Edoxaban pharmacokinetics during in vitro continuous renal replacement therapy

Abstract

Background

To evaluate the clearance of edoxaban during modeled in vitro continuous renal replacement therapy (CRRT), assess protein binding and circuit adsorption, and provide initial dosing recommendations.

Methods

Edoxaban was added to the CRRT circuit and serial pre-filter bovine blood samples were collected along with post-filter blood and effluent samples. All experiments were performed in duplicate using continuous veno-venous hemofiltration (CVVH) and hemodialysis (CVVHD) modes, with varying filter types, flow rates, and point of CVVH replacement fluid dilution. Concentrations of edoxaban and urea were quantified via liquid chromatography-tandem mass spectrometry. Plasma pharmacokinetic parameters for edoxaban were estimated via noncompartmental analysis. Two and three-way analysis of variance (ANOVA) models were built to assess the effects of mode, filter type, flow rate, and point of dilution on CL_CRRT. Linear regression was utilized to provide dosing estimations across CRRT effluent flow rates from 0.5 to 5 L/h. Optimal edoxaban doses were suggested using CL_CRRT and population non-renal clearance (CL_NR) to estimate total clearance and match the systemic AUC associated with efficacy in the treatment of venous thromboembolism.

Results

Edoxaban clearance from the CRRT circuit occurred primarily via hemofilter adsorption to the HF1400 and M150 filters at 74% and 65%, respectively, while mean percent protein binding was 41%. Multivariate analyses confirmed the lack of influence of CRRT mode, filter type, and point of dilution on the CL_CRRT of edoxaban allowing dosing recommendations to be made based on effluent flow rate. Edoxaban doses of 30-45 mg once daily were estimated to achieve target the AUC threshold for flow rates from 0.5 to 5 L/h.

Conclusion

For CRRT flow rates most employed in clinical practice, an edoxaban dose of 45 mg once daily is predicted to achieve target systemic exposure thresholds for venous thromboembolism treatment. The safety and efficacy of this proposed dosing warrants further investigation in clinical studies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12882-024-03777-7.

Introduction

Atrial fibrillation (AF) and venous thromboembolism (VTE) affect > 1 million people each in the U.S. and are major sources of cardiovascular morbidity, mortality, and excess healthcare costs [1, 2]. Renal disease is one of the most common comorbidities among patients with AF and/or VTE and is particularly problematic as it increases both the risk of developing cardiovascular disease and of associated complications such as hospitalization and death [2–4]. Acutely ill medical and surgical patients in the intensive care unit (ICU) are particularly vulnerable to new onset AF or VTE and exacerbations of existing cardiovascular disease, especially as more than half will experience acute kidney injury (AKI) in the ICU and many will require renal replacement therapy (RRT) [5, 6]. The need for RRT in the ICU is consistently associated with increased rates of AF and VTE and one year mortality rates as high as 50%. The high mortality is thought to be attributed in large part to the challenges in administering optimal preventative and therapeutic modalities given patient’s critical illness and the presence of pathophysiologic changes that alter the benefit to risk ratio of many agents. This is especially true among those requiring continuous RRT (CRRT) in the ICU given the paucity of available data to inform the pharmacokinetics (PK), pharmacodynamics (PD), safety, and efficacy of many agents that are routinely used outside the ICU setting and may have the potential to improve outcomes in critically ill patients if adequate data are generated [7–10].

One such class of agents are the direct-acting oral anticoagulants (DOACs) which have surpassed warfarin as first line therapy for stroke prevention in AF and treatment of VTE, with more than 3.5 million current DOAC users comprising over 75% of the total oral anticoagulant use in the U.S. alone [11–15]. Use of DOACs continues to increase outside the hospital setting and within the hospital setting owing to their encouraging outcomes, especially in older patients with multiple comorbidities including renal dysfunction where DOACs have demonstrated improved efficacy and safety over warfarin [16, 17]. Despite their favorable safety profile in this population, renal adjustments are complex and inappropriate dosing leading to untoward complications is unfortunately common [18–23] illustrating the need for additional and more easily applicable PK data for DOACs in this population. These agents may also have a role in more severely ill patients such as those with AKI requiring CRRT or even patients continuing home DOAC therapy in the ICU where an agent with a wide therapeutic index, decreased monitoring requirements, and few drug-drug interactions are especially advantageous as has been demonstrated throughout the COVID-19 pandemic in particular [24, 25].

