Pharmacokinetics of human antithrombin III concentrate in the immediate postoperative period after liver transplantation

Abstract

Aims

Antithrombin III (AT‐III) concentrates have been used to prevent critical thrombosis in the immediate post‐liver transplantation period without clear evidence regarding the optimal dose or administration scheme. The relationship between the AT‐III dosage and the plasma activity levels during the period was evaluated in this study.

Methods

The plasma AT‐III activity levels and clinical data obtained from patients who received liver transplantation from January 2017 to September 2018 were retrospectively analysed. A population pharmacokinetic (PK) model was developed using nonlinear mixed‐effects method and externally validated thereafter. Several dosing scenarios were simulated to maintain the plasma AT‐III activity level within the normal range using the developed PK model to search for an optimal AT‐III dosing regimen.

Results

The plasma AT‐III activity levels were best described by a single compartment model with first order elimination kinetics. The recovery of endogenous AT‐III level during the postoperative days was modelled using an E_max model. The typical values (95% confidence interval) of volume of distribution and clearance were 3.86 (3.40–4.32) L, and 0.129 (0.111–0.147) L h⁻¹, respectively. Serum albumin and body weight had significant effect on clearance and were included in the model. External validation of the proposed model demonstrated adequate prediction performance. Furthermore, simulation of previously suggested or modified dosing scenarios showed successful maintenance of AT‐III activity level within the normal range.

Conclusion

A population PK model of AT‐III concentrate was developed using data from liver recipients. Dosing scenarios simulated in our study may help establish a practical guide for AT‐III concentrate titration after liver transplantation.

What is already known about this subject

• Antithrombin III (AT‐III) concentrates have been used after liver transplantation to prevent or manage critical thrombosis.

• Pharmacokinetic models for AT‐III in normal population or patients with specific diseases have been reported, but there is little pharmacokinetic evidence for liver transplantation recipients.

What this study adds

• Pharmacokinetic characteristics of AT‐III concentrate in liver transplantation recipients are demonstrated.

• Dosing strategies for optimal AT‐III plasma activity level after liver transplantation are suggested through simulation.

1. INTRODUCTION

https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=2632 (AT‐III) is a liver‐synthesized glycoprotein that acts as a natural anticoagulant and functions as a major inhibitor of https://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=2362 and other components of the coagulation cascade such as factors IXa, Xa, XIa and XIIa, as well as plasmin.1, 2 Inactivation of coagulation factors by AT‐III is accelerated in the presence of heparin, making AT‐III a key component in heparin‐mediated anticoagulation.3 AT‐III deficiency or decreased activity may result from congenital deficiency or acquired conditions such as liver dysfunction, sepsis, cardiopulmonary bypass or major surgical procedures. Since decreased AT‐III activity level is associated with increased risk of vascular thrombosis or embolism,4, 5 AT‐III replacement is recommended for high‐risk patients, especially in the presence of thrombophilic conditions.6 Moreover, previous studies have reported positive effects of AT‐III supplementation on anticoagulation in such patient population.7, 8, 9

Hepatic artery thrombosis (HAT) is the most critical complication in the immediate postoperative period after liver transplantation, with a reported incidence of 4–15%.10 Portal vein thrombosis (PVT), which has been reported to occur in 2–7% of liver recipients, is another technical complication that may compromise graft survival.11 Both HAT and PVT may lead to acute graft failure, sepsis, and biliary complications, requiring endovascular or surgical interventions, and ultimately re‐transplantation. Risk factors of thrombotic complications include hypercoagulability, poor surgical technique, and anatomical variation.10, 12, 13 A hypercoagulable state due to the imbalance between stimulated coagulation activity and impaired fibrinolysis activity is often observed in the early postoperative period after liver transplantation.14 Preoperative prolonged hepatic dysfunction and delayed recovery of the transplanted graft may lead to depletion of anticoagulation factors. Plasma AT‐III activity levels have been reported to decrease to as low as 80% of the preoperative level in the early post‐transplant period and gradually normalize in the first 2 weeks after transplantation.3, 14, 15 Since lower plasma activity levels of AT‐III might be associated with hepatic artery thrombosis, AT‐III supplementation has been widely considered. However, the optimal range of plasma AT‐III activity level in the immediate postoperative period after liver transplantation is not known.

