,4850,5658; postoperative patients receiving LMWH12,6,16,51,5455; postoperative/critical care patients receiving UFH flushes59; and, theoretically, medical patients receiving LMWH after having received one or more preceding doses of UFH. In some settings, it may not be practical to obtain platelet counts, eg, patients receiving outpatient LMWH. Thus, less frequent platelet count monitoring may be appropriate in these patients, especially if the risk is thought to be closer to 0.1% than 1% (eg, postoperative patients receiving LMWH), and if the patient is instructed to contact the physician promptly if symptoms of venous thromboembolism occur (the most common complication of HIT).

1.1.5. For medical/obstetric patients who are receiving prophylactic-dose UFH, postoperative patients receiving prophylactic-dose LMWH, postoperative patients receiving intravascular catheter UFH flushes, or medical/obstetric patients receiving LMWH after first receiving UFH (HIT risk 0.1 to 1%), we suggest platelet count monitoring every 2 to 3 days from day 4 to day 14, or until heparin is stopped, whichever occurs first, when practical (Grade 2C).

Underlying values and preferences. This recommendation places a high value on diagnosis and early treatment of HIT to prevent sequelae, and a lower value on the burden and cost of monitoring platelet counts.

1.1.6 Platelet count monitoring when HIT is rare (< 0.1%)

In medical and obstetric patients receiving LMWH, the risk of HIT appears to be rare (< 0.1%). For example, only one possible case32 of HIT was observed among 1,167 pregnancies treated with LMWH in three studies.3133 Although fewer data exist with respect to medical patients receiving LMWH or UFH as “flushes” (eg, oncology patients with indwelling catheters),,6061 the experience of the authors is that HIT is rare in this setting.

1.1.6. For medical/obstetric patients who are only receiving LMWH, or medical patients who are receiving only intravascular catheter UFH flushes (HIT risk < 0.1%), we suggest clinicians do not use routine platelet count monitoring (Grade 2C).

Underlying values and preferences. This recommendation places a lower value on the rare diagnosis and early treatment of HIT to prevent sequelae, and a higher value on the burden and cost of monitoring platelet counts.

1.1.7 Screening for subclinical HIT antibody seroconversion

Prospective studies of HIT and HIT antibody formation12,6,16,2830,62 indicate that HIT occurs in a minority of patients who form HIT antibodies. The typical serologic finding in the patient with clinical HIT (> 95% of patients) is positive testing in both of two sensitive and complementary assays: (1) platelet activation (or “functional”) assay using washed platelets (eg, 14C-SRA, heparin-induced platelet activation assay), or (2) PF4-dependent EIA.,16 However, even though one (or both) assays are sensitive in detecting HIT antibodies, neither is completely specific for the HIT syndrome (although the functional assays are more specific than the EIA) [Table 2 ].,13 Consequently, it is easier using serology to rule out a tentative diagnosis of HIT than to confirm the diagnosis, ie, the tests have a high negative predictive value but only a moderate positive predictive value. However, the “strength” of a positive test result provides useful diagnostic information regarding the likelihood of HIT. For example, a strong positive test result (eg, > 90% serotonin release or > 2.0 absorbance units) is associated with a high likelihood ratio for HIT in patients after orthopedic surgery (approximately 100), whereas a weak positive test result (eg, 20 to 50% serotonin release or 0.50 to 0.75 absorbance units) is associated with lower likelihood ratios for HIT in this patient population (approximately 30 to 40 and 15 to 20, respectively).,1617 For patients after cardiac surgery, the corresponding likelihood ratios for “strong” and “weak” serologic results are approximately 20 and 2 to 6, respectively.17 The diagnostic interpretation of these laboratory tests must be made in the context of the clinical estimation of the pretest probability of HIT.13,17,63

Further, prospective data indicate that an increased risk of thrombosis occurs in the group of patients whose platelet count has fallen in relation to HIT antibody formation (ie, those with clinical HIT) rather than in patients who acquire HIT antibodies without a significant platelet count decline.12 In our view, it is not useful to perform HIT antibody testing in the absence of clinical indication of HIT, either by an unexpected fall in the platelet count, or an unexpected clinical event. Thus, routine platelet count monitoring, rather than routine HIT antibody studies, is most useful (and most practical) to identify patients who are at risk for thrombosis because of immunization triggered by heparin therapy.