The use of DOACs in patients with kidney disease including those requiring RRT is complicated as they are substantially renally cleared and patients with CKD stage 4 and higher and those on dialysis were excluded from randomized controlled trials, so the appropriate role of these agents remains controversial and additional data are needed to inform appropriate dosing. The current data on extracorporeal elimination of these agents is often limited to overdosage scenarios or limited by the quality and number of reported cases [26–31]. Edoxaban is the 4th DOAC and 3rd oral direct factor Xa inhibitor approved by the U.S. FDA since 2010 and is indicated for the prevention of stroke and systemic embolism in patients with non-valvular AF and for the treatment of VTE including deep vein thrombosis (DVT) and pulmonary embolism (PE) [24]. Similar to other available DOACs, edoxaban is highly renally cleared (~ 50% of total body clearance) and therefore requires a dose reduction in patients with renal impairment [32, 33]. Data suggests for patients with a creatinine clearance (CrCl) of 15–50 mL/min to reduce the dose to 30 mg qday [34]. Uniquely to edoxaban, a pre-specified subgroup analysis of the pivotal Phase 3 trial noted a non-significant decrease in relative efficacy of edoxaban in patients with AF and CrCl > 95 mL/min [35] which led to the U.S. FDA label containing a warning against using edoxaban in this population as a result of this analysis. Additionally, the physiochemical properties of edoxaban warrant consideration for CRRT including a molecular weight of 548.1 Da, lack of charge, 50% protein binding, volume of distribution of 107 L, and observed hemodialysis clearance of 101.7 L/min [32, 36, 37]. Given this possibly unique relationship with renal dysfunction, hyperdynamic function, and physiochemical properties additional PK data informing the optimal dosing of edoxaban in these CRRT scenarios are needed.

It is exceptionally challenging to manage pharmacologic dosing in patients on CRRT due to patient complexity and the variability of filtration rates, hemofilter types, and a paucity of published information on which to base scientifically sound dosing regimens. Within an in vitro model, volumes of distribution are uniformly maintained, comorbidities are absent, and the functional modalities of CRRT are the only independent variables. This tight control of confounders allows for accurate assessment and quantification of medication PK parameters within the CRRT system itself. This promotes the generalizability of the basic PK data through the clinical lens of each CRRT patient, empowering the clinician to more accurately assess the need for more advanced therapeutic monitoring and promote optimal outcomes for their patients. Further, the FDA Guidance for Industry on PK studies in patients with impaired renal function acknowledges that drug deposition during CRRT is different than that during intermittent hemodialysis (iHD) and that extrapolation from iHD to CRRT is problematic. As such, they suggest using in vitro CRRT data regarding transmembrane clearance combined with iHD clearance to estimate appropriate dosing recommendations until clinical CRRT data are available [38].

As such, the objective of this study was to evaluate the dialytic clearance of edoxaban during in vitro CRRT to provide initial guidance on dosing in this population.

Materials and methods

In vitro CRRT

In vitro CRRT was simulated using a Prismaflex 7.2 control unit (Baxter Healthcare Corporation, Deerfield, IL, USA) in continuous veno-venous hemofiltration (CVVH) and continuous veno-venous hemodialysis (CVVHD) modes using fresh 1.4m² polyarylethersulfone (PAES; Prismaflex HF1400) and 1.5m² acrylonitrile (AN69; Prismaflex M150) hemofilter sets for each experiment (Fig. 1). One liter of fresh heparinized (20 units/mL; West-ward Pharmaceutical Corp., Eatontown, NJ, USA) whole bovine blood (Densco Marketing Inc, Woodstock, IL, USA) was heated to 37 °C in a water bath and stirred continuously. Complete blood count of the fresh bovine blood demonstrated an erythrocyte count of 8.49 10⁶/µL, hemoglobin 13.6 g/dL, hematocrit 36.9%, and albumin 3.61 g/dL (Biologic Resources Laboratory, University of Illinois Chicago, Chicago, IL, USA), similar to values observed among critically ill patients with acute kidney injury undergoing CRRT [39]. The Prismaflex circuit was initially primed with 186 mL (HF1400) or 189 mL (M150) of 0.9% sodium chloride per the manufacturer’s operating instructions [40, 41]. Prior to the start of each experiment, blood was then allowed to circulate throughout the system for at least 10 min to permit adequate exposure of the hemofilter to blood proteins. The blood flow rate was fixed at 200 mL/min for all experiments while CVVH replacement fluid (PrismaSOL^® BGK 2/0; Baxter Healthcare Corporation, Deerfield, IL, USA) and CVVHD dialysate (PrismaSATE^® BGK 2/0; Baxter Healthcare Corporation, Deerfield, IL, USA) rates of 2 L/h and 4 L/h were tested with each filter type. During CVVH at 2 L/h, replacement fluid was added 100% pre-filter, 100% post-filter, and at 50% pre-/50% post-filter. During CVVH at 4 L/h, replacement fluid was added at 50% pre-/50% post-filter. All experiments were performed in duplicate in each mode, at each rate, and with each filter for a total of 24 experiments (excluding adsorption experiments).

