bs_bs_banner

© 2018 The Authors. Artificial Organs published by Wiley Periodicals, Inc. on behalf of International Center for Artificial Organ and Transplantation (ICAOT)
This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in
any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made

Zero Diffusive Sodium Balance in Hemodialysis Provided

by an Algorithm-Based Electrolyte Balancing Controller: A
Proof of Principle Clinical Study

*Uwe Kuhlmann, †Andreas Maierhofer, †Bernard Canaud, ‡Joachim Hoyer and

§Malte Gross
*Klinikum Bremen Mitte, Medizinische Klinik III, Bremen, Germany; †Fresenius Medical Care GmbH, Bad Homburg;
‡Universitätsklinik Marburg, Klinik für Innere Medizin, Nephrologie und Internistische Intensivmedizin, Marburg;
§Faculty of Mechatronics and Medical Technology, Ulm University of Applied Sciences, Ulm

Abstract: Restoring and controlling fluid volume homeo-
stasis is still a challenge in contemporary end-stage kidney
disease patients treated by intermittent hemodialysis (HD)
or hemodiafiltration (HDF). This primary target is
achieved by ultrafiltration (dry weight probing) and con-
trol of intradialytic sodium transfer (dialysate-plasma Na
gradient). The latter task is mostly ignored in clinical prac-
tice by applying a dialysate sodium prescription uniform
for all patients of the dialysis center but unaligned to indi-
vidual plasma sodium levels. Depending on the patient’s
natremia, a positive gradient gives rise to intradialytic dif-
fusive sodium load and postdialytic thirst. On the contrary,
a negative gradient may cause unwanted diffusive sodium
removal and intradialytic symptoms. To overcome these
challenges, a new conductivity-based electrolyte balancing
algorithm embedded in a hemodialysis machine with the
aim to achieve “zero diffusive sodium balance” in HD and
online HDF treatments was tested in the form of a pro-
spective clinical trial. The study comprised two phases: a
first phase with a conventional fixed-sodium dialysate
(standard care phase), followed by a phase with the
electrolyte balancing control (EBC) module activated
(controlled care phase). The results show a reduction in the
variability of the intradialytic plasma sodium concentra-
tion shift, but it is overlain by a small but statistically sig-
nificant increase in the mean plasma sodium levels.
However, no clinical manifestations were observed. This
sodium load can be explained by the design of the algo-
rithm based on dialysate conductivity instead of sodium
concentration. Furthermore, the increase in plasma so-
dium can be corrected by taking into account the potas-
sium shift during the treatment. This study showed that the
EBC module incorporated in the HD machine is able to
automatically individualize the dialysate sodium to the pa-
tient’s plasma sodium without measuring or calculating
predialytic plasma levels from previous laboratory tests.
This tool has the potential to facilitate fluid management,
to control diffusive sodium flux, and to improve intradia-
lytic tolerance in daily clinical practice. Key Words:
Conductivity—Sodium gradient—Diffusive sodium bal-
ance—Electrolyte balance—Hemodialysis—Online hemo-
diafiltration—Renal replacement therapy

Restoring and controlling fluid volume homeo-
stasis is still a challenge and clearly an unmet med-
ical need in contemporary end-stage kidney disease
doi: 10.1111/aor.13328
Received February 2018; revised June 2018; accepted July
2018.
Address correspondence and reprint requests to Malte
Gross, Faculty of Mechatronics and Medical Technology, Ulm
University of Applied Sciences, Albert-Einstein-Allee 55, Ulm
89081, Germany. E-mail: m.gross@hs-ulm.de
(ESKD) patients treated by intermittent hemodial-
ysis as renal replacement therapy (RRT) (1). Fluid
volume balance is currently achieved by combining
fluid and sodium depletion by RRT: By a so-called
“dry weight probing” approach referring to intra-di-
alytic weight loss (IDWL) (1,2), by adjusting the
dialysate-to-plasma sodium gradient (3‒5) and by
enforcing daily dietary salt and fluid restrictions (6).
The improper control of fluid volume and sodium
mass balance in RRT patients leads to chronic fluid

