Removal of urea by electro-oxidation in a miniature dialysis device: a study in awake goats

Abstract

The key to success in developing a wearable dialysis device is a technique to safely and efficiently regenerate and reuse a small volume of dialysate in a closed-loop system. In a hemodialysis model in goats, we explored whether urea removal by electro-oxidation (EO) could be effectively and safely applied in vivo. A miniature dialysis device was built, containing 1 or 2 “EO units,” each with 10 graphite electrodes, with a cumulative electrode surface of 585 cm² per unit. The units also contained poly(styrene-divinylbenzene) sulfonate beads, FeOOH beads, and activated carbon for respective potassium, phosphate, and chlorine removal. Urea, potassium, and phosphate were infused to create “uremic” conditions. Urea removal was dependent on total electrode surface area [removal of 8 mmol/h (SD 1) and 16 mmol/h (SD 2) and clearance of 12 ml/min (SD 1) and 20 ml/min (SD 3) with 1 and 2 EO units, respectively] and plasma urea concentration but not on flow rate. Extrapolating urea removal with 2 EO units to 24 h would suffice to remove daily urea production, but for intermittent dialysis, additional units would be required. EO had practically no effects on potassium and phosphate removal or electrolyte balance. However, slight ammonium releasewas observed, and some chlorine release at higher dialysate flow rates. Minor effects on acid-base balance were observed, possibly partly due to infusion of chloride. Mild hemolysis occurred, which seemed related to urea infusion. In conclusion, clinically relevant urea removal was achieved in vivo by electro-oxidation. Efficacy and safety testing in a large-animal model with uremia is now indicated.

INTRODUCTION

For decades, strategies to develop a wearable dialysis device have been explored that could offer continuous or more frequent dialysis resulting in more adequate toxin removal, steadier electrolyte concentrations, and less abrupt fluid removal (12, 17, 23, 29, 30). The key to success in developing such a device is to invent a technique to safely and efficiently regenerate and reuse a small volume of dialysate in a closed-loop system. Thus far, predominantly sorbent cartridges containing adsorbing particles for electrolyte removal and urease for enzymatic degradation of urea have been tested (3). A major drawback of urea degradation by urease is the production of ammonium that, in turn, must be adsorbed. However, development of an alternative urea removal strategy is challenging since the daily production of urea, the main waste product of nitrogen metabolism, is very high (250–400 mmol/day) (13, 31, 37), and urea is very difficult to adsorb (18, 32, 40).

Previously, we explored electro-oxidation (EO) in vitro using graphite electrodes incorporated in a sorbent cartridge containing sorbent particles for potassium and phosphate removal (39). In this technology, a current is applied to the dialysate by which urea is converted into nitrogen, carbon-dioxide, and water, either directly at the anode or indirectly in the bulk solution via the intermediate EO-product hypochlorite. A disadvantage is the generation of chlorinated byproducts. However, these can be neutralized by activated carbon (AC). We achieved clinically relevant urea degradation in blood [9.5 mmol/h (SD 1.0)] with an EO unit in series with AC (39). Chlorine release was below maximum acceptable levels as defined by the Association for the Advancement of Medical Instrumentation (AAMI) standards (1).

In the present study, we explored whether urea removal by EO could be effectively and safely applied in vivo in a hemodialysis (HD) model in goats. We incorporated an EO unit containing graphite electrodes and sorbent particles for potassium and phosphate removal in the dialysate circuit of a prototype miniature dialysis device. First, we studied efficacy of urea removal and explored whether doubling of the electrode surface area (by using 2 EO units) and increasing the dialysate flow could increase urea removal efficiency. Removal of creatinine, potassium, and phosphate by the device was also quantified. Second, we evaluated biocompatibility of EO by monitoring influence on vital parameters (calcium, magnesium, sodium, hemolysis parameters, acid-base status, iron, and glucose) by measuring the release of chlorine species and ammonium.

METHODS

Materials

A miniature dialysis device of ~1.5 kg (plus battery pack of 250 g for 3 h of dialysis) was built and provided by Nanodialysis BV (Oirschot, the Netherlands). One or two “EO unit(s)” (weight: 250 g per unit) were incorporated in the dialysate circuit, each containing 10 graphite electrodes with a cumulative electrode surface of 585 cm² per EO unit (39). Each unit contained 80 g (when using 1 unit) or 90 g (when using 2 units) poly(styrene-divinylbenzene) (PS-DVB) sulfonate beads and 40 g (when using 1 unit) or 30 g (when using 2 units) FeOOH beads for respective potassium and phosphate removal (the amounts were adjusted after the experiments with 1 unit to achieve the target potassium/phosphate removal ratio of ~2–3:1). A degassing unit and AC filters (25 g per EO unit) were placed downstream of the EO units. High-flux dialyzers (Polyflux 2H 0.2 m2; Gambro Dialysatoren GmbH, Hechingen, Germany) were used to separate blood and dialysate. Thomas pumps (Gardner Denver Thomas, Sheboygan, WI) were applied. A 72-cm double-lumen central venous catheter (CVC) was selected for blood access (Palindrome Chronic Dialysis Catheter, Covidien, Medtronic, Minneapolis, MN). Unfractionated heparin was used for anticoagulation.

Methods

Animals.

In vivo experiments were approved by the Animal Experiments Committee (Utrecht University, Utrecht, the Netherlands) and performed in accordance with national guidelines for the care and handling of animals. Healthy Dutch White goats (n = 2) were selected because these animals are docile, have easily accessible neck veins, and body weights (70–90 kg) and distribution volumes comparable to humans (38).

Sterilization and regeneration procedure.

