Heparin-induced thrombocytopenia: pathogenesis and management.

Abstract

Heparin-induced thrombocytopenia (HIT) is a transient prothrombotic disorder initiated by heparin. Its central feature is thrombocytopenia caused by antibody-mediated platelet activation. HIT can be viewed as an acquired hypercoagulability disorder, with increased thrombin generation in vivo, and increased risk for arterial and, especially, venous thrombosis. The pathogenic HIT antibodies are directed against neoepitopes on a ‘self’ protein, platelet factor 4 (PF4), that are expressed when PF4 is bound to heparin or certain other polyanions. This review focuses on studies published since 1990, and summarizes the pathogenesis, laboratory testing, frequency and clinical picture of HIT. Two treatment situations are reviewed critically: management of thrombosis complicating HIT and treatment of ‘isolated HIT’ (HIT recognized because of thrombocytopenia alone). There is potential for medicolegal risk, particularly if inappropriate therapy contributes to patient harm. Heparin-induced thrombocytopenia (HIT) can be defined as any clinical event best explained by platelet factor 4 (PF4)/heparin-reactive antibodies (‘HIT antibodies’) in a patient who is receiving, or who has recently received, heparin (or another polyanion implicated in this syndrome). Thrombocytopenia is the most common ‘event’ in HIT and occurs in at least 90% of patients, depending upon the definition of thrombocytopenia. A high proportion of patients with HIT develop thrombosis. Alternative (non-heparin) anticoagulant therapy reduces the risk of subsequent thrombosis. The diagnosis of HIT should be based upon: (a) the occurrence of one or more HIT-associated clinical events, and (b) detection of HIT antibodies in patient serum or plasma (Warkentin et al, 1998; Warkentin, 2002a) (Table I). By thus viewing HIT as a clinicopathological disorder, it follows that a patient who develops HIT antibodies during heparin treatment but who neither manifests thrombocytopenia nor other adverse events does not have HIT, but rather HIT antibody seroconversion alone. It also follows that a patient who may appear on clinical grounds to have HIT, but in whom sensitive tests for HIT antibodies are negative, also does not have HIT. Indeed, the high negative predictive value of current assays has led to recognition of various HIT-mimicking disorders (‘pseudo-HIT’) (Warkentin, 2001a). Table II lists terms used to describe HIT and its complications (Warkentin, 2001b). In this review, ‘heparin-induced thrombocytopenia’ (HIT) will be used. Figure 1 summarizes the pathogenesis of HIT. The central concept is heparin-induced generation of pathogenic antibodies of immunoglobulin (Ig)G class that recognize multimolecular complexes of PF4 and heparin (Amiral et al, 1992) on platelet surfaces, leading to platelet activation in vivo and associated thrombin generation. There is evidence that endothelial cells and monocytes can also be activated by HIT antibodies. Pathogenesis of HIT: a central role for thrombin generation. PF4/heparin complexes that can express multiple neoepitope sites bind to platelet surfaces. HIT-IgG antibodies recognize neoepitope sites on PF4, leading to formation of multimolecular PF4/heparin/IgG complexes on the platelet surface. The IgG Fc regions bind and crosslink the platelet FcγIIa receptors, resulting in platelet activation, including formation of procoagulant, platelet-derived microparticles, which provide altered membrane surfaces that support coagulation reactions. Activated platelets release additional PF4 from α-granules, leading to a vicious cycle of progressive platelet and coagulation activation. PF4 also can bind to endothelial heparan sulphate, leading to endothelial cell immunoinjury, with tissue factor expression. Monocytes also can bind PF4/heparin-IgG immune complexes, potentially leading to tissue factor expression on these cells. Ultimately, there results marked thrombin generation in vivo, which helps explain the strong association between HIT and thrombotic events. Figure adapted from Greinacher, A. & Warkentin, T.E. (2001) Treatment of heparin-induced thrombocytopenia: an Overview. In: Heparin-induced Thrombocytopenia, 2nd edn. (ed. by T.E. Warkentin & A. Greinacher) Marcel Dekker, New York, used by permission. Heparin is a glycosaminoglycan (GAG), i.e. it consists of linear polymers of repeating disaccharide subunits (Fig 2) that vary in chain length and grade of sulphation. Binding of PF4 to heparin is independent of its antithrombin-catalysing pentasaccharide region. Non-specific role of heparin and other polyanions that lead to neoepitope formation on PF4. The disaccharide units of heparin consist of alternating 1→4-linked residues of hexuronic acid (either d-glucuronic or l-iduronic acid) and d-glucosamine. The predominant disaccharide (comprising 75–90% of heparin) occurs when both ‘X’ and ‘Y’ are SO3–, i.e. the trisulphated disaccharide: [→ 4)-O-α-l-iduronic acid-2-sulphate (1→4)-O-α-d-glucosamine-2, 6-disulphate (1 →]. Heparin, the four polyvinyl structures and pentosan polysulphate are all good at causing neoepitope(s) on PF4 that are recognized by HIT antibodies. In contrast, when PF4 binds to polystyrene sulphonate, only some HIT antibodies react weakly. This may be caused by the greater distance between the negative charge (SO3–) and the carbon backbone in polystyrene sulphonate, compared with the other molecules shown (Visentin et al, 2001). PF4 is a 70-amino acid (7780 Da), platelet-specific member of the C-X-C subfamily of chemokines (Fig 3). Four PF4 molecules self-associate to form compact tetramers of globular structure (∼31 000 Da). PF4 is rich in the basic amino acids, lysine and arginine (32 and 12 residues per tetramer respectively), which form a ‘ring of positive charge’, providing the interface between the PF4 tetramer and heparin (Fig 3). PF4 is stored in platelet α-granules, where it is bound to the GAG, chondroitin sulphate. Under normal conditions, only trace levels of PF4 (∼3 ng/ml) are found in human plasma. However, heparin infusion increases PF4 levels about 15- to 30-fold for several hours, by displacing PF4 from endothelial cell surfaces. Relation of primary and secondary structure of PF4 in relation to HIT neoepitopes. (Top) A 3-dimensional representation of the PF4 tetramer is shown, indicating the two neoepitope sites identified by Li et al (2002). The ‘ring of positive charge’, formed by the lysine residues in the C-terminus of PF4 (light blue) and other lysine and arginine residues (dark blue), is also shown. (Bottom) The linear sequence of the 70-amino acid polypeptide of a single PF4 molecule is shown. Four such polypeptides combine to form the PF4 tetramer. PF4 is classified as a member of the C-X-C subfamily of chemokines, because of its cysteine10–leucine11–cysteine12 sequence. Figure adapted from Li, Z.Q., Liu, W., Park, K.S., Sachais, B.S., Arepally, G.M., Cines, D.B. & Poncz, M. (2002) Defining a second epitope for heparin-induced thrombocytopenia/thrombosis antibodies using KKO, a murine HIT-like monoclonal antibody. Blood, 99, 1230–1236. Copyright American Society of Hematology, used by permission. The immune response against PF4/heparin complexes is polyspecific, i.e. antibodies are directed against multiple neoepitope sites (Suh et al, 1998; Ziporen et al, 1998; Newman & Chong, 1999; Li et al, 2002). Suh et al (1998) showed that heparin must be in a flexible, relatively unconstrained state to react with PF4 and so create the neoepitopes. Recently, Li et al (2002) identified two neoepitopes recognized by many HIT antibodies (Fig 2). Amiral et al (2000) studied purified HIT antibodies from three patients, and showed that greater platelet-activating ability of HIT-IgG antibodies correlated with higher affinity either for PF4/heparin complexes or (in one patient) against PF4 alone. These studies are consistent with clinical studies showing that only a minority of PF4/heparin-reactive IgG are platelet activating, but that HIT typically occurs only in patients with relatively high titres of platelet-activating IgG (Suh et al, 1997; Warkentin et al, 2000). The reported affinity of HIT-IgG for PF4/heparin complexes is intermediate between that of relatively low- and high-affinity antigens (β2-glycoprotein I and tetanus toxoid respectively), although the potential for both Fab arms of HIT-IgG to recognize neoepitopes on the multimolecular PF4/heparin complexes could significantly increase affinity (Newman & Chong, 