Anti-PF4/heparin antibodies and venous graft occlusion in postcoronary artery bypass surgery patients randomized to postoperative unfractionated heparin or fondaparinux thromboprophylaxis


Correspondence: Theodore E. Warkentin, Hamilton Regional Laboratory Medicine Program, Hamilton Health Sciences, Hamilton General Site, Rm 1-270B, 237 Barton St. E., Hamilton, ON, L8L 2X2, Canada.

Tel.: +1 905 527 0271 ext. 46139; fax: +1 905 577 1421.




Anti-PF4/heparin antibodies are frequently generated after coronary artery bypass grafting (CABG) surgery, with platelet-activating IgG implicated in heparin-induced thrombocytopenia (HIT). It is controversial whether non-platelet-activating antibodies are associated with thrombosis.


To determine in post-CABG patients whether thromboprophylaxis using fondaparinux vs. unfractionated heparin (UFH) reduces the frequency of anti-PF4/heparin antibodies, and whether anti-PF4/heparin antibodies are associated with early graft occlusion.


In a pre-planned secondary analysis of a randomized control trial (RCT) comparing fondaparinux vs. UFH thromboprophylaxis post-CABG, we determined the frequency of anti-PF4/heparin antibody formation by solid-phase enzyme-immunoassay (EIA) and of platelet-activating antibodies by serotonin-release assay (SRA); the SRA and fluid-phase EIA were used to assess fondaparinux cross-reactivity. We also examined whether anti-PF4/heparin antibodies were associated with early arterial or venous graft occlusion (6-week CT angiography).


We found no significant difference in the frequency of antibody formation between patients who received fondaparinux vs. UFH (65.3% vs. 46.0%; = 0.069), and no significant fondaparinux cross-reactivity. Venous graft occlusion(s) occurred in 6/26 patients who formed ‘strong’ IgG antibodies (≥ 1.0 optical density [OD] units and ≥ 2× baseline) vs. 3/66 who did not (= 0.0139). In both unadjusted and adjusted analyses, strong postoperative (but not pre-operative) anti-PF4/heparin IgG responses were associated with a markedly increased risk of early venous (but not arterial) graft occlusion (adjusted OR, 9.25 [95% CI, 1.73, 49.43]; = 0.0093); notably, none of the three SRA-positive patients developed a venous graft occlusion.


Fondaparinux vs. UFH thromboprophylaxis postCABG does not reduce anti-PF4/heparin antibody formation. Non-platelet-activating anti-PF4/heparin IgG antibodies generated post operatively are associated with early venous graft occlusion.


Cardiac surgery, which is performed with unfractionated heparin (UFH), is associated with high frequencies (30–80%) of postoperative anti-platelet factor 4 (PF4)/heparin antibody formation [1-4]. It is generally assumed that antibody formation is primarily related to intra-operative rather than postoperative administration of UFH because antibody frequencies are high even when postoperative thromboprophylaxis with UFH is not given.

Recently, we found in a pilot randomized controlled trial (RCT) of 100 patients undergoing coronary-artery bypass grafting (CABG), in which patients received either UFH or fondaparinux for post operative thromboprophylaxis, similar frequencies of arterial and venous graft occlusion between the two study arms [5]. We now report the results of a preplanned substudy of anti-PF4/heparin antibody formation. Our substudy had two aims: first, to determine whether antibody formation, and associated risk of heparin-induced thrombocytopenia (HIT), was lower with fondaparinux compared with UFH; and second, to evaluate whether anti-PF4/heparin antibodies, already present before surgery, or generated postoperatively, predicted venous or arterial graft occlusion.

Methods and patients

Pilot RCT

‘FondaCABG’ was a single-center, blinded, pilot RCT that compared fondaparinux with UFH for the prevention of graft failure and major cardiovascular events in post-CABG patients (Clinical Trial Registration Information: identifier NCT00474591). The study was approved by the Hamilton Health Sciences Research Ethics Board; written informed consent was obtained from each study patient.

