Heparin-induced thrombocytopenia: A stoichiometry-based model to explain the differing immunogenicities of unfractionated heparin, low-molecular-weight heparin, and fondaparinux in different clinical settings

Thrombosis Research (2008) 122, 211–220

intl.elsevierhealth.com/journals/thre

REGULAR ARTICLE

Heparin-induced thrombocytopenia: A stoichiometry-based model to explain the differing immunogenicities of unfractionated heparin, low-molecular-weight heparin, and fondaparinux in different clinical settings A. Greinacher a,⁎, S. Alban b , M.A. Omer-Adam c , W. Weitschies c , T.E. Warkentin d a

Institut für Immunologie und Transfusionsmedizin, Ernst-Moritz-Arndt-Universität Greifswald, Germany Pharmazeutisches Institut, Christian-Albrechts-Universität, Kiel, Germany c Institut für Pharmazie, Ernst-Moritz-Arndt-Universität Greifswald, Germany d Department of Pathology and Molecular Medicine, Michael G. DeGroote School of Medicine, McMaster University, Hamilton, Canada b

Received 5 June 2007; received in revised form 28 September 2007; accepted 18 November 2007 Available online 8 February 2008

KEYWORDS Heparin-induced thrombocytopenia; Platelets; Heparin; Fondaparinux; Low-molecular-weight heparin

Abstract Introduction: Heparin-induced thrombocytopenia (HIT) is caused by platelet-activating antibodies that recognize platelet factor 4 (PF4)/heparin complexes. The frequency of HIT is highly variable in different clinical settings, and is more frequent with unfractionated heparin (UFH) than with low-molecular-weight heparin (LMWH), despite the in vitro observation that HIT antibodies activate platelets similarly well with LMWH as with UFH. An important difference between UFH, LMWH, and fondaparinux is their widely differing plasma concentrations. We aimed to provide a model that included anticoagulant concentrations and PF4 availability as risk factors influencing the antiPF4/heparin immune response. Materials and methods: By photon correlation spectroscopy we determined the concentrations at which UFH, LMWH, and fondaparinux form complexes optimally with PF4.

Abbreviations: UFH, unfractionated heparin; LMWH, low-molecular weight heparin; HIT, heparin-induced thrombocytopenia; PF4, platelet factor 4; AT, antithrombin; PCS, photon correlation spectroscopy; aXaU, anti-factor Xa units; MW, molecular weight. ⁎ Corresponding author. Institute for Immunology and Transfusion Medicine, Ernst-Moritz-Arndt University, Sauerbruchstraße, 17489 Greifswald, Germany. Tel.: +49 3834 865482; fax: +49 3834 865489. E-mail address: [email protected] (A. Greinacher). 0049-3848/$ - see front matter © 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.thromres.2007.11.007

212

A. Greinacher et al. Plasma concentrations of UFH and LMWH were calculated based on ex vivo pharmacokinetic data, with information on fondaparinux and PF4 concentrations taken from the literature. Results and conclusions: The main features of our model are: optimal complex formation occurs at prophylactic-dose UFH and high PF4 levels, whereas therapeuticdose LMWH concentrations are too high for optimal complex formation; in contrast, concentrations of fondaparinux are usually below the optimal stoichiometric range. Thus, immunization should occur more often in situations with major rather than minor platelet activation, and—for a given degree of platelet activation (PF4 availability)—as: prophylactic-dose UFHNtherapeutic-dose UFHNprophylactic-dose LMWH, fondaparinuxNtherapeutic-dose LMWH. Our model provides a framework for explaining empirical observations that LMWH induces less anti-PF4/heparin antibodies than does UFH, and that anti-PF4/heparin antibodies are more often found in patients undergoing major surgery than in medical patients. © 2007 Elsevier Ltd. All rights reserved.

Introduction Immune heparin-induced thrombocytopenia (HIT) is caused by platelet-activating antibodies that recognize multimolecular complexes of platelet factor 4 (PF4) bound to heparin [1]. However, the frequency of this adverse drug reaction during use of heparin is highly variable in different clinical settings and with different types of heparins. Among surgical patients, the risk of antibody formation and clinical HIT is substantially lower with low-molecular-weight heparin (LMWH) than with unfractionated heparin (UFH) [2–5]. Based on size determination of PF4/UFH and PF4/LMWH complexes, it has been suggested that LMWH is less immunogenic than UFH because LMWH forms smaller complexes with PF4 [6]. However, it is not immediately obvious why PF4/LMWH complexes should be much less immunogenic than PF4/UFH complexes, as they still express the antigen very well, as shown by their frequent cross-reactivity (N 90%) using sensitive in vitro assays, and also form large complexes with PF4 similar to those formed with UFH, as shown by different methods of particle size determination (e.g. electrophoresis, photon correlation spectroscopy) [7,8]. Paradoxically, despite its apparent lower immunogenicity in clinical studies, in vitro addition of LMWH is even better than UFH at supporting platelet activation by HIT antibodies [9]. For this reason we routinely use LMWH for HIT antibody testing, as this yields the highest sensitivity in functional (platelet activation) assays [10]. Besides the structure of the heparin used (e.g., chain length, degree of sulfation [11]), the immune response towards PF4/heparin complexes also depends on patient characteristics. For example, anti-PF4/heparin antibody formation appears to be