1999). Despite heparin's key role in initiating HIT, its ability to produce neoepitopes on PF4 is surprisingly non-specific. Indeed, two non-heparin polyanions, pentosan polysulphate (used to treat interstitial cystitis) (Fig 2) and polysulphated chondroitin sulphate (antiarthritis agent), can lead to formation of anti-PF4/heparin antibodies, thrombocytopenia and thrombosis (Greinacher et al, 1992; Goad et al, 1994). Further, the polyanion, polyvinyl sulphonate (Fig 2), which has neither sulphate nor saccharide, also promotes neoepitope formation on PF4 (Visentin et al, 2001). Characteristic features of polyanions that support neoepitope formation include regular spacing of negative charge at 0·5 nm intervals and a minimum polymer length sufficient to span at least 40% of the circumference of the PF4 tetramer along its ring of positive charge. PF4/polyvinyl sulphonate complexes are used to detect HIT antibodies in a commercially available assay. Heparin/PF4 complexes bind to platelets by the negative charge of the highly sulphated heparin chains (Greinacher et al, 1993; Horne & Hutchison, 1998). Maximal binding of PF4/heparin complexes, and thus optimal platelet activation by HIT-IgG, occurs when PF4 and heparin are present in an optimal stoichiometric relationship, ranging from approximately equimolar to a PF4/heparin ratio of about 2:1 (Visentin et al, 1994; Horne & Hutchison, 1998). Although low-sulphated oligosaccharides, such as danaparoid and de-N-sulphated heparin, do not support very well the binding of PF4 and HIT antibodies to the platelet surface, in very high concentrations, these molecules prevent activation of platelets by HIT-IgG (Greinacher et al, 1993). This suggests that even low-sulphated oligosaccharides interact with PF4 to some extent, and in large concentrations displace PF4 from the platelet surface to the fluid phase. Platelet activation leads to PF4 release from α-granules and may increase risk of HIT, either by greater immunization risk or by promoting platelet-activating effects of HIT-IgG (Horne & Hutchison, 1998). Once platelet activation by HIT-IgG begins, the syndrome is self-exacerbating (positive feedback), because PF4 released from platelets can form additional immune complexes (Greinacher, 1995; Newman & Chong, 2000). Some HIT-IgG recognize PF4 bound to solid phase in the absence of heparin (Greinacher et al, 1994a; Newman & Chong, 1999; Amiral et al, 2000), and activate platelets in vitro without added heparin (Amiral et al, 2000; Warkentin & Kelton, 2001a). Indeed, HIT neoepitopes are on PF4 (Fig 3) and so HIT can be considered a drug-induced ‘autoimmune’ disorder. The ability of HIT-IgG to activate platelets in the absence of heparin could explain the onset of thrombocytopenia and thrombosis several days after stopping heparin, so-called ‘delayed-onset HIT’ (Warkentin & Kelton, 2001a). The sequence of events by which HIT-IgG activate platelets was recently investigated by Newman and Chong (2000) using purified HIT-IgG and platelet aggregometry. Addition of heparin to citrated platelet-rich plasma led to the release of small amounts of PF4 that bound to platelet surfaces. Subsequently, increasing amounts of HIT-IgG bound to platelets over time, in parallel with progressive platelet aggregation. Although Fc receptor-blocking monoclonal antibody prevented platelet activation by HIT-IgG, it did not stop binding of HIT-IgG to platelets. Further, platelet activation by HIT-IgG occurred even though heparin remained in considerable molar excess to PF4, i.e. the PF4:heparin ratio was lower than expected for antibody binding in an enzyme immunoassay (EIA). This suggests that the platelet surface microenvironment achieves the necessary stoichiometric relationship as PF4 is increasingly released during progressive platelet activation. This is consistent with a model of in-situ formation on platelet surfaces of the PF4/heparin/IgG immune complexes, rather than forming first in plasma. Occupancy of platelet FcγIIa receptors by PF4/heparin/IgG immune complexes leads to receptor clustering and phosphorylation. Subsequent signalling events include tyrosine kinase activity (e.g. p72syk) that phosphorylates phospholipase Cγ2 (PLCγ2), producing diacylglycerol and inositol triphosphate (for review see Denomme, 2001). PLCγ2 phosphorylation requires adequate levels of phosphatidylinositol-trisphosphate, which requires activation of platelet adenosine diphosphate (ADP) receptors. This explains the importance of using the ADP scavenger, apyrase, during platelet washing in certain assays for detecting HIT antibodies: apyrase avoids ADP-induced platelet desensitization during washing, thus retaining high sensitivity of the platelets to activation by HIT-IgG via the potentiating effects of ADP (Polgár et al, 1998). A consequence of platelet activation by HIT-IgG is the formation of procoagulant, platelet-derived microparticles (Warkentin et al, 1994a; Lee et al, 1996; Hughes et al, 2000). Warkentin and Sheppard (1999) have shown that HIT-IgG and other platelet IgG agonists (heat-aggregated IgG, platelet-activating monoclonal IgG) cause an even greater platelet procoagulant response than physiological platelet agonists such as collagen and thrombin. Evidence that platelet activation occurs in vivo in patients with HIT includes elevated P-selectin on circulating platelets (Chong et al, 1994) and increased numbers of platelet-derived microparticles (Warkentin et al, 1994a). Thrombin is irreversibly inhibited by covalent binding to its major physiological inhibitor, antithrombin. The resulting thrombin–antithrombin (TAT) complexes have a short half-life (20 min) and thus can be measured in plasma to quantify recent thrombin generation in vivo. Two studies noted greatly elevated TAT complexes in HIT patients, with median values of greater than 40 ng/ml (normal < 4 ng/ml) (Warkentin et al, 1997; Greinacher et al, 2000). These values were also much higher than seen in patients with post-operative deep vein thrombosis unrelated to HIT. Increased in-vivo thrombin generation is a general feature of hypercoagulability disorders such as congenital protein C or antithrombin deficiency that evince increased risk for venous thrombosis. Although PF4/heparin-reactive antibodies of IgM and IgA class are also frequently generated in patients with HIT, it remains controversial whether these antibody classes cause HIT in the absence of HIT-IgG antibodies. In prospective studies of HIT antibody formation post surgery, we found platelet-activating IgG antibodies in all 15 patients identified with HIT (Warkentin et al, 2000). In contrast, Amiral et al (1996a) reported finding only IgM and/or IgA antibodies in 12 of 38 patients with apparent HIT, three with pulmonary embolism. However, as pulmonary embolism can mimic clinical HIT (‘pseudo-HIT’), it remains unproven, in my opinion, that HIT can be caused by IgA or IgM antibodies. An alternative explanation could be platelet-activating IgG antibodies against other chemokines (Amiral et al, 1996b). Any of the four HIT-IgG subclasses can be identified in HIT sera, some more commonly than others (IgG1 > IgG3 > IgG2 > IgG4), with more than one subclass often seen in individual patients (Denomme et al, 1997; Suh et al, 1997). To date, no difference in clinical profile of HIT in relation to particular IgG subclasses or the added presence of IgM or IgA class antibodies has been reported. The class of Fcγ receptors found on platelets (FcγIIa) has low affinity for IgG (for review see Denomme, 2001). Thus, these receptors probably bind HIT-IgG only after PF4/heparin/IgG complexes have formed on platelet surfaces, as circulating immune complexes would be preferentially cleared by high-affinity leucocyte Fcγ receptors. Because FcγIIa receptors bear a his131/arg131 polymorphism that influences platelet activation by human IgG2 (his131 >> arg131), speculation arose that this polymorphism might influence risk for HIT. However, no consistent at-risk profile has emerged. The largest study (Carlsson et al, 1998) noted over-representation of the arg131 receptor among patients with HIT-associated thrombosis, leading the authors to suggest that less efficient clearance of immune complexes by the reticuloendothelial system might predispose to greater in-vivo platelet activation and risk of thrombosis. Heparan sulphate is a GAG found on endothelial cell surfaces that is less sulphated than heparin. Nevertheless, HIT antibodies of both IgG and IgM class recognize PF4 bound to heparan sulphate