Fig. 1.

In vitro Continuous Renal Replacement Therapy (CRRT) Simulation Models. Continuous venovenous hemofiltration (CVVH) with either pre-filter or post-filter replacement fluid (left), continuous venovenous hemodialysis (CVVHD) (center), and adsorption model (right)

Edoxaban tosylate powder (Savaysa™; Daiichi Sankyo Co., Ltd., Tokyo, Japan) was reconstituted from vials with sterile water per manufacturer’s instructions. To account for the measured bovine hematocrit, the dose of edoxaban added to the central blood reservoir was adjusted a priori to simulate the mean peak serum concentration (C_max) observed in healthy adult subjects following a single 60 mg oral dose (~ 0.33 mg/L) [42]. Urea (Sigma-Aldrich, St. Louis, MO, USA) was also added to the blood reservoir at 75 mg/L to serve as a control solute.

Drug was added to the central reservoir allowing for 1 min of equilibration (T = -1). Then, serial pre-filter blood samples were collected in K₂ EDTA vacutainer tubes (Becton Dickinson, Franklin Lakes, NJ) at 0 min, 10-, 20-, 30-, 45-, and 60-minutes post-equilibration with simultaneous post-filter blood and effluent samples collected at 10 and 30 min. Blood samples were immediately centrifuged at 1,500 x g for 10 min at 4 °C and the resultant supernatant plasma and ultrafiltrate samples were frozen at -80 °C within 30 min of collection until analysis.

Adsorption

To evaluate potential adsorption of edoxaban to the hemofilters, the initial CRRT model was modified to create a closed-circuit system using the CVVHD 2 L/h orientation as no replacement fluid is added in this orientation. Effluent was rerouted to the central blood reservoir, and 0.9% normal saline was exogenously pumped into the effluent bag via a Masterflex^® Peristaltic pump (Cole-Parmer, Vernon Hills, IL, USA) at the same rate to prevent the Prismaflex system from aborting due to the patient blood loss/gain alarm. Serial blood samples were drawn from the central reservoir at 0-, 10-, 20-, 30-, 45-, 60-, 90-, 105-, 120-, 150-, and 180-minutes, immediately centrifuged at 1,500 x g for 10 min at 4 °C, and the supernatant plasma was aliquoted and frozen at -80 °C within 30 min of collection until analysis. This process was performed in duplicate for both the HF1400 and M150 filter.

Protein binding

As the interaction between bovine proteins and edoxaban have not been investigated, we estimated the degree of bovine protein binding to edoxaban to compare to known human protein binding with edoxaban. If similar, this would allow for validation of our observed and calculated PK parameters. To assess edoxaban protein binding in bovine plasma, 4 contrived blood samples spiked at C_max were centrifuged in a fixed-angle rotor at 2,000 x g for 30 min at 37 °C using a Centrifree^® Ultrafiltration Device (Merck Millipore Ltd. Tullagreen, Carrigtwohill, Co. Cork, Ireland) with resulting total plasma and supernatant samples frozen at -80 °C within 30 min until analysis. These samples were then analyzed bioanalytically as below. The protein binding was averaged from the 4 contrived blood samples as determined by the following formula:

• % protein binding = 1 – (ultrafiltrate [edox] – plasma [edox]) x 100.

Bioanalytical procedures

Concentrations of edoxaban and urea in bovine plasma and effluent solutions were quantified via liquid chromatography-tandem mass spectrometry (LC-MS) (Keystone Bioanalytical, North Wales, PA, USA) according to previously validated methods [43]. The calibration range of the assay was linear from 0.005 to 5 mg/L (r ≥ 0.999). The precision and accuracy acceptance criteria for the quality control (QC) samples and calibration standards were ≤ 15% coefficient of variance (CV) and ± 15% relative error (RE).