Artificial Organs 2018, 0(0):1–9

2 U. Kuhlmann et al.
overload that may result in hypertension, pulmonary
edema and cardiac complications with poor out-
comes (7‒9). On the other side, too tight control of
fluid volume with high ultrafiltration rates may lead
to transitory hypovolemia and intradialytic hypo-
tensive episodes, which in turn are associated with
hemodynamic stress repetitive ischemic insults, and
multiorgan damage (10‒12).
Inter-dialytic weight gain (IDWG) is usually re-
garded as an indirect clinical indicator of patients’
adherence to salt and fluid diet restriction (13). As
shown by several studies, IDWG may be modulated
by many factors, but the foremost one being excessive
thirst occurring during the interdialytic interval. The
sensation of thirst may be triggered by several stimuli
; however, plasma sodium concentration is the most
important. (3). As soon as the plasma sodium con-
centration exceeds some individual threshold, thirst
is triggered by dipsogenic receptor cells in the brain
(14,15). Intradialytic sodium load should therefore be
avoided in order to respect this setpoint (16,17). This
is especially important since some ESKD patients
may have a lower setpoint than healthy humans (18).
On the other hand, a negative sodium balance may
cause severe intradialytic cardiovascular complica-
tions, like hypotensive episodes (19).
Considering all these facts, it has been clearly iden-
tified that individualized prescription of dialysate so-
dium is necessary to ameliorate patient perception,
to improve tolerability, and to minimize side effects
of RRT, and in the long term to reduce cardiovas-
cular disease burden and improve outcomes (20).
However, there is no general procedure on how to
find this individual prescription for each patient in
clinical practice. Ideally, to restore sodium and fluid
homeostasis in dialysis patients, the accumulated
sodium intake in the interdialytic interval should
be compensated by an equivalent sodium removal
during the dialysis session (21). Unfortunately, the
dietary sodium load is difficult to quantify (22) and
usually is not known. Keen and Gotch suggested to
set the gradient between dialysate and plasma so-
dium concentrations to zero during dialysis (17). This
would mean that sodium is removed only through
net ultrafiltration and not by diffusion. In this “zero
diffusive sodium balance” approach, the plasma so-
dium at the end of the dialysis session should be the
same as the predialytic value.
The adjustment of dialysate sodium can be done
either by aligning it manually to the predialytic
plasma sodium or automatically by a control algo-
rithm integrated into the dialysis machine.
Manual adjustment of dialysate sodium has been
assessed successfully (16,23,24). However, it requires
Artificial Organs, Vol. 0, No. 0, 2018
the determination of the plasma sodium, which is a
tedious and costly procedure not suited for routine
clinical practice. Moreover, due to the Gibbs–Donnan
effect, the potentiometrically measured plasma con-
centrations have to be corrected depending on the
protein content. Raimann et al. (25) attempted to
avoid predialytic sodium measurements using a the-
oretical model allowing prediction of sodium values
from monthly routine measurements.
Automated adjustments of dialysate sodium using
conductivity as a surrogate of sodium concentration
have been tested in the past. This has the advantage of
avoiding the determination of the Donnan factor. The
plasma conductivity can be indirectly calculated from
the dialysate conductivity values up- and downstream
of the dialyzer (26). The Diacontrol module of the
Hospal/Gambro Integra (Medolla, Italy) was the first
control algorithm based on this method in a commer-
cially available dialysis machine. It works by adjusting
the dialysate conductivity automatically by changing
the mixing ratio of water and concentrate so that the
plasma conductivity attains a user defined value at the
end of the dialysis session (27). In principle, this method
can also be used to keep the plasma conductivity con-
stant during the dialysis session in order to avoid elec-
trolyte load or loss during dialysis (28). Obviously, this
requires a previous determination of the plasma con-
ductivity. One of the drawbacks of this method is that
plasma conductivity is a much less common parameter
in clinical practice than sodium concentration. A prac-
tically usable electrolyte feedback control algorithm
should not depend on blood sampling, laboratory
analysis or nonstandard disposables. Since the Integra
dialysis machine also determines the ionic mass bal-
ance, Lambie et al. used this information to control the
sodium balance by adjusting the dialysate conductivity
manually. They could achieve improvements in intra-
dialytic weight gain and blood pressure (29). However,
to be usable in daily clinical practice, the electrolyte
adjustment should work automatically without any
human intervention or surveillance.
Our approach with the development of the EBC
