Critical Care and Resuscitation Volume 14 Number 4 December 2012 274
Crit Care Resusc ISSN: 1441-2772 1 Decem-
ber 2012 14 4 274-282
for cardiovascular stability and may be important in CRRT
given the arrhythmia incidence in one large study
approached 45%.7 Magnesium supplementation for
arrhythmia prevention has been most studied in cardiac
surgical patients,8-10 with a meta-analysis suggesting an
overall reduction in the incidence of atrial fibrillation.11
Adult humans contain about 24g (1000mmol) of magne-
sium, of which 60% is in bone and available to support
plasma levels. Around 40% (400mmol) is intracellular, of
which half is contained in skeletal muscle,3 commensurate
with its pivotal role coordinating inorganic phosphate transfer
in ATP synthesis.12 The extracellular compartment contains
only 1%–1.9% of total body magnesium (about 12mmol).
About 30% of plasma magnesium is protein bound and not
readily removed by dialysis, 5%–10% is complexed with other
anions (citrate, bicarbonate, phosphate and sulfates), and
60% is present as free magnesium ions.3,13
Plasma magnesium levels are frequently elevated in chronic
kidney disease when the creatinine clearance rate is below
30mL/min. Most studies of total magnesium balance in renal
failure have been in this context as it may, in combination
with calcium, contribute to renal osteodystrophy.13 As a
consequence, some commonly used dialysis fluids have
relatively low [Mg] relative to serum concentrations.13
The effect of CRRT on magnesium in acute kidney injury
has been less well studied. The capacity for magnesium loss
during continuous venovenous haemodiafiltration
(CVVHDF) is significant, and may potentially be exacerbated
by additional chelation when citrate is used for anticoagula-
tion. We recently described calcium ion flux in CVVHDF
using citrate or heparin for anticoagulation14 and here
present an analysis of concurrently studied magnesium.
Hypomagnesaemia is common in critical care patients1
generally and particularly in patients receiving continuous
renal replacement therapy (CRRT).2,3 Low magnesium con-
centration ([Mg]) has been associated with adverse out-
comes including increased mortality.4-6 Maintaining the
serum [Mg] in the high normal range is considered desirable
Protocols for CVVHDF with citrate and heparin anticoagula-
tion, calcium replacement, blood and circuit sampling, data
collection and calculation of electrolyte loss in the effluent
have been described previously.14 Sampling sites were
patient arterial line, prefilter (after addition of predilution),
postfilter (before postdilution), and the effluent line. [Mg] in
blood and effluent was measured by spectrophotometric
dye binding (Abbott Architect c8000; dye for magnesium
was xylidyl blue; analysed at 660nm). The laboratory
reference range for total serum [Mg] is 0.7–1.16mmol/L.
Objective: To describe magnesium flux and serum
concentrations in ICU patients receiving continuous
venovenous haemodiafiltration (CVVHDF).
Design: Samples were collected from 22 CVVHDF circuits
using citrate anticoagulation solutions (Prismocitrate 10/2
and Prism0cal) and from 26 circuits using Hemosol B0 and
heparin anticoagulation. CVVHDF prescription, magnesium
supplementation and anticoagulation choice was by the
treating intensivist. We analysed 334 sample sets consisting
of arterial, prefilter and postfilter blood and effluent.
Magnesium loss was calculated from an equation for
conservation of mass, and arterial magnesium
concentration was described by an equation for
Results: Using flow rates typical of adults receiving
CVVHDF, we determined a median half-life for arterial
magnesium concentration to decay to a new steady state
of 4.73hours (interquartile range [IQR], 3.73–7.32hours).
Median arterial magnesium concentration was 0.88mmol/L
(IQR, 0.83–0.97mmol/L) in the heparin group and
0.79mmol/L (IQR, 0.69–0.91mmol/L) in the citrate group.
Arterial magnesium concentrations fell below the reference
range regularly in the citrate group and, when low, there
was magnesium flux from dialysate to patient. Magnesium
loss was greater in patients receiving citrate.
Conclusions: Exponential decline in magnesium
concentrations was sufficiently rapid that subtherapeutic
serum magnesium concentrations may occur well before
detection when once-daily sampling was used.
Measurements should be interpreted with regard to
timing of magnesium infusions. We suggest that
continuous renal replacement therapy fluids with
higher magnesium concentrations be introduced in
the critical care setting.
