Quantification of heparin-induced TFPI release: a maximum release at low heparin dose



Aims Heparin releases tissue factor pathway inhibitor (TFPI) from the endothelium and this release may decrease after repeated high dose heparin administration. The primary aim was to investigate and quantify this phenomenon during a short low dose heparin infusion. Also, the effects of heparin on tissue plasminogen activator (t-PA) were studied.

Methods Nine healthy, nonsmoking, male volunteers (range 19–23 years) received a continuous heparin infusion (2000 IU) over 40 min. The endothelial TFPI release rate was estimated from the total TFPI concentration profile using a pharmacokinetic model.

Results   Mean  ± s.d.  total  and  free  TFPI  increased  from  62.9 ± 9.4/8.3 ± 2.1 ng ml−1 at baseline to 237.2 ± 40.9/111.0 ± 19.9 ng ml−1 after 40 min infusion. The relationship between heparin concentration (anti-IIa activity) and TFPI concentration followed a maximum effect model and a clockwise loop (proteresis) was observed. The TFPI release rate rapidly increased to maximum of 200 ± 45 µg min−1 after 17.5 min heparin infusion but did not increase further although heparin concentrations further doubled. In contrast to TFPI, t-PA antigen decreased from 5.6 ± 1.0 at baseline to 4.5 ± 1.0 ng ml−1 at the end of infusion (t = 40 min) (difference of 1.1 ng ml−1 (95% confidence interval; 0.9, 1.3).

Conclusions Our application of concentration-effect models and pharmacokinetic principles to these haemostatic variables showed that endothelial TFPI release has a maximum that is already reached at low heparin dose, corresponding with an anti-IIa activity of 0.08 IU ml−1. The relationship between anti-IIa activity and TFPI release rate showed signs of acute tolerance (clockwise loop) indicating exhaustion of endothelial TFPI pools. These findings may be of importance for the heparin dose used in conditions such as unstable angina, in which the favourable effects of heparin have been ascribed to its ability to release TFPI.


The anticoagulant effect of heparin is mediated largely through its direct interaction with antithrombin-III [1]. Part of the anticoagulant effects of heparin is indirectly via the release of TFPI from the endothelium [2]. Tissue factor pathway inhibitor (TFPI) is a potent inhibitor of the extrinsic coagulation system. TFPI is distributed in three pools in vivo: approximately 80–85% is extra-­luminal, including association with endothelial cell-­surface, presumably involving glycosaminoglycans and proteoglycans, approximately 10% circulates in plasma primarily in association with lipoproteins and a small amount in free form, and approximately 3% is found in platelets [3].

Heparin induced TFPI release is induced by displacement of TFPI from the endothelial surface and/or by a release from intravascular stores [4]. The release is reported to be immediate after intravenous heparin injection [5]. Administration of heparin in bolus of 5000 IU plus 450 IU kg−1 24 h−1 continuous infusion or after repeated dosing with 5000 IU [6] results in a decrease in release of TFPI (tachyphylaxis) [7], although this was not confirmed in all studies [8]. The decrease in TFPI release suggests that dose regimens that are effective in the treatment or prevention of thrombosis result in a depletion of intravascular TFPI pools.

Enhancement of fibrinolysis may be an additional effect of heparin [9]. Analogous to the TFPI release by heparin, effects of heparin on fibrinolysis may be caused by a release of tissue plasminogen activator (t-PA) from the endothelium. The main source of circulating t-PA is the endothelium where t-PA is stored in intracellular bodies [10]. Despite numerous studies on the effects of heparin on fibrinolysis, controversy exists on the effects of heparin on t-PA concentrations [9, 11, 12].

The primary aim of the present study was to quantify the heparin induced TFPI release and to investigate the occurrence of tolerance during a low dose heparin infusion. The effects of heparin on the plasma concentration profile of TFPI was evaluated by investigating the relationship between anti-IIa activity and TFPI concentrations using a concentration-effect model. The TFPI release rate was estimated using a pharmacokinetic model. The secondary aim was to study the effects of heparin on fibrinolysis by measuring t-PA antigen and activity and complexes with PAI-1 during the heparin infusion.