AT‐III concentrate is a human plasma‐derived agent obtained from healthy donors that is processed with modified Cohn ethanol separation and heated for viral inactivation.16, 17 Intensive anticoagulation therapy including AT‐III was reported to markedly decrease the incidence of HAT after paediatric liver transplantation.18 Kaneko et al. advocated AT‐III administration early after liver transplantation based on their results of a pilot study that it might reduce fibrin degradation product D‐dimer (FDP‐DD) levels.14 Another study showed that lower AT‐III activity was associated with portal vein thrombosis after splenectomy in patients with liver cirrhosis.19 AT‐III concentrate administration might help prevent postoperative PVT in these patients. Furthermore, AT‐III concentrate may potentially be an effective treatment for PVT in patients with liver disease.20

AT‐III concentrates have been used to prevent PVT or HAT in the early postoperative period after liver transplantation. With little evidence to guide AT‐III therapy, the dosage has been determined based on the surgeon's preference. Consequences of inadequate titration may result in potentially serious complications such as postoperative bleeding. Therefore, the relationship between the AT‐III dosage and the plasma activity levels of AT‐III during the immediate postoperative period after liver transplantation was evaluated.

2. METHODS

2.1. Study design

This study was a retrospective cohort study and the study protocol was approved by the institutional review board of Seoul National University Hospital (H‐1804‐114‐939). Informed consent was waived by the institutional review board due to the retrospective nature of the study design.

2.2. Patient selection and data collection

Adult patients (>18 years) who underwent living or deceased donor liver transplantation at Seoul National University Hospital from January 2017 to September 2018 were identified. Patients who received AT‐III concentrates postoperatively and had records of pre‐ and postdose AT‐III activity levels were included in the analysis. Paediatric liver transplant patients were excluded due to separate postoperative anticoagulation protocol regarding smaller vessel size and complex vascular reconstruction.21, 22 Patients with incomplete data or who underwent liver re‐transplantation were excluded as well to avoid potential bias. Patients who underwent liver transplantation from January 2017 to May 2018 were included in the model development dataset, whereas patients who underwent liver transplantation between June 2018 and September 2018 were included in the external validation dataset.

Data were extracted from electronic medical records from April to June of 2018 through retrospective chart review and clinical data warehouse. Baseline patient characteristics including age, sex, weight, diagnosis, donor status (living or deceased), graft‐recipient weight ratio and concomitant administration status of heparin were recorded. The dose and administration time of AT‐III, as well as the plasma AT‐III activity level and the sampling time were retrieved from the electronic medical records.

2.3. Antithrombin III administration protocol

During the first 7 days after liver transplantation, 500 international units (IU) of AT‐III were infused every 6 h per protocol unless there were evidence of postoperative bleeding or hypotension requiring vasopressors. Since there is insufficient evidence regarding AT‐III concentrate dosing for liver transplantation recipients, this protocol is based on clinician's experience and preference. The initial dose of AT‐III concentrate was administered as soon as the patient was transferred to intensive care unit after the operation.

2.4. Antithrombin III measurement protocol

Baseline (predose) plasma AT‐III activity level was the plasma AT‐III activity level routinely measured at the end of liver transplantation in the operating room. Subsequent sampling was performed from postoperative day (POD) 1, just before the first AT‐III concentrate administration of the day. Additional samples were collected from POD 2 to POD 7, as clinically required. Exact time for each drug administration and blood draw were recorded.

Plasma AT‐III activity level was measured with chromogenic assay using ACL‐TOP 750 CTS (Instrumentation Laboratory, Lexington, MA, USA) based on factor Xa as the enzymatic source. Factor Xa forms a complex with functionally active AT‐III when added to the plasma. When the chromogenic substrate (HemosIL Liquid Antithrombin, Instrumentation Laboratory) is added, the residual factor Xa generates color which is inversely proportional to the active AT‐III level.