Crit Care Resusc 2012; 14: 274–282
Magnesium flux during continuous venovenous
haemodiafiltration with heparin and citrate anticoagulation
Matthew Brain, Mike Anderson, Scott Parkes
and Peter Fowler
Critical Care and Resuscitation Volume 14 Number 4 December 2012275
Citrate anticoagulation used Prismocitrate 10/2 (Gambro)
as the predilution fluid and Prism0cal (Gambro) as the
dialysate. Those not anticoagulated with citrate received
Hemosol B0 (Gambro) with or without heparin anticoagula-
tion (for simplicity in reporting results this group is
described as heparin). Compositions of fluids are shown in
This observational study was approved by the Human
Research Ethics Committee (Tasmania) Network. Treating
physicians had access to all results; however, a different
laboratory code meant results did not appear with routine
In our institution, total [Mg] is routinely measured once
daily and the target concentration is 0.9–1.0mmol/L.
Magnesium is typically administered as an intravenous
infusion in doses of 20 mmol (occasionally up to 40 mmol)
over a period of 2 hours to avoid hypotension associated
with faster infusion rates.
If administered, parenteral nutrition (PN) provided 5–
15mmol of magnesium per day according to an estimate of
patient requirements. No attempt was made to quantify
enteric magnesium intake or loss.
Exponential decay analysis
After infusion, the arterial [Mg] rises to a peak followed by
a curve demonstrating exponential decay to a baseline. A
similar decay is observed after starting a new CRRT circuit
with a high–normal arterial [Mg].
This exponential decay is described by the general equa-
tion y=B+C.e(-k.t), where y is the concentration at time t
(hours), B is the baseline concentration that the curve trends
to (an asymptote), C is the y-intercept (relative to the
baseline), e is Euler’s number, and k is the rate constant of
the decay. Half-life is the natural logarithm of 2 divided by k
Figure 1 is a circuit schematic demonstrating volumetric
flow, with sample measurements and calculations from one
We segmented the data into concentration–time
curves where the start of a curve is defined as the start of
a new haemofilter or the start of a magnesium infusion
and the end of a curve is defined as the end of a
haemofilter (filter failure or scheduled cessation) or the
start of the next magnesium infusion. We excluded
curves in which less than four data points after the peak
concentration were obtained. Observed data were sum-
marised as time above specific [Mg] cut offs. We then
analysed the individual decay curves from the peak
recorded concentration after a magnesium infusion to
the end of the curve.
Area under the concentration–time curve (AUC) and area
under the first moment curve (AUMC) were calculated from
time zero (not peak concentration as for decay curves) to
last observation, using PK Functions for Microsoft Excel
(Usansky J et al, Department of Pharmacokinetics and Drug
Metabolism, Allergan, Irvine, Cal, US). -AUC and -AUMC
were determined as area above the calculated baseline
(AUCobservations AUCbaseline) and (AUMCobservations AUMCbaseline),
respectively.15 Mean residence time (MRT) and mean k was
determined from the ratio of -AUC/-AUMC.16,17 The area
under the solute flux (point concentration in mmol/Lflow
in L/h) versus time curve (flux–time integral [FTI]) yields the
total amount (mmol) of magnesium that passed the sam-
pling point over the analysed time period.
Table 1. Electrolyte composition of fluids used in
this study and magnesium concentrations of other
fluids commonly used in continuous renal
replacement therapy (CCRT)*
CCRT fluid Electrolyte
Used in this study
Prismocitrate 10/2Trisodium citrate10
In widespread use (not used in this study)[Mg] (mmol/L)
Fresenius multiPlus (1.0mmol HPO4
Fresenius Ci-Ca K2/K4
Gambro Phoxilium (1.2mmol HPO4
Hospal Prism0cal B22
Hospal Hemosol L0/LG2/LG4
Baxter Monosol haemofiltration solution
*Data from manufacturers’ product information.
