Is tranexamic acid exposure related to blood loss in hip arthroplasty? A pharmacokinetic–pharmacodynamic study

Abstract

Aims

Tranexamic acid (TXA) is an antifibrinolytic agent, decreasing blood loss in hip arthroplasty. The present study investigated the relationship between TXA exposure markers, including the time above the in vitro threshold reported for inhibition of fibrinolysis (10 mg l⁻¹), and perioperative blood loss.

Methods

Data were obtained from a prospective, double‐blind, parallel‐arm, randomized superiority study in hip arthroplasty. Patients received a preoperative intravenous bolus of TXA 1 g followed by a continuous infusion of either TXA 1 g or placebo over 8 h. A population pharmacokinetic study was conducted to quantify TXA exposure.

Results

In total, 827 TXA plasma concentrations were measured in 166 patients. A two‐compartment model fitted the data best, total body weight determining interpatient variability in the central volume of distribution. Creatinine clearance accounted for interpatient variability in clearance. At the end of surgery, all patients had TXA concentrations above the therapeutic target of 10 mg l⁻¹. The model‐estimated time during which the TXA concentration was above 10 mg l⁻¹ ranged from 3.3 h to 16.3 h. No relationship was found between blood loss and either the time during which the TXA concentration exceeded 10 mg l⁻¹ or the other exposure markers tested (maximum plasma concentration, area under the concentration–time curve).

Conclusion

In hip arthroplasty, TXA plasma concentrations were maintained above 10 mg l⁻¹ during surgery and for a minimum of 3 h with a preoperative TXA dose of 1 g. Keeping TXA concentrations above this threshold up to 16 h conferred no advantage with regard to blood loss.

What is Already Known about this Subject

• Preoperative tranexamic acid (TXA) reduces blood loss after hip arthroplasty.

• The minimal effective therapeutic plasma concentration of TXA is believed to range from 5 mg l⁻¹ to 10 mg l⁻¹.

• The optimal duration of TXA administration in hip arthroplasty is unclear.

What this Study Adds

• A single preoperative intravenous 1 g dose of TXA maintains TXA concentrations above 10 mg l⁻¹ for a minimum of 3 h.

• Keeping TXA concentrations above10 mg l⁻¹ up to 16 h confers no advantage with regard to blood loss in hip arthroplasty.

Introduction

Total hip arthroplasty surgery leads to hyperfibrinolysis 1, which can contribute to excessive bleeding. Tranexamic acid (TXA), as designed by the International Union of Pharmacology (IUPHAR), is a synthetic antifibrinolytic agent known to be effective in reducing blood loss and the need for red blood cell transfusion in orthopaedic surgery 2.

Based on in vitro and in vivo studies, the minimal effective therapeutic plasma concentration of TXA is believed to range from 5 mg l⁻¹ to 10 mg l⁻¹ 3.

On the basis of pharmacokinetic (PK) studies, various dosage regimens have been proposed for TXA, depending on the drug concentrations targeted and the period of time during which these concentrations have to be maintained 3, 4, 5, 6, 7, 8. However, the relationship between blood TXA concentration and blood loss has never been demonstrated. The optimal TXA concentration to be targeted is therefore unknown.

In hip replacement surgery, fibrinolytic activation begins just after the start of surgery and is maintained postoperatively up to 18 h 1. It is unclear whether keeping TXA concentration levels above the minimal therapeutic threshold for a prolonged time after surgery confers an advantage in terms of blood loss.

We performed a PK–pharmacodynamic (PD) analysis on the basis of data from a randomized study in hip replacement surgery comparing a perioperative infusion of TXA over 8 h following an initial preoperative bolus with a single preoperative bolus of TXA alone 9. We first conducted a population PK study to quantify patient exposure to TXA and to identify significant covariates for exposure. We then studied the relationship between drug exposure and blood loss in individual patients to evaluate whether the duration of TXA administration and the dose administered in the context of total hip arthroplasty could be optimized.

Methods

Study design

The PeriOpeRative Tranexamic acid in hip arthrOplasty (PORTO) study was conducted in accordance with the ethical principles stated in the Declaration of Helsinki, Good Clinical Practice and relevant French regulations regarding ethics and data protection. The protocol and amendments were approved by the relevant central independent ethics committee (Comité de Protection des Personnes Sud‐Est1). This study is registered at the French National Agency for Medicines and Health Products Safety (ANSM; ref: 130908A‐21), EudraCT (ref: 2013–000791‐15) and ClinicalTrials.gov (no. NCT02252497).