2.5. Population pharmacokinetic analysis

Nonlinear mixed‐effects method was used to develop a population pharmacokinetic (PK) model using the NONMEM software version 7.4 (ICON plc, Dublin, Ireland). The model parameters were estimated using the first‐order conditional estimation with interaction method. The endogenously produced AT‐III levels were incorporated in the model as follows (Figure 1);

Figure 1.

Final model scheme for the population pharmacokinetic analysis of AT‐III abbreviations: AT‐III, antithrombin III; AT3_max, maximum increased antithrombin III level by endogenous production; CL, clearance; T₅₀, time to reach half of the maximum antithrombin III level by endogenous production; V, volume of distribution, η, interindividual variability; θ, typical population parameter values

Observed AT‐III activity level (%) = Increased AT‐III activity level from exogenous AT‐III concentrate (%) + Endogenous AT‐III activity level (%).

Single and multiple compartment models were evaluated to determine the basic structural model. The interindividual variability (IIV) was modelled exponentially to the model parameters, and the residual variability was modelled with various error models (additive, proportional and combined error models).

A previous study of AT‐III metabolism showed that AT‐III was produced at a relatively constant rate in humans.23 The elimination or distribution of endogenous AT‐III was not affected by the addition of exogenous AT‐III.24 Therefore the recovery of endogenous AT‐III activity was modelled independent of the administration of exogenous dose of AT‐III in this study. The recovery of endogenous AT‐III activity level after liver transplantation was modelled using an E_max model as follows;

$$\text{Endogenous}\mathrm{AT}‐\mathrm{III}\text{activity level}\left(\%\right)=\text{Predose}\mathrm{AT}‐\mathrm{III}\text{activity level}+\frac{\mathrm{AT}{3}_{\max }•\mathrm{POD}}{{\mathrm{T}}_{50}+\mathrm{POD}},$$

where the AT3_max is the maximum recovered AT‐III activity level during PODs and the T₅₀ is the time to recover half of AT3_max level.14 Every subject had a predose AT‐III activity level, so the observed value was used for the model. However, we could not accurately estimate the parameters (AT3_max and T₅₀) from our dataset because there were no patients who can represent the natural recovery course of endogenous AT‐III activity level (i.e. those who did not received AT‐III concentrate after the liver transplantation). Instead of estimating the AT3_max and T₅₀, both parameters were fixed based on several assumptions. First, we assumed that the fully recovered AT‐III level will be about 100%. The AT3_max was set as 70, because the mean baseline (predose) AT‐III level was about 30%. Secondly, the T₅₀ was initially fixed at 96 h (4 days) based on the natural endogenous AT‐III recovery profile observed in a published clinical study and sensitivity analysis was performed on T₅₀ to identify a reasonable parameter value.14

After the development of structural model and the error model, the effect of various covariates (age, sex, weight, diagnosis, donor status [living or deceased], graft–recipient weight ratio, estimated glomerular filtration rate calculated by modification of diet in renal disease study equation, albumin and platelet count) on the PK parameters were evaluated through the covariate screening. The potential covariates were screened by plotting against the empirical Bayesian estimates of PK parameters. A generalized additive model was also used for covariate screening. The continuous covariates were centred at the median values and modelled to the PK parameters using a power model. The categorical covariates were modelled to the PK parameters using indicator variables. The covariate selection was performed considering its statistical significance, reduction in the magnitude of IIV and precision of parameter estimates.

A likelihood ratio test, graphical evaluation, visual predictive checks (VPCs) were performed to assess the goodness of fit of model and the prediction performance of model. The precision of parameter estimates was evaluated by bootstrapping. The model parameters were estimated using 1000 bootstrap‐resampled data and its median and 95% confidence interval (CI) were compared to the original parameter estimates. The prediction performance of model for external validation dataset was evaluated by VPCs and by calculating percentage prediction errors (PE) as follows:

$${\mathrm{PE}}_{\mathrm{ij}}\left(\%\right)=\frac{\left({\text{Cobs}}_{\mathrm{ij}}-{\text{Cipred}}_{\mathrm{ij}}\right)}{{\text{Cipred}}_{\mathrm{ij}}}•100,$$

where the Cobs_ij is the observed plasma AT‐III activity level in sample j from patient i, and Cipred_ij is the individual predicted AT‐III activity level for that sample by the model. The series of PEs and absolute PEs (APE) were graphically presented. To evaluate bias and precision of the model prediction, the median PE (MDPE) and median APE (MDAPE) were calculated, respectively. If the MDPE value was between −20 and 20% and MDAPE value was <30%, the prediction performance of the PK model was regarded as clinically acceptable, based on the performance evaluation criteria of a typical target‐controlled infusion model.25