module and its algorithm is different. We aim to
achieve a fully automated control of the dialysate
sodium during HD and HDF sessions while achiev-
ing zero diffusive sodium balance and thus keeping
plasma tonicity almost constant.
PATIENTS AND METHODS
Electrolyte balancing algorithm and controller
The electrolyte balancing control (EBC) algo-
rithm is based on the dialysate-side mass balance
 AUTOMATED SODIUM CONTROL IN HEMODIALYSIS
equation for a specific solute. The mass flow for a
specific solute into the patient is the sum of the mass
flow into the dialyzer and the reinfusion flow into
the blood stream due to the Online-HDF post or
predilution:
Jin = Qdial cdi + Qs cdi
where Qdial is the dialysate flow at the dialyzer
inlet, Qs the Online-HDF substitution flow and cdi
the concentration of the solute in the fresh dialysate.
On the other hand, the mass flow out of the patient
is equal to the product of the dialysate flow at the
dialyzer outlet and the spent dialysate concentration
cdo:
Jout = (Qdial + Qs + Qf )cdo
where Qf is the net ultrafiltration rate. In the case
of an HD treatment, Qs is zero. Thus the total mass
balance ΔJ = Jin − Jout is:
ΔJ = Qd (cdi − cdo ) − Qf cdo ,
where Qd = Qdial + Qs is the total fresh dialysate
flow generated by the machine. The mass balance ΔJ
then describes the total amount of solute per time
unit transferred into or out of the patient. A positive
value of ΔJ indicates a net transfer into the patient.
Note that the mass balance can be split into two
components: the effective “diffusive component”
Qd(cdi − cdo) and the net ultrafiltration component
Qfcdo. The latter quantifies the mass flow removed
by ultrafiltration. Small solutes, as electrolytes, do
not undergo significant sieving across the dialyzer
membrane; hence their concentrations in the effluent
ultrafiltrate can be considered close to those in the
blood plasma. However, due to the Gibbs–Donnan
effect, the sodium concentration in the ultrafiltrate
is a few percent smaller than in plasma water (30).
Because the ultrafiltration volume is much smaller
than the sodium distribution volume, we expect the
effect of this misbalance on plasma sodium to be
negligible. In order to keep the plasma concentration
constant, an active electrolyte balancing should thus
assure a zero diffusive balance:
ΔJdiff = Qd (cdi − cdo ) = 0.
The zero-balance controller works by adjusting
the mixing ratio of water and concentrate and thus
cdi until ΔJdiff becomes zero. In the dialysis machine
used in the clinical trial, the control algorithm was im-
plemented as a proportional-integral (PI) controller.
3
The proportional part tries to keep the actual value
of ΔJdiff close to zero while the integral part keeps
track of the cumulated diffusive mass balance and
compensates an eventual imbalance by an oppositely
directed shift of cdi.
Since the dialysate conductivity is governed by the
electrolyte concentration where sodium is the dom-
inant cation, in a first approximation conductivity is
an easy and reliable proxy of dialysate sodium con-
centration (27). A principal objective of our study
was to test whether a zero diffusive electrolyte bal-
ance also implicates a zero diffusive sodium balance.
In the described trial, two Fresenius Medical Care
(Bad Homburg, Germany) 5008 dialysis machines
were equipped with special software allowing de-
termination and control of conductivity balance. No
changes in the hardware were made; the dialysate
conductivities up- and downstream of the dialyzer
were measured by the built-in conductivity cells of
the Online Clearance Monitor (OCM). An automatic
calibration of the conductivity cells is assured in reg-
ular intervals to compensate eventual deviations.
While sharing the conductivity cells, the electrolyte
balancing algorithm works independently from the
OCM module and used neither the clearance nor the
plasma Na calculated by OCM. A schematic view of
the EBC is shown in Fig. 1.
Patients and study parameters
Twenty informed ESKD patients were initially
included, 16 of them remained until the end of the
study. Seven patients were female and nine male.
Five patients had a known residual diuresis. Their
urinary excretion ranged from 900 to 1700 mL/day.
One patient was anuric, in the remaining 10 patients
the residual kidney function was unknown. Further
patient data are summarized in Table 1.
The trial was approved by the Ethical Committee
of Philipps-Universität Marburg. In order to be in-
cluded in the study, the patients had to meet the fol-
lowing inclusion criteria:
• Vascular access by arteriovenous graft or fistula.
• At least 6 months on chronic hemodialysis
• Not scheduled for transplantation for the duration
of the study
• No participation in a concurrent clinical study
• No severe vascular access problems
• No amputations
• No active systemic infections like HBV, HCV,
HIV, or tuberculosis
• No severe cardiovascular instability
• No unstable angina pectoris
Artificial Organs, Vol. 0, No. 0, 2018
4 U. Kuhlmann et al.