especially common after major surgery, with immunization rates as high as 30% to 75% when UFH is administered in association with orthopedic and cardiac surgery, respectively [12–14]. In contrast, antibodies occur less frequently in medical patients even when they receive UFH in higher (therapeutic) doses, such as for treatment of venous thrombosis [15]. It is known that complexes between PF4 and UFH, or between PF4 and LMWH, are formed optimally at certain stoichiometric concentrations in vitro [7,11,16]. However, little attention has been paid to consideration of those clinical conditions under which these stoichiometric ratios are reached in vivo. UFH and the different LMWHs represent complex heterogeneous and polydisperse mixtures of glycosaminoglycans. The reported molecular masses represent only the mean value and the characteristic range. Furthermore, the percentage of molecules containing the antithrombin (AT)-binding site ranges from 20–50% in UFH preparations [17] and from 8– 17% with the different LMWHs [18]. It is only this minor proportion of the AT-binding site-containing heparin molecules that are measured by the usual laboratory assays. However, heparin molecules of both high and low AT affinity can bind to other proteins, such as PF4. This binding of heparin molecules to other proteins, either intravascularly or within the s.c. tissue, reduces its anticoagulant bioavailability. While the bioavailability of the ATbinding molecules for UFH (only 20–30%) differs markedly from that of LMWH (~80–95%), the bioavailability of the non-AT-binding heparin molecules is less well-known. For this study we assumed that the bioavailability of heparin molecules at a given molecular weight would be the same, regardless of whether they contained the AT-binding side or not. In

Model to explain immunization in HIT contrast, fondaparinux, a pentasaccharide modeled after the AT-binding region of heparin, has a defined molecular mass of 1728 Da and binds tightly and nearly completely (98.6%) to AT and with a bioavailability approaching 100%. In this study we focus on the impact of three factors—type of heparin (UFH vs LMWH vs fondaparinux), plasma anticoagulant concentrations (prophylactic- vs. therapeutic-dose), and PF4 release/ availability (minor vs major platelet activation) as three risk factors affecting the PF4/polysaccharide stoichiometry and, therefore, the anti-PF4/heparin immune response. Based on in vitro studies on complex formation of PF4, UFH, LMWH and fondaparinux, and ex vivo studies on the concentrations reached after s.c. injection of the anticoagulants, we describe a model which provides some insights into why anti-PF4/heparin antibodies are formed less frequently during LMWH therapy in comparison to UFH therapy, and why these antibodies are most often generated in patients receiving these glycosaminoglycan anticoagulants following major surgery. We propose that the ratios of the concentrations of the particular anticoagulant and of PF4 under different clinical situations are of major importance for influencing the risk of developing HIT. Our model also leads to several testable hypotheses regarding likelihood of immunization in different clinical circumstances. Material and methods

213 (n = 16), the pharmacokinetics and pharmacodynamics of four different heparins with known specific activities (aXaU/mg), i.e. UFH, a medium-molecular-weight heparin (MMWH) with an average molecular weight of 10.5 kDa and a narrow MW range (9.5–11.5 kDa), and two LMWH preparations, certoparin and enoxaparin, were compared. After s.c. administration of 9000 aXaU of either heparin, the prolongation of the activated partial thromboplastin time (APTT), the anti-factor Xa (aXa) activity and the antithrombin (aIIa) activity were determined at 11 time points spread over 24 h after injection. We calculated the total weight concentrations from these ex vivo measured anticoagulant activities by means of the specific activities (IU/mg) as specified for each of the heparins. As a basis for this calculation, we assumed that both high and low AT affinity heparin molecules will be absorbed to similar extents and that their ratio will be maintained. In the case of fondaparinux, the plasma concentrations (µg/ mL) after s.c. injection of 2.5 mg (prophylaxis) and 7.5 mg

Table 1 Size of complexes of unfractionated heparin, low-molecular weight heparin, fondaparinux, with platelet factor 4, as measured by photon correlation spectroscopy UFH, MW 15,000 Da, 150 IU/mg IU

µg/mL

0 0.1 0.2 0.5 1 2 4 8 16

0 0.66 1.3 3.3 6.6 12.2 24.4 48.8 97.6

d hyd, nm (± SD) 4.9 19.8 23.6 56.8 29.5 24.9 21.7 4.9 5.7

(1.2) (6.4) (23.5) (50.1) (31.8) (22.5) (21.6) (0.9) (1.3)

Reviparin, MW 3900 Da, 105 aFXa-U/mg We used photon correlation spectroscopy (PCS) to determine the concentrations at which UFH, LMWH, and fondaparinux form complexes with PF4 (ChromaTec, Greifswald, Germany), as described [8]. A fixed concentration of filtered (0.22 µm membrane filter; MillexGP, Ireland) PF4 (10 µg/mL, final concentration) in buffer (0.05 M NaH2PO4, 2% NaN3, pH 7.5) was titrated with increasing amounts of UFH (mean MW, 15,000 Da, 150 IU/ mg; Braun, Melsungen, Germany), the LMWH, reviparin (Clivarin; Knoll, Ludwigshafen, Germany), mean MW 4000 Da, 105 antifactor Xa U/mg), and fondaparinux (MW 1728 Da, 700 anti-factor Xa U/mg; GlaxoSmithKline, London, UK). Particle size was determined by PCS (Zetasizer 3000HS; Malvern Instruments, UK). From these titrations, we determined the optimal ratio at which the largest complexes were formed, as well as the PF4polysaccharide ratios at which complexes reached 50% of maximum size. We then used these ratios to model the concentration ranges at which PF4/polysaccharide complexes form optimally at different PF4 concentrations. In separate experiments we used different concentrations of PF4 and heparin, but maintained the same PF4/heparin ratio, and showed that this did not affect binding of anti-PF4/heparin antibodies as long as the ratio was maintained (data not shown). The gravimetric concentrations of UFH and LMWH were calculated on the basis of anticoagulant concentrations found in the plasma after subcutaneous (sc) injection and the specific anti-factor Xa anticoagulant activities (aXaU/mg) of the corresponding heparins. For this, data from a pharmacokinetic study were used [19,20]. In this volunteer study

aFXa-U 0 0.025 0.05 0.1 0.2 0.4 0.8 1.6

c µg/mL 0 0.2 0.4 0.8 1.6 3.2 6.4 12.8

d hyd nm (± SD) 4.7 18.6 42.3 20.6 12.7 4.3 7.5 6.6

(1.3) (16.3) (21.0) (27.0) (3.7) (0.7) (2.9) (2.2)

Fondaparinux, MW 1728 Da µg/ml

c µg/mL

d hyd nm (± SD)

0 0.1 0.5 1 3 10 50 100 200

0 0.1 0.5 1 3 10 50 100 200

8.1 19.0 26.7 27.2 21.5 15.1 10.2 4.8 5.1

(1.7) (20.7) (0.1) (13.4) (0.1) (21.2) (14.6) (2.5) (3.4)

UFH = unfractionated heparin; MW = molecular weight; Da = Dalton; IU = international units; d hyd nm (±SD) = hydrodynamic diameter of particles in nm; SD = standard deviation. The data represent mean ± SD of n = 5–20 measurements.