2.6. Simulations to optimize AT‐III dose regimen

Using the population PK model, Monte Carlo simulations were performed to explore the plasma AT‐III activity levels upon various dosing scenarios. The change of endogenous AT‐III activity level during the PODs, and the increment of AT‐III activity level by administration of AT‐III concentrate were simulated. The simulated dose regimens included a published clinical study dose regimen (1500 IU per day intermittently, PODs 0–2) and several alternative dose regimens required to fall within normal AT‐III activity level range (80–120%) were explored.14

2.7. Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY.

3. RESULTS

Data from 198 liver transplant patients were included in the analysis. Data from 161 patients were included in the model development dataset and used to develop the population PK model and perform internal validation. Data from 37 patients were included in validation dataset and used to perform external validation for the model. Patient characteristics are summarized in Table 1 and were similar between the model development dataset and the validation dataset except for body weight and diagnosis. There were no patients who were administered heparin during the liver transplantation or postoperative periods.

Table 1.

Patient characteristics

	Model development (n = 161)	Validation (n = 37)	P‐value
Age (y)	56 (19–76)	58 (22–76)	.314
Height (m)	1.65 (1.36–1.81)	1.67 (1.42–1.78)	.991
Weight (kg)	64 (41–106)	58 (42–78)	.002
Graft‐recipient weight ratio	1.09 (0.39–4.11)	1.23 (0.76–1.78)	.105
Baseline AT‐III activity level (%)	30 (7–83)	29 (12–56)	.525
eGFR (mL/min)	94.9 (6.8–250.4)	106.3 (28.6–174.4)	.933
Serum albumin (mg/dL)	3.1 (1.1–4.8)	3.4 (1.9–4.2)	.213
Platelet count (× 106/L)	67 (12–551)	58 (22–170)	.265
Sex			.684
Male	112 (69.6)	27 (73)
Female	49 (30.4)	10 (27)
Diagnosis			.003
Hepatitis B liver cirrhosis	32 (19.9)	1 (2.7)
Hepatitis C liver cirrhosis	3 (1.9)	0 (0.0)
Alcoholic liver cirrhosis	38 (23.6)	6 (16.2)
Hepatocellular carcinoma	65 (40.4)	25 (67.6)
Hepatitis	4 (2.5)	0 (0.0)
Others	19 (11.8)	5 (13.5)
Donor status			.546
Living donor	137 (85.1)	30 (81.1)
Deceased donor	24 (14.9)	7 (18.9)

Data are expressed as median (range) or n (%)

eGFR, estimated glomerular filtration rate

3.1. Population PK analysis results

To develop a population PK model and to externally validate the model, 516 and 185 plasma AT‐III activity level‐time data were used, respectively. The median (interquartile range, IQR) number of samples per patient were 3 (2–4) including one baseline sample, and the median (IQR) sampling time was 15 h (0–68 h) after the first dose. Most of the samples were taken during the first 5 days after liver transplantation (Figure S1). The sampling time points were regular and evenly distributed as the dosing intervals were regular (Figure S2).

The plasma AT‐III activity levels were best described by a single compartment model with first order elimination kinetics. The model was parameterized in terms of the AT3_max, T₅₀, the volume of distribution (V) and the clearance (CL). The IIV was modelled exponentially to the_, V, and CL (Figure 1). A proportional error model best described the residual variability.

The typical value of CL and V were 0.129 L h⁻¹ and 3.86 L, respectively. The IIV (coefficient of variation, %) of CL and V were 66.6, and 39.9%, respectively. Several T₅₀ values were tested in sensitivity analysis, and the objective function value of model was lowest when the T₅₀ values were fixed at 72 h after the first dose (Table S1). Therefore, the T₅₀ value in the final model was fixed at 72 h.