In each experiment, a new sterile dialyzer and tubing were used. Prior to each experiment, the EO units were sterilized by 1 h of exposure of the internal fluidic circuit to 1 liter 5% hydrogen peroxide (m/m) and 0.2% peracetic acid (m/m) in 120 mM NaCl, 1.2 mM CaCl₂, and 0.45 mM MgCl₂, followed by 12 h of exposure to 1 liter 120 mM NaOH, 20% ethanol (vol/vol), in a sterile environment throughout. This procedure was shown to be effective in eliminating bacteria and endotoxins. Subsequently, pH was adjusted to 7.4–8.0 using 1 liter 120 mM NaCl, 50 mM HCl, followed by rinsing and equilibration of the EO units with 3 liters rinsing solution (pH 7.4–8.0) containing 105 mM (1 unit) or 95 mM (2 units) NaCl, 15 mM (1 unit) or 25 mM (2 units) NaHCO₃, 1.2 mM CaCl₂, and 0.45 mM MgCl₂. These concentrations were based on prior in vitro and in vivo experiments, showing no net electrolyte release at these concentrations (39). Equilibration was achieved by recirculating the rinsing solution in the blood circuit at 110 ml/min while draining the solution from the dialysate circuit at 40 ml/min and measuring calcium in the effluent (equilibration of calcium is slower than that of other relevant electrolytes). The rinsing procedure was continued until inlet and outlet calcium concentrations were equal (usually after circulation of 3 liters of rinsing solution).

Experimental procedure.

Sixteen experiments were performed in two goats (eight per goat). Goats were temporarily sedated (with detomidine and propofol) to insert a CVC in the jugular vein. A catheter was placed in the auricular artery for arterial pressure measurement and withdrawal of arterial blood gases.

To prevent a decrease of systemic potassium and phosphate plasma concentrations, we infused 1 M KCl and 0.05 M Na₂HPO₄/0.25 M KH₂PO₄ solution as described (38). To achieve higher urea concentrations, we administered an oral bolus of 30 g urea (dissolved in syrup) to the goats the evening and morning before the experiment in the first 3 experiments. However, the goats progressively disliked urea, and therefore we switched to intravenous administration of urea via the venous lumen of the CVC in the remaining experiments [200 mmol/h (0.67 liters of 0.3 M urea)] during the first hour before starting dialysis, followed by continuous infusion of 100 mmol/h (0.33 l/h of 0.3 M urea during dialysis). Spontaneous urea concentrations were 6.5 mM (SD 1.4). After dietary or intravenous administration of urea, urea concentrations were 11.0  mM (SD 2.2) and 13.1 mM (SD 1.9), respectively (Fig. 1). NaCl (0.65%–0.9%) was infused to balance urea-induced osmotic diuresis (0.65% NaCl in the experiments with 1 EO unit and in 4 of 9 experiments with 2 EO units, and 0.9% NaCl in 5 of 9 experiments with 2 EO units). The dialysate circuit was filled with ~100 ml rinsing solution. Blood lines were connected, and blood was pumped (110 ml/min) across the dialyzer (Fig. 2, experimental setup).

Fig. 1.

Plasma urea concentration without urea administration (n = 16), after dietary urea administration (n = 3), and after intravenous urea infusion (n = 13). *P < 0.001 vs. no urea administration.

Fig. 2.

Experimental setup. AC, activated carbon; EO, electro-oxidation. [Modified from Wester et al. (38), with permission from Oxford University Press.]

After 30 min habituation of the goat and equilibration of blood with the dialysate compartment, the dialysate pump was started, and dialysate was recirculated during 3 h over the EO units in counter-current direction with blood in the dialyzer (38).

To assess whether urea removal was dependent on cumulative electrode surface or dialysate flow, we performed experiments with 1 EO unit (n = 7) at a dialysate flow of 40 ml/min and 2 EO units in parallel at a dialysate flow of 40 ml·min⁻¹·unit⁻¹ (n = 5) or 70 ml·min⁻¹·unit⁻¹ (n = 4) (70 ml/min was the maximal dialysate flow that could be achieved with the pumps). A current of 3A was applied per unit because previous research showed that this leads to efficient urea degradation while keeping chlorine release low (39). A voltage of 3.5 V was applied across each EO unit. Every 30 min, the electrode polarity was inversed to eliminate deposits (e.g., Mg salts) that may form on the electrodes (11). Blood and dialysate samples were taken hourly from the blood circuit (down- and upstream of the dialyzer), from the dialysate circuit (down- and upstream of the EO units), and from the arterial line. [Urea], [Creatinine], [K⁺], [PO₄³⁻], [total Ca], [total Mg], [Na⁺], [${\text{HCO}}_3^{-}$], and [glucose] were monitored in plasma and dialysate. [LDH] was only measured in plasma. [NH₄⁺] was only measured in plasma in the last six experiments at the start and end of the experiment. In the first two experiments, dialysate [Fe²⁺] and [albumin] appeared to be below detection limit and were therefore only measured in plasma in the remaining experiments ([Fe²⁺] at the start and end of the experiment and [albumin] hourly). Hemoglobin was monitored hourly and leukocytes and thrombocytes at the start and end of the experiment (note that leukocyte and thrombocyte counts at the start and end were lacking in 2 and 4 experiments, respectively). Blood gasses were taken hourly, except in one experiment where the arterial line could not be placed. Chlorine measurements were performed hourly in dialysate in the experiments with 2 EO units. At the end of each experiment, the catheters were removed. Before, during, and after the experiments, vital parameters (heart rate, blood pressure, temperature, and blood oxygen saturation) were monitored (38).

Anticoagulation strategy.

After insertion of the CVC, a bolus of unfractionated heparin was given (10,000 IU), followed by continuous infusion of ~3,500 IU/h. The activated clotting time (ACT) was measured every hour and the heparin dose adjusted, aiming for an ACT > 500 s. One hour before removal of the CVC, heparin infusion was stopped (38).