1.1.7. In patients who receive heparin, we recommend against routine HIT antibody testing in the absence of thrombocytopenia, thrombosis, heparin-induced skin lesions, or other sequelae of HIT (Grade 1C).

1.1.8 When should HIT be suspected?

Retrospective and prospective studies suggest that > 90% of patients with clinical HIT have a platelet count fall > 50% during their heparin treatment.2,12 In those patients who are recognized with lesser degrees of platelet count decline, almost all are identified because of thrombotic complications or other sequelae, such as heparin-induced skin lesions or acute systemic reactions following IV bolus UFH.17 The pretest probability of HIT should also be influenced by the temporal features of the platelet count fall and by the likelihood of other possible alternative diagnoses to explain the thrombocytopenia.17

A diagnosis of HIT should be considered when thrombocytopenia (defined subsequently) occurs with a temporal pattern consistent with heparin-induced immunization, ie, platelet count fall that begins 5 to 10 days (or thrombocytopenia that occurs 7 to 14 days) after starting a course of heparin therapy (first day of heparin = day zero), or when thrombosis or other sequelae of HIT occur in patients treated (or recently treated) with heparin.17 The pretest estimation of the probability of HIT is also influenced by the pattern of the platelet count fall and by the likelihood of other possible alternative diagnoses to explain the thrombocytopenia.17 The strong association between HIT and thrombosis indicates that HIT should be suspected, and a platelet count drawn (and compared with previous values), in a patient who acquires symptomatic venous or arterial thrombosis during or within several days after receiving heparin treatment.

Approximately two thirds of HIT patients evince typical-onset HIT, ie, the platelet count begins to fall 5 to 10 days after starting heparin,2425 although thrombocytopenic levels (eg, ≥ 50% fall or to < 150 × 109/L) are usually not reached until a few days later (approximately 7 to 14 days after beginning heparin). In approximately 25 to 30% of patients, the platelet count falls abruptly on beginning a course of heparin.,24 Such rapid-onset HIT occurs in patients who have recently been exposed to heparin (within the previous 100 days),2425 and represents abrupt-onset of platelet activation in a patient who has residual circulating HIT antibodies related to the recent prior heparin exposure.

In at most 3 to 5% of patients, the onset of thrombocytopenia begins several days after heparin has been stopped (delayed-onset HIT).6466 This last syndrome, which was reported in late 2001, is consistent with a transient autoimmune nature of HIT, as it has been shown that such patients have PF4/heparin-reactive antibodies that can activate platelets even in the absence of heparin.64

The majority of postoperative patients who acquire HIT sustain an otherwise unexplained ≥ 50% fall in the platelet count from the postoperative peak during the second week following surgery.2 This reduction occurs on a background of the normal pattern of a rising platelet count expected between postoperative days 4 to 14 (transient postoperative thrombocytosis).12 Thus, in postoperative HIT, the serial platelet counts form an “inverted v” as the initial platelet count recovery that begins about 2 to 3 days following surgery transforms unexpectedly to a falling platelet count a few days later.12,7 In contrast, in medical patients, the platelet count fall begins or accelerates from day 5 onwards, usually without a preceding profile of a rising platelet count.4 On occasion, the platelet count declines by < 50% even though the clinical and serologic findings otherwise strongly suggest HIT-associated thrombosis.12,15

Although there are less data on an appropriate definition of HIT applicable to medical patients,4 it appears that a proportional (50%) fall in platelet count beginning between days 4 to 14 of heparin therapy is appropriate. In our opinion, such a threshold avoids trivial platelet count declines that might be detected if an absolute threshold, such as 150 × 109/L, is used to define thrombocytopenia, especially as transient thrombocytosis does not often occur in medical patients.

We are making a strong recommendation regarding thrombocytopenia in HIT because there is good evidence that a proportional fall in platelet count of ≥ 50% is superior to an absolute threshold of 150 × 109/L for the detection of HIT, at least in postoperative patients (improved sensitivity for HIT without loss of diagnostic specificity).2,7 However, no single definition of thrombocytopenia is appropriate in all clinical situations.