The study was conducted according to the principles of the ‘Declaration of Helsinki’ and in accordance with the Guideline for Good Clinical Practice. The study protocol was approved by the Committee on Medical Ethics of the Leiden University Medical Center.


Nine nonsmoking, healthy male volunteers (age range 19–23 years) participated in this study. All subjects had a normal lipids profile (plasma cholesterol range 3.3–4.3 mmol l−1) and were not taking other medication. All subjects were included after each gave written informed consent.

Study days

After an overnight fast and 1 h resting, subjects received on a single occasion a continuous heparin infusion over 40 min (2000 IU). Heparin was obtained from Akzo Nobel (Oss, The Netherlands) and prepared for administration by the pharmacy of the Leiden University ­Medical Center (Leiden, The Netherlands) following European Pharmacopoeia standards. Blood sampling was performed three times predose and following the start of the infusion at 5, 10, 15, 20, 25, 30, 35, 40 (= stop infusion), 45, 50, 55, 65, 100, and 160 min after the start of the infusion. The subjects were in a supine position and fasted during the entire experiment.

Blood sampling

At each time-point two blood samples were taken. For TFPI, APTT and anti-IIa assay blood was sampled in tubes containing 0.105 m citrate (Becton & Dickinson, Plymouth, UK). Samples for t-PA assays were collected using prechilled Stabilyte® vacutainer tubes (Biopool, Umea, Sweden) containing 0.5 ml acid anticoagulant in which the blood is immediately mixed and brought to pH 5.9 in order to stabilize t-PA activity. Following collection, the samples were immediately stored at ice-water until centrifuged within 30 min at 4 °C (5000 g for 10 min). Samples were snap-frozen in methanol with dry ice and stored at −40 °C until analysis.


All assays were done at TNO-Gaubius Laboratories, The Netherlands, using commercially available assays from Diagnostica Stago, Asnières, France for the TFPI assays and from Biopool, Umeå Sweden for the t-PA and anti-IIa activity assays.

In the two enzyme-immunoassays for free and total TFPI (Asserachrom®) a monoclonal antibody is used directed towards the binding site for factor Xa. It thus catches TFPI not complexed to factor Xa. For the free TFPI, a tagging antibody is used which reacts with the fraction of TFPI that is not binding to lipids and thus represents the free, nonlipid-bound active TFPI. For the total TFPI a catching antibody to the free and lipid-bound fraction of TFPI is used, thus representing the total TFPI not bound to Factor Xa. Lipoprotein-bound TFPI concentration was determined by subtracting the free TFPI from total TFPI. The coefficient of variation of the intra- and interassay variability for total TFPI was 4.2% and 6.8% and the assay range was 4–230 ng ml−1. The intra- and interassay variability for free TFPI was 3.9 and 4.4% and the assay range was 2–70 ng ml−1.

t-PA activity was determined using Chromolize® t-PA assay with an intra- and interassay variability smaller than 10%, a lower limit of quantification (LOQ) of 0.05 IU ml−1 and a reference range of 0.094–1.789 IU ml−1. t-PA antigen was determined using Imulyse® t-PA assay, with an intra-assay variability smaller than 10% and an interassay variability smaller than 12%, a LOQ of 1.5 ng ml−1 and a reference range of 3.0–13.0 ng ml−1. t-PA/PAI-1 complex was determined using TintElize® t-PA/PAI-1 complex assay with an intra- and interassay variability smaller than 7.5% and a LOQ of 0.2 ng ml−1.