Details of the study design have been reported previously [5]. Briefly, patients were eligible if ≥18 years of age and were scheduled for on-pump CABG surgery with at least one free aorto-coronary bypass graft. Exclusion criteria included the need for long-term anticoagulation, heparin/fondaparinux allergy, unacceptably high bleeding risk or contraindication to postoperative computed tomography (CT) angiography. Patients randomized to fondaparinux received 2.5 mg subcutaneously (s.c.) once daily and a heparin placebo s.c. twice daily, whereas those randomized to heparin received a fondaparinux placebo s.c. once daily and heparin 5000 U s.c. twice daily (double-dummy design). Injections were started once chest drains were removed and were continued throughout hospitalization. After hospital discharge, the fondaparinux group received daily fondaparinux 2.5 mg s.c. once daily until day 30 postsurgery whereas the heparin group received fondaparinux placebo s.c. until day 30 postsurgery.

The primary efficacy outcome (main study) was the composite, graft patency (per cardiac CT angiography), death, myocardial infarction and stroke. Safety outcomes included bleeding events and re-exploration for bleeding. Graft patency was assessed at 47 days (mean).

Blood sampling for anti-PF4/heparin antibodies

A total of 928 blood samples were collected (mean, 9.3 samples/patient), as follows (day of surgery = day 0): pre-operatively; daily postoperatively until discharge (day of last sample pre-discharge, median day 5 [IQR, 5, 6; range 3–15]); at the postdischarge follow-up visit (median day 48 [IQR, 40, 56]; range 29–116). Plasma was obtained from EDTA-anticoagulated tubes, except for preoperative and predischarge samples, which were serum (serum and EDTA-plasma yield similar anti-PF4/heparin antibody test results [6]).

Anti-PF4/heparin antibodies

We performed the platelet serotonin-release assay (SRA) [7] and class-specific anti-PF4/heparin EIAs (IgG, IgA and IgM) [8]. We also performed high heparin inhibition (using 100 IU mL−1 UFH). Serial plasma sets were tested on the same day to minimize inter-assay variability.

Classification of immune responses

We evaluated anti-PF4/heparin responses hierarchically, as follows [9]:

  • Group A (platelet-activating IgG) comprised patients who developed (vs. pre-operative baseline) a positive SRA (≥20% serotonin-release at 0.1–0.3 IU mL−1 UFH) with inhibition at 100 IU mL−1 UFH.
  • Group B (‘strong’ non-platelet-activating anti-PF4/heparin IgG): ≥ 2-fold increase in reactivity (by optical density [OD] units) in the EIA-IgG (vs. preoperative baseline), with a maximum OD ≥ 1.00 units and inhibition > 50% by high-dose heparin.
  • Group C (‘weak’ non-platelet-activating anti-PF4/heparin IgG): ≥ 2-fold increase in reactivity (by OD units) in the EIA-IgG (vs. pre-operative baseline), but maximum OD 0.450–0.999 units and inhibited > 50% by high-dose heparin.
  • Group D (non-platelet-activating anti-PF4/heparin IgA and/or IgM): ≥ 2-fold increase in reactivity (by OD units) in the EIA-IgA and/or the EIA-IgM (vs. pre-operative baseline), with maximum OD ≥0.450, and inhibited >50% by high-dose heparin.
  • Group E (miscellaneous immune responses not meeting the above criteria): ≥ 2-fold increase in reactivity (by OD units) and ≥ 0.450 in any EIA (vs. pre-operative baseline) but not inhibited by high-dose heparin; or ≥ 2-fold increase in reactivity and inhibited by high heparin but maximum OD < 0.450 units.
  • Group F (no immune response among evaluable patients): none of criteria A–E were met.
  • Group G (non-evaluable): no criteria for seroconversion met; however, no blood samples available between postoperative days 7–70 (thus, seroconversion could not be excluded).

The above hierarchically ordered categories are mutually exclusive. Thus, a patient who formed ‘strong’ (but non-platelet-activating) IgG and IgA antibodies would be classified as Group B, based on the IgG reaction pattern.