From the covariate screening, several covariates showed a significant association with CL and V (Table S2). The patient's body weight, donor status, sex, serum albumin and diagnosed disease had a significant association with CL. Serum albumin and body weight were included in the final model considering their statistical significance, reduction in the magnitude of IIV and precision of parameter estimates. The reduction of IIV were highest when the effect of albumin was included in the model. The patient's body weight was highly correlated with sex, so only the body weight was included in the model to avoid multicollinearity. The effect of body weight on CL was modelled by a power model using 0.75 for exponents because body weight has a well‐described scientific background and because the allometric relationship had been successfully applied in previous studies.26, 27 The effect of donor status was not significant when the effect of albumin was included in the model. The precision of parameters was unacceptable when diagnosed disease were included in the model. The screened covariates did not result in decrease of IIV of V so the effect of covariates was considered only for CL (Table S2). The formulas for final model are as follows:

$$\mathrm{Y}=f\left(\mathrm{\theta };\mathrm{t}\right)+f\left(\mathrm{\theta };\mathrm{t}\right)•\mathrm{\epsilon }$$

$$f\left(\mathrm{\theta };\mathrm{t}\right)=\frac{\mathit{\text{Dose}}}{V}·e^{-\frac{\mathit{CL}}{V}·t}+\text{endogenous}\mathrm{AT}‐\mathrm{III}\text{activity level}$$

$$\mathrm{CL}\left(\mathrm{L}/\mathrm{h}\right)=0.129•{\left(\text{Albumin}/3\right)}^{-2.26}•{\left(\text{Weight}/70\right)}^{0.75}•\exp \left(\mathrm{\eta }\right)$$

$$\mathrm{V}\left(\mathrm{L}\right)=3.86•\exp \left(\mathrm{\eta }\right)$$

$$\text{Endogenous}\mathrm{AT}‐\mathrm{III}\text{activity level}\left(\%\right)=\mathrm{Pre}‐\text{dose}\mathrm{AT}‐\mathrm{III}\text{activity level}+\frac{70•\mathrm{POD}}{72+\mathrm{POD}}$$

Based on goodness‐of‐fit plots, VPCs, and bootstrapping results (Figure 2, Figure 3, and Table 2), the established final model was robust and showed adequate prediction performance. The high observed AT‐III levels (levels >150%) were under predicted on a population level, but those were well predicted on an individual level (Figure 2a and Figure 2b). The proposed final model also showed good prediction performance for external validation dataset based on VPCs, PEs and APEs (Figure 3 and Figure 4). The MDPE and MDAPE were 8.1 and 26.2% respectively, and the prediction performance of PK model for external validation dataset was regarded as clinically acceptable.

Figure 2.

Goodness of fit plots for the final model

Figure 3.

Visual predictive check plot for model development dataset A, and external validation dataset B, circles represent observed antithrombin III plasma activity levels. The lines represent the median (red) and the 5th and 95th percentiles (blue) of the observed plasma activity levels. The shaded areas represent the 95% confidence intervals for the median (red) and the 5th and 95th percentiles (blue) of the simulated plasma activity levels

Table 2.

Final population pharmacokinetic model parameters

Parameters	Model estimate		Bootstrap result
	Estimate (%RSE)	95% CI	Median	95% CI
CL (L/h)	0.129 (7)	(0.111 to 0.147)	0.129	(0.111 to 0.149)
Serum albumin on CL	‐2.26 (30)	(−3.58 to −0.94)	−2.18	(−3.84 to −0.82)
V (L)	3.86 (6)	(3.40 to 4.32)	3.84	(3.42 to 4.41)
AT3max (%)	Fixed at 70		Fixed at 70
T50 (h)	Fixed at 72		Fixed at 72
IIV
IIV of CL (%)	66.6 (11)	(47.8 to 83.4)	66.2	(47.2 to 85)
IIV of V (%)	39.9 (23)	(12.6 to 56.9)	39.6	(15.5 to 60.9)
Residual variability (ε)
Proportional residual error (%)*	12.8 (13)	(9.5 to 16.1)	12.5	(9.1 to 15.8)

Abbreviations: AT3_max, maximum increased antithrombin III level by endogenous production; CI, confidence interval; CL, clearance; IIV, interindividual variability (coefficient of variation, %); RSE, relative standard error (%); T₅₀, Time to reach half of the maximum antithrombin III level by endogenous production, V, volume of distribution

^* Y = f(θ; x) + f(θ; x) • ε.