Anti-IIa activity assay (Spectrolyse® Heparin) inter- and intra-assay variability was smaller than 10% and LOQ was 0.01 IU ml−1. The anti-IIa activity assay was calibrated against the 1st International Standard from the National Institute for Biological Standards and Control (Hertfordshire, United Kingdom) (code 85/600). The APTT was determined on a STA® coagulation analyser (Roche Diagnostics, Mannheim, Germany) using STA APTT reagent (Roche Diagnostics). Inter- and intra-assay variability was smaller than 5%.

Calculations and statistics

Heparin concentration was assessed by measuring anti-IIa activity, as anti-IIa activity is not influenced by TFPI [13]. The relationship between anti-IIa activity and free or total TFPI concentration was investigated using a linear and a maximum effect (Emax) concentration effect model (NONMEM Version V, NONMEM Project Group, UCSF, San Francisco, CA, USA, using first order conditional estimation with the interaction option). The Emax model was regarded superior as the minimum value of the objective function of the linear vs the Emax model was 1186 vs 1168 (a difference of 18, P < 0.001). With NONMEM all data are analysed collectively while individual information is preserved. NONMEM provides information on the accuracy (by its standard error; SE) and on the interindividual variability (by the coefficient of variation; CV) of the population estimate. A CV of 0% means that no improvement in fit is obtained when parameter values are allowed to vary between subjects. This should not be interpreted as proof for equality of parameter values between subjects but is a reflection of the lack of information on this parameter. Average predicted profiled were calculated using individual empirical Bayes estimates.

The endothelial release rate of TFPI was calculated using the following equation that describes the change in concentration (dCt/dt) as a function of infusion rate (Rinf ; here endogenous TFPI release), drug concentration (CL ), clearance (Ct ) and volume of distribution (Vd) [14, 15]:


Multiplying the equation by Vd gives the following equation:


The Rinf can be split into a baseline endogenous TFPI secretion (R) and the secretion induced by heparin (I  ) between two consecutive time points. The equation can now be written to describe the release rate induced by heparin:


where Ct is the average of the total TFPI concentration at two consecutive times points and dCt/dt is the change in concentration between two timepoints. The clearance was assumed to be 630 ml min−1 (noncompartmental TFPI clearance) and the Vd was set at 18 l (both values were derived from a study with exogenously administered (recombinant) TFPI [16]). Baseline endogenous secretion was calculated by multiplying clearance with the baseline endogenous TFPI concentration.

The cumulative TFPI secretion was calculated by multiplying the release rate with the time period.

The lipoprotein-bound TFPI concentration was calculated as the total minus the free TFPI concentration. The lipoprotein-bound fraction was calculated as the lipoprotein-bound TFPI concentration divided by total TFPI. Testing of significance of the decrease in lipoprotein-bound TFPI fraction was done using Student's t-test. Values are given as mean ± s.d. or mean plus s.e.mean and interindividual coefficient of variation (CV) when determined with NONMEM.


Heparin increased total and free TFPI plasma concentrations from 62.9 ± 9.4 and 8.3 ± 2.1 ng ml−1 prior to infusion to 237.2 ± 40.9 and 111.0 ± 19.9 ng ml−1 after 40 min infusion (Figure 1). Anti-IIa activity, a measure for heparin concentration, increased from 0.05 IU ml−1± 0.02 to 0.19 IU ml−1 ± 0.08 after 40 min infusion ­(Figure 1). APTT increased from 40 ± 4 s to 99 ± 29 s during the heparin infusion.

Figure 1.

Total (●) lipoprotein-bound and free (○) TFPI concentration profiles (a) and anti-IIa activity profile (b) during continuous intravenous (i.v.) heparin infusion (2000 IU) from 0 to 40 min.

Total and free TFPI concentrations were plotted against heparin concentration as assessed with anti-IIa activity. The relationship between anti-IIa activity and free and total TFPI concentration appeared to be linear at lower heparin concentrations but the curves tended to flatten at higher heparin concentrations (Figure 2). As a consequence, the concentration-effect relationship was best described using a maximum effect (Emax) concentration-effect model (P < 0.001, when compared with a linear model). If modelled using a maximum effect model the TFPI concentration goes with the following equation:

Figure 2.