Critical Care and Resuscitation Volume 14 Number 4 December 2012276
Calculation of magnesium loss
Total magnesium flux from the patient to the effluent was
calculated in two ways. Method 1 takes the median (across
all measurements) result of an equation for
conservation of mass from each sample set:
where Mgloss is in mmol/h, Q=volumetric
flow rate (L/h), eff = effluent, dial =
dialysate, PBP=pre blood pump fluid, and
Method 2 uses the FTI at each sampling
site to determine the total amount of
magnesium to transit that point in the
circuit. Conservation of mass then dictates
loss over the total time of the circuit
Mgloss=FTIpre – FTIpost+FTIPBP+FTIrep
where FTIpre and FTIpost are integrals of the
pre- and postfilter flux–time curves, and
FTIPBP and FTIrep are integrals of magnesium
added via predilution (none in citrate) and
postdilution over curve life. The result is
the amount of magnesium lost over the
curve’s duration and can be indexed to 24
hours to roughly estimate mean daily loss
(assuming a similar sampling and replace-
The curve of best fit for exponential decay
from the observed peak concentration was
found using the Solver generalised reduced
gradient nonlinear module (Frontline Sys-
tems) contained in Excel 2010 (Microsoft
Corporation) to solve for the smallest root
mean square error (RMSE) by varying the
constants of the exponential decay equation
B, C, and k. The normalised RMSE (NRMSE=
RMSE/[range of observed concentrations]),
coefficient of variation (CVRMSE=RMSE/mean
concentration), coefficient of determination
(R2), and significance (F-distribution) are
reported. Four curves that did not follow a
magnesium infusion demonstrated decay
better described by a linear equation. Efflu-
ent dose and half-life were compared using
linear regression. Statistical comparisons
were made in Intercooled Stata, version 9
(StataCorp) and R, version 2.15.1 (R Founda-
tion for Statistical Computing) by two sam-
ple Wilcoxon rank-sum tests with continuity
correction for ties, and Wilcoxon signed-rank test with conti-
nuity correction where appropriate. Significance was set at
Figure 1. Schematic of circuit with sample measurements from one
[Mg]=magnesium concentration. The diameter of the large tube reflects the blood compartment
volumetric flow (L/h) with variations in diameter reflecting volume shifts from predilution,
ultrafiltration and postdilution. The increasing tube diameter from dialysate to effluent reflects
the addition of ultrafiltrate to the dialysate path. Typically, blood enters the circuit at 200mL/min
(12L/h) and is further expanded by the addition of about 1.5–2L of predilution before entering
the haemofilter. Manipulation of transmembrane pressure within the filter forces fluid from the
blood path to the dialysate–effluent path and is controlled to remove all pre- and postdilution
and any additional fluid removal. Example measurements and derived fluxes after a 20mmol
magnesium infusion in a patient receiving citrate anticoagulation are included and are presented
as raw measurements in the top graph, magnesium flux in the blood path (bottom left), and
magnesium flux in the effluent path (bottom right). Note magnesium concentration falls from
arterial to prefilter but is similar between pre- and postfilter sites, whereas magnesium flux is
unchanged (in citrate) between arterial and prefilter sites and falls before the postfilter site. Note
also that effluent [Mg] falls below the dialysate concentration, implying net absorption when
arterial concentrations are low.
Critical Care and Resuscitation Volume 14 Number 4 December 2012277
Patient characteristics and data
Patient parameters, outcomes and filter life are described in
Table 2. In total, 334 sample sets each comprising an
arterial, prefilter, and postfilter blood and an effluent
sample were collected from 13 consecutive ICU patients
treated with CVVHDF as part of their ICU stay. Hemosol B0
was used in 26 circuits with or without heparin anticoagula-
tion. Citrate anticoagulation was used in 22 circuits using
Prismocitrate 10/2 and Prism0cal fluids.
Raw magnesium concentrations
Figure 2 demonstrates range and median concentrations
from each sample site across all patients studied. The median
arterial [Mg] in the heparin group was 0.88mmol/L (inter-
quartile range [IQR], 0.83–0.97mmol/L; n=180) and in the
citrate group, 0.79mmol/L (IQR, 0.69–0.91mmol/L; n=151;
Wilcoxon W=8351; P v heparin<0.0001; two-tailed).
Arterial magnesium concentration–time decay curve
Forty curves with four or more arterial magnesium samples
in the decay phase (time zero defined as peak [Mg] after
starting a new filter or administration of a magnesium
infusion) were identified for decay curve analysis. Of these,
19 were from patients receiving citrate anticoagulation and
21 from patients using Hemosol B0 with or without heparin
anticoagulation. Concentration–time curve parameters are
described in Table 3 and an example from one patient is
provided in Figure 1. Four curves did not demonstrate
exponential decay over the sampling period but demon-
strated a smaller RMSE when described with a linear
equation. Inspection of the range and slope (Table 3) of
these curves suggest that concentrations at time zero were
too near steady state in three cases, limiting decay defini-
tion. In the fourth case, a rising slope is observed in
association with PN administration however this may be
explained by noise in the data.
Of 26 magnesium infusions administered to the patients,
four were excluded due to insufficient sample points after
the infusion. PN was administered to 12 patients in the
citrate group and to five in the heparin group. The total
quantity (mmol) of magnesium provided in PN over the
course of the curve is shown in Table 3. Thirteen curves
sampled a magnesium infusion in the citrate group and
nine in the heparin group. Figure 3 displays the time the
arterial [Mg] remained between stratified ranges after a
magnesium infusion. Thirty-eight observations were
<0.7mmol/L in the citrate group but only two <0.7mmol/L
were recorded in the heparin group. In the citrate group,
Table 2. Patient characteristics, continuous renal replacement therapy (CRRT) duration and number of filters
GI loss, acidosis
Cardiac failure, sepsis
Hepatic failure, CCF
CKD, fluid removal
AAA=abdominal aortic aneurysm. AKI=acute kidney injury. AMI=acute myocardial infarction. APACHE=Acute Physiology, Age and Chronic Health
Evaluation. APO=acute pulmonary oedema. CCF=congestive cardiac failure. CKD=chronic kidney disease. GI=gastrointestinal. LDH=low-dose heparin.