Laboratory measurements.

All electrolyte measurements were performed at the hospital laboratory of the University Medical Center Utrecht. Urea, creatinine, potassium, phosphate, sodium, calcium, magnesium, bicarbonate, albumin, ammonium, lactate dehydrogenase (LDH), and iron concentrations were analyzed with an AU 5800 routine chemistry analyzer (Beckman Coulter, Brea, CA) using ion selective electrodes and spectrophotometry. Arterial blood gas analyses were performed using a blood gas analyzer (Rapidlab type 1265; Siemens Medical Solutions Diagnostics B.V., Breda, the Netherlands). Free and total chlorine concentrations were measured using a ChloroSense Chlorine Meter (CS 100 Palintest, Gateshead, UK) (38).

Calculations, analyses, and statistics.

Removal from plasma was calculated using the following formula:

$${\left({\text{A}}_{\text{p}}\right)}_{t1\to t2}=\left[{\left({\text{C}}_{\text{pi}}–{\text{C}}_{\text{po}}\right)}_{t1}+{\left({\text{C}}_{\text{pi}}–{\text{C}}_{\text{po}}\right)}_{t2}\right]/2\times \text{Q}\times \Delta t\times \left(1-\text{Ht}\right),$$

and removal from dialysate was calculated using:

$${\left({\text{A}}_{\text{d}}\right)}_{t1\to t2}=\left[{\left({\text{C}}_{\text{di}}–{\text{C}}_{\text{do}}\right)}_{t1}+{\left({\text{C}}_{\text{di}}–{\text{C}}_{\text{do}}\right)}_{t2}\right]/2\times \text{Q}\times \Delta t\mathrm{.}$$

Removal is presented as the mean value of removal from plasma and dialysate:

$${\text{A}}_{\text{m}}=\left({\text{A}}_{\text{p}}+{\text{A}}_{\text{d}}\right)/2.$$

Plasma clearance was calculated using the following formulas:

$$Forurea:{\text{Cl}}_{t1\to t2}={\left({\text{A}}_{\text{d}}\right)}_{t1\to t2}/\left\{\left[{\left({\text{C}}_{\text{pi}}\right)}_{t1}+{\left({\text{C}}_{\text{pi}}\right)}_{t2}\right]/2\right\}$$

$$Forcreatinine:{\text{Cl}}_{t1\to t2}={\left({\text{A}}_{\text{m}}\right)}_{t1\to t2}/\left\{\left[{\left({\text{C}}_{\text{pi}}\right)}_{t1}+{\left({\text{C}}_{\text{pi}}\right)}_{t2}\right]/2\right\},$$

where A is amount removed from plasma (A_p) or dialysate (A_d) or average removal from plasma and dialysate (A_m), Cl is plasma clearance (ml/min), C_pi is inlet plasma concentration (i.e., upstream of dialyzer), C_di is inlet dialysate concentration (i.e., upstream of the EO unit), C_po is outlet plasma concentration (i.e., downstream of dialyzer), C_do is outlet dialysate concentration (i.e., downstream of the EO unit), t is time after start of dialysate pump (min), Q is blood or dialysate flow (l/min), Δt is t2 − t1, and Ht is hematocrit.

Note that urea, potassium, and phosphate removal was calculated using only dialysate data, and urea clearance was calculated using only removal from dialysate since some of the outlet plasma concentrations were influenced by their infusion via the venous lumen of the CVC. Moreover, urea is also removed from the cell compartment to an unknown extent (8).

Total plasma calcium concentration (mM) was corrected for albumin (g/l):

$${\left[\text{Ca}\right]}_{\text{corrected}}={\left[\text{Ca}\right]}_{\text{measured}}+0.02\times \left(\text{normal}–\text{measured plasma}\left[\text{albumin}\right]\right)$$

Anion gap (mEq/l) was calculated using [Na⁺], [Cl⁻], and [${\text{HCO}}_3^{-}$] (mM) and corrected for albumin (g/l) using the following formula:

$$\text{Anion\hspace{0.17em}gap}=\left[{\text{Na}}^{+}\right]-\left(\left[{\text{Cl}}^{-}\right]+\left[{\text{HCO}}_3^{-}\right]\right)+0.25\times \left(\text{normal}-\text{measured plasma}\left[\text{albumin}\right]\right)$$

[for normal albumin concentration, average albumin concentration in the goats before intervention (34.2 g/l) was used.]

Data are shown as means (SD). Linear interpolation was used for 12% of the arterial blood gas measurements (7 out of 60) where data were missing.

Statistical significance was determined using Student’s (paired) t-test, Wilcoxon matched-pairs signed rank tests, and linear regression analysis as appropriate. For comparison of absolute removal, two-way ANOVA for repeated measures with post hoc correction using Tukey’s multiple comparisons test was applied.

RESULTS

Efficacy

Urea removal.

Urea removal and clearance were 8.4 mmol/h (SD 1.3) and 11.9 ml/min (SD 1.1), respectively, using 1 EO unit (dialysate flow 40 ml/min; Fig. 3, A and B) and remained stable during consecutive hours. Use of 2 units in parallel doubled the removal of urea [15.9 mmol/h (SD 2.2)] and caused a 1.6-fold increase in urea clearance [19.5 ml/min (SD 2.8)]. Increase of the dialysate flow to 70 ml/min per unit did not increase urea removal [15.8 mmol/h (SD 3.4); P = 0.77 vs. flow 2 × 40 ml/min] or urea clearance [21.0 ml/min (SD 2.5); P = 0.53 vs. flow 2 × 40 ml/min]. Urea removal was dependent on urea plasma concentration (Fig. 3, C and D) whereas there was no relation between urea concentration and clearance.