The relationship between anti-IIa activity and free (○) and total (●) TFPI concentration. Each marker represents the anti-IIa activity and the free or total TFPI concentration for a sample-point before, during and after the heparin infusion. The consecutive sample points are connected with a line and the arrows indicate the time-course. The concentration-effect relationship between the anti-IIa activity (heparin concentration) and the TFPI concentration was not linear and was best described using a maximum effect (Emax) concentration-effect model (P < 0.001). However, the quality of fit was hampered by the occurrence of proteresis (a clockwise loop). Bold line: modelled total TFPI.


where E0 is the effect at zero anti-IIa activity, and EC50 the anti-IIa activity at which 50% of Emax was reached. According to the Emax model, estimated total TFPI level at zero anti-IIa activity (E0) was 58 ng ml−1 (s.e. mean 6, interindividual coefficient of variation (CV) 24%), with a Emax of 292 ng ml−1 (s.e. mean 31, CV 0%) and an anti-IIa activity at which 50% of Emax was reached (EC50) of 0.22 IU ml−1 (s.e. mean 0.05, CV 30%). Estimated Free TFPI E0 was 14 ng ml−1 (s.e. mean 3, CV 46%), with an Emax of 124 ng ml−1 (s.e. mean 15, CV 0%) and EC50 of 0.15 IU ml−1 anti-IIa activity (s.e. mean 0.04, CV 33%).

When concentrations went up, the concentrations of total and free TFPI were higher than at similar concentrations during the down-slope of the curve. This clockwise loop (proteresis) was observed in the individual graphs and also appeared in the average profile.

The TFPI release rate increased rapidly after starting the heparin infusion and reached a maximum of 200 ± 45 µg min−1 after 17.5 min infusion (Figure 3). The release rate remained approximately at that level during the rest of the heparin infusion and decreased after stopping the infusion. Analogously, the relationship between anti-IIa activity and TFPI release rate reached a plateau at 0.08 ± 0.05 IU ml−1 anti-IIa activity and did not increase further, although anti-IIa activity increased further to 0.19 ± 0.08 IU ml−1 during the heparin infusion (Figure 4). The relationship between anti-IIa activity and the total TFPI release rate showed proteresis; at the upward part of the curve the TFPI release was higher than when concentrations went down. From the release rate it was calculated that the 2000 IU heparin infusion induced a total release of 5.7 ± 1.4 mg TFPI.

Figure 3.

The total TFPI release rate over time profile before, during and after the heparin infusion (from t = 0–40 min). Each marker represents the TFPI release rate with error bars (s.d.) that was calculated from the total TFPI concentration of two consecutive sample-points using a pharmacokinetic model. The negative outcome at t = 48 min is the result of a larger decrease (fluctuation) in TFPI concentration than can be described by clearance alone.

Figure 4.

The relationship between the anti-IIa activity and TFPI release rate before, during and after the heparin infusion. Each marker represents the average anti-IIa activity between two sample-points and the estimated TFPI release rate between these two sample-points. The consecutive estimations of    TFPI release rate are connected with a line and the arrows indicate the time-course.

The lipid bound TFPI concentration decreased from 54.6 ± 9.7 ng ml−1 to 45.0 ± 9.8 ng ml−1 in the first 5 min of infusion (nonsignificant after Bonferoni correction) and increased thereafter to 126.3 ± 23.8 ng ml−1 at 40 min (Figure 5). At baseline, 87% ± 4% of total TFPI was lipoprotein-bound. In the first 10 min of the infusion, this fraction decreased to 58% ± 8% and decreased slightly further to 53% ± 3% at the end of infusion.

Figure 5.