FIG. 1. Schematic view of the electrolyte balancing controller (EBC). Inlet and outlet dialysate conductivities cdi and cdo are measured
by conductivity cells and used together with the total dialysate flow Qd and ultrafiltration rate Q f as input for the EBC. The controller
adjusts the dialysate inlet sodium concentration in order to achieve zero diffusive sodium balance.

In all treatments Fresenius SK-F 313/1 concentrate was TABLE 1 . Patient data at the time of the first treatment
used to prepare the dialysate. Using the standard mixing after inclusion into the study
ratio, this concentrate yields the following dialysate com-
position: 138 mmol/L Na+, 3 mmol/L K+, 1.5 mmol/L Ca2+, Mean Minimum Maximum
0.5 mmol/L Mg2+, and 1 g/L glucose. Bicarbonate was sup- Age , years 61.3 38.6 77.6
plied in liquid form in order to avoid concentration variations
Height, cm 168.7 155.0 189.0
due to the dissolution of dry bicarbonate. Canisters taken
BMI, kg/m 2 26.3 17.1 33.6
from a single production batch were used for both the dial-
ysate concentrate and the bicarbonate solution. Polysulfone TBW, L 37.9 26.5 50.3
high-flux hollow-fiber dialyzers of the Fresenius Medical Overhydration, L 3.1 0.0 6.1
Care FX-type were used in all treatments. Depending on Hb, g/dL 10.9 9.6 13.2
the prescribed blood flow, sizes from FX60 up to FX100
Body mass index (BMI) was calculated from height and weight,
were chosen. All patients were dialyzed three times per total body water (TBW), and overhydration were measured by
week. Because the electrolyte balancing algorithm does not the Body Composition Monitor (BCM). Hemoglobin (Hb) was
explicitly depend on the substitution rate Qs, we expect it to determined by laboratory analysis.
work in both HD and HDF treatments. Since online HDF
is a very common treatment modality and the most chal- rate, and mode as well as the dialyzer type were not
lenging modality in terms of fluid and electrolyte balancing changed between phase A and B. Since the treat-
today, we focused on HDF in our clinical study. From a total
ment mode in phase A did not differ from the treat-
of 126 analyzed treatments, HDF postdilution was the treat-
ment mode before the study, a wash-out phase was
ment mode in 112 sessions, HDF predilution in 10 sessions.
In two treatments the mode was switched from HDF post not required between A and B. In both phases dial-
to predilution during the sessions. In another two sessions, ysate and ultrafiltration flow and the conductivity
HD was the treatment mode. values in the affluent and effluent dialysate were
The study was conducted in two phases: In phase registered continuously by the dialysis machines.
A (standard care phase), the dialysate sodium con- In order to determine the blood-side electrolyte
centration was set to the fixed value of 138 mmol/L. balance, blood samples were drawn from the arte-
This was the common standard value used for all rial blood line at the beginning and the end of each
patients before and after the study. In phase A, treatment, respectively. The blood samples were
four dialysis sessions were planned. Subsequently, analyzed immediately by direct potentiometry for
the patients passed to phase B (controlled care plasma Na+, K+, Ca2+, Cl−, and glucose using an
phase) which also comprised four treatments; this ABL800 FLEX blood gas analyzer (Radiometer,
time, however, with active EBC. In this phase, the Copenhagen, Denmark).
dialysate sodium was set to 138 mmol/L at the be- As supporting information the dialysis dose Kt/V
ginning of the session. Then, the EBC was activated was measured by OCM. Before each session the total
to achieve “zero diffusive balance” by adjusting the body water (TBW) volume was determined by bio-
dialysate sodium concentration. All other parame- impedance using a Fresenius Medical Care Body
ters, such as blood flow, dialysate flow, substitution Composition Monitor (BCM). The obtained TBW

Artificial Organs, Vol. 0, No. 0, 2018

 AUTOMATED SODIUM CONTROL IN HEMODIALYSIS 5

was used as the urea distribution volume V needed TABLE 2 . Results from the standard care phase and the
as input for the OCM module. controlled care phase

Standard Na Controlled Na P
Statistical analysis
From the 20 patients initially included in the study ΔNap, mmol/L −1.0 ± 1.4 1.3 ± 0.9 <0.001
two left after the standard care phase for personal ΔKp, mmol/L −0.82 ± 0.41 −0.85 ± 0.34 0.83 (n.s.)
reasons, another had to be excluded because of se- Napre, mmol/L 137.1 ± 2.3 137.2 ± 2.2 0.67 (n.s.)
vere hyponatremia. One patient repeatedly suffered Kpre, mmol/L 4.78 ± 0.54 4.90 ± 0.47 0.49 (n.s.)
from cramps in the standard care phase and had to Nad, mmol/L 138.1 ± 0.1 142.2 ± 2.2 <0.001
be treated with saline infusions. Thus, 16 patients Qd, mL/min 396 ± 75 408 ± 84 0.31 (n.s.)
were included in the statistical analysis. Blood and Q b, mL/min 280 ± 57 293 ± 64 0.13 (n.s.)
dialysate flows as well as substitution rates and fresh Vsub, L 13.45 ± 3.76 13.68 ± 4.35 0.76 (n.s.)
dialysate sodium concentrations as set by the user or VUF, mL 2306 ± 1019 2495 ± 1075 0.23 (n.s.)
the control algorithm were averaged over the active Tdial, min 251 ± 23 254 ± 23 0.17 (n.s.)
treatment time. Together with the electrolyte data TBW, L 37.8 ± 7.4 37.7 ± 7.3 0.39 (n.s.)
from the blood samples, this led to a collection of Kt/V (OCM) 1.30 ± 0.36 1.40 ± 0.40 0.02
per-session data.
Mean ± SD.
Patient wise averaging over all sessions in each Dialysate Na was calculated from its concentration in the con-
study phase led to patient-specific mean values. In centrates and the concentrate-water mixing ratio of the dialysis
order to check whether the electrolyte balancing had machine. Dialysate Na and flow values were time-averaged over
the duration of each session.
a statistically significant effect on a specific quantity,
the group averages were compared by a two-sided
paired t-test (MS Excel 2010), where the averages for
each patient in the standard care and controlled care 6.0
phases, respectively, formed a pair. Standard devia- controlled Na
tions were compared by the two-sided Fisher’s F-test 4.0 standard Na
(31).
For the plasma electrolytes the difference between 2.0
the postdialytic and predialytic concentrations were
[mmol/l]