Details of the present double‐blind controlled study design are presented elsewhere [9]. Briefly, consecutive patients aged over 18 years undergoing scheduled cementless primary unilateral total hip arthroplasty surgery through a posterior approach at the University Hospital of Saint Etienne, France, were enrolled after giving their written informed consent to participate in the study. Exclusion criteria were pregnancy or breast‐feeding, hip fracture less than 3 months previously, contraindication for venous thromboprophylaxis with apixaban, chronic anticoagulation therapy, absence of French public health insurance coverage, and contraindication for TXA treatment (previous or current arterial or venous thrombosis, a creatinine clearance (CrCL) below 15 ml min⁻¹ or previous seizure). Exclusion criteria for patients with arterial or venous thrombosis were amended in June 2014 (2 months after study initiation, by which time six patients had been enrolled) to exclude only those with acute thrombosis, after modification of the French National Agency for Medicines and Health Products summary of product characteristics for TXA.

Patients were randomized in a 1:1 ratio into one of the two treatment groups. After the start of anaesthesia, patients in both groups received an unblinded 1 g instantaneous intravenous bolus of tranexamic acid (Exacyl, Sanofi Aventis, France). This was followed immediately by an intravenous infusion of either 1 g of tranexamic acid over 8 h (bolus‐plus‐infusion group) or a matching placebo (0.9% saline; bolus group). The correct administration of the treatment from the infusion pump was assessed by collecting the volumes delivered at various time points. The decision on whether to transfuse red blood cells was based on French guidelines 10. Use of a cell salvage device was not allowed. Postoperatively, apixaban 2.5 mg twice a day per os was given for thromboprophylaxis, initiated 24 h after the end of surgery and continued for 5 weeks.

PK analysis

Sample acquisition and drug assay

Sampling times for TXA concentrations were optimized using WinPOPT software (version 1.2). Venous blood samples were drawn into lithium heparin‐coated tubes at the following times: 3 min and 20 min after the preoperative bolus of TXA, at the end of surgery, and 3 h and 8 h after the preoperative bolus of TXA. The samples were then stored at −80 °C pending analysis. All the doses were given at the same time. Plasma TXA concentrations were measured by a validated liquid chromatography method, coupled with tandem mass spectrometry detection using an isotopically labelled internal standard. This procedure has been shown to be rapid, sensitive and accurate, and was linear over the concentration range of 0.8–200 mg l⁻¹. The intra‐ and inter‐day precision values were below 11.5% and accuracy was better than 9.6% 11.

Model development and evaluation

Data analysis was performed using MONOLIX modelling software (version 4.3, release 3, Lixoft). TXA concentrations were analysed using the following nonlinear mixed‐effect model framework:

$${\mathit{Obs}}_{\mathit{ij}}=F\left(t_{\mathit{ij}},{\varphi }_i\right)+\left(a+b\times F\left(t_{\mathit{ij}},{\varphi }_i\right)\right)\times {\epsilon }_{\mathit{ij}}$$

where Obs_ij denotes the observed data measured for patient i at time j; the function F(t_ij, ϕ_i) corresponds to the concentration predicted by the model for patient i at time j with individual PK parameters ϕ_i; and parameters a and b are, respectively, the constant and proportional components of the error model with ε_ij~ N (0, 1).

The stochastic approximation expectation maximization algorithm was used to estimate the maximum likelihood of the model 12. The parameters of the model were assumed to be log‐normally distributed. We built our model according to a stepwise procedure, initially identifying the best structural model for TXA pharmacokinetics without covariates (base model) by estimating one‐, two‐ and three‐compartment PK models.

In a second step, we examined the effect of covariates on TXA exposure. The covariates tested were selected according to the pharmacological and chemical properties of TXA 3. They comprised total body weight (Wt), ideal body weight (IBW) using the Devine formula 13, lean body weight (LBW) using the Janmahasatian formula 14, CrCL calculated according to the Cockcroft–Gault formula 15, using either Wt or IBW, and finally the Chronic Kidney Disease–Epidemiology Collaboration (CKD‐EPI) formula 16. Covariates were tested with allometric scaling according to the following equation, using clearance (CL) as an example:

$${\mathit{CL}}_i={\mathit{CL}}_{\mathit{pop}}\times {\left(\frac{{\mathit{\text{CrCL}}}_i}{{\mathit{\text{median}}}_{\mathit{\text{CrCL}}}}\right)}^{{\theta }_{\mathit{\text{CrCL}}}}\times e^{{\eta }_i}$$

where CL_i and CrCL_i are, respectively, the individual values of CL and CrCL for patient i, CL_pop is the estimated typical population value of CL, θ_CrCL is the estimated effect factor for CrCL, and ηi is the random effect for patient i.