214 (therapy) were used as reported in the “Summary of Products Characteristics” (state: 07.12.2006) [21]. PF4 concentrations released by certain numbers of platelets were taken from the literature [22,23].

Results Constituent ratios allowing PF4-polysaccharide complex formation We observed that optimal concentrations of UFH, LMWH, and fondaparinux for complex formation with 10 µg/mL PF4 differed greatly among these oligo- or polysaccharides (Table 1). For UFH we found the peak for complex formation at 3.3 µg/mL (= 0.5 U/mL), with 50% of maximal complexes formed at 1.3 µg/mL (lower concentration) and 6.6 µg/mL (higher concentration). For the LMWH, reviparin, the peak for formation of the largest complexes was observed at 0.4 µg/mL (= 0.05 U/mL), with 50% of maximal complex size formed at 0.2 µg/mL (lower concentration) and 0.8 µg/mL (upper concentration). For fondaparinux, we found complexes only half as large as those formed between UFH and PF4. However, these larger complexes were formed at a broad concentration range between 0.5 and 3.0 µg/mL, with 50% of maximal complex size formed at 0.1 µg/mL and 10 µg/mL (Table 1). Thus, whereas the lower and upper concentration boundaries at which large complexes were formed (as judged by 50% of maximal complex size) for UFH and LMWH differed by about four- to five-fold, the range was much larger for fondaparinux for which the corresponding lower and upper concentration boundaries differed by approximately 100-fold. We then used these concentrations to model the range of ratios at which PF4 and the various glycosaminoglycans (UFH, LMWH, fondaparinux) form clusters at different concentrations of each reactant (lines in Fig. 1). The scale of the figures was set to a maximum of 4 µg/mL to facilitate comparison between the different heparin types, although the concentration of LMWH when given in therapeutic dose is much higher (~10 µg/mL). In vivo concentrations of UFH, LMWH, and fondaparinux In a study of healthy volunteers, s.c. injection of 9000 aXaU UFH resulted in a mean peak concentration of 0.26 (±0.12) aXaU/mL (corresponding to ~ 1.46 g µg/mL total heparin); however, interindividual variations were large (Table 2). Especially in patients, even wider interindividual variations can be expected due to the pronounced nonspecific binding of UFH, e.g., to inflammatory proteins [24]. Due to its short half-life of about 1– 2 h, nearly no UFH was detectable 12 h after s.c. injection. The two LMWHs assessed (certoparin and enoxaparin) had specific aXa-activities of about 94 and 100 aXaU/mg, yielding calculated mean peak concentrations of 7.66 µg/mL and 8.20 µg/ mL for 9000 aFXa-U (given by s.c. administration) of certoparin and enoxaparin, respectively. This corresponds approximately to 2.5 µg/mL and 3.6 µg/mL at a prophylactic dose of 3000 aFXa-U and 4000 aFXa-U, respectively (Table 2). Calculation of these concentrations was feasible, as LMWHs show a better dose– effect relationship than UFH [25] and the linearity for certoparin has been established [26]. These LMWHs have a longer half-life than UFH and considerable aFXa activities were still measurable after 10 h and after 24 h (Table 2). The calculated range of total UFH and LMWH concentrations reached in plasma during prophylactic-dose anticoagulation are given as shaded area in Fig. 1. Fondaparinux administered by the s.c. route is completely and rapidly absorbed (bioavailability ~100%). After a single s.c. injection