calculated in order to determine the net electrolyte 0.0

∆Na

shift during the treatments. In the following, a posi-

tive electrolyte shift ΔXp (X = Na or K) denotes an
increase in the plasma level of this specific electrolyte.
RESULTS
Table 2 summarizes the results. Averages over
all 16 patients who participated in both phases are
given. Figs. 2‒4 show the plasma electrolyte data on
a patient-specific basis.
Fig. 2 shows the plasma sodium concentration shift
in both phases for each of the patients, Fig. 3 the
plasma potassium concentration shift, respectively.
Fig. 2 also shows that the controlled care phase
lead to a positive plasma sodium concentration shift
for 15 out of 16 patients, whereas in the standard care
phase only nine show a positive plasma sodium shift.
Averaging the sodium shift over all 16 patients, the
mean sodium shift was 1.3 mmol/L with a standard
deviation of 0.9 mmol/L when electrolyte control
was active. On the contrary, the fixed sodium dialy-
sate led to a mean sodium shift of −1.0 mmol/L with
a standard deviation of 1.4 mmol/L. This difference is
highly significant (P < 0.001). Moreover the standard
-2.0
-4.0
-6.0
Patient 1-16
FIG. 2. Plasma sodium shift (Mean ± SD) during dialysis for
each patient. In the standard care phase (squares) dialysate
sodium was kept constant at Nad = 138 mmol/L, in the
controlled phase (diamonds) the zero diffusive electrolyte
balancing control was active.
deviation in the controlled care phase is smaller than
that of the standard phase. However, this difference
is just not significant (P = 0.052).
The cause of the positive sodium shift in the ses-
sions with active electrolyte control can be deduced
from the mean dialysate sodium concentration Nad as
shown in Table 2. Nad is the sodium concentration pro-
duced by the mixing system of 5008 dialysis machine,
averaged over the time of the dialysis session. In the
standard care phase, Nad stayed very close to the pro-
grammed fixed level of 138 mmol/L with negligible
variation. In the controlled care phase, however, the

Artificial Organs, Vol. 0, No. 0, 2018

6 U. Kuhlmann et al.

0.5 Fig. 4 shows the mean predialytic plasma sodium

levels for each patient and both phases. With a few
0.0 exceptions, the sodium levels vary only little from ses-
sion to session for a specific patient. There is no obvi-
-0.5 ous change from the standard care to the controlled
care phase. The average over all patients of the mean
[mmol/l]

-1.0 predialytic sodium increases only marginally from

∆K

137.1 ± 2.3 mmol/L for the fixed sodium treatments

-1.5
to 137.2 ± 2.2 mmol/L with activated EBC. This in-
controlled Na crease was not significant (P = 0.6).
-2.0
standard Na Kt/V as measured by OCM was higher in the
phase with active electrolyte balance (Table 2). This
-2.5
Patient 1-16 increase was significant (P = 0.02).

FIG. 3. Plasma potassium shift (Mean ± SD) during dialysis for DISCUSSION
each patient in standard care phase (squares) and controlled
phase (diamonds).
The results show a significant mean increase in
the plasma sodium when the electrolyte balance
145.0 control was active. Since the algorithm achieved
zero conductivity balance in a number of in vitro
tests, there is no reason that this was not the case
140.0 in in vivo conditions. The algorithm uses conduc-
tivity measurements at the dialyzer inlet and outlet
to determine the dialysate-side diffusive electrolyte
[mmol/l]