Figure 4.

Prediction errors A, and absolute prediction errors B, of the final population pharmacokinetic model when applied to external validation dataset red line denotes median prediction error (A) and median absolute prediction error (B). Hairline denotes proposed acceptable ranges for median prediction error and median absolute prediction error

3.2. Simulated plasma AT‐III activity levels and proposed dosing regimens

Various dosing scenarios were simulated using the final population PK model to explain natural course of endogenous AT‐III recovery after liver transplantation and to suggest optimal AT‐III concentrate dosing regimens. The simulated endogenous AT‐III activity level gradually normalized during the postoperative period, but it required >7 days after liver transplantation (Figure 5a). The simulated serum AT‐III activity level using regimen from a published clinical study (1500 IU per day intermittently, from POD 0 to POD 2) was successful in maintaining normal AT‐III activity level range after the second dose (Figure 5b).14 A modified intermittent dose regimen (2000 IU at POD 0, followed by 1000 IU per day for POD 1 and 2) was also successful in maintaining normal AT‐III activity level range. Moreover, the 95% prediction interval for simulated concentrations of the modified regimen was narrower than that of the previous regimen (Figure 5c). A continuous infusion regimen following a loading dose (loading dose, 2000 IU for 1 h; maintenance dose, 3000 IU for 72 h) was also successful in maintaining normal AT‐III activity level in the early postoperative period after liver transplantation (Figure 5d).

Figure 5.

Simulated plasma AT‐III activity levels according to various dosing scenarios line, median prediction; area, 95% prediction interval; hairline, normal AT‐III activity level range (80–120%). A, Change of endogenous AT‐III activity level during the postoperative periods; B, 1500 IU per day intermittently, for postoperative days 0–2; C, 2000 IU per day for postoperative day 0, followed by 1000 IU per day for days 1 and 2; D, a continuous IV infusion regimen after a loading dose (loading dose, 2000 IU for 1 h; maintenance dose, 3000 IU for 72 h). AT‐III, antithrombin III; IU, international units

4. DISCUSSION

A PK model of AT‐III for liver transplantation recipients was developed from this retrospective analysis. External validation using extra dataset showed that the new model successfully predicted plasma activity levels of AT‐III. In addition, simulations of potential dosing regimens to maintain clinically acceptable AT‐III activity level were performed.

The median baseline AT‐III activity level in liver transplantation patients were 30 and 29% respectively for the model development and validation datasets, which are much lower compared to severe sepsis or hereditary AT‐III deficiency patients (Table 1).28, 29 Of the 198 patients included in the analysis, 195 patients had preoperative levels of AT‐III activity and the median value was 44% (IQR 27.5–67). Our result are similar to a previous report that showed AT‐III activity levels may decrease to as low as 80% of the preoperative level in the immediate postoperative period after liver transplantation.14 Decrement in plasma AT‐III activity level can be attributed to the consumption of endogenous AT‐III during the surgical procedure. Moreover, prolonged hepatic dysfunction and delayed recovery may contribute to the deficiency in the early post‐transplantation period.3, 15, 30

The model developed in our study agrees with previous published PK models for AT‐III. For the basic structural model, one‐ or 2‐compartment model with linear elimination process were used to describe AT‐III concentration according to sparse or rich PK samples, respectively.27, 28, 29, 31 In our study, body weight and albumin were included in the PK model as covariates. The body weight showed a positive correlation with the CL of AT‐III, similar to the result observed in the population PK model developed for paediatric patients.27 Low serum albumin was associated with low AT‐III concentration in critical patients with disseminated intravascular coagulation due to increased vascular permeability in those patients.32 In our study, the serum albumin showed a negative correlation with CL, which results in low AT‐III concentration in the patients with low serum albumin. The vascular permeability can be also increased in postoperative patients and that can be one reason for the effect of serum albumin on CL of AT‐III.33

The body weight and the diagnosis were the factors significantly different between patients in the model development and the model validation datasets. However, the effect of body weight on the PK of AT‐III was considered and included in the developed final model. Besides, the difference in diagnosis is of no particular importance in this model, taking into account that most patients with hepatocellular carcinoma have underlying cirrhosis34 and that baseline AT‐III levels are similar between two groups. Also, the prediction performances of PK model for the external validation dataset was acceptable (Figure 4).

