An In Vitro System for Studying Toxin Transport Across Hemodialysis Membranes

ABSTRACT

Background

Healthy kidneys regulate acid–base balance and electrolytes, produce hormones, and clear excess fluids and uremic toxins. Progressive kidney function loss in chronic kidney disease leads to end‐stage kidney disease (ESRD), where fluid and toxin accumulation causes severe complications and can be fatal. ESRD requires renal replacement therapy; kidney transplantation is preferred; however, limited availability of donor organs makes hemodialysis (HD) the most used treatment option. In the last years, many studies had focused on development of new membranes with improved toxin removal, however, these studies implement various experimental conditions making direct comparison between membranes challenging. Here, we develop an in vitro system for evaluating various membranes.

Methods

We performed toxin transport studies using commercially available hollow fiber membranes (FX1000, Fresenius) as well as newly developed mixed matrix membrane (MMM) hollow fibers, both assembled in a mini dialyzer. Creatinine (Cr), hippuric acid (HA) and indoxyl sulfate (IS) were used as model uremic toxins and their removal by the membranes from human plasma and full human blood under a range of experimental conditions was studied.

Results

We mainly implemented dialysate recirculation, which allowed for monitoring of the mass balance of toxins inside the mini dialyzer and for obtaining better understanding about the membrane fouling and concentration polarization phenomena. Dialysance normalized by membrane effective area was used as a key parameter for assessing the performance of the membranes.

Conclusions

Our findings are in good agreement to various literature studies for small and larger scale dialyzers suggesting that the proposed in vitro system can be used for comparison of membranes for HD.

Development of an in vitro hemodialysis (HD) system based on mini dialyzers for evaluating hemodialysis membranes under physiologically relevant conditions.

Abbreviations

ACP

activated carbon particles

Aeff

effective surface area

CKD

chronic kidney disease

CL

clearance

Cr

creatinine

CWF

clean water flux

DL

dialysance

ESRD

End‐Stage Renal Disease

HA

hippuric acid

HD

hemodialysis

HF

hollow fiber

IOF

inside‐out filtration

IS

indoxyl sulfate

MMM

mixed matrix membrane

NMP

N‐methylpyrrolidone

OIF

outside‐in filtration

PBUT

protein‐bound uremic toxin

PES

polyethersulfone

PVP

polyvinylpyrrolidone

Rc

recirculation

RRT

renal replacement therapy

SEM

scanning electron microscopy

SP

single pass

TMP

transmembrane pressure

1. Introduction

Kidneys have a very important role in our body balancing acid–base and electrolytes, producing hormones, and eliminating extra fluids and metabolic products. Nephrons are the functional and structural basic unit of the kidney, which filter blood and selectively reabsorb or excrete substances to maintain homeostasis. When the nephrons are damaged, kidney function is lost gradually, culminating to the fifth stage of chronic kidney disease (CKD), called kidney failure or End‐Stage Renal Disease (ESRD). Consequently, fluids and uremic toxins accumulate in blood causing several problems and, eventually, death. ESRD patients, to survive, need Renal Replacement Therapy (RRT) in the form of kidney transplantation, although this is not always an option due to low donor kidney availability and incompatibility issues. Therefore, the most used RRTs are peritoneal and hemodialysis (HD) [1]. The latter is typically performed three times per week for approximately 4 h per session and involves a dialyzer containing thousands of hollow fibers (HFs) that act as semipermeable membranes to filter the blood. During HD, blood flows through the fiber lumen while dialysate flows in the counter‐current direction, on the outside of the HFs, enabling the removal of excess fluids, electrolytes, and uremic toxins by diffusion and convection.

HD removes mainly water‐soluble and some middle molecular weight uremic toxins effectively. However, it does not adequately remove protein‐bound uremic toxins (PBUTs), therefore many studies focus on developing new membranes for improving their removal without albumin loss. On one hand, many studies are performed in vitro and report application of a variety of membranes with various characteristics (polymers, surface area, transport properties, etc.) and employ various experimental conditions, making comparison between membranes and between experimental conditions really challenging [2, 3]. On the other hand, most clinical HD studies report results of commercial large size dialyzers having various types of HFs and employ dialysis machines with various settings (e.g., blood flow rates, dialysate flow rates, treatment durations, and anticoagulation protocols), further complicating direct comparison of membranes. Moreover, the current ISO guidelines (ISO 8637‐1:2024 and ISO 8637‐2:2018) for testing dialyzers for clinical use describe experimental conditions suitable for full‐scale dialyzers (surface area A _eff typically around 1.5–2.0 m² for human therapy) and cannot be easily implemented to evaluate newly developed HFs produced in small scale. In our opinion, there is a clear need of having a facile in vitro system for evaluating the toxin transport of new HF membranes and for comparing to clinical studies [4]. Computational models, such as those by J. Yu et al., could further complement experimental work by simulating transport mechanisms in hollow‐fiber dialyzers [5].

This study aims to address the above‐mentioned need by developing an in vitro system using mini dialyzers (effective membrane surface area of 12.4 cm²), operated with human blood plasma and full human blood. To achieve this, we investigate the removal of three representative uremic toxins: creatinine (a small water‐soluble) and two PBUTs, namely hippuric acid (~40% bound to albumin) and indoxyl sulfate (> 90% bound to albumin) [6] under the following relevant HD conditions:

1. Single pass (SP) versus recirculation (Rc) of the dialysate fluid.

2. Various plasma and dialysate fluid flow rates (Reynolds number (Re) for both plasma/blood and dialysate in laminar regime (~30–180) as for clinical dialyzers [7]).

3. Toxin removal from human plasma and full human blood.

4. Blood/plasma flow; inside‐out filtration (IOF) versus outside in filtration (OIF).

(IOF: blood/plasma flows inside the lumen of the fibers while dialysate fluid flows around the fibers. OIF: blood/plasma flows around the fibers, while dialysate fluid flows inside the lumen of the fibers).

The mini dialyzer used here requires low total blood volume (~15–30 mL) [8] and can be suitable for in vivo studies with small animal models (i.e., rat), maintaining the extracorporeal circuit volumes below ~10% of the total blood volume to avoid hemodynamic instability [9]. This type of mini dialyzer has been used earlier for assessing blood compatibility of experimental membranes [10] and for performing in vivo studies with small animals [11, 12, 13]. However, to the best of our knowledge, our study is the first one focusing on development and implementation of this system for performing a comprehensive investigation of toxin removal by membranes.

We firstly study mini dialyzers with HFs from commercially available dialyzer FX‐1000 (Fresenius Medical Care, Bad Homburg, Germany). Then, the optimal conditions set there are used for investigating the toxin removal for Mixed Matrix Membranes (MMM) developed in our lab. The MMM are dual layer membranes combining diffusion and adsorption and in earlier studies have shown to remove effectively PBUT for human plasma [3, 14, 15]. Here we investigate for the first time their application for removing PBUT from full human blood. All results are reported based on toxin removal per membrane surface area (mg m⁻²) and dialysance, DL, estimated for blood/plasma and dialysate respectively (DL_p, DL_d in mL min⁻¹ m⁻²), represented by the slope of the total removal (mL m⁻²) against time (min). DL is the toxin removal normalized to the concentration gradient across the membrane. The ratio of DL_p/DL_d allows assessing important phenomena influencing toxin removal, such as toxin membrane interaction, membrane fouling and concentration polarization. Membrane fouling is mainly caused by interaction of blood components with the membrane leading to pore blockage and reduced transport across the membrane. It is generally considered irreversible phenomenon. Concentration polarization is caused by accumulation of molecules close to the membrane surface (either on blood or dialysate side) due to insufficient mixing there. This leads to creating a boundary layer with lower mass transfer coefficient (mass transfer limitations) which could even induce higher membrane fouling at the membrane surface. Concentration polarization is considered reversible phenomenon since can be minimized when local mixing close to the membrane is increased [16]. A DL_p/DL_d ratio close to 1 suggests that most of the toxins are removed to dialysate via diffusion. DL_p/DL_d > 1, indicates accumulation of the toxin to the membrane due to concentration polarization phenomena and/or toxin adsorption to the membrane. Finally, our results are compared to literature studies using small and larger size dialyzers.