Fig. 3.

Urea removal using 1 or 2 electro-oxidation (EO) units at a dialysate flow of 1 × 40 (n = 7), 2 × 40 (n = 5), or 2 × 70 (n = 4) ml/min. A: cumulative urea removal in 3 h. B: urea clearance. C and D: relation between cumulative urea removal and plasma urea concentration (solid line) and between urea clearance and plasma urea concentration (dashed line) using 1 unit (C) or 2 EO units (D). ***P < 0.001, **P = 0.003 vs. flow 1 × 40 ml/min.

Creatinine removal.

Creatinine removal and clearance were 97 µmol/h (SD 6) [at plasma creatinine concentrations of 59 µM (SD 4)] and 27.9 ml/min (SD 2.5), respectively, using 1 EO unit (dialysate flow 40 ml/min; Fig. 4, A and B) and remained stable during consecutive hours. Doubling the electrode surface by using 2 units in parallel caused a 1.5-fold increase in creatinine removal [155 µmol/h (SD 16) at plasma creatinine concentrations of 60 µM (SD 7)] and clearance [42.8 ml/min (SD 3.4)]. In contrast to urea removal and clearance, creatinine removal and clearance were dialysate flow-dependent. Increase of the dialysate flow to 70 ml/min resulted in a creatinine removal of 210 µmol/h (SD 40) and clearance of 62.6 ml/min (SD 6).

Fig. 4.

Creatinine removal using 1 or 2 electro-oxidation (EO) units at a dialysate flow of 1 × 40 (n = 7), 2 × 40 (n = 5), or 2 × 70 (n = 4) ml/min. A: cumulative creatinine removal in 3 h. B: creatinine clearance. ***P < 0.001 and $$P = 0.006 vs. flow 1 × 40 ml/min; $$$P < 0.001 and **P = 0.004 vs. flow 2 × 40 ml/min.

Potassium and phosphate removal.

Details on potassium and phosphate removal have been discussed in our previous publication (38). Novel in the present study is that we also applied two EO units. Cumulative potassium removal in 3 h was 7.8 mmol (SD 3.1) [98 μmol/g (SD 39) sorbent] using 1 EO unit (Fig. 5A) at average plasma potassium concentrations of 4.5 mM (SD 0.6). When using 2 units in parallel, cumulative potassium removal in 3 h was 13.2 mmol (SD 1.2) [73 μmol/g (SD 7) sorbent] at a dialysate flow of 40 ml/min and 22.5 mmol (SD 4.9) [125 μmol/g (SD 27) sorbent] at 70 ml/min at average plasma potassium concentrations of 4.7 mM (SD 0.2) and 4.6 mM (SD 0.1), respectively. Cumulative phosphate removal in 3 h was 5.3 mmol (SD 1.0) [132 μmol/g (SD 25) sorbent] using 1 unit (Fig. 5B) at average plasma phosphate concentrations of 2.5 mM (SD 0.5). When using 2 units in parallel, cumulative phosphate removal in 3 h was 6.5 mM (SD 1.0) [108 μmol/g (SD 17) sorbent] at a dialysate flow of 40 ml/min and 10.0 mmol (SD 1.4) [166 μmol/g (SD 23) sorbent] at 70 ml/min at average plasma phosphate concentrations of 2.1 mM (SD 0.6) and 2.0 mM (SD 0.3), respectively.

Fig. 5.

A: cumulative potassium removal in 3 h using 1 or 2 electro-oxidation (EO) units at a dialysate flow of 1 × 40 (n = 7), 2 × 40 (n = 3), or 2 × 70 (n = 3) ml/min. B: cumulative phosphate removal in 3 h using 1 or 2 EO units at a dialysate flow of 1 × 40 (n = 7), 2 × 40 (n = 3), or 2 × 70 (n = 3) ml/min. ***P < 0.001 vs. flow 1 × 40 ml/min; **P = 0.027 and $$P = 0.01 vs. flow 2 × 40 ml/min.

Safety Analysis

Influences on vital parameters.

Vital parameters remained practically stable, except for a rise in heart rate (Table 1).

Table 1.

Vital parameters before, during, and after dialysis

	Before Dialysis	During Dialysis	After Dialysis
SBP, mmHg	119 (22)	123 (14)	119 (19)
DBP, mmHg	76 (14)	87 (14)*	84 (13)
Heart rate, beats/min	67 (26)	84 (24)†	98 (24)‡
Oxygen saturation, %	95 (4)	97 (5)	97 (2)

Values are means (SD). DBP, diastolic blood pressure; SBP, systolic blood pressure.

^*P < 0.05,

^†P < 0.01,

^‡P < 0.001 vs. after dialysis.

Calcium, magnesium, and sodium.

Plasma calcium concentrations showed a limited decrease of, on average, 0.04 mM (SD 0.08) (P = 0.04) across the dialyzer, whereas dialysate calcium concentrations remained stable (Fig. 6, A and B). Systemic plasma calcium concentrations decreased from 2.2 mM (SD 0.2) to 2.1 mM (SD 0.2) (P = 0.003) (Fig. 6C). Plasma and dialysate magnesium concentrations showed a limited decrease of, on average, 0.04 mM (SD 0.04) (P = 0.002) across the dialyzer and 0.02 mM (SD 0.05) (P = 0.006) across the EO unit, respectively (Fig. 6, D and E). Systemic magnesium concentrations decreased from 0.84 mM (SD 0.06) to 0.78 mM (SD 0.06) (P = 0.003; Fig. 6F). Plasma sodium concentrations were stable across the dialyzer (Fig. 6G), but dialysate sodium concentrations slightly decreased across the EO unit [0.67 mM (SD 0.40), P < 0.001; Fig. 6H]. Systemic plasma sodium concentration increased from 146 mM (SD 1.9) to 148 mM (SD 2.5) (P = 0.03; Fig. 6I).