Pharmacokinetic analysis

Pharmacokinetic parameters for edoxaban were estimated from observed pre-filter plasma concentrations via noncompartmental analysis in Phoenix WinNonlin Version 8.2 (Certara USA Inc., Princeton, NJ, USA). Reported parameters included: maximum plasma concentration (C_max), last observed plasma concentration (C_last), half-life (t_1/2), apparent volume of distribution (V_d), clearance (designated as CL_CRRT), and the area under the concentration-time curve (AUC_0−∞ and AUC_{0 − last}) as determined via the linear up-log down method. As in vitro experiments were performed over a period of one hour, AUC_{0 − last} was multiplied by 24 to approximate AUC_0 − 24. Estimated edoxaban and urea removal via the CRRT filter were calculated as follows:

• sieving coefficient (SC) = (2 * C_uf)/ (C_pre + C_post).

• saturation coefficient (SA) = (2 * C_dialysate) / (C_pre + C_post).

Where C_uf is the concentration in the ultrafiltrate, C_pre is the concentration from the pre-filter sampling port, C_dialysate is the concentration in the dialysate, and C_post is the concentration from the post-filter sampling port [44–46].

Adsorption was calculated as the difference between the total amount of edoxaban added to the system and the total amount recovered in the dialysate and plasma after 180 min using the following equation at each sampling time point [47]:

• Adsorption (%) = Ʃ1- [(dose of edoxaban added at time zero) / (concentration of edoxaban * measured volume in central reservoir)] x 100.

Dose optimization

As the edoxaban AUC has been associated with efficacy in the treatment of VTE over trough concentrations and recognized by the FDA as the PK metric of interest, it was utilized as the target parameter when estimating optimal CRRT dosing [48]. Since no specific thresholds have been identified, data from the Phase 3 ENGAGE-AF trials and associated population PK (popPK) analyses were utilized [48–50]. Based on popPK analysis, the mean predicted AUC in AF patients given 60 mg of edoxaban with normal renal function (CrCl > 80 mL/min) and body weight ≥ 60 kg were 1.739 mg · h/L. Therefore, optimal dosing was calculated to match an AUC of 1.739 mg · h/L. We calculated our AUC via the equation AUC = Dose / CL_T. Total clearance (CL_T) was the sum of renal clearance (CL_R) plus non-renal clearance (CL_NR). Clearance by CRRT in vitro (CL_CRRT) was substituted for renal clearance (CL_R) and added to non-renal clearance (CL_NR) to estimate CL_T. A previous analysis of 11 edoxaban studies with over 1100 subjected determined the predicted CL_NR/F to be 19.6 L/h from an average subject (~ 80 kg) with a CrCl of 100 mL/min [51]. The model predicted average absolute bioavailability of edoxaban was 58.3% which was imputed for F and CL_NR/F was assumed to be constant and residual renal function to be negligible [52]. We performed calculations for CRRT effluent flow rates from 0.5 to 5 L/h in 0.5 L/h increments to generate total daily dose estimations which were then converted to optimal dosing regimens by rounding to the nearest 15 mg based on the available tablet strength of edoxaban [32].

Statistical analysis

Data are presented as mean (± SD) or with 95% confidence intervals (95% CI). Continuous data were compared via Student’s t-test or Mann-Whitney U as appropriate. One- and two-way analysis of variance (ANOVA) models with Tukey’s post-hoc tests were built to evaluate significant differences in mean CL_CRRT according to CVVH point of dilution within and between each filter type, respectively. Then, three-way ANOVA models were fit using CL_CRRT as the outcome to evaluate the interaction between CRRT mode, filter type, and effluent flow rate. ANOVA-generated means and 95% CI of CL_CRRT were then used to estimate optimal total daily doses (TDD) of edoxaban during CRRT. Finally, multiple linear regression via backwards stepwise analysis was used to correlate flow rate with mean CL_CRRT while adjusting for covariates (CRRT mode, filter type, point of dilution, and effluent flow rate) and predict optimal dosing regimens across flow rates from 0.5 to 5 L/h. Model performance was assessed via the adjusted R² value. Collinearity was assessed via tolerance and variance inflation factor. A P value of ≤ 0.05 was considered statistically significant in the final model. A post hoc power calculation was performed for the linear regression model using the statistically significant covariate(s), a type 1 error rate of 0.05, observed R², and number of experiments (N = 24). All statistical analyses were performed using SPSS^® Version 26 (IBM Corp, Armonk, NY, USA).