Covariates were kept in the model if they improved the fit, reduced interpatient variability and decreased the objective function, calculated as –2log likelihood, by at least 3.84 compared with the previous model (χ2, P < 0.05 for one degree of freedom).

Model evaluation and selection were based on visual inspection of the goodness of fit and a visual predictive check (VPC). The goodness of fit was determined by plotting the observations vs. the population predictions of the model, the normalized prediction distribution errors (NPDE) vs. time, and the NPDE vs. predictions 17. The VPC was generated by performing 1000 simulations of the parameters of the subjects from the final model. The precision of the model to describe the observed data was evaluated by inspecting the distribution of the simulated concentrations compared with the observations.

PK simulations

To characterize the effect of significant covariates on TXA pharmacokinetics, we generated PK simulations from our population PK model using Mlxplore software (version 2016R1, Lixoft). Graphs of the results were generated using R software (version 3.2.2) with the ggplot2 package (version 2.1.0).

PK/PD analysis

The PD efficacy outcome analysed for TXA was perioperative blood loss. This outcome was based on haemoglobin (Hb) balance, assuming that blood volume on postoperative day 5 was the same as before surgery 18. The following formula was used:

$$\begin{array}{l}\text{Blood}\text{loss}\left(\mathrm{ml}\right)=1000\times \mathrm{Hb}\text{loss}/\text{Hbpre}\\ \text{with}\mathrm{Hb}\text{loss}=\left(\text{Hbpre}–\mathrm{Hb}5\right)\times \mathrm{BV}+\mathrm{Hbt}\end{array}$$

where Hbpre is the preoperative Hb concentration (g l⁻¹), Hb5 is the Hb concentration (g l⁻¹) on postoperative day 5, BV is the blood volume estimated according to Nadler's formula, taking gender, Wt and height into account 19, and Hbt is the total transfused Hb (g). We considered one unit of red cells transfused to contain 52 g of Hb [personal communication from the French National Blood Service (Etablissement Français du Sang), Saint Etienne, France].

To examine the relationship between TXA exposure and efficacy, we tested the correlation between perioperative blood loss and various TXA exposure markers estimated by our population PK model for each patient. The primary exposure marker tested was the time during which TXA concentration was above the 10 mg l⁻¹ threshold. Other markers were the area under the plasma drug concentration–time curve (AUC), TXA concentration immediately after the preoperative injection (C_max) and TXA concentration at the end of surgery. Sensitivity analyses were performed to assess the robustness of the results. The first studied the effect of setting the minimal therapeutic level of TXA at two other concentrations, 5 mg l⁻¹ and 20 mg l⁻¹. The second studied the effect of the duration of TXA exposure relative to surgery. For this analysis, four exposure markers were tested: the time during which the TXA concentration was above 10 mg l⁻¹ after surgical incision, the time during which the TXA concentration was above 10 mg l⁻¹ after skin closure, the AUC up to skin closure and the AUC from skin closure to infinity.

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 20.

Results

Patients

Between April 2014 and December 2015, 286 patients were screened for participation in the study, of whom 168 were randomized, 84 to the bolus‐plus‐infusion group and 84 to the bolus group. One patient in the bolus group withdrew his consent after randomization and before administration of the study drug; PK data were missing for one patient in the bolus‐plus‐infusion group. Finally, data from 166 patients were analysed, 83 in each group. Demographic variables did not differ between the groups (Table 1). The median interval between the preoperative TXA bolus and surgical incision was 33 min (range, 15–71 min). Overall, patients experienced a median blood loss of 876 ml (range, 0–2071 ml). Blood loss did not differ between the treatment groups. Two patients experienced symptomatic bilateral distal venous thrombosis up to day 45, both in the bolus group. No other vascular events or seizures occurred and no patients died.

Table 1.

Baseline patient characteristics and perioperative blood loss.