2. Materials and Methods

2.1. Membranes

Two types of hollow fiber membranes were studied here. HFs obtained from commercial dialyzer FX‐1000 (Fresenius Medical Care, Bad Homburg, Germany) and MMM HFs developed in our lab [3]. For the MMM, two polymer solutions were prepared, one without particles and another with activated carbon particles (ACP); these were then coextruded using a spinneret with dry jet spinning technique to obtain a MMM dialysis hollow fiber. Polymer solutions were composed by Polyvinylpyrrolidone (PVP), Polyethersulfone (PES) and N‐methylpyrrolidone (NMP) as solvent, all products from BASF (Germany). Table 1 presents the compositions. The pump speeds used were: 0.8 mL/min for particle‐free solution, 0.9 mL/min for the solution with ACP and 0.5 mL/min for bore solution. The speed of the fiber pulling wheel was 7.2 m/min. The collected hollow fibers were washed for 3 days in Milli‐Q water. Afterwards, fibers soaked in 5% glycerol solution for at least, 24 h to preserve the pores when drying. Finally, fibers were stretched and dried at room temperature overnight. Milli‐Q water was flushed through the fibers at 760 mmHg for 2 h to ensure the complete elimination of glycerol prior to transport studies.

TABLE 1.

Compositions of the three polymer solutions used for spinning MMM hollow fibers used in this study.

Solutions	PES	PVP 90	PVP K30	NMP	H2O	ACP a
Particle‐free	16%	2.5%	2.5%	79%	—	—
Particle‐based	14%	1.4%	—	84.6%	—	60%
Bore	—	—	—	50%	50%	—

^a ACP percentage in respect to the total weight of solid material (polymer and ACP).

2.2. Fiber Characterization

2.2.1. Scanning Electron Microscopy (SEM)

Dried fibers were fractured in liquid nitrogen to obtain a clean cross‐section cut, sputter‐coated with Au, and imaged by SEM to assess wall architecture and ACP dispersion. Fiber dimensions (lumen diameter, outer diameter, layer thicknesses) were determined from these pictures.

2.2.2. Ultrafiltration Coefficient (K _UF) [17, 51]

The water permeance of the fibers and build quality of the mini dialyzers was tested using a clean water flux (CWF) set‐up in dead‐end mode. For these water transport experiments, a first pre‐compaction step of 760 mmHg transmembrane pressure (TMP) was performed for 30 min. Permeated ultrapure water was measured at different TMP values, ranging from 375 to 760 mmHg. The water mass measured with the scale in grams is considered volume (V) in milliliters and then divided by the time (t) in hours and the effective membrane surface area of the module (A _eff) in m² to obtain the clean water flux (mL m⁻² h⁻¹), following Equation (1).

$$\text{Clean water flux}\left(\mathrm{CWF}\right)=\frac{V}{t\times A_{\mathrm{eff}}}$$ (1)

The ultrafiltration coefficient (K _UF, mL m⁻² h⁻¹ mmHg⁻¹) was calculated as the slope of the linear fit of the flux (mL m⁻² h⁻¹) versus TMP (mmHg) graph. All mini dialyzers were tested with water to assure they were all in the same range of K _UF before being applied to studies with blood and plasma.

2.3. Assembling Mini Dialyzers

The mini dialyzers were built by using two 3‐way connectors (QST‐6 push‐in T‐connector, Festo, Esslingen, Germany) and 3 plastic tubes of about 5 cm each in length and 6 mm diameter. The mini dialyzers were built by firstly connecting the plastic tubes to the connectors (see Figure 1, white tubes for the outer sections, orange tube for the middle section). The resulting distance between the further ends of the two T‐connectors should be 6.5 cm (length where the fibers are not encapsulated by epoxy glue), defining the section where transport across the membranes occurs. 31 HF and 20 HF (FX1000, Fresenius, Bad Homburg v. d. Höhe, Germany) were put inside the filters for IOF and OIF measurements, respectively. Both ends of the filter were sealed using a two‐component epoxy glue (Klium, Hasselt, Belgium) that was applied around the fibers, completely filling the white tubes (see Figure 1). After at least 1 day of curing at room temperature, the excess length of both ends of the filters was cut with a scalpel, exposing the open fibers. This way, the only way that fluid can flow, is through the lumen of the fibers.

FIGURE 1.

Assembling the mini dialyzer. [Color figure can be viewed at wileyonlinelibrary.com]

The mini dialyzers with the MMM were prepared similarly as to the dialyzers with the commercial membranes. Before performing blood or plasma experiments, the mini dialyzers were washed with water for 2 h at 1 bar to ensure the complete removal of glycerol. All mini dialyzers used in this study had an average A _eff = 12.4 cm² (this value was used for the normalization of the transport data). The MMM dialyzers contained 12 HF per mini dialyzer, due to larger fiber size compared to FX1000. The effective membrane surface area of exchange of the mini dialyzers has been calculated using Equation (2).

$${\text{𝐴}}_{\mathrm{eff}}\left(\text{mini dialyzer}\right)=2\text{𝜋𝑟𝑛𝐿}$$ (2)

whereas r = radius of the fiber, distance between lumen‐center and the fiber's selective layer (selective layer of IOF fibers is in the internal surface and for OIF in the external surface), n = number of fibers inside mini‐dialyzer, L = effective membrane length, length between both ends of the mini‐dialyzer.

2.4. Dialysis Experiments

Two set‐ups have been used in this study: one from Convergence and the other from Repligen (see Figure 2). Both set‐ups consist of peristaltic pumps, pressure sensors and software for system control, allowing precise adjustments of flow rates and transmembrane pressure (TMP). Convergence set‐up was primarily used for SP and Rc experiments, It consisted of 4 pumps enabling 2 parallel HD experiments, using one pump for plasma/blood and another one for dialysate fluid. Full human blood experiments were conducted in Rc mode using the Repligen system. Both systems include 2 balances, one for dialysis fluid, and another one for plasma, to monitor volume exchange across the membrane. For our experiments, flow rates were selected to minimize TMP during the dialysis sessions and avoid convection across the membrane.

FIGURE 2.

Graphic representation of SP and Rc experiments (top) and Convergence (bottom‐left) and Repligen (bottom‐right) set‐ups. [Color figure can be viewed at wileyonlinelibrary.com]

Before transport studies, the blood/plasma tubing was rinsed with a saline solution (0.9 wt% NaCl) containing 200 I.U mL⁻¹ of heparin sodium, while the dialysate fluid tubing was flushed with Milli‐Q water for 1 h to wet the fibers and eliminate any air in the system. To prevent air from entering the system during connection to the blood container, the system was temporarily stopped. The tubing was then connected to the blood container, and the system was restarted, creating an interface between the saline solution and the blood. This interface was carefully flushed out before connecting the outlet of the blood side, ensuring a fully primed and air‐free circuit prior to starting the experiment.