Assessment of cross-reactivity

Cross-reactivity of antibodies against fondaparinux was tested in two ways, by fluid-phase EIA and by SRA. For the fluid-phase EIA, we adapted a previously reported method [10], as described [9]. We used the following concentrations of PF4 (10 μg mL−1 [20% biotinylated] and drug (UFH, 0.6 IU mL−1; enoxaparin, 0.5 anti-factor Xa U mL−1; fondaparinux, 0.1, 0.3, 0.5 and 1.2 μg mL−1) because preliminary experiments using positive HIT sera showed maximal reactivity at these (final) concentrations of UFH and LMWH. Known positive and negative HIT sera were used as controls. For assessment of fondaparinux cross-reactivity, we studied patients with strong reactions in the EIA-IgG (patients from categories A and B [n = 27]); for comparison, we also tested these samples in the fluid-phase EIA for reactivity against UFH and LMWH.

Blood samples that yielded a positive SRA were also tested for cross-reactivity against fondaparinux by comparing serotonin release in the presence of fondaparinux (0.1, 0.3, 0.5, 1.2 and 100 μg mL−1) with serotonin release in the presence of UFH (0.1, 0.2, 0.3, 1.0 and 100 IU mL−1), in duplicate. Fondaparinux cross-reactivity was defined as a ≥ 20% increase in serotonin release in the presence of any concentration of fondaparinux, compared with buffer control.

Definition of HIT

HIT was defined as a > 30% platelet count fall and/or a clinically evident, objectively documented thrombotic event bearing a temporal association with SRA seroconversion (subclinical graft occlusions detected through study CT angiography were not considered HIT-defining events).

Antiphospholipid antibodies

We tested for anti-β2glycoprotein I IgG using QUANTA Lite β2GPI IgG ELISA; and anticardiolipin IgG using QUANTA Lite ACA IgG III (both from INOVA Diagnostics, Inc., San Diego, CA, USA).

Anti-PF4/heparin antibodies and graft occlusion

For each of the three immunoglobulin classes (IgG, IgA and IgM), we tested for a possible relationship between the presence of anti-PF4/heparin antibodies and the risk of graft occlusion. Quantitative analyses (using OD values) was performed for the pre-operative and highest postoperative antibody levels. In addition, we performed a qualitative assessment of antibody levels based upon seroconversion, using the hierarchical seroconversion categories described earlier, as another measure of a possible effect of postoperative antibodies. Separate analyses were performed for arterial and venous grafts: saphenous vein grafts were classified as ‘venous’ grafts whereas all other grafts (left and right internal mammary, radial artery and so forth) were regarded as ‘arterial’ grafts.

Data analysis and statistics

Categorical variables are reported using frequency (%) and continuous variables are expressed as mean (standard deviation [SD]) or median (interquartile range [IQR]). For the quantitative analyses, we used the OD values obtained after subtracting the OD values obtained at 100 IU mL−1 heparin from the corresponding values obtained in the absence of added high heparin. For comparisons of reactivities in the fluid-phase EIA, we used paired and unpaired t-tests, as appropriate (Microsoft® Excel® 2008 for Mac Version 12.3.2).

We pre-specified our ‘primary’ serologic analysis to be a comparison of the frequency of antibody formation for groups A + B + C (i.e. patients who formed IgG antibody responses) between patients who received postoperative fondaparinux vs. those who received postoperative UFH. Our ‘secondary’ serological analysis was a comparison for serological groups A + B + C + D + E (any antibody response) between these two patient groups. Our hypothesis was that there would be a reduced frequency of anti-PF4/heparin antibody formation among the patients who received postoperative anticoagulation with fondaparinux vs. UFH.

The significance of any difference between the treatment groups (fondaparinux vs. UFH) in the proportion of patients who developed anti-PF4/heparin antibodies (of different antibody classes), or in the proportion of patients who developed graft occlusion with different categories of antibody status, was assessed using Fisher's exact test. Odds ratios (and 95% confidence intervals [CIs]) were estimated using Firth's penalized maximum likelihood logistic regression to reduce bias caused by separation, which is a problem that occurs when the binary dependent variable is perfectly separated by a single covariate.

We evaluated for a relationship between antibodies and graft occlusion using two statistical approaches: one with each patient as the unit of the analysis (defined as occluded if at least one graft was occluded), and the other with each graft as the unit of the analysis.