Patient characteristics	Bolus group (n = 83)	Bolus‐plus‐infusion group (n = 83)	Total (n = 166)
Female gender, no. (%)	39 (47)	47 (56)	86 (52)
Age, years	68 (24–94)	66 (26–93)	66 (24–94)
Height, cm	167 (142–185)	165 (148–185)	165 (142–185)
Total body weight, kg	76 (45–122)	73 (37–125)	75 (37–125)
Body mass index, kg m −2	28 (18–44)	27 (16–44)	27 (16–44)
Body mass index ≥30 kg m −2 , no. (%)	25 (30)	27 (33)	52 (31)
Ideal body weight, kg	61 (41–79)	57 (41–79)	59 (41–79)
Lean body weight, kg	52 (31–79)	49 (27–75)	50 (27–79)
Preoperative creatinine, μmol ml −1	74 (48–272)	70 (36–156)	72 (36–272)
Preoperative CrCL WT , ml min −1 a	88 (15–168)	86 (31–204)	86 (15–204)
Preoperative CrCL IBW , ml min −1 a	65 (17–124)	69 (23–142)	69 (17–142)
Preoperative CKD‐EPI, ml min −1 b	85 (17–129)	88 (28–122)	88 (17–129)
Interval from preoperative TXA bolus to surgical incision, min	34 (16–66)	32 (15–71)	33 (15–71)
Duration of surgery, min	68 (29–149)	70 (37–135)	69 (29–149)
Perioperative blood loss, ml	900 (0–2074)	868 (155–1827)	876 (0–2074)

Data are presented as the median (range) or number (%). TXA, tranexamic acid

^aCreatinine clearance estimated by the Cockcroft–Gault equation using, as body size descriptor, total body weight (CrCL_Wt) or ideal body weight (CrCL_IBW) 15

^bGlomerular filtration rate estimated by the Chronic Kidney Disease–Epidemiology Collaboration (CKD‐EPI) formula 16. Perioperative blood loss comprises intra‐ and postoperative blood loss up to day 5

Data sampled

A total of 827 TXA concentrations were measured for the 166 patients analysed, from 3 min to 8 h after the preoperative TXA bolus. Sampling times are presented in Table S1. No concentration was below the 0.8 mg l⁻¹ level of quantification of TXA. Figure S1 shows the plot of the observed TXA concentrations vs. time by treatment group. At the end of surgery, all patients had TXA concentrations above the therapeutic target of 10 mg l⁻¹. At the end of the 8 h infusion of TXA or placebo, concentrations of TXA were higher in the bolus‐plus‐infusion group than in the bolus group. The percentages of patients with TXA concentrations above 5 mg l⁻¹, 10 mg l⁻¹ and 20 mg l⁻¹ were, respectively, 100%, 99% and 59% in the bolus‐plus‐infusion group, and 55%, 15% and 0% in the bolus group (P < 0.001; chi‐square test for difference between groups).

Population PK modelling

The structural model best fitting the data was a two‐compartment model. Interpatient variability was estimated for CL, central volume of distribution (Vc), intercompartmental clearance (Q) and peripheral volume of distribution (Vp), with a covariance between CL and Vc. Residual variability was described by a proportional error model.

The covariate analysis showed that interpatient variability in Vc was best explained by Wt. Interpatient variability in CL was best explained by CrCL, using Wt as the body size descriptor. Table 2 displays the PK parameter estimates of the final model. Interpatient variability in Vc and CL in the final covariate model decreased by 16% and 27%, respectively, from the values estimated by the base model. Inclusion of covariates resulted in a reduction in objective function of 130 points. The median terminal half‐life of TXA was 2.3 h. The VPC showed that the observed data were well within the 90% prediction interval of the simulated data (Figure 1), confirming the good predictive properties of the model. The goodness‐of‐fit plots by treatment group for our final population PK model showed no apparent bias in model prediction (Figure S2).

Table 2.

Parameter estimates of the final population pharmacokinetic model.