Fig. 6.

Influences on calcium, magnesium, and sodium concentration at the inlet (In) and outlet (Out) of the dialyzer (plasma) and up- (In) and downstream (Out) of the electro-oxidation (EO) unit (dialysate) (n = 16). Measurements were performed each hour, and per experiment the concentrations In and Out were averaged. A and B: plasma and dialysate calcium concentration. C: plasma calcium concentration. D and E: plasma and dialysate magnesium concentration. F: plasma magnesium concentration. G and H: plasma and dialysate sodium concentration. I: plasma sodium concentration.

Influences on hemolysis parameters, leukocytes, and thrombocytes.

Hb across the dialyzer and systemic Hb remained stable during the experiments (Fig. 7, A and B). LDH did not change across the dialyzer (Fig. 7C) but increased systemically in the experiments where urea was infused (Fig. 7D; P = 0.05). In two of these experiments, visible hemolysis occurred with pink discoloration of plasma. Of note, we also observed visible hemolysis in one of two experiments where only urea was infused, and no dialysis (or EO) was performed. Leukocyte number increased during the experiment from 7.7 (SD 2.0) to 8.9 (SD 1.8) × 10⁹/liter (Fig. 7E; P = 0.002). Thrombocyte number remained stable (Fig. 7F).

Fig. 7.

A: Hb concentration at the inlet (In) and outlet (Out) of the dialyzer (n = 16). B: Hb concentration (n = 16). C: plasma lactate dehydrogenase (LDH) concentration at the In and Out of the dialyzer (n = 16). D: plasma LDH concentration (n = 16 all experiments, n = 3 oral urea, and n = 13 IV urea). E: leukocyte number at the start and end of the experiment (n = 14). F: thrombocyte number at the start and end of the experiment (n = 12). In A, C, E, and F, measurements were performed each hour, and per experiment the concentrations In and Out or Start and End were averaged.

Influences on acid-base status.

Arterial blood pH at the start of the experiments was 7.41 (SD 0.03) (1 unit) and 7.43 (SD 0.05) (2 units) and did not change during the experiments (Fig. 8A). Plasma bicarbonate concentrations showed a tendency to gradually decrease from 25.7 mM (SD 2.7) to 23.3 mM (SD 1.1) (P = 0.04) for arterial bicarbonate and from 27.6 mM (SD 3.1) to 25.6 mM (SD 2.5) (P < 0.001) for venous bicarbonate using 1 unit and from 26.8 mM (SD 2.2) to 25.4 mM (SD 2.0) (P = 0.13) for arterial bicarbonate and from 28.0 mM (SD 2.2) to 26.0 mM (SD 2.2) (P = 0.08) for venous bicarbonate using 2 units (Fig. 8, B and C). The systemic decrease in bicarbonate concentration was accompanied by a decrease in Pco₂ from 40.6 mmHg (SD 3.4) to 34.6 mmHg (SD 5.0) (P = 0.03 using 1 unit) and from 40.2 mmHg (SD 5.7) to 37.3 mmHg (SD 3.9) (P = 0.01 using 2 units) (Fig. 8D). Plasma bicarbonate concentrations remained stable across the dialyzer using 1 unit and increased across the dialyzer from 26.9 mM (SD 2.1) to 27.8 mM (SD 2.5) using 2 units (P < 0.001, Fig. 8E). Dialysate bicarbonate concentrations showed a tendency to decrease across the EO unit from 28.3 mM (SD 2.5) to 27.6 mM (SD 2.4) using 1 unit (P = 0.08) and from 29.1 mM (SD 2.1) to 28.5 (SD 2.0) using 2 units (P = 0.01; Fig. 8, E and F).

Fig. 8.

Influences on acid-base status. A: arterial pH [1 electro-oxidation (EO) unit: n = 7; 2 EO units: n = 8]. B: arterial bicarbonate (1 EO unit: n = 7; 2 EO units: n = 8). C: venous bicarbonate (1 EO unit: n = 7; 2 EO units: n = 9). D: arterial Pco₂ (1 EO unit: n = 7; 2 EO units: n = 8). E: plasma bicarbonate at the inlet (In) and outlet (Out) of the dialyzer (1 EO unit: n = 7, 2 EO units: n = 9). F: dialysate bicarbonate concentration upstream (In) and downstream (Out) of the EO unit (1 EO unit: n = 7; 2 EO units: n = 9). In E and F, measurements were performed each hour, and per experiment the concentrations In and Out were averaged.

Plasma anion gap showed a tendency to increase in time in the systemic circulation (P = 0.18 using 1 unit, and P = 0.07 using 2 units; Fig. 9A), whereas a tendency to decrease was observed across the dialyzer (P = 0.06 for 1 unit, and P < 0.001 for 2 units; Fig. 9B). Lactate did not increase systemically (Fig. 9C). Plasma chloride concentration showed a tendency to gradually increase in the systemic circulation from 109 mM (SD 3) to 111 mM (SD 2) (P = 0.005) using 1 unit and from 109 mM (SD 2) to 112 mM (SD 3) (P = 0.08) using 2 units (Fig. 9D) and across the dialyzer using 2 units from 112 (SD 1.5) to 111 (SD 1.5) (P = 0.03; Fig. 9E). Plasma chloride concentration remained stable across the dialyzer using 1 unit (Fig. 9E). Dialysate chloride concentration increased across the EO unit from 120 mM (SD 3) to 121 mM (SD 3) (P = 0.03) using 1 unit and from 121 mM (SD 2) to 122 mM (SD 2) (P = 0.03) using 2 units (Fig. 9F).