1.1.8. For patients receiving heparin, or who have received heparin within the previous 2 weeks, we recommend excluding a diagnosis of HIT if the platelet count falls by ≥ 50%, and/or a thrombotic event occurs, between days 4 to 14 following initiation of heparin, even if the patient is no longer receiving heparin therapy when thrombosis or thrombocytopenia have occurred (Grade 1C).

1.1.9 Special situation: anticoagulant prophylaxis and platelet count monitoring after cardiac surgery

The risk of symptomatic venous thrombosis is relatively low in patients after cardiac surgery, even when no antithrombotic prophylaxis is administered (although subclinical deep venous thrombosis [DVT] can be detected in 20% of patients).67 To our knowledge, there are no formal studies proving that routine anticoagulant prophylaxis either with UFH or LMWH is safe and effective following cardiac surgery. Many cardiac surgery centers give antithrombotic prophylaxis with UFH (North America more than Europe) or LMWH (Europe more than North America). Even if anticoagulant prophylaxis is not routinely administered, individual patients after cardiac surgery may receive anticoagulants because of a prosthetic valve or unexpected complications such as atrial fibrillation, thrombotic stroke, or prolonged immobilization.

The risk of HIT antibody formation is especially high in the population after cardiac surgery, ranging from 35 to 65% by days 7 to 10, even when postoperative anticoagulant prophylaxis with heparin is not administered.16,5355,68 More importantly, the absolute risk of clinical HIT in such patients who receive UFH following surgery ranges from 1 to 3%.6,5355,69 Finally, this patient population has a relatively high burden of atherosclerosis, and appears to be at a disproportionately higher risk for life- and limb-threatening arterial complications, compared with other patient populations.7

A nonrandomized trial5455 reported a lower frequency of HIT with LMWH use, compared with UFH use, following cardiac surgery. However, there were differences in the patient population that led to one or the other drug being administered. Further, HIT antibodies resulting from UFH therapy frequently cross-react with LMWH, and since patients after cardiac surgery receiving LMWH have invariably received UFH during cardiac surgery, there is the potential for HIT to occur more frequently with LMWH in this patient population than in other clinical settings.

Thus, given the known high risk of HIT in this patient population, we believe that monitoring for HIT is especially important if UFH or LMWH is used.69 A practical problem in monitoring for HIT after postcardiac surgery is that major hemodilution occurs both during, and in the first several days following, cardiac surgery. This perioperative platelet count decrease typically attains its nadir 2 days following surgery. However, HIT is rare in the first 4 days following cardiac surgery, even in patients who have received heparin during the precardiac surgery period. This is because HIT resulting from heparin exposure during angiography or for treatment of acute coronary syndrome is infrequent (< 1%), whereas postoperative dilutional thrombocytopenia occurs universally. Thus, it is difficult on clinical grounds to distinguish the occasional case of HIT beginning soon after cardiac surgery (in which immunization resulted from preoperative heparin exposure). In contrast, HIT is a relatively likely explanation for a platelet count fall ≥ 50% that begins from postoperative day 5 onwards. This is because the circumstances of cardiac surgery are a frequent stimulus for HIT antibody generation, and because the typical onset of HIT (beginning 5 to 10 days after cardiac surgery) coincides with the time period in which the platelet count typically is rising to thrombocytotic levels following perioperative hemodilution. Accordingly, in patients after cardiac surgery, a fall in the platelet count of ≥ 50% from the highest postoperative value that occurs between postoperative days 4 to 14 should be considered HIT unless proven otherwise (day of cardiac surgery = day zero).2,54

1.1.9. For postoperative cardiac surgery patients, we recommend excluding a diagnosis of HIT if the platelet count falls by ≥ 50% (and/or a thrombotic event occurs) between postoperative days 4 to day 14 (day of cardiac surgery = day zero) [Grade 1C].

HIT is a prothrombotic condition that is associated with increased in vivo thrombin generation (as evidenced by the presence of elevated levels of thrombin-antithrombin complexes70) and thus can be considered an acquired, hypercoagulability syndrome.13 However, unlike other acquired hypercoagulability syndromes (eg, antiphospholipid antibody syndrome, malignancy-associated thrombosis), HIT is transient, with recovery of platelet counts to normal levels within days or weeks, and disappearance of the pathogenic HIT antibodies within weeks or a few months.,24 Thus, there is important potential benefit (over the risk) of optimal antithrombotic management over the relatively brief period of the patient’s life in which this paradoxical adverse event has occurred.