on endothelium, leading to speculation that high levels of PF4 during acute HIT could focus immunoinjury to the endothelium (Visentin et al, 1994). Kwaan and Sakurai (1999) observed hyperplastic endothelial cells, together with immunoglobulin deposition within platelet thrombi and proliferative endothelial cells in ischaemic tissues obtained from patients with HIT. Two reports suggested that HIT-IgG can activate monocytes in the presence of PF4 (Arepally & Mayer, 2001; Pouplard et al, 2001). Moreover, these activated monocytes expressed tissue factor and generated procoagulant activity. Heparin is not required for PF4 binding to monocytes, which is mediated by surface proteoglycans such as chondroitin sulphate. Exactly how heparin triggers immunization remains uncertain. Bacsi et al (1999, 2001) found that PF4/heparin complexes stimulate T cells isolated from patients with HIT. However, little evidence exists for immune memory in HIT. First, HIT antibodies often become undetectable within weeks after an episode of HIT (Warkentin & Kelton 2001b). Second, HIT antibodies are usually not restimulated when such a patient with previous HIT is re-exposed to heparin (Pötzsch et al, 2000) and, if antibodies are regenerated, they are not formed more quickly than 5 d following re-exposure (Warkentin & Kelton 2001b; Lubenow et al, 2002a). Various animal models have been described for HIT (for review see Warkentin, 2002b). Only one model, reported by Reilly et al (2001), recapitulates several key clinical and laboratory features of HIT. These investigators developed a novel murine model employing double-transgenic FcγRIIA/hPF4 mice, i.e. mice with platelets bearing human FcγRIIa and human PF4 (mice lack platelet Fcγ receptors, and murine PF4 is not recognized by HIT antibodies). When these mice were treated with a HIT-mimicking murine monoclonal antibody (named KKO) that recognizes hPF4/heparin and then given heparin, the mice developed severe thrombocytopenia and fibrin-rich thrombi in multiple organs, including the pulmonary vasculature. Only a minority of patients who form HIT antibodies develop thrombocytopenia or other sequelae of HIT (Amiral et al, 1995, 1996c; Warkentin et al, 1995). Thus, assays that are more sensitive at detecting HIT antibodies are less specific for clinical HIT, even if they have high specificity for detecting antibodies. Tests for HIT antibodies can be classified as platelet activation assays and PF4-dependent antigen assays (Table I). Historically, standard platelet aggregation assays were first used to detect platelet-activating HIT antibodies (for review see Warkentin, 2001b). However, greater sensitivity, specificity and test throughput is achieved using ‘washed’ platelets. Platelet aggregometry. A standard platelet aggregometer can detect aggregation of platelets (prepared as citrated platelet-rich plasma) in the presence of patient plasma and heparin (Chong et al, 1993; Warkentin & Greinacher, 2001). However, this method is relatively insensitive for clinical HIT (sensitivity 35–85%) (Greinacher et al, 1994b; Warkentin & Greinacher, 2001). Inability to perform multiple assays limits specificity, as relatively few reaction conditions and controls can be tested. False-positive reactions can result from acute-phase reactants such as fibrinogen that cause a patient plasma-dependent increase in platelet aggregation in the absence of HIT antibodies (Warkentin & Greinacher, 2001). Platelet activation using washed platelets. Platelets that are washed and resuspended in divalent cation-containing buffer have increased sensitivity and specificity for detecting platelet-activating HIT antibodies, compared with conventional aggregometry. Donor selection is important, as platelet responsiveness to HIT-IgG varies among normal donors (Warkentin et al, 1992). A variety of platelet activation endpoints can be used, including release of radioactive serotonin (Warkentin et al, 1992), visual assessment of platelet aggregation (Greinacher et al, 1991), or generation of platelet-derived microparticles detected by flow cytometry (Lee et al, 1996). The assays are performed in microtitre wells, so hundreds of reactions can be assessed simultaneously. This permits study of many reaction conditions, e.g. platelet