A. Greinacher et al. of 2.5 mg to young, healthy subjects, a mean peak plasma concentration of 0.34 µg/ml is observed 2 h after s.c. injection which then decreases with a half-life of 17 h (“Summary of Products Characteristics”[21]). Following once daily dosing, steady state of plasma levels is obtained after 3 to 4 days with a 1.3-fold increase in the mean peak concentration and the mean area under the curve. In patients undergoing hip replacement surgery receiving 2.5 mg sc/ day, the mean maximal concentration is 0.39 µg/mL decreasing to 0.14 µg/mL after 24 h. In hip fracture patients (associated with their increased age), the fondaparinux steady state maximal plasma concentration is 0.50 µg/mL, decreasing to 0.19 µg/mL after 24 h. However, in plasma, about 98.6% of fondaparinux molecules are bound to AT with high affinity [27], which results in the plasma concentrations of free fondaparinux being less than 0.01 µg/mL. When given in therapeutic doses, the concentrations of UFH are estimated to range from about 1 to 4.6 µg/mL (i.e. about 0.2–0.9 IU/ml) as based on 1.5–2.5-fold prolongation of APTT, or 1.3–3.5 µg/mL as based on the corresponding recommended target activities of 0.3–0.7 IU/mL [28,29] (upper shaded area in Fig. 1a). In the case of LMWH, peak aXa-activities with therapeutic dosing vary somewhat between different LMWHs, but all correspond to a concentration of approximately 9–11 µg/ mL (i.e., above the range shown in Fig. 1b). The mean steady state maximal concentration of fondaparinux after once daily application of 5 mg is 1.41 µg/mL, decreasing to 0.52 µg/mL after 24 h [30], corresponding to 0.034 and 0.007 µg/mL free fondaparinux, respectively. Plasma concentrations of PF4 After a bolus of UFH, plasma concentrations of PF4 reach approximately 8 nM [22], which corresponds to 0.25 µg/mL PF4 (leftmost arrows in Figs. 1a and 1b [the MW of the PF4 tetramer is 31,600]). These quantities reflect the pool of PF4 released from endothelial or platelet surface reservoirs after heparin application, but do not consider PF4 released from activated platelets. When platelets are activated, they release PF4 from their alpha granules, and complete activation of 250 × 109/L platelets will generate a plasma PF4 concentration of about 200 nM PF4 [23] which corresponds to 6–7 µg/mL. Fig. 1 therefore shows PF4 concentrations up to 5 µg/mL (rightmost arrows in Figs. 1a and 1b), i.e., a situation of near-maximal platelet activation and PF4 release. Figs. 1a and 1b indicate by arrows the approximate amount of PF4 which will be released by activation of certain numbers of platelets. Model of PF4-UFH, -LMWH, and -fondaparinux complex formation As shown in Fig. 1 the main features of our model are: - with therapeutic-dose UFH, complex formation with PF4 should be expected primarily in situations of major platelet activation resulting in very high concentrations of PF4. - with prophylactic-dose UFH, complex formation with PF4 should be expected primarily in situations of moderate to major platelet activation resulting in increased concentrations of PF4. - for a given amount of PF4 availability, complex formation with PF4 is more likely to occur with prophylactic-dose UFH compared with therapeutic-dose UFH. - therapeutic-dose LMWH is too high for optimal formation of PF4/LMWH complexes. - in prophylactic-dose administration of LMWH, complex formation with PF4 should primarily be expected in case of major platelet activation resulting in high concentrations

Model to explain immunization in HIT

215

Figure 1 a: Optimal formation of PF4-UFH complexes at differing plasma concentrations of UFH and PF4. The X-axis shows platelet factor 4 (PF4) concentrations. At the top of the figure, the approximate amounts of PF4 maximally released from varying degrees of platelet activation are shown. The Y-axis gives concentrations of unfractionated heparin (UFH) as calculated from an ex vivo study. These concentration ranges during prophylactic-dose and therapeutic-dose UFH application are given as shaded areas with the most often reached concentrations indicated by darker green and less frequently reached concentrations indicated in lighter green. The range of concentration ratios at which optimal (dotted line) or 50% of optimal (lower and upper solid line) complex formation occurred in vitro are superimposed. This shows that with therapeutic-dose UFH administration, complex formation between PF4 and UFH should be expected primarily in situations of major platelet activation resulting in high concentrations of PF4. In prophylactic-dose administration of UFH, complex formation with PF4 is more likely to occur compared with therapeutic-dose UFH, although also more often in the case of increased PF4 concentrations. Please note that the range of optimal concentrations for complex formation do not take into account molecular weight-dependent differences in bioavailability of UFH molecules. It is therefore likely that the optimal in vivo ratios differ somewhat from the optimal in vitro ratios, with complex formation occurring at even lower UFH plasma concentrations. b: Optimal formation of PF4-LMWH complexes at differing plasma concentrations of LMWH and PF4. The X-axis shows platelet factor 4 (PF4) concentrations. At the top of the figure, the approximate amounts of PF4 maximally released from varying degrees of platelet activation are shown. The Y-axis gives concentrations of low-molecular-weight heparin (LMWH). The expected concentration range during prophylacticdose LMWH application is given as shaded area with the most often reached concentrations indicated with darker green, and the less frequently reached concentrations indicated in lighter green. The concentration range for therapeutic-dose LMWH is at about 10 µg/mL (above the range shown in the figure). The range of concentration ratios between PF4 and LMWH at which optimal (dotted line) or 50% of optimal (lower and upper solid line) complex formation occurred superimposed. This shows that in therapeutic-dose LMWH little if any complexes are formed. In prophylactic-dose LMWH complex formation with PF4 should primarily expected in case of major platelet activation resulting in high concentrations of PF4 and during time periods when LMWH reaches its lowest concentration before the subsequent dose is given (compare with Table 2). Nevertheless, it is likely that the in vivo concentrations— particularly for therapeutic-dose LMWH are far greater than the concentrations needed for optimal complex formation. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

216

A. Greinacher et al.

Table 2

Plasma concentrations of different heparins obtained in a study with healthy volunteers Measured in 16 healthy volunteers after single s.c. injection

Calculated

Heparin type applied dose

UFH 9000 aFXa-U = 51.4 mg

MMWH 9000 aFXa-U = 56.6 mg

Certoparin 9000 aFXa-U 95.7 mg

Enoxaparin 9000 aFXa-U 90.0 mg

Certoparin 3000 aFXa-U 31.9 mg

Enoxaparin 4000 aFXa-U 40.0 mg

aFXa-U/mg µg/aFXa-U MW Cmax aXa aXaU/ml aXaU/ml SD µg/ml c(5 h) aXa aFXa-U/ml µg/ml c(10 h) aXa aFXa-U/ml µg/ml c(24 h) aXa aFXa-U/ml µg/ml Time parameters Tmax t(1/2)

175 5.72 13000

159 6.29 10500

94 10.67 6000

100 10.00 4500

94 10.67 6000

100 10.00 4500

0.26 0.12 1.46

0.41 0.14 2.61

0.72 0.17 7.66

0.82 0.18 8.20

0.24 0.17 2.55

0.36 0.18 3.64

0.17 0.98

0.35 2.19

0.63 6.69

0.71 7.14

0.21 2.23

0.32 3.17

0.03 0.18

0.11 0.67

0.29 3.10

0.36 3.64

0.10 1.03

0.16 1.62

0.00 0.00

0.00 0.00

0.032 0.34

0.042 0.42

0.01 0.11

0.02 0.19

2.84 1.92

3.63 2.28

3.59 3.56

3.69 4.21

3.59 3.56

3.69 4.21

s.c. = subcutaneous injection; UFH = unfractionated heparin; aXa-U = anti-factor Xa units; MMWH = median molecular weight heparin; U = units; Cmax = peak concentration; Tmax = time until Cmax; t(1/2) = elimination half-life; c(5 h) = concentration 5 h after injection; c(10 h) = concentration 10 h after injection; c(24 h) = concentration 24 h after injection.

of PF4 and during time periods when LMWH reaches its lowest concentration, i.e., before the subsequent dose is given.