Results

In vitro CRRT

Mean (± SD) pre-filter PK parameters of edoxaban in bovine plasma during CRRT stratified by mode, filter type, effluent flow rate, and point of replacement fluid dilution are displayed in Table 1, and respective plasma concentration-time profiles via LC-MS are shown in Fig. 2. Average (± SD) C_max and C_last values observed across the 24 experiments were 0.34 ± 0.06 mg/L (≤ 3% difference from targeted value) and 0.09 ± 0.02 mg/L, respectively. Clearance rates (CL_CRRT) estimated via noncompartmental analyses were similar to effluent flow rates and increased 1.6-fold as flow rates increased from 2 to 4 L/h (P = 0.90). The CL_CRRT ranged from 1.7 to 4.1 L/hr depending on effluent rate, replacement fluid strategy, and filter type (Table 1). The mean (± SD) apparent Vd was larger than the volume of the blood reservoir at 1.94 ± 0.45 L, likely due in large part to adsorption of edoxaban to the hemofilters as discussed below. Together these parameters resulted in an average AUC_0−∞ of 0.264 ± 0.07 mg · h/L and extrapolated AUC_0 − 24 of 3.64 ± 0.82 mg/L. Table 2 displays mean (± SD) SC and SA values. The mean edoxaban SC during CVVH with the M150 filter was 0.667 and 0.673 with the HF1400 filter (P = 0.977) and mean SA during CVVHD was 0.610 with M150 and 0.614 with HF1400 (P = 1.00). The average SC across all CVVH was 0.676 and SA across CVVHD was 0.624 (P = 0.852). The overall mean SC/SA was 0.623 across all CRRT modes, filter types, effluent flow rates, and points of replacement fluid dilution tested. The SC/SA decreased as flow rates increased from 2 L/hr to 4 L/hr except for the M150 filter, where the SC remained relatively constant (Table 2).

Table 1.

Mean (± SD) bovine plasma pharmacokinetic parameters of edoxaban during in vitro CRRT determined via noncompartmental analyses

CRRT modality	Cmax (mg/L)	t1/2 (h)	Vd (L)	CLCRRT (L/h)	AUC∞ (mg·h/L)	AUClast (mg·h/L)	AUC24 (mg·h/L)
CVVH	HF1400
2 L/h, 50/50%	0.34 ± 0.02	1.08 ± 0.03	2.35 ± 0.16	2.37 ± 0.09	0.278 ± 0.01	0.134 ± 0.01	3.229 ± 0.19
2 L/h, 100/0%	0.31 ± 0.05	0.90 ± 0.00	2.40 ± 0.43	2.93 ± 0.50	0.228 ± 0.04	0.125 ± 0.02	3.012 ± 0.56
2 L/h, 0/100%	0.28 ± 0.06	1.29 ± 0.23	2.59 ± 0.31	2.19 ± 0.15	0.300 ± 0.02	0.129 ± 0.01	3.100 ± 0.24
4 L/h, 50/50%	0.28 ± 0.04	0.60 ± 0.08	2.28 ± 0.36	4.08 ± 0.09	0.161 ± 0.00	0.110 ± 0.01	2.645 ± 0.25
CVVH				M150
2 L/h, 50/50%	0.31 ± 0.02	0.98 ± 0.18	1.82 ± 0.26	2.10 ± 0.66	0.329 ± 0.10	0.172 ± 0.04	4.122 ± 0.91
2 L/h, 100/0%	0.27 ± 0.01	0.93 ± 0.00	2.08 ± 0.18	2.40 ± 0.20	0.274 ± 0.02	0.145 ± 0.01	3.493 ± 0.32
2 L/h, 0/100%	0.38 ± 0.04	0.88 ± 0.02	1.57 ± 0.04	1.92 ± 0.09	0.342 ± 0.02	0.189 ± 0.01	4.532 ± 0.16
4 L/h, 50/50%	0.39 ± 0.09	0.52 ± 0.03	1.81 ± 0.57	3.87 ± 1.07	0.176 ± 0.05	0.131 ± 0.04	3.159 ± 0.09
CVVHD				HF1400
2 L/h	0.36 ± 0.04	0.87 ± 0.00	1.89 ± 0.21	2.38 ± 0.28	0.278 ± 0.03	0.155 ± 0.02	3.722 ± 0.37
4 L/h	0.38 ± 0.02	0.55 ± 0.00	1.74 ± 0.15	3.49 ± 0.40	0.189 ± 0.02	0.136 ± 0.01	3.261 ± 0.30
CVVHD				M150
2 L/h	0.40 ± 0.04	0.87 ± 0.09	1.37 ± 0.22	1.71 ± 0.08	0.385 ± 0.02	0.218 ± 0.03	5.220 ± 0.64
4 L/h	0.39 ± 0.04	0.49 ± 0.02	1.32 ± 0.23	2.94 ± 0.43	0.226 ± 0.03	0.173 ± 0.03	4.145 ± 0.68

Fig. 2.