Parameter	Population mean (% RSE)	Interpatient variability, % (% RSE)
CL (l h −1 ) = θ1 × (CrCL/85) θ2	–	22 (6)
Θ1	5.32 (8)	–
Θ2	0.548 (8)	–
Vc (l) = θ3 × (Wt/70) θ4	–	32 (7)
Θ3	6.09 (3)	–
Θ4	1 (fixed)	–
Q (l h −1 )	23.3 (4)	23 (26)
V P (l)	9.55 (3)	28 (10)
Covariance between CL and Vc	0.46 (17)	–
Proportional residual variance (%)	13 (5)	–

RSE, relative standard error; CL, clearance; CrCL, creatinine clearance; Q, intercompartmental clearance; Vc, central volume of distribution; Vp, peripheral volume of distribution; Wt, total body weight (kg)

CrCL (ml min⁻¹) was estimated by the Cockcroft–Gault equation, using Wt as the body size descriptor 15

Figure 1.

Visual predictive checks for the pharmacokinetic model. The 5th, 50th and 95th prediction intervals from the simulated concentrations of tranexamic acid (TXA) are plotted against time, with the observed data superimposed. Left panel: bolus group; right panel: bolus‐plus‐infusion group. y‐axes are on a log‐scale

PK simulations

Simulations of the TXA concentration–time course for the two regimens investigated in the study are presented in Figure 2. They illustrate the effect over time of Wt on C_max, and that of CrCL on drug exposure.

Figure 2.

Simulations of the concentration–time course of tranexamic acid (TXA). Panels A and C show predicted concentrations of TXA after a 1 g bolus. Panels B and D show predicted concentrations of TXA after a 1 g bolus followed by an additional infusion of 1 g over 8 h. Panels A and B illustrate the effect of total body weight on maximum plasma TXA concentration in a patient with a creatinine clearance (CrCL) of 85 ml min⁻¹ and weighing 50 kg (red line), 70 kg (green line) or 100 kg (blue line). Panels C and D illustrate the effect of CrCL on TXA exposure over time in a patient weighing 70 kg with a CrCL of 30 ml min⁻¹ (red line), 85 ml min⁻¹ (green line) or 150 ml min⁻¹ (blue line). y‐axes are on a log‐scale

PK/PD analysis

Estimation of TXA exposure by the population PK model indicated that the median time during which the TXA concentration was above 10 mg l⁻¹ was 8.7 h (range 3.3–16.3 h), being longer in the bolus‐plus‐infusion group (10.3 h) than in the bolus group (5.6 h; P < 0.01; Table 3). The AUC was also greater in the perioperative bolus‐plus‐infusion group (P < 0.01). Median TXA concentrations were similar in the two groups after injection of the preoperative TXA bolus (C_max), but the median TXA concentration at the end of surgery was higher in the bolus‐plus‐infusion group (P < 0.01; Table 3).

Table 3.

Estimated tranexamic acid (TXA) exposure.

	Bolus group (n = 83)	Bolus‐plus‐infusion group (n = 83)	Total (n = 166)	P
Time above threshold, h	5.6 (3.3–11.8)	10.3 (8.2–16.3)	8.7 (3.3–16.3)	<0.01
AUC 0–∞ , mg h l −1	197 (104–358)	348 (183–686)	268 (104–686)	<0.01
C max , mg l −1	148 (65–311)	163 (63–510)	155 (63–510)	0.09
TXA concentration at end of surgery, mg l −1	29 (16–51)	41 (24–80)	35 (16–80)	<0.01

Data are presented as the median (range). Time above threshold signifies the time during which TXA plasma concentration was above 10 mg l⁻¹. AUC_0–∞, area under the plasma drug concentration–time curve from 0 to infinity; C_max, maximum plasma concentration of TXA after the preoperative bolus

No relationship was found between blood loss and the time during which the TXA concentration was above 10 mg l⁻¹ (Figure 3). There was also no relationship between blood loss and any of the other exposure markers tested (AUC, C_max, TXA concentration at the end of surgery). Sensitivity analyses indicated similar results when two other thresholds were considered (time above 5 mg L⁻¹ and time above 20 mg l⁻¹). The duration of TXA exposure relative to surgical incision or skin closure similarly did not affect the results (Table S2 and Figure S3).

Figure 3.

Relationship between tranexamic acid (TXA) exposure and blood loss. Pink triangles indicate individual predictions of TXA exposure in the bolus group, blue circles indicate those in the bolus‐plus‐infusion group. The red line represents the regression line of the relationship with its 95% confidence interval (shaded blue region). AUC, area under the plasma drug concentration–time curve; C_max, maximum plasma concentration of TXA after the preoperative bolus; p, P‐value for the slope coefficient; r², square of the correlation coefficient. The time above threshold was the time during which the plasma TXA concentration was above 10 mg l⁻¹

Discussion

With the dosage regimens of TXA used in the present study of patients undergoing hip arthroplasty, no relationship was found between plasma TXA exposure and blood loss. An optimal plasma concentration–time profile could not be identified.