Discussion We present a model which provides a framework to explain empirical observations that LMWH induces less anti-PF4/heparin antibodies than does UFH [4,5], and that anti-PF4/heparin antibodies are found more often in patients undergoing major surgery than in medical patients [5]. We suggest that in most clinical settings with little platelet activation, UFH or LMWH are present in considerably higher concentrations than required for optimal complex formation with PF4. In major surgery there is greater release of PF4 from activated platelets and, as shown in Fig. 1, this increases the chance that complex formation would occur at a given concentration of UFH or LMWH. Further, the higher concentrations of PF4 will also lead to more antigen being formed. To define the range of UFH and LMWH concentrations reached in vivo after s.c. injection, we measured the ex vivo anticoagulant activities of UFH and LMWH in a pharmacokinetic study and calculated the concentrations by means of the specific anticoagulant activities (IU/mg) of the used heparins. Our model transferred the in vitro data of the PCS experiments to the in vivo situation

assuming that the molecular weight distribution of UFH molecules being absorbed into the vascular system after s.c. injection would be the same as in the UFH preparation. Under these assumptions, optimal antigen formation at prophylactic-dose UFH concentration would occur at a concentration of about 1 µg/mL PF4—corresponding to the release of PF4 from about 40 to 50 × 109 /L platelets (Fig. 1). This is very close to the usual decrease in platelet counts seen after major surgery [2]. However, the molecular composition of UFH present in plasma differs from that in vitro. In particular, the very large molecules do not reach the circulation (due to sequestration within the s.c. tissues), so that the mean molecular weight of the UFH molecules being available in vivo for binding to PF4 is likely to be lower than that of UFH used in vitro [31]. Correspondingly, the optimal concentrations of UFH (measured as µg/mL) for complex formations in vivo are likely to be somewhat lower than the ones given in the figure which were calculated from the in vitro PCS experiments, as in vivo relatively more molecules of smaller molecular weight are present per µg (i.e. the molar concentration is higher). However, this would not change the basic implications of the model, namely that lower heparin concentrations and higher PF4 concentrations provide the optimal concentration range for PF4/heparin complex formation.

Model to explain immunization in HIT According to our model, patients undergoing minor surgery should generate a lesser immune response as compared to patients undergoing major surgery, despite receiving the same UFH in the same dose. Indeed, this is precisely what we recently observed in a randomized trial of post-trauma thromboprophylaxis in which the incidence of antiPF4/heparin antibodies was 5-fold higher in patients undergoing major surgery as compared to minor surgery [32]. Our model also indicates that immunization should be more frequent in case of prophylactic-dose UFH as compared to therapeutic-dose UFH, at least in situations of comparable PF4 concentrations. In fact, HIT was more frequent with prophylactic-dose UFH compared with therapeutic-dose UFH in a prospective study in medical patients [33], with HIT occurring in 5 (1.4%) of 360 patients receiving UFH in prophylactic dose but in none of the 298 patients receiving UFH in therapeutic dose. This comparison is especially noteworthy given that both prophylactic-and therapeutic-dose UFH was administered by s.c. injection in this study [33]. It would be interesting to compare the immunization rates in patient groups receiving the same UFH dose by either i.v. and s.c. routes: based on our model, during s.c. application, more antiPF4/heparin antibodies should occur, as many UFH molecules are already retained by the subcutaneous tissue barrier, leading to a greater chance of achieving stoichiometrically optimal concentrations of PF4 and UFH. This observation runs counter to the intuitive (and common) assumption that higher doses of anticoagulant should be more likely to trigger an immune response. However, it is wellknown that in vitro at high concentrations of heparin (10–100 U/mL), the effects of HIT serum on platelet activation are inhibited, a laboratory phenomenon that has been used for more than two decades to optimize the specificity of diagnostic testing for HIT antibodies [34]. There are two major patients groups known to have very high frequencies of anti-PF4/heparin antibody formation even though very high doses of heparin are given. One is the situation of patients undergoing cardiopulmonary bypass (CPB) surgery. However, although very high PF4 levels are likely to occur, it is unknown precisely what heparin concentrations trigger immunization, as the heparin levels range from very high (during CPB itself) to very low (post-protamine reversal of heparin). The second group comprises patients on cardiac assist device, who also have high frequencies of anti-PF4/ heparin antibodies reported ~ 40–50% [35,36]. However, in these patients, major platelet activation occurs due to blood-device interactions at the same time that they are receiving therapeutic doses of