It is well known that anticoagulants such as heparin may affect AT‐III activity level. However, anticoagulants were not routinely administered during and after liver transplantation due to concerns of bleeding. The effect of anticoagulants were excluded from the model as patients did not receive any additional anticoagulation therapy.

Various dosing scenarios aiming for a normal AT‐III activity level range (80–120%) were simulated. The simulated AT‐III activity level gradually normalized during the postoperative period (Figure 5a), and the administration of AT‐III concentrate using the published clinical study regimen (1500 IU per day from POD 0 through 2) was helpful in restoring serum AT‐III activity level in the early postoperative period (Figure 5b). The simulated serum AT‐III profiles were similar to a previous study,14 which showed that the FDP‐DD level and the platelet concentrate transfusion rate were significantly lower in the AT‐III treated group compared to the non‐treated group, supporting the utility of AT‐III supplementation in the early postoperative period. Of the simulated dosing scenarios, a modified intermittent dose regimen or a continuous infusion regimen were found to be more adequate in maintaining the serum AT‐III activity level within the normal range with a narrower prediction interval (Figure 5c and 5d).

AT‐III concentrates have been used in liver transplant patients, but the dosage or administration strategy has relied on clinician's experience or preference. The regimen used in our institution was 500 IU of AT‐III concentrates every 6 h for 7 days after liver transplantation. Plasma AT‐III activity levels lower than normal range may increase the risk of thrombotic complications such as HAT or PVT.3, 18, 20 However, higher levels of AT‐III activity may result in surgical site bleeding. In this regard, the suggested AT‐III concentrate dosing regimens based on our PK model may be helpful for optimal AT‐III therapy.

The current study has several limitations. First, the target AT‐III activity level for liver transplantation patients is still unclear. Effective AT‐III activity level ranges have been proposed, but no consensus has been reached due to insufficient clinical evidence.14, 18 Therefore, the optimal AT‐III activity level and the dose–response relationship of AT‐III in liver transplantation patients require further study. Second, our PK model was based on sparse sampling data. Using serial samples obtained from a prospective PK study may result in a more complex and accurate PK model as described in other groups.28, 31 However, our model used actual patient data and the performance of PK model was clinically acceptable based on precision and prediction performances (Figures 3 and 4). Finally, we excluded data from paediatric liver recipients. Post‐transplantation HAT incidence is higher in paediatric patients compared to adult patients due to the smaller vessel size and complex vascular reconstruction.21, 22 Rather than extrapolating our data to the paediatric population, development of a population PK model using data from the paediatric liver transplantation population should be considered.

In conclusion, our study described a population PK model of human AT‐III concentrate in patients who received liver transplantation. The simulated dosing regimens from this PK model may serve as a practical guide for maintaining optimal AT‐III activity level after liver transplantation.

COMPETING INTERESTS

There are no competing interests to declare.

CONTRIBUTORS

Conceptualization and methodology: H.G.R., K.Y.

Data extraction: B.R.K.

Data analysis and validation: J.O.

Writing of manuscript: B.R.K, J.O,

Supervision: H.G.R., K.Y.

Supporting information

Table S1 Sensitivity analysis result for T₅₀

Table S2 Covariate screening result

Figure S1 Distribution of plasma antithrombin III levels in the liver transplantation patients during postoperative days

Figure S2 Distribution of plasma antithrombin III levels in the liver transplantation patients during dosing intervals

Kim BR, Oh J, Yu K‐S, Ryu HG. Pharmacokinetics of human antithrombin III concentrate in the immediate postoperative period after liver transplantation. Br J Clin Pharmacol. 2020;86:923–932. 10.1111/bcp.14200

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.