The mechanism of this hypercoagulability state is multifactorial, and includes the following: (1) in vivo platelet activation,71with formation of procoagulant, platelet-derived microparticles7274 caused by occupancy and cross-linking of platelet Fc receptors75by in situ formation of PF4/heparin/IgG immune complexes;76 (2) expression of tissue factor on endothelial cells that have become activated because HIT antibodies recognize PF4 bound to endothelial heparan sulfate;7778 and (3) expression of tissue factor by monocytes activated by HIT antibodies.7980 Neutralization of the anticoagulant effects of heparin by PF4 released from activated platelets may explain “heparin resistance” that is commonly observed in HIT.

Marked in vivo thrombin generation helps explain several clinical aspects of HIT, including its association with venous and arterial thrombosis, the occurrence of decompensated (hypofibrinogenemic) disseminated intravascular coagulation in 5 to 10% of HIT patients, and the risk for progression of DVT to venous limb gangrene (or, less often, “classic” nonacral coumarin-induced skin necrosis) in some patients with HIT who are treated with warfarin or other vitamin K antagonists (VKAs).8186 These coumarin-induced necrosis syndromes result from a disturbance in procoagulant-anticoagulant balance during VKA therapy: warfarin treatment results in severe acquired reduction in protein C, while at the same time it fails to control thrombin generation.8182 Finally, recognition of the role for in vivo thrombin generation in HIT provides a rationale for current therapies that emphasize reduction of thrombin generation,,11,70 either via direct inhibition of thrombin (eg, argatroban, lepirudin, bivalirudin) or by inhibiting factor Xa (eg, danaparoid, fondaparinux).

In making recommendations for the management of HIT, we have chosen to combine the approach to patients with “isolated HIT” and HIT-associated thrombosis. There are three reasons for this approach. First, from the point of view of pathophysiology, patients with isolated HIT and HIT-associated thrombosis have similar disease processes, as shown by platelet count nadirs (median, approximately 50 to 60 × 109/L for each group), and similar elevations of thrombin-antithrombin complexes. Second, the time course of thrombosis in HIT is a continuum, with approximately equal numbers of patients being recognized with symptomatic thrombosis (1) during the initial period of a falling platelet count, (2) after crossing a threshold defining thrombocytopenia but while heparin treatment remains ongoing, and (3) after discontinuation of heparin because of thrombocytopenia.12,9 Third, and most importantly, among patients who are recognized as having isolated HIT (subsequently confirmed serologically), and who are managed by simple discontinuation of heparin, or substitution of heparin by warfarin, the risk of symptomatic thrombosis ranges from 25 to 50%, including an overall risk of fatal thrombosis of approximately 5%.12 These event rates resemble those in other clinical situations in which antithrombotic management is generally considered mandatory (eg, after hip fracture).

Unlike hip fractures, however, the diagnosis of HIT may not be initially clear, especially since HIT might not be the only potential explanation for thrombocytopenia and/or thrombosis in patients receiving heparin. Thus, it is important to emphasize that the recommendations we have made are appropriate for patients in whom the diagnosis of HIT is strongly suspected (or “confirmed” by strong positive test results for HIT antibodies). In clinical settings in which HIT is considered unlikely, it may be appropriate to continue heparin or (in settings of antithrombotic prophylaxis) to administer usual prophylactic doses of an alternative anticoagulant, eg, prophylactic-dose recombinant hirudin (15 mg bid SC),87fondaparinux (2.5 mg qd SC),88or danaparoid (750 U bid or tid SC, where available).8990 Scoring systems to help physicians estimate the pretest probability of HIT have been developed.12,17,91