Pre-filter plasma concentration-time profiles via liquid chromatography-tandem mass spectrometry (LC-MS) of edoxaban during in vitro CRRT in each mode, at each rate, and using each point of dilution with the M150 filter (left) and HF1400 filter (right). Mean values are displayed with error bars representing standard deviations

Table 2.

Mean (± SD) sieving coefficient (SC) and saturation coefficient (SA) values of edoxaban during in vitro CRRT

CRRT modality	HF1400	M150
CVVH	SC	SC
2 L/h, 50/50%	0.69 ± 0.12		0.66 ± 0.06
2 L/h, 100/0%	0.71 ± 0.05		0.68 ± 0.03
2 L/h, 0/100%	0.69 ± 0.04		0.66 ± 0.03
4 L/h, 50/50%	0.61 ± 0.05		0.67 ± 0.07
CVVHD	SA		SA
2 L/h	0.74 ± 0.13		0.63 ± 0.06
4 L/h	0.65 ± 0.08		0.59 ± 0.03

Effect of mode, filter type, effluent flow rate, and point of dilution on CLCRRT

Analysis of the effect of point of replacement fluid dilution on CL_CRRT across the 16 CVVH experiments did not reveal any significant impact, although mean CL was ~ 1 L/h higher in the 50/50% dilution group (3.25 ± 0.98 L/h) compared to the 0/100% (2.14 ± 0.27 L/h, P = 0.074) and 100/0% (2.30 ± 0.19 L/h, P = 0.131) likely driven by the inclusion of the 4 L/h effluent rates in the average of the 50/50% dilution group. Two-way ANOVA also did not demonstrate any significant interactions between CRRT mode, filter type, and/or point of dilution although the numerically higher mean CL in the 50/50% dilution group observed on one-way ANOVA was most pronounced (mean difference 1.31 L/h, P = 0.087) in CVVH mode with the HF1400 filter. There were also no significant two-way interactions between effluent flow rate and either CRRT mode or filter type, although flow rate alone significantly increased CL_CRRT from 2.18 L/h at an rate of 2 L/h to 3.60 L/h at 4 L/h (mean difference 1.416, 95%CI 0.971–1.860, P < 0.001) regardless of mode or filter type. Finally, analysis for the effect of CRRT mode, filter type, and effluent flow rate on CL_CRRT demonstrated no significant two-way interactions between CRRT mode and filter (P = 0.428), filter and effluent flow rate (P = 0.745), or CRRT mode and effluent flow rate (P = 0.245). The three-way interaction between CRRT mode, filter, and effluent flow rate was also non-significant (P = 0.968) with an adjusted R² of 0.703, while effluent flow rate again was the only significant covariate affecting CL_CRRT. The estimated marginal means and 95% CI for CL_CRRT generated from these ANOVAs as a function of CRRT mode, filter type, and effluent flow rate are displayed in Supplementary Table 1.

Adsorption

The mean (± SD) percent of the initial drug concentration not remaining in the closed circuit over the course of the 3-hour experiment and therefore assumed to be adsorbed to the hemofilter was substantial and differed significantly between the HF1400 and M150 filter types as expected at 74.2 ± 5.6% and 64.7 ± 9.4% (P < 0.001). Adsorption peaked over the first 30 min during which 35.4% and 41.1% of the available edoxaban was adsorbed to the HF1400 and M150 filter, respectively, after which adsorption plateaued corresponding to the minimal circulating edoxaban concentrations.

Protein binding

Protein binding of edoxaban in bovine plasma across the 4 contrived samples tested was 41.3 ± 4.6% and therefore congruent with the measured bovine albumin concentration of 3.6 g/dL, the inverse of the overall mean SC/SA observed of 0.377 (0.623^− 1), and consistent with the protein binding range of 40–59% previously reported in human subjects, notwithstanding differences in ultracentrifugation methods [53].