The significance of the difference in the distribution of antibody levels between the two groups of patients (those with and without graft occlusion) was assessed using Wilcoxon's rank sum test. Odds ratios (95% CIs) were calculated from univariate logistic models with graft patency as the dependent variable and antibody variables as the independent variable. When the graft was the unit of the analysis, the generalized estimating equation (GEE) was used to adjust for multiple graft occlusion data collected from a single patient. We assumed that the correlation between all grafts' occlusion status within each patient was equal (exchangeable correlation matrix).

To confirm that there was an independent association between anti-PF4/heparin antibodies and graft occlusion, we performed multiple regression analyses, with adjustment for the following variables: sex, age, study drug (fondaparinux vs. UFH thromboprophylaxis), cardiopulmonary bypass time, body mass index, diabetes, hyperlipidemia and current smoking status. We excluded serologically non-evaluable patients (Group G, n = 3) from the multivariate analyses. We performed the following comparisons: (i) groups A + B (‘strong’ IgG seroconverters) vs. groups C + D + E + F (all other evaluable serological responses); (ii) groups A + B vs. groups D + E + F (i.e. excluding weak IgG responses); and (iii) groups A + B + C vs. groups D + E + F (strong and weak IgG responses vs. all other evaluable responses).

A two-sided P-value < 0.05 was considered statistically significant. Statistical analyses were performed using SAS 9.1 (SAS Institute Inc, Cary, NC, USA).


Anti-PF4/heparin seroconversion: categorical analysis

Table 1 presents the categorical analysis of seroconversion data, and compares the patients randomized to postoperative UFH vs. patients who received postoperative fondaparinux. We found (in contrast to our hypothesis) that there was a trend to more frequent formation of IgG antibodies (groups A + B + C [primary outcome]) with fondaparinux compared with UFH, beginning already on day 7 (i.e. at a time when equivalent durations of postoperative UFH and fondaparinux had been given), 10/49 (20.4%) vs. 6/50 (12.0%); = 0.2869 (Fisher's exact test), and even more pronounced at the end of the study: 32/49 (65.3%) vs. 23/50 (46.0%); = 0.069. Only three patients, however, tested positive in the SRA (with UFH): 2 patients in the fondaparinux group, and 1 in the UFH group; none reacted in the SRA with fondaparinux (Fig. 1). No difference was found regarding the incidence of all antibodies (groups A+B+C+D+E [secondary outcome]), including anti-PF4/heparin IgM and IgA, between the two groups.

Table 1. Serologic results: categorical analysis
Serological groupUFH (n = 50)Fondaparinux (n = 49) P *
  1. The SRA-positive sera yielded strong positive results in the EIA-IgG (1.891, 2.000, and 2.306 OD units), i.e. these three SRA-positive patients also met Group B criteria. Although 100 eligible patients were recruited, one patient died prior to randomization, did not receive any study medication and was not included in the analyses. *Fisher's exact test. SRA, serotonin-release assay; OD, optical density.

Group A (SRA positive)1 (2.0%)2 (4.1%)0.62
Group B (strong IgG, i.e. ≥1.0 OD units; ↑2× baseline)10 (20.0%)14 (28.6%)0.36
Group C (weak IgG, i.e. 0.45–0.99 OD units; ↑2× baseline)12 (24.0%)16 (32.7%)0.38
A + B + C (IgG response, primary endpoint)23 (46.0%)32 (65.3%)0.07
Group D (IgA and/or IgM)3 (6.0%)3 (6.1%)1.00
Group E (miscellaneous weak)10 (20.0%)4 (8.2%)0.15
A + B + C + D + E (any antibody response, secondary endpoint)36 (72.0%)39 (79.6%)0.48
Group F (no immune response)12 (24.0%)9 (18.4%)0.62
Group G (non-evaluable)2 (4.0%)1 (2.0%)1.00
Figure 1.

Studies of fondaparinux cross-reactivity: platelet activation assays. Left panel. Three sera yielded > 20% serotonin release at pharmacologically relevant unfractionated heparin (UFH) concentrations (peak release at 0.2–0.3 U mL−1). In contrast, none of the sera yielded serotonin release above buffer control (0 μg mL−1) at any of the pharmacologically relevant fondaparinux concentrations. The experiment is representative of two performed.