Different HD settings were used during this study (see Table 2).

TABLE 2.

Transport results across the FX1000 and MMM mini dialyzers.

Membrane	FX1000					FX1000	MMM
Medium	Plasma					Full blood	Plasma	Full blood
Plasma/blood flow rate (mL min−1)	1	1	1.5	0.5	25	1.5	1	1.5
Dialysate fluid flow rate (mL min−1)	10	10	2.25	25	0.5	2.25	10	2.25
Type of dialysate flow	Sp	Rc				Rc	Rc
Mode of operation	IOF				OIF	IOF	IOF
Total Cr removed (mg m−2)	3360 ± 330	2374 ± 385	1883 ± 603	1061 ± 274	1196 ± 149	1079 ± 194	—	—
DLp Cr (mL min−1 m−2)	305 ± 73	228 ± 65	192 ± 34	133 ± 49	162 ± 39	166 ± 2	—	—
DLd Cr (mL min−1 m−2)	—	155 ± 37	215 ± 37	77 ± 14	111 ± 39	115 ± 54	—	—
DLp/DLd ratio	—	1.5 ± 0.4	0.9 ± 0.2	1.7 ± 0.6	1.5 ± 0.1	1.7 ± 0.8	—	—
Total HA removed (mg m−2)	1765 ± 451	1344 ± 40	1290 ± 231	897 ± 261	933 ± 166	507 ± 37	2381 ± 374	1539 ± 183
DLp HA (mL min−1 m−2)	156 ± 53	142 ± 21	85 ± 21	76 ± 36	79 ± 14	38 ± 8	123 ± 17	100 ± 16
DLd HA (mL min−1 m−2)	—	169 ± 15	66 ± 8	56 ± 8	54 ± 8	26 ± 6	16 ± 11	a
DLp/DLd ratio	—	0.8 ± 0.1	1.3 ± 0.3	1.3 ± 0.5	1.5 ± 0.1	1.5 ± 0.1	—	—
Total IS removed (mg m−2)	229 ± 130	207 ± 18	200 ± 146	91 ± 54	166 ± 27	14 ± 3	437 ± 107	104 ± 59
DLp IS (mL min−1 m−2)	41 ± 33	28 ± 3	26 ± 15	16 ± 14	22 ± 5	—	58 ± 13	12 ± 7
DLd IS (mL min−1 m−2)	—	17 ± 2	8 ± 1	6 ± 2	7 ± 1	1.4 ± 0.3	a	a
DLp/DLd ratio	—	1.2 ± 0.5	3.1 ± 1.2	2.6 ± 1.9	3.2 ± 0.6	—	—	—
	A	B	C	D	E	F	G	H

Note: A, B: 3.2 Effect of single‐pass versus recirculation; B–D: 3.3 Effect of different plasma/dialysate fluid flow rates; C, F: 3.4 Effect of different toxin medium; D, E: 3.5 Effect of Inside‐out versus Outside‐in filtration; G, H: 3.6 Dialysis experiments using Mixed Matrix Membranes. For all IOF experiments (FX1000), 31 fibers were used for a total surface area of 12.4 cm², while for OIF experiments (FX1000), 29 fibers were used for a total surface area of 12.4 cm². For MMM experiments, 11 fibers were used for a total surface area of 12.4 cm².

^a Below the detection limit, therefore calculations were not possible to make.

2.4.1. Dialysate Fluid

Human plasma and full human blood experiments were conducted to evaluate the efficiency of hemodialysis in two configurations: Recirculation (Rc) and Single pass (SP). Recirculation mode of the dialysate fluid allows for better tracking of toxin removal (e.g., diffusion, adsorption), because the toxin concentration in the dialysate fluid as well as in the plasma can be measured in time. This way, the amount of toxins lost in plasma can be compared to the amount of toxins gained in the dialysate fluid. The experiments were designed to simulate clinical conditions and were performed using a Convergence crossflow set‐up.

In Rc, 100 mL of dialysate fluid was prepared using the following composition: 93.84 parts of Milli‐Q water, 3.94 parts of NaHCO₃ and, 2.22 parts of HCl 1 M. The dialysate fluid inlet and outlet of the mini dialyzer was connected to the same beaker containing the dialysate fluid. For single‐pass configuration, 3 L of dialysate fluid were prepared following the ratio of the recirculation configuration. In this single‐pass configuration, the mini‐dialyzer's dialysate fluid inlet was placed in a container with the 3 L of dialysate fluid, and the outlet was placed in a different container. In this configuration the mass of the fluids could not be tracked using scales. Unless stated otherwise, all experiments were done in recirculation mode.

2.4.2. Human Plasma Protocol

Plasma samples were obtained from healthy donors from Sanquin (Deventer, The Netherlands). Prior to use, plasma was filtered using Greiner bio‐one Easy Strainer 70 nm (Greiner Bio‐One, Greiner bio‐one Easy Strainer 70 nm) and a control sample of pristine plasma was taken. The plasma was then spiked with either PBUTs (HA and IS) or Cr. The concentration of PBUTs in the plasma was 110 mg L⁻¹ for HA and 40 mg L⁻¹ for IS, and 1 g L⁻¹ for Cr, to simulate real ESRD patients. After that, the mini dialyzer is added to the system and washed with the same solutions. When the mini dialyzer is wet and free of air bubbles, the plasma and dialysate fluid are connected to the system. The experimental time starts (t = 0), when the front of the plasma flow reaches the plasma container. During the experiments, 0.6 mL samples of plasma and dialysate fluid were taken at times t = 0, t = 30 min, t = 1 h, t = 2 h, t = 3 h and t = 4 h, to determine the uremic toxin concentration changes in both solutions during the dialysis session.

To prepare the samples for analysis, they were diluted four times in Milli‐Q water, heated at 95°C for 15 min, and subsequently cooled in ice for 10 min. Then, all samples were filtered using Amicon Ultra 0.5 mL 10 K filters using a centrifuge at 18 407 rcf for 15 min and subsequently transferred into vials and analyzed using HPLC (Jasco, Tokyo, Japan). Toxin removal was assessed using HPLC and a NanoDrop spectrophotometer. HPLC was used for the analysis of HA via UV–Vis detection and IS via fluorescence detection, ensuring selective and sensitive quantification [18]. Cr was measured separately using the NanoDrop (UV–Vis) to avoid potential interference in the HPLC UV–Vis detector. From the t = 0 sample, two different samples were prepared: t0 and t0n. The only difference between these two samples is that the t0n sample, is not heated. Therefore, proteins present in this plasma sample were not denaturated and can be used to calculate the amount of toxins that were bound to the proteins.