2.1 Nonheparin anticoagulants for HIT

Table 3 lists five agents that can be considered for treatment or prevention of HIT-associated thrombosis.9296 Pharmacokinetic information, including site of organ clearance for these anticoagulants, is also listed. Of these drugs, only two (argatroban, lepirudin) are approved for treatment of HIT in the United States.9293 Another agent, bivalirudin, which is approved for anticoagulation during percutaneous coronary interventions (PCIs), has been used off-label to a limited extent in HIT.94,9798 A fourth agent, danaparoid, was recently withdrawn from the US and UK markets, but is approved for treatment and prevention of HIT-associated thrombosis in Canada, continental Europe, Australia, New Zealand, and Japan, and presently remains available in these countries.95A fifth agent, fondaparinux, was recently introduced into the US market.96 This pentasaccharide inactivates factor Xa in an antithrombin-dependent manner and does not cross-react in vitro with HIT antibodies.,99102 Therefore, theoretically, it should be effective for HIT, although its reported use in this indication to date is minimal.103

The evidence for the efficacy of nonheparin anticoagulants for HIT is not based on large prospective randomized trials, due to the overall infrequent occurrence of HIT and the clinical heterogeneity of affected patients. Indeed, only one randomized trial104has been performed in HIT; this open-label study compared danaparoid (plus warfarin) with dextran (plus warfarin). In addition, several retrospective cohort studies105108 have been reported assessing danaparoid therapy. In contrast, prospective cohort studies (generally with historical controls) have been performed for the two direct thrombin inhibitors (DTIs), lepirudin109113 and argatroban.114115 Among these prospective cohort studies, the primary efficacy end point was a composite end point consisting of new thrombosis, limb amputation, and all-cause mortality. This end point may overestimate the occurrence of new apparent thrombosis or thrombosis growth, as deaths and limb amputations could be related to clinical factors already established when an alternative anticoagulant therapy is begun.5

Antihirudin antibodies are commonly generated during treatment with lepirudin116118; reports of anaphylaxis in patients reexposed to lepirudin (as high as 1 in 625 in patients re-exposed to lepirudin)119 led the European Agency for the Evaluation of Medicinal Products in a public statement (October 2002) to recommend that nonhirudin anticoagulants be considered in patients who have previously been exposed to lepirudin. Treatment of HIT-associated thrombosis

Table 4 summarizes the results of the efficacy and major bleeding end points for the lepirudin109113 and argatroban114115 prospective cohort groups of patients with HIT complicated by thrombosis, including their respective historical control data. The initial prospective studies utilizing the DTIs, lepirudin and argatroban, showed that new thrombosis occurred in 10.1% and 19.4% of patients receiving lepirudin and argatroban, respectively, and the composite end point occurred in 21.3% and 43.8% of patients receiving lepirudin and argatroban, respectively. Compared with their respective historical controls, these results corresponded to relative risk reductions (RRRs) of 63% and 44% for lepirudin and argatroban, respectively. Later trials showed better outcomes with both agents: the reported thrombosis rate declined from 10.1 to 6.1% with lepirudin, and from 19.4 to 13.1% with argatroban. A large postmarketing study113 with lepirudin showed an even lower incidence of thrombosis (5.2%).

Significant differences in the entry criteria and conduct of the trials occurred. For example, patients entered into the lepirudin trials needed to be positive for HIT antibodies, whereas argatroban patients were entered based on a clinical diagnosis (only 65% of patients were shown to have HIT antibodies in the Arg-911 study, and the data for the Arg-915 study are not reported). Moreover, patients received lepirudin for 12.1, 13.5, and 14 days (mean values of three lepirudin trials for HIT-associated thrombosis), but argatroban only for 5.9 to 7.1 days (means of the Arg-911 and Arg-915 trials, respectively). A greater percentage of patients in the lepirudin trials were transitioned to a VKA, compared with patients in the argatroban trials (at least 83% vs 62%). Particularly as observation periods in the studies were relatively long (35 days and 37 days for lepirudin and argatroban, respectively), the longer duration of lepirudin therapy, and the greater likelihood of transition to VKA, could explain its greater apparent efficacy.