Frequency of HIT

No patient met criteria for HIT.

Reactivity of Anti-PF4/heparin IgG antibodies with PF4/fondaparinux complexes

Figure 2 shows the results of cross-reactivity studies of 27 patients with strong-positive results (>1.0 units) in the (solid-phase) EIA-IgG (UFH, n = 11; fondaparinux, n = 16). Mean reactivity was significantly greater (mean ± SD) in the presence of enoxaparin (1.558 ± 0.654 OD units) and UFH (1.068 ± 0.701 OD units) than in the presence of any of the fondaparinux concentrations (e.g. 0.399 ± 0.491 OD units at 0.5 μg mL−1 fondaparinux) (= 0.0006 for fondaparinux vs. UFH; < 0.0001 vs enoxaparin) (Fig. 2A). When the data are shown as a ratio against background binding (in the absence of drug), only two patients, one each who received UFH and fondaparinux, exhibited evidence of in vitro cross-reactivity against fondaparinux, and only at one fondaparinux concentration (0.5 μg mL−1) (Fig. 2B).

Figure 2.

Studies of fondaparinux cross-reactivity: fluid-phase EIA. (A) The figure shows mean and individual reactivities for 27 patient sera (that tested ≥1.0 OD units in the solid-phase EIA) in the fluid-phase EIA at the following conditions: no drug; unfractionated heparin (UFH), 0.6 U mL−1; low-molecular-weight heparin (LMWH) (enoxaparin), 0.5 U mL−1; and fondaparinux (at the 4 concentrations shown, in μg mL−1). Reactivities of corresponding positive (heparin-induced thrombocytopenia [HIT] serum) and negative internal controls are also shown. (B) The data are shown as ratios of optical density (OD) in the presence of the drug to OD in the absence of the drug. In this analysis, only 2 of 27 sera showed cross-reactivity at one concentration (0.5 μg mL−1) of fondaparinux.

Anti-PF4/heparin antibodies and graft occlusion

Ninety-five of the 100 patients underwent angiographic assessment of graft patency. We saw no relationship between the pre- and postoperative antibody status and arterial graft occlusion (not shown). We also saw no relationship between pre-operative antibodies and venous graft occlusion. However, we found a significant association between postoperative anti-PF4/heparin IgG antibodies and venous graft occlusion. When using each patient as the unit of analysis, one or more venous graft occlusions occurred in 6/26 patients within antibody groups A+B (i.e. ‘strong’ IgG antibodies) vs. 3/66 patients within antibody groups C + D + E + F (= 0.0139). When using each graft as the unit of analysis, occlusions occurred in 7/67 venous grafts within antibody groups A + B vs. 3/163 venous grafts within antibody groups C + D + E + F (= 0.0086).

Expressed as odds ratios, ‘strong’ levels of IgG antibodies formed postoperatively (i.e. groups A + B) were significantly associated with venous graft occlusion, both by categorical analysis and by quantitative analysis, and irrespective of whether we used the patient (Table 2) or the graft (Table 3) as the unit of analysis. For all analyses, the relationship was consistently seen with ‘strong’ IgG antibodies (i.e. EIA-IgG >1.0 OD units and > 2× baseline) in the categorical analyses, and by the IgG quantitative analysis (Tables 2 and 3). Notably, however, none of the three patients with platelet-activating antibodies by SRA had venous graft occlusion.