217 heparin. This fits our model, and would be expected to be complicated frequently by antibody formation, as both PF4 and UFH are present in high concentrations (Fig. 1). In contrast to UFH, LMWH has a bioavailability exceeding 80%. Although the ex vivo ratio of the aXa- to the aIIa-activity is somewhat higher than that in vitro [31], the in vivo molecular weight distribution of LMWH resembles more closely the in vitro molecular weight distribution and the molar concentrations remain quite similar. Accordingly, for LMWH, the in vitro PCS experiments should apply more directly to the in vivo situation. Furthermore, in our in vitro experiments we used a LMWH, reviparin, with a low median molecular weight and a narrow range of molecular weight distribution. Therefore, the in vitro data obtained with this LMWH likely should also reflect the in vivo situation reached with other LMWHs of the same or even higher mean molecular weight reasonably. According to our model, maximal LMWH concentrations that are reached using prophylactic doses are too high to allow optimal formation of complexes over a broad range of PF4 concentrations (Fig. 1). According to our model, LMWH would not be expected to immunize well at all unless massive amounts of PF4 were present. However, it is important to consider that LMWH concentrations decline between successive LMWH injections given every 12 h (North America) or 24 h (Europe), and at the concentration nadir is most likely to reach the range that could favor complex formation, and resulting immunization and, perhaps, also the development of HIT. The model further suggests that antibody formation would be even less likely with therapeutic-dose, compared with prophylactic-dose LMWH. As with UFH, all situations causing increased release of PF4 should increase the immunization rate. Interestingly, we found anti-PF4/heparin antibodies occurring more frequently in patients receiving LMWH after knee surgery as compared to hip surgery (p b 0.001) [37]. In knee surgery, there is evidence that the prothrombotic risk seems to be higher than in hip surgery: In a recent metaanalysis of 10 venographic studies involving 5796 patients, the prevalence of asymptomatic DVT after total hip replacement was 13.2% (95%CI, 12.2%–14.2%) and after total knee replacement it was 38.1% (95%CI, 35.5%–40.8%) [38]. It therefore seems plausible that more platelets are activated, and PF4 released, in knee than in hip surgery. In our model, such a difference in PF4 availability between these two orthopedic surgery populations would account for a greater risk of immunization in knee surgery patients. This could be tested by measuring PF4 levels in blood samples of these two patient groups.

218 Fondaparinux presents the opposite situation than LMWH, as it is more likely to be present at concentrations below the optimal range needed for forming complexes with PF4. This is because fondaparinux is given at a much lower dose than UFH and LMWH, whether considered either in molar or weight-based terms. Furthermore, more than 98% of the molecules are tightly bound to AT, whereas the fractions of UFH and LMWH with high affinity for ATcan be displaced from binding to AT by competing proteins such as PF4 [39]. Thus, the total amount of fondaparinux available for binding to PF4 in vivo should be much lower than that anticipated as being optimal from in vitro binding studies. Nevertheless, fondaparinux binds to PF4 and can induce the epitope on PF4 leading to anti-PF4/ heparin antibody formation [40,41]; however, the number of PF4/fondaparinux complexes is much lower and their size much smaller than with UFH or LMWH [8]. These few PF4/fondaparinux complexes do not support formation of sufficient immune complexes to permit platelet activation in the presence of anti-PF4/heparin antibodies. Indeed, only one case of HIT has been reported among fondaparinux-treated patients. And this case represented an unusual situation in which the antibodies generated were capable of activating platelets in the absence of any glycosaminoglycan, fondaparinux included, and is most likely an example for a drug induced, HIT-like autoimmune reaction [42]. However, to date, no data on anti-PF4/heparin antibody formation during treatment with therapeutic-dose fondaparinux have been reported. Our model has several limitations. First, we assessed the stoichiometry of PF4-polysaccharide interactions under in vitro conditions. After s.c. injection, heparin must cross the s.c. barrier, and considerable quantities-especially of UFH—fail to be absorbed into the intravascular compartment. We do not know exactly which subfractions of UFH (and, to a lesser extent, LMWH) will penetrate into the intravascular compartment. Furthermore, the PCS experiments required a plasma-free setting. In plasma, LMWH and especially UFH will also bind to proteins other than PF4, thus in vivo complex formation between UFH, LMWH, and PF4 may occur at somewhat lower PF4 concentrations than anticipated by the model. As outlined before, most of the fondaparinux would bind to AT and will therefore not be available for binding to PF4. Furthermore, the model is based on plasma levels of PF4 and heparin. Very likely the concentrations of these two constituents of the PF4/glycosaminoglycan complexes differ within certain microenvironments (e.g., the surfaces of platelets, endothelial cells, or immune competent cells). These effects

A. Greinacher et al. could be relevant to formation of PF4/heparin complexes and therefore for the anti-PF4/heparin immune response. The model also does not take into account the differences between LMWH preparations. E.g., the injection of 3500 aXa-IU of bemiparin, the smallest LMWH, results in peak activity of 0.32 aXa-IU/ml, whereas 4500 aXa-IU of tinzaparin, the largest LMWH, result in 0.24 aXa-IU [43]. Also, pharmacokinetics of LMWHs differ dependent on renal function. Especially in elderly patients the concentrations of LMWHs might be higher than the ones found in healthy volunteer studies [44]. Unfortunately, it is not feasible to isolate the circulating UFH and LMWH molecules from the circulation in high enough quantities to further validate our model, e.g. by performing the PCS experiments or PF4/heparin EIA studies with these ex vivo purified molecules. Therefore, the concentration ranges given in our model should be seen as rough approximations. Nevertheless, when we applied our model to available data on the frequencies of anti-PF4/heparin antibody formation and development of clinical HIT in different patient populations, the empirically obtained in vivo data appear to fit the model. In summary, based on in vitro data of complex formation between PF4 and heparin and ex vivo data on the in vivo concentrations and pharmacokinetics of these anticoagulants, we present a model which takes into account the different plasma concentrations of the glycosaminoglycan anticoagulants (UFH, LMWH, or fondaparinux) and PF4, in terms of explaining differing immunization rates under different clinical situations. This model appears to explain empirical data on immunization rates available from clinical studies assessing the frequency of anti-PF4/heparin antibodies. The model might also be used to estimate the risk of inducing anti-PF4/ heparin antibodies of other new sulfated glycanbased anticoagulants that interact with PF4.