activation at various heparin concentrations, in the presence of platelet Fc receptor-blocking monoclonal antibody and so forth, which optimizes specificity. HIT antibodies produce a characteristic reaction profile: maximal activation at 0·1–0·3 IU/ml heparin that exceeds the buffer control, minimal activation at 100 U/ml heparin, and inhibition by Fc receptor-blocking monoclonal antibody. Unfortunately, washed platelet activation assays are technically demanding, and performance varies widely among laboratories (Eichler et al, 1999). Another limitation is that about 2–3% of patient samples contain immune complexes or platelet-activating human leucocyte antigen (HLA) alloantibodies and so yield indeterminate results, i.e. platelet activation occurs at all heparin concentrations tested. Solid-phase EIA. Two EIAs are commercially available that detect antibodies of the three major immunoglobulin classes (IgG, IgM, IgA) against PF4 bound either to heparin (Asserachrom, Stago, France) (Amiral et al, 1992) or polyvinyl sulphonate (GTI, Brookfield, WI, USA) (Visentin et al, 2001). The former assay utilizes recombinant PF4, whereas the latter obtains PF4 from outdated platelets. Research laboratories that perform ‘in-house’ PF4/heparin-EIAs have the option to detect antibodies of just the IgG class, which increases specificity for clinical HIT by avoiding detection of non-pathogenic IgA and IgM antibodies (Lindhoff-Last et al, 2001). Recently, a rapid antigen assay has been developed (Meyer et al, 1999) that appears to have operating characteristics (sensitivity–specificity trade-offs) intermediate between the commercial EIAs and a washed platelet activation assay (Eichler et al, 2002a). Fluid-phase EIA. Newman et al (1998) developed a fluid-phase antigen assay that avoids the problems of protein (antigen) denaturation inherent in solid-phase assays. This assay may give a lower rate of false-positive reactions and, unlike solid-phase EIA's, is useful for assessing in-vitro crossreactivity of HIT-IgG against low-molecular-weight heparin (LMWH) and heparinoids. In general, testing is performed when HIT is clinically suspected. In this context, washed platelet activation assays and antigen assays have similar high sensitivity for diagnosis of HIT. Indeed, negative testing by two sensitive and complementary assays (e.g. PF4-dependent EIA and washed platelet activation assay) essentially rules out HIT (Warkentin, 2001a, 2002a). Test sensitivity is significantly less using standard platelet aggregometry (Greinacher et al, 1994b). Diagnostic specificity is greater with the washed platelet activation assays compared with EIAs, as the latter are more likely to detect clinically insignificant antibodies (Warkentin et al, 2000). The clinician's estimate of the probability of HIT (‘pretest probability’), together with the type of assay used and its quantitative result (generating a ‘likelihood ratio’), determines the ‘post-test probability’ of HIT (Bayesian model) (Warkentin & Heddle, 2003). For example, a strong-positive washed platelet activation assay is associated with a much higher likelihood ratio for HIT than a weak-positive EIA (about 20–50 vs 2–3). Thus, for a patient with a pretest probability of HIT of about 50%, the associated post-test probabilities range from about 65% to > 95% respectively. Very-strong-positive assay results are characteristic of delayed-onset HIT (Warkentin & Kelton, 2001a). Figure 4A illustrates a generic ‘iceberg’ showing the relationship between antigen assays, washed platelet activation assays, thrombocytopenia and HIT-associated thrombosis (Lee & Warkentin, 2001). This model is consistent with the following observations: (i) only a subset of PF4/heparin-reactive antibodies have platelet-activating properties (Amiral et al, 2000; Warkentin et al, 2000); (ii) washed platelet activation assays have greater diagnostic specificity for clinical HIT than antigen assays; and (iii) thrombosis is not associated with HIT antibody formation that does not result in a significant platelet count fall (Warkentin et al, 1995). (A) Schematic ‘iceberg’ model showing the relationship between HIT antibodies detected by antigen assay [enzyme immunoassay (EIA)], washed platelet activation assay [serotonin