2.4.3. Full Human Blood Protocol

The following protocol has been developed for this in vitro system and has been extensively tested with full human blood. Full human blood samples were obtained from the donor service of University of Twente (the Netherlands) from healthy donors. The research does not fall within the scope of the Dutch Medical Research Involving Human Subjects Act. Blood was collected from pseudonymized healthy volunteers in 10 mL blood collection tubes containing Heparin‐lithium as anticoagulant (BD, Franklin Lakes, USA) at the University of Twente. Informed consent was obtained from all volunteers, and the used blood collection procedure was approved by the local Medical Research Ethics Committee (METC Twente, reference K11‐23). As soon as the blood sample was received, heparin sodium solution (20.000 I.U. mL⁻¹) was added to obtain a concentration of 667 I.U mL⁻¹ (2 mL of heparin solution to 60 mL of blood), and the hematocrit was measured subsequently. Coagulation time was tested before each experiment (> 1000 s). Then, the blood was centrifuged at 937 rcf for 15 min to separate the plasma from the blood cells. At that point, the control sample is taken from the plasma to determine the baseline toxin concentrations. Then, to achieve the hematocrit of ESRD patients (~35% [19]), the ratio of plasma and red blood cell fraction was calculated and adjusted to this required percentage.

Before adding the cell fraction back to the plasma, the uremic toxins (Cr, IS and HA) were added to the plasma, to achieve predetermined toxin concentrations and to prevent hemolysis. The calculated amount of blood cells was then added back to the plasma, and the hematocrit was measured again to assure that it is ~35%.

A 50 mL falcon tube containing the spiked blood was placed on a nutating mixer (Fisherbrand Nutating Mixer) with three needles on the upper part of the tube, one for the inlet, one for the outlet and one for taking the samples with 1 mL syringes, see Figure 3.

FIGURE 3.

Graphic representation of the blood reservoir used in full human blood experiments in this study. [Color figure can be viewed at wileyonlinelibrary.com]

Samples of blood and dialysate fluid (0.6 mL) were taken at timepoints t = 0, t = 20 min, t = 40 min, t = 80 min, t = 120 min, t = 160 min, t = 200 min and t = 240 min and stored in 1.5 mL Eppendorf tubes. The Eppendorf tubes were centrifuged at 937 rcf for 15 min to separate plasma from blood cells. The plasma was taken out to different Eppendorf's tubes. The plasma samples were diluted four times in different Eppendorf's with Milli‐Q water (100 μL plasma +300 μL Milli‐Q water). Like the human plasma protocol, two samples were made, t0 (time 0) and t0n (time 0 with not‐denaturized proteins) from the t = 0 plasma sample. All samples, but the t0n sample, were heated at 95°C and cooled in ice for 10 min, subsequently. Afterwards, all samples (t0n included) were filtered using Amicon Ultra 0.5 mL 10 K filters using a centrifuge at 18 407 rcf for 15 min to obtain a clear solution eligible for further analysis of the toxin concentration. Toxin removal was assessed using HPLC and a NanoDrop spectrophotometer. HPLC was used for the analysis of HA via UV–Vis detection and IS via fluorescence detection, ensuring selective and sensitive quantification [18]. Cr was measured separately using the NanoDrop (UV–Vis) to avoid potential interference in the HPLC UV–Vis detector. These methods provided precise concentration readings pre‐ and post‐dialysis, allowing accurate calculation of solute removal efficiency.

2.4.4. Toxin Removal

In this study, the toxin removal normalized to the membrane surface area was estimated as total toxin removal (mg m⁻²), dialysance and/or clearance (mL min⁻¹ m⁻²). Clearance and dialysance from plasma/blood were estimated as follows:

$$\text{Dialysance},\mathrm{DL}=\frac{\frac{X_{\mathrm{p}}}{t·A_{\mathrm{eff}}}}{\left(C_{\mathrm{p}}-C_{\mathrm{d}}\right)}$$ (3)

where X _p is the amount of toxins removed from human plasma (or human full blood) and X _d (replaces X _p in the equation when DL_d is being calculated) is the amount of toxins transported to the dialysis fluid (mg) after a certain time t (min). A _eff is the mini dialyzer effective surface area (in m²). C _p and C _d (both in mg mL⁻¹) are the concentrations of the toxins in the plasma or blood and in the dialysis fluid, respectively. If there is no adsorption of toxins on the membranes, it is expected that DL_p/DL_d = 1. If there is toxin adsorption on the membranes, DL_p/DL_d > 1. In conventional hemodialysis, the dialysis fluid is constantly refreshed (C _d = 0); therefore, Equation (3) can be simplified to Equation (4) and corresponds to the toxin clearances normalized to the membrane surface area in the plasma/blood and dialysis fluid.

$$\text{Clearance},\mathrm{CL}=\frac{\frac{X_{\mathrm{p}}}{t·A_{\mathrm{eff}}}}{C_{\mathrm{p}}}$$ (4)

2.5. Statistics

All data is presented as a mean ± standard deviation. One‐way ANOVA and student's t‐test were used as appropriate. This statistical analysis was performed using GraphPad Prism (version 10.4.1). A p value of < 0.05 was considered significant.

3. Results

3.1. Membrane Characterization

Figure 4 presents typical SEM images of the hollow fiber membranes studied here. The Fresenius FX1000 has a spongy and homogeneous porous structure and a selective layer in the fiber lumen. It has inner diameter of ~190 μm and outer diameter of ~260 μm (see Figure 4B1 and B2). The dual layer MMM contains macrovoids in both the inner selective (particle‐free) and in the outer layer (containing the ACP particles, see Figure 4A1, A2). Compared to FX1000, it has larger dimensions with an inner diameter of 432 ± 5 μm and a total diameter of 690 ± 4 μm. The K _UF of the membranes was estimated to be 60 mL m⁻² h⁻¹ mmHg⁻¹ for the FX1000, and 41 mL m⁻² h⁻¹ mmHg⁻¹ for the MMM.

FIGURE 4.

Typical SEM images of MMM and FX1000 membranes. A1 shows the cross‐section of the MMM. A2 shows a close‐up of the same fiber where both membrane layers are shown. B1 shows a cross‐section of a Fresenius FX1000 HF and B2 shows a close‐up of the same single layer HF. [Color figure can be viewed at wileyonlinelibrary.com]

3.2. Single Pass Versus Recirculation of Dialysate Fluid on Uremic Toxin Removal

In this section, we compare the uremic toxin removal by the FX1000 mini dialyzers when using SP versus Rc of dialysis fluid. In SP, fresh dialysate passes through the dialyzer once and is discarded. This is employed in clinical settings for maximizing concentration gradient between the patient's blood and the dialysate fluid. In Rc dialysate flow, a fixed volume of dialysate is recirculated. If the dialysate volume is not large enough, toxin accumulation can occur there in time resulting in lowering the concentration gradient across the membrane over time. The investigation of Rc is very relevant in the context of an in vitro system since it allows of monitoring closely the mass balance of the blood components/toxins and the kinetics of toxin removal. The latter can indicate membrane fouling and concentration polarization phenomena [15, 20, 21, 22] which are important for comparing HF membranes.

FX1000 HFs were used here in IOF mode with 1 mL min⁻¹ of human plasma flow and 10 mL min⁻¹ dialysate flow using, for the latter, either SP or Rc mode (see Table 2). Figure 5 presents the kinetics of removal of Cr, HA and IS (A, C and D respectively) and their total removal toxins after 4 h (B, D and E respectively). Table 2 summarizes the results. Figure S1 (Appendix S1) presents the total toxin removal over time by the mini dialyzers.

FIGURE 5.