Limb amputation represents a relatively “hard” end point. Comparing limb amputation rates among the trials, there is a lower amputation rate among patients who received lepirudin, compared with argatroban (12 of 214 patients [5.6%] vs 51 of 373 patients [13.7%]) when comparing the three combined Heparin-Associated Thrombocytopenia (HAT) studies and Arg-911/915 study event rates shown in Table 4. Further, the RRR values for limb amputation were 38 to 51% for lepirudin (compared with historical controls), but were – 8 to – 36% for argatroban, ie, the limb amputation rates were higher than the corresponding historical controls. The explanation for this difference in limb amputation rates between the lepirudin and argatroban studies is not known. However, one plausible reason is that the combination of shorter treatment duration in the argatroban trials, compared with the lepirudin studies (5.9 to 7.1 days vs 12.1 to 14 days), combined with the greater potential of argatroban and VKA to prolong the international normalized ratio (INR), may have led to early cessation of argatroban, with the potential for progression of limb thrombosis (and venous limb ischemia and gangrene) in patients with active HIT. Our recommendations for managing DTI-VKA overlap are discussed later in section 2.2.

Recombinant hirudin (including lepirudin) has been shown to be superior to UFH in randomized clinical trials (RCTs) of acute coronary syndrome and angioplasty.120In contrast, similar evidence for efficacy of univalent DTIs, such as argatroban, in similar patient populations is not available.121

Although bivalirudin appears to be promising as a treatment for HIT, based on case series,9798 the absence of historical or contemporaneous control data, and the uncertainty regarding the numbers of patients who had clinical HIT in some of the studies, we provide weak recommendation (grade 2C). Compared with lepirudin and argatroban, bivalirudin offers some significant pharmacologic advantages (short half-life, enzymic metabolism, low immunogenicity, minimal effect on INR prolongation).

Table 5 shows studies that have evaluated danaparoid as treatment of HIT complicated by thrombosis. Danaparoid was studied in a randomized open-label study104 that compared danaparoid (plus warfarin) against dextran-70 (plus warfarin). Patients received danaparoid without prior testing for in vitro cross-reactivity against HIT antibodies. This study showed a significantly lower progression of thrombosis rate (12.0% vs 52.9%) among the 25 patients who received danaparoid, compared with the 17 control patients. No patients had major bleeding.

Additional corroborating evidence for the efficacy of danaparoid in HIT includes a comparison between lepirudin and danaparoid for treatment of HIT-associated thrombosis that used identical inclusion/exclusion criteria, and that analyzed patients with HIT diagnosed in the same laboratory during the identical time period.105 Thus, unlike the prospective cohort studies of lepirudin and argatroban that utilized historical controls, this evaluation included contemporaneous controls. The study suggested that danaparoid and lepirudin have similar efficacy for treatment of HIT-associated thrombosis (9.4% thrombosis rate with danaparoid, 7.9% thrombosis rate with lepirudin), but with significantly less major bleeding observed with danaparoid (2.5% vs 10.4%, respectively; p < 0.05).105

A retrospective evaluation of danaparoid vs ancrod (defibrinogenating snake venom) in one medical community showed a significantly lower thrombotic event rate in patients treated with danaparoid.106 (Ancrod has been removed from the market.)

Certain of the pharmacokinetic features of danaparoid, such as its long half-life, lack of effect on the INR, and its potential for SC administration make it an appropriate choice for an otherwise uncomplicated patient with venous thromboembolism in whom eventual overlap with oral anticoagulants is required. Danaparoid does not cross the placenta,95 and thus should be safe for management of pregnant patients with HIT.

Fondaparinux has some pharmacologic similarities with danaparoid. Both have anti-factor Xa activity, either exclusively (fondaparinux, anti-Xa:anti-IIa ratio > 100) or predominantly (danaparoid, anti-Xa:anti-IIa ratio = 22). Both fondaparinux and danaparoid have long half-lives for their anti-factor Xa activities (17 h and 25 h, respectively), and both show either absent (fondaparinux) or generally negligible (danaparoid) in vitro cross-reactivity with HIT antibodies. All of these features of fondaparinux indicate that at least theoretically it should be useful for treating patients with HIT. As fondaparinux is marketed in a prophylactic-dose regimen (2.5 mg qd SC) for prevention of thrombosis after orthopedic surgery, this suggests that it also may be appropriate for prevention of thrombosis in its low-dose regimen in non-HIT situations in which the physician would prefer not to administer heparin, eg, a thrombocytopenic patient in whom HIT is nevertheless judged to be unlikely. However, the minimal data supporting the efficacy of fondaparinux in HIT and other thrombocytopenic situations precludes us from making any recommendation.