Dose optimization

All 4 applicable covariates (CRRT mode, filter type, flow rate, and point of dilution) were entered into the multiple linear regression model and only effluent flow rate was significant and therefore retained, demonstrating an 0.828 L/h increase in CL_CRRT for every one unit increase in effluent flow rate (95%CI 0.533–1.123, P < 0.001) with an adjusted R² of 0.702. This model demonstrated 99.9% power at an alpha of 0.05. The regression equation (CL_CRRT = 0.663 L/h + (0.828 · effluent flow rate (L/h)) was then used to make predictions for CL_CRRT and estimate optimal dosing recommendations for edoxaban during CRRT across simulated flow rates from 0.5 to 5 L/h (Table 3). Predicted edoxaban CL_CRRT ranged from 1.7 L/h to 7.6 L/h at effluent flow rates from 0.5 to 5 L/h, comprising just 8 to 28% of total body clearance in this model, considerably less than the typical ~ 50% contribution in patients with normal renal function. Total body clearance (CL_T) ranged from 21 to 27 L/h (Table 3). This resulted in a relatively narrow range of clearance CL_CRRT and subsequently CL/F values which allowed for a dose recommendation of 45 mg to achieve the target AUC threshold associated with efficacy in VTE treatment at all effluent flow rates > 0.5 L/h. Notably, the predicted exposure after the 30 mg edoxaban dose typically utilized for initial dose reductions in patients with renal dysfunction would have resulted in an AUC below this threshold at all effluent flow rates except 0.5 L/h. Additionally, the 45 mg has been studied in previous clinical trials and is readily commercially available in 15 mg tablets.

Table 3.

Optimal dosing recommendations of edoxaban according to CRRT effluent flow rate

CRRT effluent flow rate (L/h)	CLCRRTa/Fb (L/h)	CLNR/F (L/h)	CLT/F (L/h)	Goal AUC (mg · h/L)	Optimal total daily dose (mg)	Optimal dosing regimen (once daily)
0.5	1.7	19.6	21.3	1.739	37.0	30
1	2.4	19.6	22.0	1.739	38.2	45
1.5	3.0	19.6	22.6	1.739	39.3	45
2	3.7	19.6	23.3	1.739	40.5	45
2.5	4.3	19.6	23.9	1.739	41.6	45
3	5.0	19.6	24.6	1.739	42.7	45
3.5	5.6	19.6	25.2	1.739	43.9	45
4	6.3	19.6	25.9	1.739	45.0	45
4.5	6.9	19.6	26.5	1.739	46.2	45
5	7.6	19.6	27.2	1.739	47.3	45

^a Predicted via the linear regression equation: CL_CRRT = 0.663 L/h + (effluent flow rate (L/h) * 0.828)

^b Free fraction determined from 1 – observed protein binding = [1–0.37] = 0.63

*AUC = area under the curve, CL_CRRT = clearance from CRRT circuit, CL_NR = non-renal clearance, CL_T = total estimated clearance

Discussion

To our knowledge this is the first study to evaluate the transmembrane clearance (CL_TM) of edoxaban during CRRT and provides the first set of in vitro PK information for which to guide dosing. These results suggest that a dose of 45 mg of edoxaban may be adequate to achieve target exposure thresholds associate with efficacy during CRRT with effluent flow rates > 0.5 to 5 L/h. Importantly, CRRT removed edoxaban more efficiently in this study compared to iHD in previously published data but not so efficiently that it would approach a clearance level equivalent to that of patients with a CrCl > 95 mL/min where the drug is contraindicated. Previous published data suggests the edoxaban is not removed by iHD [31]. These findings further underscore the importance of thorough in vitro PK investigation to inform both clinical practice and future studies of agents with potential benefit in complex patients and as a reminder against attempting to extrapolate PK data across varying modalities of RRT or patient populations of varying levels of drug clearance.

In addition to providing clinicians with the first set of data for which to guide dosing of edoxaban during CRRT, our study has several other notable strengths. Primarily, our methodology for assessing CL_CRRT employed a rich PK sampling scheme and thorough statistical analyses which significantly improved our ability to accurately estimate drug removal during CRRT. The majority of previous studies employ a single sample design and attempt to estimate CL_CRRT by multiplying SC or SA derived from a single time point by the flow rate [54–57]. These methods falsely assume SC and SA are static over time and that CL_CRRT is directly proportional to flow rate across the continuum of CRRT settings. Moreover, the methods used for calculating SC, SA, and CL_CRRT have varied dramatically throughout the literature, even among the same authors/groups across different studies [45, 58–64], especially with regards to the influence of point of dilution during CVVH. Therefore, our optimal dosing regimens were generated exclusively via noncompartmental analyses given it is a more rigorous method compared to calculating the CL_CRRT via PK equations as performed in previous studies.