Toxin removal (mL m⁻²) in time (min) by FX1000 fibers, using SP or Rc of dialysate fluid (A: Cr, C: HA, and E: IS). Also in this figure, the total toxin removed after 4 h of hemodialysis (mg m⁻²) via SP or Rc (B: Cr, D: HA, and F: IS, N = 3). The *indicantes that there is a significant difference between the two data sets (p < 0.05). [Color figure can be viewed at wileyonlinelibrary.com]

For Cr, application of SP achieved higher removal than with Rc over a 4 h experiment. With SP, ~3360 mg m⁻² of Cr was removed versus ~2374 mg m⁻² with Rc. Plasma dialysance is also consistent with those results, showing a DL_p of ~305 mL min⁻¹ m⁻² in SP versus ~228 mL min⁻¹ m⁻² in Rc. In Rc mode, the dialysate‐side dialysance DL_d was ~155 mL min⁻¹ m⁻², yielding a DL_p/DL_d ratio of about 1.5, suggesting that some Cr remained in the circuit or adsorbed to the membrane. For SP, since the dialysate is constantly refreshed, a DL_p/DL_d ratio cannot be calculated.

For protein‐bound toxins like HA and IS, removal depends on the plasma–dialysate concentration gradient established by the free (unbound) toxin fraction. In Sp, HA total removal was slightly higher (1765 ± 451 mg m⁻², DL_p ~156 mL min⁻¹ m⁻²) than in Rc (1344 ± 40 mg m⁻², DL_p ~142 mL min⁻¹ m⁻²) though not significantly (p = 0.259). For IS, removal was low in both modes due to its strong albumin binding, with no significant difference between Sp (~229 ± 130 mg m⁻², DL_p ~41 mL min⁻¹ m⁻²) and Rc (~207 ± 18 mg m⁻², DL_p ~21 mL min⁻¹ m⁻²).

3.3. Effect of Different Plasma/Dialysate Fluid Flow Rates on Uremic Toxin Removal

In this section we present how different plasma and dialysate flow rates affect toxin removal from FX1000 membranes. Three plasma/dialysate flow conditions were compared: (1) 0.5/25 mL min⁻¹ (low plasma flow, high dialysate flow, used also in a previous study [23]), (2) 1/10 mL min⁻¹ (we set as “baseline”, used in section 3.2 and in earlier studies [15]), and (3) 1.5/2.25 mL min⁻¹ (higher plasma flow, lower dialysate flow). Figure 6 ((A1) for DL_p and (A2) for DL_d), (B1–2) and (C1–2) shows the removal kinetics for Cr, HA and IS respectively, while Figure 6 (D1–3) shows the total removal of toxins after 4 h. Table 2 (columns B, C and D) present the results and Figure A.2 (Appendix S1) presents total toxin removal over time for Cr (A), HA (B) and IS (C).

FIGURE 6.

Toxin removal (mL m⁻²) in time (min) by FX1000 fibers at different flow rates for Cr (A1: DL_p, A2: DL_d), HA (B1: DL_p, B2: DL_d), and IS (C1: DL_p, C2: DL_d). Also shown in this figure is the total toxin removed after 4 h at flow rates of 0.5/25, 1/10 and 1.5/2.25 mL min⁻¹ (D1, 2 and 3, N = 3). The *indicantes that there is a significant difference between the two data sets (p < 0.05). [Color figure can be viewed at wileyonlinelibrary.com]

For Cr, the low plasma flow rate (0.5/25 mL min⁻¹) resulted in the lowest total removal (~1061 mg m⁻², DL_p 133 ± 49 mL min⁻¹ m⁻²), while plasma flow rates above 1 lead to similar removal results (for 1.5/2.25, removal of ~1883 mg m⁻², DL_p 192 ± 34 mL min⁻¹ m⁻² and for 1/10, removal of ~2374 mg m⁻², DL_p 228 ± 65 mL min⁻¹ m⁻²). In both cases, DL_p/DL_d ~1 suggesting that the majority of Cr removed from plasma ended up in the dialysate.

The removal of HA and IS followed a similar pattern to Cr, but with generally lower values than Cr. When of 1/10 mL min⁻¹ and 1.5/2.25 mL min⁻¹, were applied the removal was for both toxins, higher than when 0.5/25 mL min⁻¹ was applied.

3.4. Toxin Removal From Human Plasma Versus Human Full Blood

Here we compare toxin removal by FX1000 mini dialyzers using human blood plasma versus full human blood. For direct comparison we implemented the same flow for blood/plasma and dialysate during 4‐h experiments. Each experiment was done using blood from a different donor. Figure 7 (A1, B1 and C1) shows kinetics of toxin removal for Cr, HA and IS, while Figure 7 (A2, B2 and C2) shows total toxin removal after 4 h. Table 2 (columns C and F, plasma and blood) show total toxin removal, DL_p, DL_d and DL_p/DL_d ratios. Figure A.3 (Appendix S1) shows the total toxin removal over time for Cr (A), HA (B) and IS (C).

FIGURE 7.

Toxin removal (mL m⁻²) in time (min) in plasma (black) and full blood (red) configurations by FX1000 for Cr (A1), HA (B1), and IS (C1). Also shown in this figure is the total toxin removed after 4 h (Cr: A2, HA: B2, IS: C2, N = 4). The *indicantes that there is a significant difference between the two data sets (p < 0.05). [Color figure can be viewed at wileyonlinelibrary.com]

The Cr removal for full blood was lower than in plasma, although not significantly (~1079 ± 194 mg m⁻² in full blood vs. ~1883 ± 603 mg m⁻² in plasma, p = 0.44). Cr is water soluble and distributes in total body water (including inside and outside RBCs) [24]. The presence of cells can create a diffusion barrier. The dialysance DL_p for full blood was lower but not significantly different to that in plasma (~166 ± 2 mL min⁻¹ m⁻² in blood, vs. ~192 ± 34 mL min⁻¹ m⁻² in plasma). Interestingly, the dialysate‐side clearance DL_d was significantly lower in blood (~115 ± 54 mL min⁻¹ m⁻²) than in plasma (~215 ± 37 mL min⁻¹ m⁻²). A DL_p/DL_d of 1.7 ± 0.8 for blood (vs 0.9 ± 0.2 for plasma) suggests that there was some interaction with the membrane, most likely enabled by the presence of blood cells creating a diffusion barrier.

The HA removal from full blood was also much lower than from human plasma (~507 ± 37 mg m⁻², DL_p of 38 ± 8 mL min⁻¹ m⁻² vs. ~1290 ± 231 mg m⁻² and DL_p of 85 ± 21 mL min⁻¹ m⁻²). The DL_d is also lower in blood than in plasma (~26 ± 6 vs. 66 ± 8 mL min⁻¹ m⁻²). For both toxins the ratio DL_p/DL_d > 1 indicating again toxin membrane interaction and/or polarization phenomena, as observed for Cr.

Similar findings were found for IS. In plasma‐only dialysis, 200 ± 142 mg m⁻² of IS was removed whereas in full blood, removal was much lower: 14 ± 3 mg m⁻² although the difference is not significant (p = 0.27) due to large deviation between donors. The dialysance values show similar trends: plasma‐only DL_p was 26 ± 15 mL min⁻¹ m⁻² for IS, while in full blood it was very low (below the detection limit). DL_d for IS in full blood was 1.4 ± 0.3 mL min⁻¹ m⁻² versus 8 ± 1.4 mL min⁻¹ m² in plasma. The DL_p/DL_d ratio in plasma was ~3.1, again indicating IS membrane interaction and/or polarization phenomena. During the blood experiments, we observed an increase of the pressure of the system in time (see Figure A.3, Appendix S1) and a decrease of removal rate, after approximately 2 h (see Figure A.3, Appendix S1), probably due to the interaction of blood components with the membranes. Despite the use of anticoagulation, clots formed or got trapped in the lumen of the fibers over time, reducing the effective membrane surface and increasing the flow resistance.