135.0
Na pre

balance. However, a zero conductivity difference be-

130.0 controlled Na
standard Na
125.0
Patient 1-16
FIG. 4. Predialytic plasma sodium concentration (Mean ± SD)
for dialysis with dialysate Na concentration fixed at 138 mmol/
L (squares) and electrolyte controlled treatments (diamonds).
mean value is 142.2 mmol/L with a standard deviation
of 2.2 mmol/L. This shows that the control algorithm
indeed actively adapts the dialysate sodium level to
the plasma concentration of the patients. The average
Nad, however, is substantially higher than the average
predialytic plasma Na, which was 137.2 mmol/L. This
results in a diffusive sodium transfer into the patient.
In the two HD treatments, the average sodium shifts
were 0.5 and 1.5 mmol/L, respectively. These values
are within the range of the HDF results.
In contrast to the plasma sodium concentration
shift, the plasma potassium concentration shifts
were not affected either by fixed sodium dialy-
sate or controlled electrolyte levels (Fig. 3). In the
standard care phase, the mean potassium shift was
−0.82 ± 0.41 mmol/L, in the controlled care phase
−0.85 ± 0.34 mmol/L. This difference is not signifi-
cant (P = 0.83).
tween dialyzer inlet and outlet does not forcibly im-
plicate a zero sodium difference since sodium may
be exchanged by other ions. Small differences in the
sodium concentrations between dialyzer inlet and
outlet may sum up during the treatment to a signifi-
cant deviation from the goal of zero sodium balance.
The second most abundant cation in the dialysate
is potassium. Since the potassium balance during
dialysis is always negative, a zero conductivity bal-
ance automatically requires a positive sodium load
in order to compensate for the potassium loss when
keeping the dialysate conductivity constant. On the
anion side, the plasma bicarbonate concentration is
usually much lower than that in the fresh dialysate.
This causes an exchange of plasma anions, especially
chloride, against bicarbonate.
In order to analyze the influence of ions other than
sodium on the conductivity balance in more detail,
a retrospective analysis using the measured plasma
electrolytes was done. Since the balancing algorithm
is based on conductivity, it cannot distinguish be-
tween different ions. Instead of plasma sodium alone
other ions also contribute to the total mass balance.
Taking only potassium into account for simplicity, the
plasma potassium balance ΔKp has to be added to
the sodium balance ΔNap:
𝜎k
ΔNap,corr = ΔNap + ΔKp .
𝜎Na