Fig. 9.

Influences on acid-base status. A: venous anion gap [1 electro-oxidation (EO) unit: n = 7; 2 EO units: n = 9]. B: plasma anion gap at the inlet (In) and outlet (Out) of the dialyzer (1 EO unit: n = 7; 2 EO units: n = 9). C: arterial lactate concentration (1 EO unit: n = 7; 2 EO units: n = 8). D: venous chloride concentration (1 EO unit: n = 7; 2 EO units: n = 9). E: plasma chloride concentration at the In and Out of the dialyzer (1 EO unit: n = 7; 2 EO units: n = 9). F: dialysate chloride concentration at the In and Out of the dialyzer (1 EO unit: n = 7; 2 EO units: n = 9). In B, E, and F, measurements were performed each hour, and per experiment the concentrations In and Out were averaged.

Iron.

Iron did not significantly change across the dialyzer or systemically during the experiments (Fig. 10).

Fig. 10.

A: plasma iron concentration at the inlet (In) and outlet (Out) of the dialyzer (n = 16). Measurements were performed at the start and end of each experiment, and per experiment the concentrations In and Out were averaged. B: plasma iron concentration at the start and end of each experiment (n = 16).

Glucose.

Plasma and dialysate glucose concentrations decreased across the dialyzer from 5.5 mM (SD 1.6) to 5.0 mM (SD 1.6) (P < 0.001; Fig. 11A) and EO unit from 5.7 mM (SD 1.6) to 5.2 mM (SD 1.7) (P < 0.001; Fig. 11B), respectively. Systemic plasma glucose concentration remained stable during the experiment (Fig. 11C).

Fig. 11.

A: influences on plasma and dialysate glucose concentration at the inlet (In) and outlet (Out) of the dialyzer (n = 16). B: influences on dialysate glucose concentration upstream (In) and downstream (Out) of the electro-oxidation (EO) unit (n = 16). C: plasma glucose concentration (n = 16). In A and B, measurements were performed each hour, and per experiment the concentrations In and Out were averaged.

Chlorine species.

Chlorine concentrations in the dialysate downstream of the AC filter were below the maximum allowed level of 0.1 mg/l, according to AAMI standards for conventional HD (1) at dialysate flow rates of 40 ml/min. At dialysate flow rates of 70 ml/min (4 experiments), total dialysate chlorine was >0.1 mg/l in 0, 2, 4, and 4 out of 4 measurements per experiment with a maximal bound chlorine of 0.22 mg/l and maximal free chlorine of 0.05 mg/l.

Ammonium.

Limited release of ammonium into plasma was observed [Fig. 12A; 54 µM (SD 30) at the inlet and 387 µM (SD 68) at the outlet of the dialyzer; P = 0.03; 148 µmol (SD 48) ammonium per removed mmol urea per h]. This did not result in a systemic increase in plasma ammonium during the experiment [Fig. 12B; 36 µM (SD 22) at the start vs. 58 µM (SD 31) at the end; P = 0.25].

Fig. 12.

A: plasma ammonium concentration at the inlet (In) and outlet (Out) of the dialyzer (n = 6). Measurements were performed at the start and end of each experiment, and per experiment the concentrations In and Out were averaged. B: plasma ammonium concentration at the start and end of each experiment (n = 6).

DISCUSSION

In this study, clinically relevant urea removal was achieved with a miniature dialysis device in awake nonuremic goats by use of EO. The biocompatibility of this technique is a point of concern, as ammonium release was detected and dialysate chlorine concentrations were above acceptable levels at higher dialysate flow rates.

Our study shows that a miniature dialysis device containing an EO unit with graphite electrodes can remove ~8.4 mmol urea/h at a plasma [urea] of ~12 mM. This amount is comparable to that reported previously in vitro (9.5 mmol/h at a higher blood [urea] of 18.6 mM) (39). As shown in our in vitro report, absolute urea removal increased with increasing urea concentration. In our in vitro report, we suggested several measures to increase urea removal. First, we hypothesized that increasing the electrode surface area would result in more urea removal. In the present study, we showed that doubling the electrode surface area indeed doubled the urea removal rate (15.9 mmol/h). Second, we suggested that a higher dialysate flow rate would result in more urea removal. However, the current experiments showed that increasing the dialysate flow rate from 40 to 70 ml/min did not increase urea removal (15.8 mmol/h). We submit that reduced dialysate electrode-contact time, resulting from an increased dialysate flow rate, compromised the urea oxidation efficiency and, thereby, overall urea degradation (30, 39). Third, increasing the applied current is expected to increase urea removal. However, in our in vitro report, we showed that a current of 3A (0.5 A/dm² electrode), as applied in the current study, resulted in the lowest release of chlorine species per removed mole of urea (39).

To compensate for the total production of urea from catabolism of dietary protein, ~250–400 mmol of urea will have to be removed daily (13, 31, 37). Extrapolation of the urea removal that we obtained by using two EO units in parallel during 3 h would result in removal of ~380 mmol per day if the device were used continuously 24 h/day. To achieve similar time averaged urea plasma clearance as thrice weekly 4-h HD [12.3 ml/min (9)] or continuous ambulatory peritoneal dialysis (PD) [9 ml/min (10)], the 2-unit device (urea clearance: 19.5 ml/min) should be used for at least ~15 h/day or 11 h/day 7 days per week, respectively. However, several factors preclude the use of our device for an extended time up to 24 h/day. First, our two-unit device weighs ~2 kg (including battery), which may be too heavy for frail patients. Second, the lack of a reliable vascular access that can be used continuously is a bottleneck since conventional long-term venous access solutions, such as tunneled CVCs, carry the risk of considerable blood loss and air embolism in case of accidental disconnection and are associated with a high incidence of bloodstream infections (15). To allow the use of EO for intermittent HD, it will be necessary to further increase the cumulative electrode surface of the EO units. For example, for daily 8-h nocturnal HD, the electrode surface area would have to be increased by two- to threefold.