release assay (SRA)], thrombocytopenia and HIT-associated thrombosis. Although the antigen assay is more sensitive for detecting HIT antibodies, it is less specific for clinical HIT than is the washed platelet activation assay. (B) Multiple iceberg model of HIT. The risk of HIT antibody formation and for developing HIT depends upon the type of heparin and the patient population receiving heparin. Figures adapted from Lee, D.H. & Warkentin, T.E. (2001) Frequency of heparin-induced thrombocytopenia. In: Heparin- induced Thrombocytopenia, 2nd edn. (ed. by T.E. Warkentin & A. Greinacher) Marcel Dekker, New York, used by permission. Figure 4B also illustrates how the heparin preparation and patient population influences the frequency of HIT. These and other factors are summarized in Table III. A meta-analysis of four randomized trials performed in the 1980s showed that unfractionated heparin (UFH) from bovine lung was more likely to cause HIT than heparin derived from porcine intestinal mucosa (Lee & Warkentin, 2001). More recently, a randomized trial by Francis et al (2003a) found that beef lung UFH was also more likely to be associated with HIT antibody formation after heart surgery. Greater polysaccharide chain length and degree of sulphation of bovine lung heparin could explain its higher immunogenicity. HIT is more common with UFH than LMWH. In post-operative orthopaedic surgery, in patients receiving heparin prophylaxis, UFH is about threefold more likely to be associated with formation of HIT-IgG antibodies and also about threefold more likely to cause thrombocytopenia when HIT-IgG are formed, i.e. overall, UFH is about 8–10 times more likely to cause HIT (Warkentin et al, 1995, 2000, 2003; Warkentin & Sigouin, 2002). In a non-randomized comparison of UFH and LMWH given after cardiac surgery, Pouplard et al (1999) also observed HIT more often with UFH. HIT antibodies were formed more often in medical patients receiving UFH compared with LMWH (Lindhoff-Last et al, 2002). HIT is less common in medical and obstetrical patients than in surgical patients (Lee & Warkentin, 2001). Fausett et al (2001) reported that none of 244 pregnant women developed HIT during use of UFH, even though HIT occurred in 10 of 244 (4%) non-pregnant patients who received UFH (P = 0·0014). This was especially striking given that the pregnant patients received heparin for longer (61·7 vs 10·5 d; P = 0·0001). Lepercq et al (2001) observed no HIT among 624 pregnancies managed with LMWH. Severin and Sutor (2001) reviewed the literature that described HIT in 27 children. Patients were either neonates/infants or (pre)adolescents, with no cases observed in children between 3 and 7 years of age. This bimodal age distribution probably reflects the high-risk periods for children to receive heparin. In my opinion, neonatal HIT is not established, as antigen and washed platelet activation assays have not been used in this population. Warkentin and Sigouin (2002) found that women are more likely to develop HIT: odds ratio 3·3 (95% confidence interval 1·1–10·2). Although women were no more likely to form HIT antibodies, they were more likely to develop HIT when antibodies were formed, possibly because they formed higher levels of HIT-IgG. Although it is widely assumed that previous heparin therapy increases the subsequent risk of HIT, no evidence supports this contention (Girolami et al, 2003). Indeed, patients with a prior history of HIT can safely receive heparin following the disappearance of HIT antibodies (vide infra). The only unequivocal link between previous heparin use and risk of HIT is the syndrome of ‘rapid-onset HIT’, in which a recent immunizing exposure to heparin explains an abrupt platelet count fall, if heparin is given prior to disappearance of the antibodies (Warkentin & Kelton, 2001b; Lubenow et al, 2002a). Thrombocytopenia is the central feature of HIT: the median platelet count nadir is about 55 × 109/l (Fig 5). In at least 85–90% of patients, the platelet count falls below 150 × 109/l (Warkentin, 2001c). In the remaining patients, HIT is recognized either because of a substantial fall in the platelet count (e.g. 50% or more) or because of clinical events, such as thrombosis or skin lesions at heparin injection sites, that