Table 2. Association between anti-PF4/heparin antibodies and venous graft occlusion: each patient as the unit of analysis
SerologyAt least one venous graft occluded N (%) (N = 9)No venous graft occluded N (%) (N = 83) P * OR (95% CI)
Postoperative, categorical analysis
Group A (SRA+)0 (0%)3 (3.6%)1.001.21 (0.01, 13.99)
Group A + B (strong IgG)6 (66.7%)20 (24.1%)0.015.75 (1.49, 26.09)
Group A + B + C (all IgG)7 (77.8%)45 (54.2%)0.292.54 (0.63, 14.22)
Group A + B + C + D (IgG, IgA, and/or IgM)7 (77.8%)51 (61.4%)0.481.89 (0.47, 10.62)
Group A + B + C + D + E (any immune response)9 (100%)63 (75.9%)0.206.13 (0.72, 803.43)
Postoperative, quantitative analysisMedian (IQR)Median (IQR)  
  1. IQR, interquartile range; CI, confidence interval; OR, odds ratio; SRA+, serotonin-release assay-positive. For the nine patients with one or more venous graft occlusions (i.e. each patient as the unit of analysis), antibody groups were B (n = 6), C (n = 1) and E (n = 2), whereas for the 86 patients with no venous graft occlusions, antibody groups were A (n = 3), B (n = 17), C (n = 25), D (n = 6), E (n = 12), F (n = 20) and G (n = 3). *P-values for the categorical analyses were calculated using Fisher's exact test, and for the post-/pre-operative quantitative analyses were determined using Wilcoxon's rank sum test. OR (95%CI) were calculated from the Firth's penalized maximum likelihood logistic regression in the categorical analysis and the logistic regression in the quantitative analysis. For these OR determinations, two P-values (not shown in the Table) were significant: the P-value for the OR of 5.75 for the categorical analysis of Group A+B was 0.01, and the P value for the OR of 5.38 for the postoperative IgM quantitative analysis was 0.02.

IgG1.045 (0.549, 1.295)0.356 (0.093, 0.818)0.042.75 (0.94, 8.07)
IgA0.071 (0.033, 0.220)0.038 (0.016, 0.114)0.451.41 (0.30, 3.79)
IgM0.149 (0.057, 0.417)0.100 (0.037, 0.178)0.305.38 (1.37, 28.16)†
Pre-operative, quantitative analysis
IgG0.025 (−0.003, 0.053)0.024 (0.004, 0.076)0.711.61 (0.00, 170.84)
IgA0.002 (0.001, 0.013)0.006 (0.001, 0.020)0.491.00 (0.00, 1186.90)
IgM0.068 (0.048, 0.078)0.066 (0.032, 0.121)0.911.53 (0.04, 5.89)
Table 3. Association between anti-PF4/heparin antibodies and venous graft occlusion. Analysis of all grafts (each graft as individual unit of analysis): odds ratio (OR) for graft occlusion (controlled for correlated outcome of occluded graft within patient)
SerologyOR for graft occlusion (95% CI) P
  1. OR, odds ratio; CI, confidence interval; SRA+, serotonin-release assay-positive. For the 10 venous grafts that were occluded (i.e. each graft as the unit of analysis), antibody groups were B (n = 7), C (n = 1) and E (n = 2), whereas for the 228 venous grafts that were not occluded, antibody groups were A (n = 7), B (n = 53), C (n = 64), D (n = 12), E (n = 37), F (n = 47) and G (n = 8). *NA, not available (estimate for OR was N/A when the GEE algorithm failed to converge due to sparse non-zero data).

Postoperative, categorical analysis
Group A (SRA+)NA
Group A + B (strong IgG)6.25 (1.52, 25.69)0.01
Group A + B + C (all IgG)3.15 (0.64, 15.40)0.16
Group A + B + C + D (all IgG, IgA, and/or IgM)NA
Group A + B + C + D + E (any immune response)NA
Postoperative, quantitative analysis
IgG2.38 (1.12, 5.08)0.02
IgA1.04 (0.34, 3.20)0.94
IgM9.09 (2.75, 30.01)<0.01
Pre-operative, quantitative analysis
IgG0.47 (0.00, 473.95)0.83
IgA0.01 (0.00, 6873.89)0.48
IgM0.95 (0.22, 4.13)0.95

Table 3, which presents the analyses using each individual graft as the unit of analysis, shows also that postoperative IgM levels were strongly associated with venous graft occlusion. However, only four patients developed seroconversion to ‘strong’ IgM class antibodies (>1.0 OD units), and although two of these patients developed three venous graft occlusions, these two patients also exhibited concomitant ‘strong’ IgG seroconversion. Thus, our data do not point to any independent pathogenic role of IgM class antibodies.