Creatinine removal showed a different pattern than urea removal. Creatinine was completely removed from the dialysate when one unit or two units were used, also at a higher dialysate flow rate, and was therefore flow-dependent (in contrast to urea removal). Creatinine removal and clearance were primarily determined by equilibration of creatinine across the dialyzer membrane, which was not complete using a dialyzer with a surface area of 0.2 m² (dialysate concentrations downstream of the dialyzer were lower than plasma concentrations upstream of the dialyzer at all dialysate flow rates) and may therefore be further increased by using a dialyzer with a higher surface area. Of note, the absolute amount of creatinine that was removed is relatively low (at maximum 210 µmol/h) compared with the daily creatinine production [~8–17 mmol (19)] because of low (nonuremic) plasma creatinine concentrations (range: 45–76 µM) in the healthy goats. In vitro studies performed earlier with the single unit device showed that creatinine removal up to ~1 mmol/h could be achieved at “uremic” concentrations (unpublished observation, Fig. 13). To match time-averaged plasma clearances for conventional HD and PD [9.4 ml/min and 4.3 ml/min (10), respectively], the 2-unit device (50 g of AC) would need to be used for at least 25 h/week or 12 h/week, respectively. Alternatively, the number of units and the cumulative dialysate flow could be increased.

Fig. 13.

Cumulative creatinine removal in vitro circulating uremic plasma (n = 3; mean [creatinine] 734 (132) µM) at a blood flow of 110 ml/min and a dialysate flow of 40 ml/min using 1 electro-oxidation unit.

Clinically relevant potassium and phosphate removal was achieved with 2 units (17.9 mmol/3 h and 8.2 mmol/3 h, respectively, as compared with a required daily removal of ~45 mmol potassium and ~15 mmol phosphate) (21, 33, 35). In comparison, potassium and phosphate removal with conventional HD is ~53 mmol (2) and ~15–30 mmol per 4-h session (22, 36), respectively, and with PD ~29–42 mmol/day (24, 41) and ~7.2 mmol/day (6), respectively. Potassium and phosphate removal using 1 EO unit were in the same range as the previously reported removal using sorbent units without EO (38). Consequently, potassium and phosphate removal appear not to be influenced by EO. Increasing the amount of potassium sorbent from 80 to 180 g (×2.25) and the amount of phosphate sorbent from 40 to 60 g (×1.50) resulted in a comparable increase in removal (a 2.29-fold increase in potassium removal and a 1.55-fold increase in phosphate removal). To match the potassium and phosphate removal during a conventional HD session, the amounts of the potassium and phosphate sorbent should be further increased to ~533 g and ~110–220 g, respectively, and to match PD, only the amount of the potassium sorbent should be increased to ~292–422 g.

Although these in vivo results showed that clinically relevant removal of urea, potassium, and phosphate can be achieved with our prototype miniature device that combines EO and sorbent technology, several adverse effects were noted. They can be divided into EO-related events, sorbent-induced changes, and effects probably induced by the specific animal model.

Specific EO-induced adverse effects included the generation of chlorine by-products and ammonium. Although dialysate chlorine concentrations remained well below the maximum levels defined in the AAMI standards for dialysate (bound chlorine <0.10 mg/l, free chlorine <0.50 mg/l) at a dialysate flow rate of 40 ml/min, concentrations exceeding these levels were detected at a dialysate flow rate of 70 ml/min. This should be avoided, as increased levels of chlorine by-products may cause hemolysis (34). The explanation for the increased chlorine levels at higher dialysate flow rates is probably that the contact time of the dialysate with the AC was too short for sufficient adsorption to occur (14). To counteract this problem, the amount of AC that functions as an electron donor for reductive dechlorination (39) could be increased. Alternatively, the current density (i.e., current per unit surface area of the electrode) could be decreased while increasing the total electrode surface area to maintain urea removal rate constant, since we observed in vitro that this strategy virtually abolishes chlorine release (unpublished observation).

Ammonium was also generated in the EO unit, probably due to hydrolysis of urea (20), and released into the dialysate at a rate of 148 (48) µmol ammonium per removed mmol urea per hour. At a urea removal rate of ~16 mmol per hour, this would amount to a production of ~7 mmol during a 3-h treatment. This amount is low compared with the daily amount of ammonia/ammonium that is converted into urea by the liver (~500–800 mmol) and resembles the production by healthy kidneys under basal conditions (~60–80 mmol/day) of which ~50% is secreted and the remaining enters the systemic circulation (37). Ammonium release did not result in a systemic increase in ammonium concentration, presumably because of fast hepatic conversion into urea and possibly also increased ammonium excretion by the healthy kidneys (25, 26). For proper evaluation of the influence of EO-induced ammonia/ammonium release on systemic ammonium concentrations in end-stage kidney disease, experiments should be repeated in animals with end-stage kidney disease.

Nevertheless, the direct release of ammonium into the systemic circulation may be harmful, in particular for the brain, although other organs and tissues may also be affected (7) and should preferably be avoided. Consequently, strategies should be developed to prevent the generation of ammonium in the EO unit or to prevent its release into the circulation by incorporating additional cation exchanger in the circuit downstream of the EO unit.