Artificial Organs, Vol. 0, No. 0, 2018

 AUTOMATED SODIUM CONTROL IN HEMODIALYSIS
Since the specific conductivity of potassium is
higher than that of sodium, ΔKp has to be cor-
rected by the conductivity ratio σK/σNa. Because the
conductivities of single ions cannot be measured
independently of their counter ions, the conductivi-
ties of KCl and NaCl solutions have to be used in-
stead. σK/σNa depends only slightly on the actual ion
strength and can be deduced from in vitro measure-
ments: Tura et al. (32) found conductivities of 12.93
and 15.92 mS/cm, for a 140 mmol/L sodium chloride
or potassium chloride solution respectively. This
leads to σK/σNa = 1.231. An independent source (33)
yields a very similar value of σK/σNa = 1.215, using lin-
ear extrapolation of the conductivity data for NaCl
and KCl. For the calculation of ΔNap,corr a value of
σK/σNa = 1.22 was used. The result is shown in Fig. 5.
The mean plasma Na+ shift is reduced to 0.3 mmol/L
with a standard deviation of 0.8 mmol/L. This cor-
rected shift is still significantly different (P = 0.006)
from the plasma Na+ shift of −1.0 mmol/L in the
standard care phase. The standard deviation is also
significantly smaller than in the standard phase (P =
0.02). The remaining shift may be due to the concen-
tration change of other ions contributing to the total
conductivity which were not measured by the blood
gas analyzer. Another cause of the remaining sodium
imbalance may be the intake of food or beverages by
the patients during the dialysis session.
In spite of the positive sodium load, the mean
predialytic plasma sodium did not increase sig-
nificantly when the electrolyte control was active.
This is in agreement with the setpoint hypothesis,
which claims that all patients keep their predialytic
plasma sodium levels nearly constant despite in-
tradialytic sodium shifts or salt and water intake in
6.0
4.0 controlled Na
2.0
[mmol/l]
∆Nacorr
0.0
-2.0
-4.0
-6.0
Patient 1-16
FIG. 5. Plasma sodium shift (Mean ± SD) with active electrolyte
control for each patient, corrected for potassium removal.
7
the interdialytic interval (17). Normally, according
to the setpoint hypothesis a sodium load would be
compensated by an increased fluid intake after the
dialysis session. However, such an effect could not
be observed. The predialytic TBW as measured by
the BCM did not significantly change between the
phases. Presumably the TBW increase was too small
to be noticed. Since most patients still had residual
diuresis, increased urinary excretion may have com-
pensated an increasing water intake. The observed
difference in the sodium shifts between the standard
care (−1.0 ± 1.4 mmol/L) and controlled care phase
(1.3 ± 0.9 mmol/L) was 2.3 mmol/L (Table 2). Since
the predialytic plasma sodium level in the standard
care phase was 137.1 mmol/L, an increase in the
TBW by only 1.4% would be sufficient to keep the
plasma sodium constant in the controlled care phase.
This value is far below the statistical variance of the
TBW measurements. Additionally, several patients
had a non-negligible residual diuresis. This may have
helped to eliminate an eventual overhydration.
Treatments in the controlled care phase achieved
a significantly higher Kt/V compared to the standard
care phase. An obvious reason for this cannot be
found. Eventually this could be due to the combined
effect of slightly higher Qb, Qd, Vsub, VUF, and Tdial as
well as lower total body water volumes in the con-
trolled care phase.
CONCLUSION
The study analyzed the effects of a new conductiv-
ity based EBC algorithm in comparison to a standard
treatment with fixed dialysate sodium on plasma so-
dium concentrations. The control algorithm relied
on the conductivity values measured in the dialysate
before and after the dialyzer, using them as a sur-
rogate for the sodium concentration. The controller
was designed with the aim to achieve a zero diffusive
sodium balance by adjusting the concentrate-water
mixing ratio in the preparation of the fresh dialysate.
However, due to the lack of specificity for individual
ions, the zero diffusive sodium balance was not fully
achieved. A possible explanation for this deviation
is that the negative potassium balance was compen-
sated by a positive sodium balance. The bias in the
convective mass balance due to the Gibbs-Donnan
effect may also have contributed to the sodium shift.
From a clinical perspective, the electrolyte bal-
ancing control module has proved its reliability to
reduce significantly the variability of intradialytic
plasma sodium concentration shifts provided that
the influence of ions other than sodium on conduc-
tivity is compensated. This compensation could be
Artificial Organs, Vol. 0, No. 0, 2018
8 U. Kuhlmann et al.
achieved by a kinetic model estimating the concen-
trations of all ions other than sodium in the spent di-
alysate. However, this improved approach has to be
validated in clinical trials.
The module has the capability to achieve approx-
imate isonatremic dialysis and to prevent excessive
thirst and weight gain due to sodium overload. At
the same time, the EBC module has the potential to
avoid the side effects of low dialysate sodium and to
improve tolerability of dialysis in preventing dialytic
symptoms (e.g. cramps, hypotensive episodes). In
contrast to existing sodium balancing methods, the
new algorithm-based controller works in a fully au-
tomated and self-adapting way, without the need for
manual intervention. No blood samples have to be
taken because the predialytic plasma sodium concen-
tration is not required as an input. Since the algorithm
only relies on the conductivity difference between
fresh and spent dialysate, absolute conductivity er-
rors cancel out. Thus, the algorithm compensates for
miscalibrations of the conductivity cells, deviations
from the calibrated concentrate-water mixing ratio
and fluctuations of concentrate composition (34).
Although the zero-diffusive-balance approach may
not be adequate for patients with pronounced disor-
ders of electrolyte homeostasis, for example severe
hyponatremia, it should provide an adequate option