To quantify exposure to TXA, we first conducted a population PK study. Our final model was a two‐compartment model with Wt best describing interpatient variability in Vc, and CrCL best describing interpatient variability in CL. The PK parameter values estimated by this population model were in accordance with the pharmacological and chemical properties of the drug and the results of the initial PK studies conducted with TXA 3, 21. These studies showed that TXA is eliminated unchanged by glomerular filtration, with an elimination half‐life of about 2 h. One study in cardiac surgery observed a rise in TXA concentrations with increasing creatinine levels, and suggested that infusion of TXA during surgery should be reduced in patients with renal impairment 22. However, the two population PK models that were described in adults before our study did not identify CrCL as a covariate for CL, probably because they did not include patients with renal impairment 5, 6. Different descriptors of body weight were tested in our covariate analysis. LBW has been advocated as the most suitable descriptor to quantify the influence of body composition, in particular in obese patients 23. Although our study included patients with a wide range of body weights (from very severely underweight to very severely obese patients), use of LBW did not improve the fit of the model compared with Wt. Thus, Wt was chosen as it is more practical for routine use. To illustrate the effect of Wt and CrCL on plasma TXA exposure, we performed several simulations (Figure 2). These showed that: (i) C_max decreased with increasing Wt, and (ii) the period during which TXA concentrations were above the therapeutic threshold increased with renal impairment. For example, the time above threshold was similar for a patient with severe renal failure receiving a single bolus of TXA and for a patient with a CrCL of 150 ml min⁻¹ receiving a bolus plus an infusion of TXA.

Our PK analysis indicated that the additional infusion of 1 g of TXA over 8 h was effective in maintaining concentrations above the therapeutic target of 10 mg l⁻¹ at the end of this infusion in 99% of patients (Figure S1). As a result, the period during which TXA was above the desired target ranged from 3.3 h to 16.3 h (Table 3). However, we found no relationship between this exposure marker and blood loss (Figure 3). Maintaining concentrations of TXA above 10 mg l⁻¹ for a longer period than that observed with a single preoperative bolus of 1 g therefore does not seem justifiable.

A limitation of the present study was that the duration of the perioperative administration of TXA might not have been sufficient to observe a relationship between TXA concentration and outcome. This is unlikely, however, as another study in hip arthroplasty found no additional benefit of postoperative infusion of TXA over 18 h, using doses of TXA similar to those administered in our study 24. On the contrary, the duration of exposure to TXA may have been sufficient. At the end of surgery, all patients had TXA concentrations above the therapeutic target of 10 mg l⁻¹. In hip arthroplasty, fibrinolytic activity is maintained throughout the first postoperative day 1. Yet, inhibition of fibrinolytic activation, mediated by increased inhibition of tissue plasminogen activator, can also begin during surgery, resulting in a fibrinolytic shutdown that peaks on the day after surgery 25. If this transition from hyperfibrinolysis to a hypofibrinolytic state occurs soon after hip arthroplasty surgery, it could explain the absence of a relationship between blood loss and the duration of exposure to TXA above the therapeutic target.

Another limitation was the level of the target concentration we considered. This was based on the lowest minimal effective therapeutic plasma concentration of TXA reported, ranging from 5 mg l⁻¹ to 10 mg l⁻¹ 3. An in vitro study suggested that 10 mg l⁻¹ of TXA inhibited fibrinolysis by 80%. 26 This threshold was used to implement a dosage regimen of TXA in hip arthroplasty that was shown to be effective in reducing blood loss 4, 27. Other authors have proposed that higher TXA concentrations, from 20 mg l⁻¹ up to 150 mg l⁻¹, are maintained during cardiac surgery, based on data from in vitro studies but without studying the PK/PD relationship at these concentrations 5, 6, 22. Our sensitivity analyses indicated that the results were unchanged when considering a 5 mg l⁻¹ or 20 mg l⁻¹ threshold. The efficacy, in terms of reducing blood loss, of maintaining TXA concentrations during the perioperative period at levels above the target concentration we studied remains unknown.