For Cr, the pressure increase was relatively low (< 0.5 bar) in the first half of the treatment but increased steadily after 120 min, reaching after 4 h, ~1.5 bar. In the case of HA and IS, we also observed pressure increase within the 4 h of experiment. All these findings again indicate the interaction of blood components with the membrane, as described earlier. After 2 h, we can clearly see that DL_p of Cr and HA from blood decreases (in contrast to plasma DL_p) due to higher interaction of the membrane with the blood components and the corresponding membrane fouling. After 2 h, DL_p of Cr from plasma and blood is 174 ± 52 mL min⁻¹ m⁻² and 185 ± 8 mL min⁻¹ m⁻², respectively. As well as for HA, the DL_p from plasma and blood is 80 ± 27 mL min⁻¹ m⁻² and 51 ± 16 mL min⁻¹ m⁻², respectively (see Table A.1).

3.5. Effect of Inside‐Out Versus Outside‐In Filtration Modes on Uremic Toxin Removal

In this section, we compare the removal of uremic toxins from the human plasma using two different directions of flow (IOF vs. OIF). These results have been reported earlier by our group [23] and are used again here to highlight the transport phenomena of IOF versus OIF in the context of this in vitro system. In these experiments, the flow rates were also adapted to minimize transmembrane pressure and minimize volume flow across the membrane. For IOF, the experiments were run with a low plasma flow (0.5 mL min⁻¹) and high dialysate flow (25 mL min⁻¹), whereas OIF used a high plasma flow (25 mL min⁻¹) and very low dialysate flow (0.5 mL min⁻¹). In both cases, we used mini dialyzers with same total membrane surface area and ran the experiments for 4 h.

Figure 8 (A1, B1 and C1) shows the kinetics of removal while Figure 8 (A2, B2 and C2) shows the total removal after 4 h for Cr, HA and IS respectively. Table 2 (columns D and E) shows total toxin removal, DL_p, DL_d and DL_p/DL_d ratio values. Figure A.4 (Appendix S1) shows removal over time of Cr (A), HA (B) and IS (C). These results were published previously by our group [23].

FIGURE 8.

Toxin removal (mL m⁻²) in time (min) in IOF (black) and OIF (red) configurations by FX1000 for Cr (A1), HA (B1), and IS (C1) in plasma. Also shown in this figure is the total toxin removed after 4 h for toxin (Cr: A2, HA: B2, IS: C2, N = 3). [Color figure can be viewed at wileyonlinelibrary.com]

OIF resulted in slightly higher Cr removal (~1196 vs. ~1061 mg m⁻², although not significantly (p = 0.57)) and plasma dialysance (~162 vs. ~133 mL min⁻¹ m⁻²) compared to IOF. Despite the low dialysate flow, OIF yielded to high dialysate dialysance (~111 vs. ~77 mL min⁻¹ m⁻²). The DL_p/DL_d ratios (~1.5–1.7) was similar for both filtration modes.

HA removal was also comparable in IOF and OIF (~897 vs. ~933 mg m⁻², not significant (p = 0.88)), with nearly identical plasma (~76–79 mL min⁻¹ m⁻²) and dialysate (~54–56 mL min⁻¹ m⁻²) dialysance.

IS removal was slightly higher in OIF (~166 vs. ~91 mg m⁻², but not significantly (p = 0.15)). DL_p was similar for both (~22 vs. ~16 mL min⁻¹ m⁻²), while DL_d remained low in both (~7 vs. ~6 mL min⁻¹ m⁻²). The high DL_p/DL_d for both modes (3.2 in OIF, 2.6 in IOF) suggests IS interaction with the membrane, as reported earlier.

3.6. Dialysis Experiments Using Mixed Matrix Membranes

Figures 9 and 10 and Table 2 (G, H) present the results of removal of HA and IS by the MMMs from human plasma and full human blood, respectively. HA total removal was 2381 ± 374 mg m⁻² by the MMM versus 1339 ± 40 mg m⁻² by FX1000 and IS total removal was 437 ± 107 mg m⁻² by the MMM compared to 207 ± 18 mg m⁻² by the FX1000. During the plasma experiments, no pressure increase was observed in the MMM dialyzers, maintaining TMP at 0 bar.

FIGURE 9.

Comparison of toxin removal (mL m⁻²) in time (min) between FX1000 and MMM in human plasma over time in Rc‐IOF configuration for HA (A1) and IS (B1). Also shown in this figure is the total toxin removed after 4 h for toxin (HA: A2, IS: B2, N = 3). [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 10.

Comparison of toxin removal (mL m⁻²) in time (min) between FX1000 and MMM in full human blood over time in Rc‐IOF configuration for HA (A1) and IS (B1). Also shown in this figure is the total toxin removed after 4 h for toxin (HA: A2, IS: B2, N = 3). The *indicantes that there is a significant difference between the two data sets (p < 0.05). [Color figure can be viewed at wileyonlinelibrary.com]

The total removal and DL of HA and IS from full human blood by the MMM is also higher compared to the FX1000 at the same flow conditions (1.5 mL min⁻¹ for blood, and 2.25 mL min⁻¹ for dialysate fluid), see Figure 10. For HA, the total removal by the MMM was 1539 ± 183 mg m⁻² versus 507 ± 37 mg m⁻² with FX1000 whereas the DL_p for MMM was 100 ± 16 mg min⁻¹ m⁻² compared to 38 ± 8 mg min⁻¹ m⁻² for FX1000. For IS, the total removal was 104 ± 59 mg m⁻² by MMM versus 14 ± 3 mg m⁻² by FX1000. The DL_p for MMM was ~12 mg min⁻¹ m⁻² but for the FX1000 it was very difficult to estimate accurately due to low IS removal there. During the full blood experiments with MMM, no pressure increase was observed probably due to larger inner diameter of the MMM fibers combined to low membrane fouling there.

4. Discussion

The aim of this study was to develop and validate a controlled in vitro mini‐dialyzer system for investigating toxin transport across hollow fiber dialysis membranes under clinically relevant conditions. By using a small‐scale platform with defined surface area, controlled plasma/blood and dialysate fluid flow rates, and dialysate recirculation, we can keep track of the mechanism of which the toxins are being removed throughout the experiment. Using creatinine as a model small water‐soluble toxin and hippuric acid and indoxyl sulfate as representative PBUTs, we aimed to (i) assess the impact of operational parameters on dialysance, (ii) compare plasma versus full human blood behavior, and (iii) to assess the usability of this platform with newly developed dialysis membranes like MMM. The results provide insights into transport phenomena at laboratory scale while enabling meaningful comparison with clinical dialyzer perfor‐mance.

4.1. Effect of Single Pass Versus Recirculation of Dialysate Fluid on Uremic Toxin Removal

We first validated the mini dialyzer with FX1000 HFs using two different dialysate modes: Rc and SP. In SP, fresh dialysate was continuously supplied to the mini dialyzer; in Rc, the same dialysate was recirculated. Cr removal was significantly higher in SP (~63%) due to the maintained concentration gradient across the membrane, whereas in Rc, the gradient decreased over time, reducing removal efficiency, as expected [25]. In contrast, the removal of PBUTs was similar in both modes, as their transport is limited by protein binding, which keeps the free toxin concentration in plasma, and thus the concentration gradient across the membrane, low [26, 27].