for the majority of the ESKD population. Taking the
zero-diffusive-balance controller as a starting point,
one could envisage extending it by the option to set
a well-defined deviation from the zero-balance goal.
This would allow it to treat patients with electrolyte
disorders in the future.
Conflicts of Interest: The authors A. Maierhofer and
B. Canaud are employees of Fresenius Medical Care,
M. Gross is a former employee of this company. U.
Kuhlmann has received honorary fees for scientific
consulting for FMC. The study was financed by
Fresenius Medical Care.
REFERENCES
1. Flythe JE, Brookhart MA. Fluid management: the chal-
lenge of defining standards of care. Clin J Am Soc Nephrol
2014;9:2033–5.
2. Sinha AD, Agarwal R. Setting the dry weight and its car-
diovascular implications. Semin Dial 2017;30:481–8.
3. Santos SFF, Peixoto AJ. Revisiting the dialysate sodium
prescription as a tool for better blood pressure and inter-
dialytic weight gain management in hemodialysis patients.
Clin J Am Soc Nephrol 2008;3:522–30.
4. Penne EL, Levin NW, Kotanko P. Improving volume sta-
tus by comprehensive dietary and dialytic sodium man-
agement in chronic hemodialysis patients. BloodPurif
2010;30:71–8.
5. Munoz Mendoza J, Sun S, Chertow GM, Moran J, Doss
S, Schiller B. Dialysate sodium and sodium gradient in
Artificial Organs, Vol. 0, No. 0, 2018
maintenance hemodialysis: a neglected sodium restriction
approach? Nephrol Dial Transplant 2011;26:1281–7.
6. Penne EL, Sergeyeva O. Sodium gradient: a tool to individ-
ualize dialysate sodium prescription in chronic hemodialy-
sis patients? BloodPurif 2011;31:86–91.
7. Flanigan M. Dialysate composition and hemodialysis hy-
pertension. Semin Dial 2004;17:279–83.
8. Hecking M, Moissl U, Genser B, et al. Greater fluid over-
load and lower interdialytic weight gain are independently
associated with mortality in a large international hemodi-
alysis population. Nephrol Dial Transplant 2018.
9. Santos SFF, Peixoto AJ. Sodium balance in maintenance
hemodialysis. Semin Dial 2010;23:549–55.
10. London GM. Ultrafiltration intensification for achieve-
ment of dry weight and hypertension control is not always
the therapeutic gold standard. J Nephrol 2011;24:395–7.
11. Ok E, Levin NW, Asci G, Chazot C, Toz H, Ozkahya M.
Interplay of volume, blood pressure, organ ischemia, resid-
ual renal function, and diet: Certainties and uncertainties
with dialytic management. Semin Dial 2017;30:420–9.
12. Odudu A, McIntyre C. Volume is not the only key to hy-
pertension control in dialysis patients. NephronClin Pract
2012;120:c173–7.
13. Chou JA, Kalantar-Zadeh K. Volume balance and intradi-
alytic ultrafiltration rate in the hemodialysis patient. Curr
Heart Fail Rep 2017;14:421–7.
14. McKinley MJ, Johnson AK. The physiological regulation
of thirst and fluid intake. News Physiol Sci 2004;19:1–6.
15. Bourque CW. Central mechanisms of osmosensation and
systemic osmoregulation. Nat Rev Neurosci 2008;9:519–31.
16. Moret KE, Beerenhout CH, Kooman JP. Variations in
predialytic plasma conductivity in dialysis patients: ef-
fect on ionic mass balance and blood pressure. ASAIO J
2011;57:53–61.
17. Keen ML, Gotch FA. The association of the sodium “set-
point” to interdialytic weight gain and blood pressure in
hemodialysis patients. Int J Artif Organs 2007;30:971–9.
18. Martinez-Vea A, García C, Gaya J, Rivera F, Oliver JA.
Abnormalities of thirst regulation in patients with chronic
renal failure on hemodialysis. Am J Nephrol 1992;12:73–9.
19. Petitclerc T, Jacobs C. Dialysis sodium concentration:
what is optimal and can it be individualized? Nephrol Dial
Transplant 1995;10:596–9.
20. Basile C, Lomonte C. It is time to individualize the dialy-
sate sodium prescription. Semin Dial 2016;29:24–7.
21. Locatelli F, Di Filippo S, Manzoni C. Relevance of the
conductivity kinetic model in the control of sodium pool.
Kidney Int Suppl 2000;76:S89–95.
22. Maduell F, Navarro V. Valoración de la ingesta de sal en
hemodiálisis. Nefrologia 2001;21:71–7.
23. Murisasco A, France G, Leblond G, et al. Sequential so-
dium therapy allows correction of sodium-volume balance
and reduces morbidity. Clin Nephrol 1985;24:201–8.
24. de Paula FM, Peixoto AJ, Pinto LV, Dorigo D, Patricio
PJM, Santos SFF. Clinical consequences of an individu-
alized dialysate sodium prescription in hemodialysis pa-
tients. Kidney Int 2004;66:1232–8.
25. Raimann JG, Thijssen S, Usvyat LA, Levin NW, Kotanko
P. Sodium alignment in clinical practice–implementation
and implications. Semin Dial 2011;24:587–92.
26. Petitclerc T, Goux N, Hamani A, Béné B, Jacobs C.
Biofeedback technique through the variations of the dialy-
sate sodium concentration. Nefrologia 1997;17:50–5.
27. Bosetto A, Bene B, Petitclerc T. Sodium manage-
ment in dialysis by conductivity. Adv Ren Replace Ther
1999;6:243–54.
28. Chevalier L, Tielemans C, Debelle F, et al. Isonatric
dialysis biofeedback in hemodiafiltration with online
regeneration of ultrafiltrate in hypertensive hemodialy-
sis patients: a randomized controlled study. BloodPurif
2016;41:87–93.
 AUTOMATED SODIUM CONTROL IN HEMODIALYSIS
29. Lambie SH, Taal MW, Fluck RJ, McIntyre CW. Online
conductivity monitoring: validation and usefulness in a
clinical trial of reduced dialysate conductivity. ASAIO J
2005;51:70–6.
30. Lysaght MJ. An experimental model for the ultrafiltra-
tion of sodium ion from blood or plasma. BloodPurif
1983;1:25–30.
31. Bogle W, Hsu YS. Sample size determination in comparing
two population variances with paired-data: application to
bilirubin tests. Biom J 2002;44:594.
32. Tura A, Sbrignadello S, Mambelli E, Ravazzani P,
Santoro A, Pacini G. Sodium concentration measurement
9
during hemodialysis through ion-exchange resin and con-
ductivity measure approach: in vitro experiments. PLoS
ONE 2013;8:e69227. https://doi.org/10.1371/journal.
pone.0069227.
33. Lide DR, editor. CRC Handbook of Chemistry and
Physics, 84th Edition. Boca Raton, FL: CRC Press,
2003–2004.
34. Gul A, Miskulin DC, Paine SS, et al. Comparison of pre-
scribed and measured dialysate sodium: a quality improve-
ment project. Am J Kidney Dis 2016;67:439–45.
Artificial Organs, Vol. 0, No. 0, 2018