Changes induced by the sorbents included minor decreases in systemic plasma calcium and magnesium concentrations. As both the plasma calcium and magnesium concentrations decreased slightly after flowing through the dialyzer, some adsorption of these substances must have occurred by the PS-DVB sulfonate beads contained in the EO/sorbent unit. This can be easily prevented by prerinsing the sorbents at somewhat higher calcium and magnesium concentrations. Of note, sodium release, which could in theory occur because of exchange of sodium for other cations by the PS-DVB sulfonate beads, was prevented by prerinsing the sorbents at 120 mM [Na⁺], as previously reported (38). A minimal decrease in sodium concentration (<1 mmol) occurred across the dialyzer, suggesting even minimal sodium adsorption by the PS-DVB sulfonate beads. Consequently, we should use a slightly higher sodium concentration for prerinsing the sorbents. Plasma and dialysate glucose concentrations decreased moderately across the dialyzer and EO/sorbent unit, respectively, whereas systemic glucose concentration did not change during the experiments. Removal of glucose by the device is probably due to adsorption by activated carbon, which may be prevented by prerinsing at a physiological glucose concentration. In vitro experiments have shown that electro-oxidation of glucose is negligible (data not shown).

Several changes in plasma chemistry could not be explained by parallel changes occurring in plasma flowing through the dialyzer, and they must therefore be related to the animal model using nonuremic goats. First, the systemic plasma sodium concentration increased, despite a tendency of the plasma sodium concentrations to decrease across the dialyzer. This very mild systemic increase in plasma sodium concentration may be related to a high electrolyte-free water clearance inherent to urea-induced osmotic diuresis (5, 19, 27). Second, a small but significant decrease in plasma bicarbonate concentration of 1–2 mM was observed, which contrasted with a minimal (<1 mM) increase in plasma bicarbonate across the dialyzer. This reduction in plasma bicarbonate concentration was accompanied by an increase in plasma chloride of a similar magnitude. The most likely explanation for this combination (hyperchloremic metabolic acidosis) is the intravenous infusion of solutions that exclusively contained chloride as an anion (16), although minor chloride release across the dialyzer (only in experiments with 2 units) probably related to the exchange of phosphate and organic acids for chloride in the FeOOH beads, may have played a role. No influence on pH was observed, which was due to a coinciding decrease in arterial Pco₂, probably because of stress-induced hyperventilation which increased during the experiment while the effect of the sedatives wore off. Third, hemolysis was observed in vivo. This could be attributed to urea infusion, since visible hemolysis and LDH increase occurred exclusively in experiments with urea infusion. Importantly, hemolysis also occurred in an experiment with urea infusion in which no dialysis and EO were performed, indicating that hemolysis was not EO-related. In agreement with our observation in goats, it has been shown that urea infusion also induces hemolysis in humans (4, 27). Finally, the leukocyte number increased during the experiments. This may be due to “catecholamine-induced leukocytosis,” a phenomenon characterized by a rapid and transient increase in leukocytes observed during acute stress, presumably reflecting demargination of leukocytes from the marginal pool (39). In agreement with such an acute stress reaction, diastolic blood pressure and heart rate increased during treatment in these awake animals. Similar leukocyte and cardiovascular responses were observed in animals treated by the device without activating EO (38) and in HD experiments using a conventional HD machine (unpublished observation, Fig. 14), suggesting that this stress reaction was not induced by EO, the sorbents, or other materials in the device. However, it has not been excluded that dialyzable substances derived from the device may provoke an acute inflammatory response with a presentation indistinguishable from that of an acute stress reaction. To exclude this, the proinflammatory capacity of dialysate treated by the device should be studied in vitro.

Fig. 14.

Leukocyte number at the start and end of treatment with a conventional hemodialysis machine (n = 3).

In summary, the present study shows that a device combining EO, sorbents, and AC can remove relevant amounts of urea, potassium, and phosphate in a nonuremic large-animal model. However, many challenges have to be overcome to reach the stage of application of the device in humans. First, a uremic large-animal model has to be developed, as the current model involves infusion of relatively large volumes of solutions containing electrolytes and urea that obscure the true effects of the device on electrolyte and acid-base parameters and hemocompatibility. Second, little is known about potentially toxic oxidative by-products, other than chlorines and ammonium, that may be formed during electro-oxidation (38). Research is underway to define the chemical constituents of dialysate of uremic subjects exposed to electro-oxidation and to test the biocompatibility of such fluids in in vitro experiments (28). Third, the changes in the concentration of bicarbonate in plasma flowing through the dialyzer are small and equivocal. Apparently, very little bicarbonate is released from the current device, which would prevent correction of the metabolic acidosis present in uremic patients needing dialysis, and this problem remains to be solved. Finally, the system would require upscaling to develop a portable device that can be used for nocturnal or daily home dialysis. However, this process will also increase the generation of known toxic by-products of electro-oxidation, such as chlorines or ammonium, and strategies will have to be developed to prevent the production or release of these substances.

GRANTS

This study was financially supported by the Dutch Kidney Foundation (Grant no. NT 12.05) and the European Commission (Nephron+ Grant FP7-ICT-2009-4 and WEAKID, Horizon 2020 Research and Innovation program, Grant agreement no. 733169).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

M.W., J.A.J., F.S., and K.G.F.G. conceived and designed research; M.W., F.S., D.H.M.H., T.W.M.v.B., and K.R.D.V. performed experiments; M.W., M.K.v.G., and K.G.F.G. analyzed data; M.W., M.K.v.G., J.A.J., F.S., and K.G.F.G. interpreted results of experiments; M.W. and M.K.v.G. prepared figures; M.W. and K.G.F.G. drafted manuscript; M.W., M.K.v.G., J.A.J., F.S., W.H.B., M.C.V., and K.G.F.G. edited and revised manuscript; M.W., M.K.v.G., J.A.J., F.S., D.H.M.H., T.W.M.v.B., K.R.D.V., W.H.B., M.C.V., and K.G.F.G. approved final version of manuscript.