The association between postoperative development of IgG antibodies and venous graft occlusion remained evident when we controlled for potential confounders by multiple regression analyses (each patient as the unit of analysis). Once again, group A + B patients (i.e. those with ‘strong’ IgG immune responses) had a markedly increased risk of venous graft occlusion, irrespective of whether weak IgG responders were included in the comparison group (i.e. antibody groups A + B vs. C + D + E + F: OR = 9.25 [95% CI, 1.73, 49.43]; = 0.0093]) or not (i.e. antibody groups A + B vs. D + E + F: OR = 8.14 [95% CI, 1.11, 59.55]; = 0.0390). However, when weak IgG responders were combined with the strong IgG responders, the association was no longer significant (i.e. antibody groups A + B + C vs. D + E + F: OR = 2.95 [95% CI, 0.51, 16.92]; = 0.2260]).

A possible contribution of antiphospholipid antibodies in explaining graft occlusions was excluded by negative testing for anticardiolipin IgG and anti-β2GPI antibodies in all patients who developed a venous and/or arterial graft occlusion.


This study found a trend to an increased frequency of anti-PF4/heparin IgG seroconversion with fondaparinux compared with UFH: 32/49 (65.3%) vs. 23/50 (46.0%); P = 0.069. We also found that high levels of anti-PF4/heparin IgG antibodies formed during the postoperative period were associated with an increased risk for venous, but not arterial, graft occlusion.

Previous studies of postorthopedic surgery thromboprophylaxis found that fondaparinux was at least as likely as enoxaparin to be associated with the formation of anti-PF4/heparin antibodies [9, 11]. The present study suggests that an immunization effect of fondaparinux might be generalizable for other patient populations undergoing major surgery thromboprophylaxis.

However, in spite of being immunogenic, fondaparinux shows a low frequency of cross-reactivity with anti-PF4/heparin antibodies. None of the three patients in the present study whose serum tested SRA-positive showed evidence for fondaparinux increasing the platelet-activating potential of these antibodies (Fig. 1). Moreover, 27 sera that reacted strongly positive in the solid-phase EIA-IgG showed a markedly lower cross-reactivity profile for fondaparinux (vs. UFH and LMWH) when examined by the fluid-phase EIA (indeed, only two of these 27 sera cross-reacted with fondaparinux in the fluid phase EIA) (Fig. 2). The much lower frequency of cross-reactivity with fondaparinux vs. UFH or LMWH is consistent with previous studies [9, 12] and supports the hypothesis that fondaparinux given for post-cardiac surgery thromboprophylaxis might therefore reduce the risk of HIT compared with UFH or LMWH. However, this hypothesis requires testing in appropriately powered clinical trials.

We found a strong association between postoperative formation of high levels of anti-PF4/heparin IgG antibodies and the risk for a venous graft occlusion. This differs from a previous study by Gluckman et al. [1] of 368 patients (297 of whom underwent multidetector CT coronary angiography at 6 months post-surgery), which found no association. This study differs in certain key respects from this previous report. First, we assessed graft occlusion at approximately 47 days (median) post-CABG [5], whereas assessment was performed at 6-month follow-up by Gluckman et al. Thrombotic mechanisms are likely to play a greater role in early graft occlusion whereas accelerated intimal hyperplasia, which will be unaffected by a putative prothrombotic effect of anti-PF4/heparin IgG antibodies, is the major underlying mechanism of later graft failure [13]. Thus, early graft failures may be more sensitive to the effects of anti-PF4/heparin IgG antibodies than graft failure measured at 6 months. Second, we obtained our in-hospital blood samples somewhat later, median, day 5 [IQR, 5, 6] vs. day 4 [IQR, 3, 4] as measured by Gluckman et al., and thus we could have more readily detected seroconversion events. In our study, the association between postoperative IgG formation and venous graft occlusion was robust, regardless of which analysis we performed. There also appeared to be a ‘dose effect’, as only patients with strongly reactive IgG antibodies showed an association with venous graft occlusion.