Our results clearly show that SP offers increased removal for small, highly water‐soluble solutes like Cr. However, the application of Rc enables mass balance estimation, reduces dialysate use, increases detection sensitivity, and allow to keep track of how toxins are removed. In fact, application of Rc is particularly suitable for studying membrane fouling and toxin adsorption of PBUTs [28]. Finally, the estimation of dialysance in both modes allows normalization for concentration gradient differences, enabling better comparison of phenomena like membrane adsorption, fouling, and concentration polarization.

4.2. Effect of Different Plasma/Dialysate Fluid Flow Rates on Uremic Toxin Removal

Here, we varied both plasma and dialysate flow rates to assess the effects of system hydrodynamics to the toxin removal. Lowering dialysate flow from 10 to 2.25 mL min⁻¹ reduced Cr clearance by ~26% and HA by ~13%. The removal of IS (highly PBUT) was less affected, in agreement with literature for large dialyzers [29, 30, 31]. When plasma flow was reduced to 0.5 mL min⁻¹, even with high dialysate flow (25 mL min⁻¹), dialysance lowered significantly: Cr by ~50%, HA by ~40%, and IS by ~60%. Increasing plasma flow to 1 mL min⁻¹ (with 10 mL min⁻¹ dialysate) resulted in double Cr removal, showing that plasma flow rate is the limiting factor. In addition, at the lowest dialysate flow rates of 1.5/2.25 mL min⁻¹, a gradual increase in dialysance over time was observed (Figure 6A1). This behavior is attributed to time‐dependent improvements in effective mass transfer, likely due to stabilization of flow distribution within the fibers and reduction of local mass transfer resistances. Under these conditions, the system operates closer to a dialysate‐side mass‐transfer limited regime, making it more sensitive to such effects. In contrast, at higher dialysate flow rate (1/10 mL min⁻¹), these limitations are minimized and dialysance remains stable over time.

For IS, which is highly protein‐bound (> 90%), transport maybe influenced by plasma‐side mass‐transfer resistance combined to slow kinetics of dissociation from albumin leading to an apparent plateau in removal over time, as ob‐served in Figure 6C1. Increasing the plasma flow rates improves mixing, reducing the concentration polarization and resulting to improved removal. These findings are also consistent with clinical and in vitro studies showing that higher plasma flow improves solute clearance by reducing concentration polarization near the membrane and improving mixing [5, 32]. For application of low plasma flow rates, in particular, the higher viscosity and protein content increase the risk of protein adsorption, further reducing membrane removal efficiency [33].

4.3. Effect of Inside‐Out Versus Outside‐In Filtration Modes on Uremic Toxin Removal

Here, we compared the removal of the Cr, HA and IS toxins using IOF and OIF modes under identical conditions and same effective surface area. Results showed no significant differences in toxin clearance between the two modes, in agreement to previous work [15, 23, 34]. Costa et al. [35] also showed similar clearance for both IOF and OIF modes using commercially available dialyzers. While IOF relies on low plasma flow and high dialysate flow, OIF can provide more uniform plasma contact with the membrane and requires lower dialysate flows. This may reduce fiber clogging and transmembrane pressure, as reported by Dukhin et al. [36], who achieved > 100 h of continuous operation. Our previously reported findings [23] support the clinical potential of OIF, especially for longer treatments, and show that it can be implemented in existing mini‐dialyzer designs without any modification and still achieve comparable results.

4.4. Toxin Removal From Human Blood Plasma and Full Human Blood

Here, we compared toxin removal by FX1000 and MMM from full human blood versus human plasma using 1.5 mL min⁻¹ for both full blood and plasma and 2.25 mL min⁻¹ for dialysate. Toxin removal from blood using large commercial dialyzers is often reported as clearance, CL. Here, we use DL_p (for plasma) and CL (for blood) to compare our results with literature, normalized to the surface area.

For Cr, our results show that removal by FX1000 from human plasma and blood is not significantly different, even though for plasma the average is slightly higher (Table 2), with DL_p values of around ~192 mL min⁻¹ m⁻² for plasma and 166 mL min⁻¹ m⁻² for full blood. Meyer et al. found that, for similar plasma/dialysate fluid flow ratios while using two full size dialyzers, Cr clearance from plasma was between 150 and 200 mL m⁻² min⁻¹ [30]. Külz et al. also reported Cr clearance from plasma of 172 mL min⁻¹ using commercial dialyzers [37]. Regarding blood studies, Bai‐Hai Su et al. observed, using large scale dialyzers, similar in vivo (goat) and in vitro (pig blood) Cr clearance of around 142 mL m⁻² min⁻¹ [38]. Additionally, Sirich et al. also reported a Cr clearance of approximately 190 mL m⁻² min⁻¹ during a standard 4‐h HD session (in vivo) [39]. Moreover, we clearly see that the slope of DL_p of Cr and HA starts flattening after the second hour of experiment with human blood, separating from the plasma results. That fact is due to the higher fouling that occurs with the presence of blood, that also affects the increase of pressure in the system.

For PBUTs, the removal by FX1000 from full blood was much lower than from plasma, with a decrease of ~60% for HA and ~93% for IS. This difference can be attributed to several physiological and transport‐related factors. Compared to plasma, full blood introduces cellular complexity, mainly from red blood cells (RBCs), which constitute 40% to 45% of total blood volume. RBCs reduce the effective plasma volume per unit blood, can create a cell‐free layer near the membrane wall under laminar flow, and decrease the diffusion of plasma proteins and solutes [40]. We clearly saw this effect in the DL_p/DL_d ratios for Cr, which were higher for blood. For plasma, the ratio was ~1 meaning that all the cleared toxins were found in the dialysate fluid, but for blood the ratio was ~1.7, meaning that some adsorption to the membrane occurred. For PBUTs, the ratio was above 1 in plasma, with similar values for blood. For HA the DL was constant within the 4 h experiments with both human plasma and full blood, however the DL for IS was much lower and decreased in time, especially in the case of the full blood experiments, probably due to membrane fouling. Farell et al. reported higher PBUT removal for in vitro compared to in vivo. In fact, they reported HA clearance from blood of ~33 mL min⁻¹ m⁻² using a commercial dialyzer (A _eff of ~2 m²) [41] which is very close to the one we reported here (38 ± 8 mL min⁻¹ m⁻²). Camacho et al. [31] also observed that, for nocturnal dialysis using large scale dialyzers (Optiflux F160NR and F250NR with A _eff of 1.6 and 2.5 m², respectively) and similar blood/dialysate flowrate ratio, IS clearance was 12 mL min⁻¹ m⁻² (within the range of our DL_p = 14 ± 3 mL min⁻¹ m⁻²). They also reported low IS removal due to protein binding. PBUTs must diffuse through a potentially stagnant boundary layer near the membrane, especially at low flow rates [32, 42]. Recent findings also suggest that both HA and IS can be transported into RBCs [43]: HA tends to equilibrate more rapidly between the intracellular and extracellular compartments, while IS shows persistent intracellular accumulation. This means that although HA uptake by RBCs may briefly delay removal, its impact is less pronounced. In contrast, IS uptake by RBCs significantly affects removal.