Finally, we also assessed the effect of protein binding and adsorption on the clearance of edoxaban during CRRT. Although protein binding is known to be one of the most important factors affecting drug removal during CRRT [65], exceedingly few agents have available data regarding binding to bovine plasma as these animals are not typically utilized in the drug development process [65]. As the measured protein binding of edoxaban in bovine plasma in our study was smalir to edoxaban protein binding observed in healthy human volunteers, it is unlikely that this significantly altered our optimal dosing recommendations. While crystalloid solutions are often used as the vehicle for drug delivery during in vitro CRRT studies due to accessibility and low cost [47, 67, 68], these solutions lack blood proteins vital for facilitating drug-protein binding and do not allow for the formation of a protein or fibrin layer on the extracorporeal circuit and hemodialyzer membrane. Albeit the use of modern, highly biocompatible hemofilters has often made drug adsorption negligible compared to the effect of filtration, it is critical to evaluate this component of removal from the circuit especially for moderately water soluble, lipophilic drugs like edoxaban. Given that peak adsorption almost always occurs within the first 5–30 min of CRRT [55, 69, 70], our 180 min experiments allowed ample time for the deposition of blood proteins to the hemofilter and circuit in order to assess the reversibility or saturation point of edoxaban adsorption [71]. Although the adsorption rates observed in this study appear substantial, the high % adsorption is likely an artifact of the very low circulating edoxaban concentrations along with other known limitations of performing adsorption studies in vitro including the use of fresh hemofilters for each experiment and the unequal sampling periods compared to clearance experiments [55, 69–73].

Despite these strengths, our study is not without limitations. First, although we included different CRRT machines, filter types, dilution points, and flow rates, the results may not be representative of all modalities of CRRT and are limited by a finite number of samples for each experiment. Second, we assumed non-renal drug clearance to be stable when estimating dosing recommendations. This allowed for ease of calculation though in reality, this could change, especially in the setting of AKI as there are some data to suggest AKI may affect non-renal clearance [74]. Unfortunately, there are currently no practical methods or useful biomarkers to assess changes in non-renal clearance and this will remain a limitation. Third, we assumed linear PK with our linear up – log down model but given the impact of initial drug adsorption during model creation, the assumption may be at risk. While we tried to ameliorate this limitation by starting with therapeutic concentrations of drug, in the clinical setting this is likely better accounted for by continued usage of filters allowing for plasma protein hemofilter saturation and repeat steady state dosing of the medication. Fourth, our calculations and dosing recommendations are based on a lack of residual renal function. While most patients on CRRT for AKI will be without residual renal function, those that have some retained or regained function, would need to have that renal clearance accounted for [75]. Fifth, while we employed rigorous methodology to our sampling, we observed some aberrations from predicted values. For example, we would expect the SC to decrease with increased flow rates, but this was not observed with the M150 hemofilter. Instead, the SC stayed constant across the experiments. This observation could be due to inadequate sampling, unexpected variance, or undetermined filter characteristics. Lastly, as the PK of edoxaban has already been extensively described, our 1-hour PK sampling scheme in this study was designed solely to evaluate the CL of edoxaban during CRRT, and therefore, the noncompartmental PK parameters reported should be interpreted considering this. As alluded to above, there are limitations to both components of total extracorporeal removal (adsorption + CL_CRRT) of any drug with in vitro modeling. Testing at different concentrations or ensuring a saturated hemofilter could help determine whether the observed clearance was secondary to adsorption versus CL_CRRT. While we did not test different concentrations, we did allow for sampling up to 1 h which was beyond the peak adsorption observed in our experiment of 30 min which could mimic a saturated hemofilter state. Longer experiments could ensure control of this variable. Despite these limitations, in vitro estimates have been shown to be predictive of CL_CRRT in patients undergoing CRRT and provide data on a range of settings that would otherwise would not be feasible to determine in a clinical study [76].

Conclusion

This study explored the dialytic clearance of edoxaban during CRRT in a tightly controlled in vitro setting. Edoxaban was approximately 41% bound to bovine plasma and demonstrated significant adsorption to the HF1400 and M150 filters at approximately 74% and 65%, respectively. For CRRT flow rates most commonly employed in clinical practice [77], a 45 mg once daily dose of edoxaban is predicted to achieve target systemic exposure thresholds. The safety and efficacy of this proposed dosing regimens warrants further investigation in in vivo studies of patients undergoing CRRT.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

E.W. and S.B. were responsible for conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, visualization, and writing/reviewing/editing the manuscript. L.A. contributed to the conceptualization, formal analysis, investigation, and writing/reviewing/editing the manuscript. K.D. contributed to the formal analysis, investigation, and writing/reviewing/editing the manuscript.

Funding

This work was supported by an investigator-initiated research grant awarded to S.B. by Daiichi Sankyo, Inc. Other than the named authors, the study sponsor had no role in study design, data collection/analysis, decision to publish, or preparation of the manuscript.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due confidentiality agreements with the study sponsor but may be available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

This research was funded by an investigator initiated grant supported by the manufacturer of edoxaban.