Acknowledgements This work was supported by Deutsche Forschungsgemeinschaft, Gr 1096/2-4; German Federal Ministry for Education and Research (CAN04/006; NBL3 program, reference 01-ZZ0403), Forschungsverbund Molekulare Medizin, Projekt Anschubförderung “Untersuchungen zur klinischen Relevanz der Immunglobulinklassen IgG,-A,-M bei Heparin-induzierter Thrombozytopenie (HIT) und weitere Charakterisierung der Antikörper”; and is part of the Department of Cardiovascular Medicine of the Medical Faculty at the Ernst-Moritz-Arndt-University Greifswald; im Rahmen des Fellows-Programms “Life Sciences” 2007 des Alfried Krupp

Model to explain immunization in HIT Wissenschaftskollegs Greifswald und von der Alfried Krupp von Bohlen und Halbach-Stiftung.

References [1] Greinacher A, Amiral J, Dummel V, Vissac A, Kiefel V, Mueller-Eckhardt C. Laboratory diagnosis of heparin-associated thrombocytopenia and comparison of platelet aggregation test, heparin-induced platelet activation test, and platelet factor 4/heparin enzyme-linked immunosorbent assay. Transfusion 1994;34:381–5. [2] Warkentin TE, Levine MN, Hirsh J, Horsewood P, Roberts RS, Gent M, et al. Heparin-induced thrombocytopenia in patients treated with low-molecular-weight heparin or unfractionated heparin. N Engl J Med 1995;332:1330–5. [3] Greinacher A, Eichler P, Lietz T, Warkentin TE. Replacement of unfractionated heparin by low-molecular-weight heparin for postorthopedic surgery antithrombotic prophylaxis lowers the overall risk of symptomatic thrombosis because of a lower frequency of heparin-induced thrombocytopenia. Blood 2005;106:2921–2. [4] Martel N, Lee J, Wells PS. Risk for heparin-induced thrombocytopenia with unfractionated and low-molecular-weight heparin thromboprophylaxis: a meta-analysis. Blood 2005;106:2710–5. [5] Warkentin TE, Sheppard JA, Sigouin CS, Kohlmann T, Eichler P, Greinacher A. Gender imbalance and risk factor interactions in heparin-induced thrombocytopenia. Blood 2006;108:2937–41. [6] Rauova L, Zhai L, Kowalska MA, Arepally GM, Cines DB, Poncz M. Role of platelet surface PF4 antigenic complexes in heparin-induced thrombocytopenia pathogenesis: diagnostic and therapeutic implications. Blood 2006;107:2346–53. [7] Rauova L, Poncz M, McKenzie SE, Reilly MP, Arepally G, Weisel JW, et al. Ultralarge complexes of PF4 and heparin are central to the pathogenesis of heparin-induced thrombocytopenia 2. Blood 2005;105:131–8. [8] Greinacher A, Gopinadhan M, Gunther JU, Omer-Adam MA, Strobel U, Warkentin TE, et al. Close approximation of two platelet factor 4 tetramers by charge neutralization forms the antigens recognized by HIT antibodies. Arterioscler Thromb Vasc Biol 2006;26:2386–93. [9] Warkentin TE, Greinacher A. Laboratory testing for heparininduced thrombocytopenia. In: Warkentin TE, Greinacher A, editors. Heparin-induced thrombocytopenia. 4 ed. New York: Informa Healthcare USA; 2007. p. 227–60. [10] Greinacher A, Feigl M, Mueller-Eckhardt C. Crossreactivity studies between sera of patients with heparin associated thrombocytopenia and a new low molecular weight heparin, reviparin. Thromb Haemost 1994;72:644–5. [11] Greinacher A, Alban S, Dummel V, Franz G, Mueller-Eckhardt C. Characterization of the structural requirements for a carbohydrate based anticoagulant with a reduced risk of inducing the immunological type of heparin-associated thrombocytopenia. Thromb Haemost 1995;74:886–92. [12] Warkentin TE, Sheppard JA, Horsewood P, Simpson PJ, Moore JC, Kelton JG. Impact of the patient population on the risk for heparin-induced thrombocytopenia. Blood 2000;96:1703–8. [13] Warkentin TE, Sheppard JA, Moore JC, Moore KM, Sigouin CS, Kelton JG. Laboratory testing for the antibodies that cause heparin-induced thrombocytopenia: how much class do we need? J Lab Clin Med 2005;146:341–6. [14] Lee D, Warkentin T. Frequency of heparin induced thrombocytopenia. In: Warkentin T, Greinacher A, editors. Heparininduced thrombocytopenia. 4 ed. New York: Informa Healthcare; 2007. p. 67–116.