Similarly, we found no relationship between blood loss and the other exposure markers tested. In particular, C_max was not predictive of blood loss. In other words, we could not identify a target concentration to be attained with the preoperative bolus dose of TXA. As C_max depends on Wt, it appears that it is not mandatory to adjust the loading dose of TXA according to Wt. However, the scope of this suggestion is limited to the relevant range of concentrations observed with a 1 g dose of TXA in hip arthroplasty and should not be applied to patients with extremely low or high body weight values. For a more accurate assessment of a potential concentration–effect relationship, a wider range of concentrations would need to be investigated in a dose–effect study. Indeed, a recent trial in hip replacement surgery indicated that a preoperative dose of 15 mg kg⁻¹ TXA was more effective than a preoperative dose of 10 mg kg⁻¹ in reducing blood loss 28. The optimal regimen for the preoperative administration of TXA remains to be determined.

Our study was also subject to other limitations. The first was related to the unreliability of the formulas used to calculate blood loss, in that they do not ensure that the values obtained are sufficiently accurate for absolute measurement 29. These formulas also assume that blood volume is identical at the beginning and at the end of the period during which blood loss is estimated. Thus, calculating intra‐ or postoperative blood loss is subject to additional measurement bias. We could not, therefore, evaluate independently the effect of TXA on intraoperative and postoperative blood loss. A further limitation of our study was that there were too few safety events to study their relationship with TXA concentrations. TXA is associated with an increased risk of postoperative seizures 30. Most events have been reported during cardiac procedures and seem to be dose related. None has been reported in orthopaedic surgery where lower doses are used. Furthermore, it is not known whether administering a bolus instantaneously, as in the present study, which results in immediate high plasma drug concentrations, is more harmful than a slow infusion.

In conclusion, TXA interindividual PK variability was partly explained by Wt and CrCL. A 1 g preoperative dose of TXA maintained plasma TXA concentrations above 10 mg l⁻¹ during hip arthroplasty and for a minimum of 3 h. Keeping TXA concentrations above this threshold up to 16 h conferred no advantage in terms of blood loss.

Competing Interests

This research received funding from the University Hospital of Saint Etienne, France, the study sponsor (sponsor protocol number 1308015). All the investigators are employees of the funding source, which had no role in the design or conduct of the study; the collection, management, analysis or interpretation of the data; the preparation, review or approval of the manuscript; or the decision to submit the manuscript for publication. No other conflicts of interest with respect to the conduct of this research or the contents of this article are declared.

The authors would like to acknowledge the contribution of the following investigators who participated in the PORTO study: Béatrice Deygas, Mathilde Donnat, Juanita Techer, Laurent Tordella, Jean‐Noël Fort, Unité de Recherche Clinique Innovation et Pharmacologie, University Hospital of Saint‐Etienne, France (coordinating and methodology centre); Nicolas Barbe, Julie Gavory, Sylvie Passot, Laetitia Burnol, Stéphanie Sève, Jean‐Yves Bien, and Pierre Lambert (clinical investigators).

Contributors

X.D. and P.J.Z. designed the study. S.H. and J.L acquired the data. X.D., J.L., E.O. and P.J.Z carried out data analysis. X.D., J.L. and P.J.Z interpreted the data and drafted the manuscript. All authors contributed to the manuscript and critical review before submission.

Supporting information

Table S1 Sampling times for tranexamic acid

Figure S1 Observed tranexamic acid concentration–time data.

Figure S2 Goodness‐of‐fit plots

Table S2 Estimated tranexamic acid exposure

Figure S3 Relationship between tranexamic acid exposure and blood loss

Figure S3A Time above 5 mg l⁻¹

Figure S3B Time above 20 mg l⁻¹

Figure S3C Time above 10 mg l⁻¹ after surgical incision

Figure S3D Time above 10 mg l⁻¹ after skin closure

Figure S3E Area under the plasma drug concentration–time curve from 0 to skin closure

Figure S3F Area under the plasma drug concentration–time curve from skin closure to infinity

Lanoiselée, J. , Zufferey, P. J. , Ollier, E. , Hodin, S. , Delavenne, X. , and for the PeriOpeRative Tranexamic acid in hip arthrOplasty (PORTO) study investigators (2018) Is tranexamic acid exposure related to blood loss in hip arthroplasty? A pharmacokinetic–pharmacodynamic study. Br J Clin Pharmacol, 84: 310–319. doi: 10.1111/bcp.13460.