The lipoprotein-bound TFPI concentration profile over time (a) and the lipoprotein-bound fraction profile over time (b). The lipoprotein-bound concentration is the total minus the free TFPI concentration. The lipoprotein-bound fraction is the lipoprotein-bound TFPI divided by total TFPI.

During heparin infusion, the t-PA antigen concen­trations decreased from 5.6 ± 1.0 at baseline to 4.5 ± 1.0 ng ml−1 at the end of infusion (t = 40 min) (a decrease of 1.1 ng ml−1 (95% confidence interval (CI); 0.9, 1.3) (Figure 6). The t-PA activity increased from 0.45 ± 0.13–0.52 ± 0.14 U ml−1 during infusion (difference of 0.07 U ml−1 (95% CI 0.02, 0.12). t-PA/PAI-1 complex concentrations did not change (difference of −0.2 ng ml−1 (95% CI −0.9, 0.5).

Figure 6.

t-PA antigen (●), activity (s) and t-PA/PAI-1 complex (▪) profiles during continuous heparin infusion (2000 IU) from 0 to 40 min.


In the present study, the effects of short low dose heparin infusion on TFPI release were studied by relating the anti-IIa activity, a measure of heparin concentration, to the total and free TFPI concentration. Within 20 min of heparin infusion, equivalent to a dose of 1000 IU, the initial linear concentration-effect relationship between anti-IIa activity and TFPI concentration flattened to a steady state level and the profile was best described using a maximum-effect model. This flattening to steady state level occurred at an anti-IIa activity of 0.08 IU ml−1. Analogously, the TFPI release rate reached a maximum of 200 ± 45 µg min−1 after 17.5 min infusion at an anti-IIa activity of 0.8 IU ml−1. The release rate did not increase further although heparin infusion continued and anti-IIa activity doubled. Therefore, it appears that low doses of heparin (approximately 1000 IU) are already capable of attaining steady state TFPI concentration with a maximum TFPI release rate. Both the maximum effect model and the finding of a maximum release rate indicate that higher heparin doses will not lead to substantial higher TFPI concentrations.

TFPI concentration and release rate were greater when heparin concentrations went up then at similar concentrations during the down-slope of the curve. The resulting clock-wise loops (proteresis) in the concentration effect plots indicate occurrence of acute tolerance of endothelial TFPI release. This tolerance may be caused by depletion of endothelial TFPI sources. Indications for depletion have been observed by others after repeated doses, but the doses that were studied were substantially higher [6]. Given the fact that we observed maximum TFPI release during the low dose infusion it is conceivable that such a depletion may also occur at low heparin doses. Nevertheless, our observation of proteresis is complicated by the fact that it occurred shortly after stopping the infusion, when no ‘fresh’ unbound heparin appeared in the circulation. Longer or repeated low dose infusions may help to clarify this.

For the calculation of TFPI release rate we used a clearance and distribution volume derived from pharmacokinetic analysis of an infusion with human recombinant TFPI. It is conceivable that both endothelial released TFPI and recombinant TFPI have similar kinetic properties as both are full length TFPI molecules [17]. The most important difference is the fact that in the present study heparin may have influenced clearance. In rabbits, it has been demonstrated that heparin decreases the clearance of full-length TFPI [18]. This may be caused by the fact that heparin is known to bind to the basic C-terminus of TFPI [19, 20] and the C-terminus of TFPI is required for binding to hepatoma cells and subsequent clearance [21]. If heparin affects TFPI clearance in humans, the clearance used in the calculations may be to high. However, since clearance is not the major determinant of the release rate equation, the overestimation of the release rate would be minimal.

In the concentration-effect plots we used the anti-IIa activity as a representation of the heparin concentration. Approximately one third of the heparin molecules have a pentasacharide group and can bind to antithrombin [22]. Of this fraction the high molecular weight fraction has anti-IIa activity. The anti-IIa activity assay represents therefore a fraction of the total heparin concentration. Nevertheless, this assay was chosen as a measure of heparin concentration since the alternatives for monitoring heparin, anti-Xa activity and activated partial thromboplastin time assays, are both influenced by TFPI [23].