219 [15] Lindhoff-Last E, Nakov R, Misselwitz F, Breddin HK, Bauersachs R. Incidence and clinical relevance of heparininduced antibodies in patients with deep vein thrombosis treated with unfractionated or low-molecular-weight heparin. Br J Haematol 2002;118:1137–42. [16] Amiral J, Bridey F, Wolf M, Boyer-Neumann C, Fressinaud E, Vissac AM, et al. Antibodies to macromolecular platelet factor 4-heparin complexes in heparin-induced thrombocytopenia: a study of 44 cases. Thromb Haemost 1995;73:21–8. [17] Linhardt RJ, Gunay NS. Production and chemical processing of low molecular weight heparins. Semin Thromb Hemost 1999;25(Suppl 3):5–16. [18] Young E, Wells P, Holloway S, Weitz J, Hirsh J. Ex-vivo and invitro evidence that low molecular weight heparins exhibit less binding to plasma proteins than unfractionated heparin. Thromb Haemost 1994;71:300–4. [19] Alban S, Gastpar R. Plasma levels of total and free tissue factor pathway inhibitor (TFPI) as individual pharmacological parameters of various heparins. Thromb Haemost 2001;85:824–9. [20] Al Dieri R, Alban S, Beguin S, Hemker HC. Fixed dosage of low-molecular-weight heparins causes large individual variation in coagulability, only partly correlated to body weight. J Thromb Haemost 2006;4:83–9. [21] Summary of Products Characteristics http://www.emea. europa.eu/humandocs/PDFs/EPAR/arixtra/H-403-PI-en.pdf. [22] Dawes J, Pumphrey CW, McLaren KM, Prowse CV, Pepper DS. The in vivo release of human platelet factor 4 by heparin. Thromb Res 1982;27:65–76. [23] Horne MK. The effect of secreted heparin-binding proteins on heparin binding to platelets. Thromb Res 1993;70:91–8. [24] Hirsh J, Warkentin TE, Shaughnessy SG, Anand SS, Halperin JL, Raschke R, et al. Heparin and low-molecular-weight heparin: mechanisms of action, pharmacokinetics, dosing, monitoring, efficacy, and safety. Chest 2001;119(1 Suppl):64S–94S. [25] Hirsh J, Raschke R. Heparin and low-molecular-weight heparin: the Seventh ACCP Conference on Antithrombotic and Thrombolytic Therapy. Chest 2004;126(3 Suppl):188S–203S. [26] Alban S, Wolf H, Fareed J, Welzel D, Franz G. Correlation between the anticoagulant activity of a LMWH and the TFPIplasma concentration in dependence on dosage and mode of application. Ann Hematol Suppl II 1997;298 [Abstract]. [27] Paolucci F, Clavies MC, Donat F, Necciari J. Fondaparinux sodium mechanism of action: identification of specific binding to purified and human plasma-derived proteins. Clin Pharmacokinet 2002;41(Suppl 2):11–8. [28] Smythe MA, Mattson JC, Koerber JM. The heparin anti-Xa therapeutic range: are we there yet? Chest 2002;121:303–4. [29] Eikelboom JW, Hirsh J. Monitoring unfractionated heparin with the aPTT: time for a fresh look. Thromb Haemost 2006;96:547–52. [30] Donat F, Duret JP, Santoni A, Cariou R, Necciari J, Magnani H, et al. The pharmacokinetics of fondaparinux sodium in healthy volunteers. Clin Pharmacokinet 2002;41(Suppl 2):1–9. [31] Schauerte A, Welzel D, Franz G, Alban S. Is there any accordance between the in vitro and ex vivo anti factor Xa/ anti-thrombin ratio of UFH and LMWHs? Ann Hematol 1998;76(Suppl I):V59. [32] Lubenow N, Hinz P, Lietz T, Ladwig A, Jünger M, Ekkernkamp A, et al. Clinical HIT and HIT-antibody seroconversion in trauma patients receiving unfractionated heparin vs. certoparin: a randomised, double-blind study. Transfus Med Hemother 2006;33(suppl 1):7 [Abstract]. [33] Girolami B, Prandoni P, Stefani PM, Tanduo C, Sabbion P, Eichler P, et al. The incidence of heparin-induced thrombocytopenia in hospitalized medical patients treated with subcutaneous unfractionated heparin: a prospective cohort study. Blood 2003;101:2955–9.

220 [34] Sheridan D, Carter C, Kelton JG. A diagnostic test for heparin-induced thrombocytopenia. Blood 1986;67:27–30. [35] Schenk S, El-Banayosy A, Prohaska W, Arusoglu L, Morshuis M, Koester-Eiserfunke W, et al. Heparin-induced thrombocytopenia in patients on mechanical circulatory support. J Thorac Cardiovasc Surg 2006;131:1373–81. [36] Schenk S, El Banayosy A, Morshuis M, Arusoglu L, Eichler P, Lubenow N, et al. IgG classification of anti-PF4/heparin antibodies to identify patients with heparin-induced thrombocytopenia during mechanical circulatory support. J Thromb Haemost 2007;5:235–41. [37] Warkentin T, Cook R, Greinacher A, Marder V, Kelton JG. Factors influencing the anti-platelet factor 4/heparin antibody (anti-PF4/h ab) immune response: an analysis of four randomized controlled trials (RCTs) comparing enoxaparin (enox) to fondaparinux (fonda). Presented at ISTH-meeting Geneva; 2007. Abstract. [38] Quinlan DJ, Eikelboom JW, Dahl OE, Eriksson BI, Sidhu PS, Hirsh J. Association between asymptomatic deep-vein thrombosis detected by venography and symptomatic venous thromboembolism in patients undergoing elective hip or knee surgery. J Thromb Haemost 2007 [Electronic publication]. [39] Bauersachs R, Alban S. Perioperative bridging with fondaparinux in a woman with antithrombin deficiency. Thromb Haemost 2007;97:498–9.

A. Greinacher et al. [40] Warkentin TE, Greinacher A. Unfractionated, low-molecularweight heparin and risk of HIT: are medical patients different? [Inside Blood Commentary on BELZONI Investigators Group. The incidence of heparin-induced thrombocytopenia in medical patients treated with low molecular weight heparin. A prospective cohort study. Blood 2005;106:2931–2. [41] Pouplard C, Couvret C, Regina S, Gruel Y. Development of antibodies specific to polyanion-modified platelet factor 4 during treatment with fondaparinux. J Thromb Haemost 2005;3:2813–5. [42] Warkentin TE, Maurer BT, Aster RH. Heparin-induced thrombocytopenia associated with fondaparinux. N Engl J Med 2007;356:2653–5. [43] Depasse F, Gonzalez de Suso MJ, Lagoutte I, Fontcuberta J, Borrell M, Samama MM. Comparative study of the pharmacokinetic profiles of two LMWHs-bemiparin (3500 IU, anti-Xa) and tinzaparin (4500 IU, anti-Xa)-administered subcutaneously to healthy male volunteers. Thromb Res 2003;109:109–17. [44] Duplaga BA, Rivers CW, Nutescu E. Dosing and monitoring of low-molecular-weight heparins in special populations. Pharmacotherapy 2001;21:218–34.