The pathogenesis of venous graft occlusion in patients with anti-PF4/heparin IgG antibodies might differ from ‘classic’ HIT immunopathogenesis, which is primarily platelet mediated. None of the three patients with detectable platelet-activating antibodies developed a venous graft occlusion. An intriguing hypothesis to explain our findings relates to a recent finding by Kowalska and colleagues [14] that HIT antibodies are able to inhibit the generation of activated protein C (aPC) by thrombin/thrombomodulin (aPC generation is usually enhanced in the presence of PF4 [14]). Although non-platelet-activating anti-PF4/heparin antibodies have long been considered non-pathogenic, saphenous vein grafts could be especially susceptible to the effects of hypercoagulability resulting from aPC inhibition because of their venous origin and abrupt transition from low flow-pressure to high-flow pressure dynamics [15, 16]. This could also explain why we saw no association between anti-PF4/heparin antibodies and arterial graft occlusion.

We excluded a role for antiphospholipid antibodies in explaining our findings, as none of the patients who developed a venous or arterial graft occlusion tested positive for antiphospholipid antibodies. The rationale for assessing antiphospholipid antibodies is that they are associated with thrombosis and because they can yield false-positive tests in anti-PF4/heparin immunoassays [17].

The current dogma in the HIT literature is that non-platelet-activating anti-PF4/heparin antibodies cause neither thrombocytopenia nor thrombosis [18]. However, if our finding of an association between postoperative formation of anti-PF4/heparin antibodies and venous graft occlusion is real, it could be explained by at least one other mechanism (besides inhibition of aPC generation) independent of the development of classic, clinically overt HIT. Such an association could represent the effects of a common factor, such as the degree of intra- or postoperative inflammation, which could influence both seroconversion frequency as well as venous graft occlusion. In this regard, it is worth noting that one group has reported an association between a marker of inflammation and antibody formation in cardiac surgery patients [19] (however, we did not examine markers of inflammation in our current study). Moreover, Selleng et al. [20] found an association between the pre-operative presence of IgM class antibodies and subsequent non-thrombotic morbid events in post-cardiac surgery patients, which they concluded was unlikely to reflect any HIT-related pathomechanism, but which they speculated could represent an independent factor such as inflammation.

It is worth noting that the two aforementioned potential explanations for an association between anti-PF4/heparin antibody formation and venous graft occlusion have distinctly different therapeutic implications. If the explanation relates to a specific prothrombotic effect of the antibodies, such as interference with endothelial aPC generation, then in theory the use of non-heparin anticoagulation during (and possibly after) cardiac surgery, by avoiding/reducing anti-PF4/heparin antibody formation, would enhance early venous graft patency. In contrast, if the explanation is a common factor such as inflammation, then the choice of anticoagulant would be expected to play a little role in modulating graft patency, as anti-PF4/heparin antibody formation (except on occasion when it induces a patient to develop ‘true’ HIT) would be largely an epiphenomenon.

In conclusion, we found that fondaparinux thromboprophylaxis post-cardiac surgery was not associated with a reduced frequency of anti-PF4/heparin antibody formation; however, platelet-activating antibodies did not enhance platelet activation in the presence of fondaparinux, indicating that theuse of this anticoagulant in place of UFH for postoperative thromboprophylaxis might nevertheless reduce the risk of HIT. We also observed an association between the formation of anti-PF4/heparin antibodies and early venous graft occlusion, which does not appear to be mediated by platelet-activating antibodies.


The Authors thank Dr Patricia Liaw for helpful comments. The anti-PF4/heparin serology studies were supported by the Heart and Stroke Foundation of Ontario (operating grant #T6950). The Fonda CABG RCT was supported by a research grant from GlaxoSmithKline Canada (who also provided fondaparinux and placebo for the study). The sponsor played no role in the design, conduct or interpretation of the data for the anti-PF4/heparin substudy, and did not have any role in the decision to report the study findings. All of the authors had access to the complete data set.

Disclosure of Conflict of Interests

T.E.W. has received lecture honoraria from GlaxoSmithKline, Pfizer Canada, and Sanofi-Aventis, has provided consulting services to, and/or has received research funding from, Canyon Pharmaceuticals, Gen-Probe GTI Diagnostics, GlaxoSmithKline, W. L. Gore, and Paringenix, and has provided expert witness testimony relating to HIT. Dr Sun has received a research grant from Bristol-Myers-Squibb Canada. Dr Eikelboom has received grants and honoraria from GlaxoSmithKline. Hyejung Jung and Jo-Ann Sheppard report no conflicts of interest.