The removal of both PBUT studied here, HA and IS, by MMM was much higher than FX1000 in both human plasma and full blood studies showing the importance of the adsorption for improved blood detoxification (the DL_p/DL_d >> 1 indicates the high adsorption of toxins by the MMM). The removal of HA and IS from human plasma by the MMM is consistent to earlier findings comparing PBUT removal of MMM with various commercial membranes [3, 15, 44]. During the blood experiments with FX1000, the pressure of the system increased progressively indicating blood—membrane interaction, possibly changing of the protein‐binding equilibrium and/or cellular degradation in the circuit over time [45, 46, 47]. The progressive rise in pressure during the experiment, potentially inducing a Starling‐driven solvent flux across the membrane. Although the system was initially operated to minimized TMP to favor diffusion‐dominated transport, membrane fouling and partial fiber clogging may have generated a non‐negligible convective component over time. Such convective effect could contribute to additional transport of small water‐soluble toxins (e.g., creatinine), while its effect on PBUT removal is expected to be limited by their low free fraction available and intracellular partitioning in red blood cells. Moreover, local convective flow may enhance concentration polarization and protein deposition at the membrane surface, further influencing the effective mass transfer during prolonged blood operation. For the experiments with the MMM no pressure increased took place, probably due to larger fiber inner diameter and low membrane fouling due to high MMM blood compatibility [44].

4.5. Applicability of In Vitro System for Hollow Fiber Membrane Characterization

Here, we present an in vitro system for the facile investigation of toxin transport across hollow fiber membranes. It enables evaluation of membrane–solute interactions and mass‐transfer characteristics (e.g., dialysance) and permits direct, like‐for‐like comparisons between membranes. Our results can represent well the various phenomena affecting removal of toxins during HD compared to literature studies for full‐scale clinical dialyzers (Table 3) however, the system is not intended to replicate the full complexity of fluid dynamics in clinical dialyzers. The DL and total removal values for Cr, HA, and IS are within the ranges of CL reported in literature for full‐size dialyzers. For Cr, DL_p values ranging from 133 to 305 mL min⁻¹ m⁻² closely matched those reported for standard dialyzers (75–325 mL min⁻¹ m⁻²). For HA and IS, we observed reduced toxin removal in blood versus plasma and increased toxin removal in plasma at higher plasma flow rates. The DL_p values for IS from blood of our study (9–41 mL min⁻¹ m⁻²) are quite like the CL reported in clinical studies (9–15 mL min⁻¹ m⁻²). Earlier studies [48, 49] reported that the CL of an 50% protein bound toxin is expected to be two thirds lower compared to unbound one of the same size, and the CL of a 95% bound toxin is expected to be less than one tenth of an unbound one of the same size (CL around 250–200 mL min⁻¹ for small water soluble toxins like urea, 150–100 mL min⁻¹ for HA, and 50–10 mL min⁻¹ for IS). These results are consistent to our finding for plasma and blood studies. We also calculated flow velocities and Reynolds numbers (Re, see Table A.3 in the Appendix S1) for all performed experiments. Across all these conditions, velocities ranged from 0.0003 to 0.50 m s⁻¹, and Reynolds numbers ranged from 2 to 92, within the laminar regime (Re < 200). These values are in agreement with those reported for clinical dialyzers [7, 20, 50, 51, 52, 53] and support the use of the mini‐module as a relevant in vitro system.

TABLE 3.

Total removal and plasma dialysance of different membranes, with varying surface areas and flow rates.

Toxin	Total removal (mg m−2)	CL (mL min−1 m−2)	Membrane	Surface area (m2)	Flow rates (blood/dialysate fluid) (mL min−1)	Literature
Cr	—	75–325	Polyflux 2H Gambro, Baxter	0.2	20–100/30–500	Manufacturer information (available online)
	—	124–139	F8HPS—Fresenius	1.8	300–400/500	Manufacturer information (available online)
	633 ± 89	58 ± 5	F8HPS—Fresenius	1.8	324/500	[54]
	779 ± 381	61 ± 16	F8HPS—Fresenius	1.8	350/350	[55]
HA	346 ± 228	66 ± 20	F8HPS—Fresenius	1.8	300/600	[56]
	377	73 ± 7	F8HPS—Fresenius	1.8	300/700	[48]
IS	82 ± 44	14 ± 4	F8HPS—Fresenius	1.8	300/600	[56]
	47	15 ± 3	F8HPS—Fresenius	1.8	300/700	[48]
	54 ± 41	15 ± 10	PUREMA L—Medica	1.4	379–385/500	[57]
	102 ± 49	9 ± 3	F160NR—Fresenius	1.5	350/300	[58]

5. Conclusions

This study presented an in vitro system for investigating transport studies across commercially available dialysis membranes at laboratory scale in a wide range of HD conditions. We mainly implemented dialysate recirculation, which allowed estimation of the mass balance of toxins, and therefore could lead to better understanding of the membrane toxin interaction, membrane fouling and concentration polarization. Moreover, we used DL normalized by membrane effective area (mL min⁻¹ m⁻²) as the key parameter for understanding the various phenomena occurring inside the mini dialyzer.

When comparing our results to values reported in the literature for full‐scale clinical dialyzers, our system can represent well the various phenomena affecting removal of toxins during HD. The DL and total removal values for Cr, HA, and IS are within the ranges reported in literature for full‐size dialyzers. For Cr, DL_p values ranging from 133 to 305 mL min⁻¹ m⁻² closely matched those reported for CL of standard dialyzers (75–325 mL min⁻¹ m⁻²). For HA and IS, we observed similar trends: reduced toxin removal in blood versus plasma and increased toxin removal in plasma at higher plasma flow rates. Notably, DL_p values for IS from blood of our study (9–41 mL min⁻¹ m⁻²) are quite like the CL reported in clinical studies (9–15 mL min⁻¹ m⁻²). For the first time, we reported here, the superior removal of HA and IS from full human blood by the MMM compared to FX1000, showing its high potential for clinical implementation, especially in portable artificially kidney systems with low amount of dialysate.

Author Contributions

M.T.‐Y. and D.R. contributed equally to this publication; designed and conducted the experiments, analyzed the data, and wrote the manuscript. F.P. developed the MMM, conducted the MMM experiments and analyzed the data. J.d.V., B.H.L. and K.G.F.G. provided support with the design of blood experiments. O.E.M.t.B. and D.S. supervised the project and provided critical input throughout the research and manuscript preparation.

Funding

D. Ramada acknowledge the financial support of the Top Sector Life Sciences & Health (Health~Holland), NODIAL project 21OP+035. M. Torrents‐Yeste and F. Pel, acknowledge the financial support of the Dutch National growth fund program NXTGEN Hightech Biomed04 “Artificial Organs” project.

Conflicts of Interest

Dimitrios Stamatialis is the scientific advisor of the company M3 Nephron—a startup company of the University of Twente. The company obtained an exclusive license from the University of Twente for the Mixed Matrix Membrane Technology.

Supporting information

Appendix S1: aor70139‐sup‐0001‐Appendix.docx.

Torrents‐Yeste M., Ramada D., ter Beek O. E. M., et al., “An In Vitro System for Studying Toxin Transport Across Hemodialysis Membranes,” Artificial Organs 50, no. 7 (2026): 988–1006, 10.1111/aor.70139.

Data Availability Statement

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