Total TFPI concentrations increased to 237 ng ml−1 during the heparin infusion and the maximum effect model predicted a maximum TFPI concentration (Emax+ E0) of 350 ng ml−1. According to the maximum effect model, attaining the additional increase from 237 to 350 ng ml−1 would require a relatively large anti-IIa activity increase as the effect of heparin on TFPI concentration diminishes at higher anti-IIa activity levels. This observation is supported by the release rate model as endothelial TFPI production has a maximum that is reached already at low anti-IIa activity levels. Also from the maximum release rate and clearance of TFPI the maximum TFPI concentration can be predicted. Calculating the maximum attainable concentration as the maximum release rate (200 µg min−1) divided by clearance [15], a value of 317 ng ml−1 is obtained. This corresponds well (difference of 10%) with the predicted maximum concentration from the maximum effect model (350 ng ml−1). It appears that both models are concordant in their predictions. This implies that the applied models are both able to accurately describe the TFPI concentration profiles during heparin infusion.

During heparin infusion, a small decrease in t-PA antigen concentrations and an increase in t-PA activity were observed. Although no control experiment was ­performed to study the concentration profiles without heparin, the small fluctuations observed in t-PA antigen, t-PA activity and t-PA/PAI-1 complex can be ascribed to the diurnal fluctuations observed in earlier studies [24].

Interestingly, during the first 5 min of infusion the lipoprotein bound fraction decreased while total TFPI did not increase. This indicates that apart from inducing a release from the endothelium, heparin increases the free TFPI fraction by direct release from lipoproteins. This is concordant with a previous in vitro observation that heparin prevents complex formation between lipoproteins and TFPI [25]. Both lipoproteins and heparin are known to bind to the basic C-terminus of TFPI [19, 20], therefore heparin may compete with lipoproteins for the same binding site on TFPI. During the heparin infusion the lipoprotein-bound TFPI fraction decreased rapidly further, predominantly within the first 10 min of infusion from 87% to approximately 50% and changed only slightly further during heparin infusion. This relatively large free fraction may be the result of prevention of complex formation between free TFPI and lipoproteins by heparin. Despite the presence of heparin, after a transient decrease, there was a large increase in the concentration of lipoprotein associated TFPI. One possibility is that some free TFPI released by heparin may associate with lipoproteins. Adequate control experiments would, however, be needed to rule out in vitro effects of heparin in the free TFPI assay which may explain the apparent decrease and increase in lipoprotein associated TFPI.

In cardiovascular disease, a local excess of tissue factor may serve as the ‘trigger’ of the extrinsic pathway and induce thrombin generation [26]. TFPI inhibits tissue factor and thus prevents clot formation. The beneficial effects of heparin administration in certain conditions have been ascribed to the ability of heparin to release TFPI and prevent clot formation [26–28]. Our findings indicate that heparin doses well below the normal clinical dose already achieve effective maximum TFPI release and thus may be effective in this respect. This has not been studied yet in patients with coronary artery disease. To investigate whether this maximum TFPI also occurs in elderly patients with cardiovascular disease, studies are currently under way to determine the heparin induced TFPI release in patients with clinical atherosclerosis and age-matched controls.

In this study, we quantified the heparin-induced release of TPFI by using concentration-effect modelling and a pharmacokinetic model. The TFPI concentration appeared to reach a plateau, and likewise the endothelial TFPI release reached a maximum at low heparin doses and corresponding low anti-IIa activity of 0.08 IU ml−1, which is well below the therapeutic range of 0.2–0.4 U ml−1 anti-IIa activity [22]. We observed signs of acute tolerance in the relationship between anti-IIa activity and both total TFPI concentration and release rate, suggesting that at already low heparin doses an acute exhaustion of heparin-releasable TFPI pools may occur.