MOD=multiorgan dysfunction. WBH=weight-based heparin. *Patient number refers to the sequential order of enrolment in the study. †Duration refers to
total time in hours of venovenous haemodiafiltration; filters refers to the total number of haemofilters consumed over therapy course. ‡Patients 8 and 9
received low-dose heparin initially and were later changed to citrate; their data have been divided and analysed in each group. §Two anticoagulation
methods were performed sequentially and filter hours are divided respectively. ¶Patients 1 and 10 had periods without any circuit anticoagulation due
to coagulopathy; during this time, their continuous venovenous haemodiafiltration treatment otherwise continued as per our LDH protocols with
Critical Care and Resuscitation Volume 14 Number 4 December 2012278
the median time that the arterial [Mg] spent below
0.7mmol/L was 4hours (IQR, 0–8hours).
Elimination half-life and mean residence time
The median half-life for decay to baseline for all derived
magnesium curves was 4.73hours (IQR, 3.73–7.32hours).
When confined only to the curves where a magnesium
infusion was administered, the result
was not significantly different
4.68hours; IQR, 4.32–7.09hours)
and heparin (median, 5.14hours;
IQR, 3.42–7.37hours; W=67; P=
0.9274). There was no relationship
between the effluent flow rate and
half-life (R2=0.048; P=0.24).
Median residence time (MRT) was
calculated from -AUC and -AUMC
for arterial [Mg] (Table 4). There was
no significant difference between
the two groups, with a median
heparin MRT of 5.90hours (IQR,
4.11–8.49hours) and a median cit-
rate MRT of 6.56hours (IQR, 4.93–
7.91hours; Pvheparin=0.84; W=
206). From the MRT, an estimated
elimination half-life can be derived;
for citrate, the half-life is 4.09hours
and for heparin, 4.54hours.
Baseline of the magnesium
concentration–time decay curve
Table 3 reports the minimum arterial
[Mg] for each curve and the first
constant of the corresponding
equation is the calculated baseline.
The calculated baselines of the
derived decay curves correlated well
with the minimum arterial [Mg]
from each curve (B=1.0262C
0.03, where C is the minimum arte-
rial [Mg]; R2=0.8335). The median
of the minimum arterial [Mg] was
significantly lower in the citrate
group (0.68mmol/L; IQR, 0.63–
0.71mmol/L) compared with the
heparin group (0.82mmol/L; IQR,
0.78–0.88 mmol/L; W = 216.5;
proportional to the arterial [Mg].
For heparin, y= 1.644x 0.724;
R2=0.43, and for citrate, y =2.265x 0.716; R2=0.71,
where y is magnesium lost to the patient in mmol/h and x
is the arterial [Mg]. As a marker of model validation, using
flux–time integrals to calculate magnesium leaving the
filter blood compartment (FTIpre – FTIpost; Table 4) correlated
well with magnesium appearing in effluent (FTIeff – FTIdial;
R2= 0.946; P <0.001).
Figure 3. Time after a magnesium infusion at stratified arterial magnesium
Figure 2. Range and median magnesium concentrations by patient and
sampling site for 13 patients
Critical Care and Resuscitation Volume 14 Number 4 December 2012279
Magnesium loss was greater in patients receiving citrate
than heparin. When calculated by conservation of mass for
each sample set, the median loss from heparin circuits was
0.72mmol/h (IQR, 0.53–0.97mmol/h) or 17.28mmol/day.
Median loss from citrate circuits was 1.09mmol/h (IQR,
0.80–1.41mmol/h; Wilcoxon W=20151.5; P<0.001), or
26.2mmol/day. When calculated using flux–time integrals
(indexed to 24 hours), the values for heparin were
13.07mmol/day (IQR, 10.17–18.79mmol/day) and for cit-
rate, 25.17mmol/day (IQR, 18.65–30.70mmol/day; P v
Table 4 and Figure 2 reveal the postfilter [Mg] frequently
to be greater than the prefilter concentration. The effect is
small but significant (citrate: median postfilter [Mg],
0.03mmol/L greater than prefilter [Mg] [P=0.004; Wil-
coxon W =5722.5]; heparin: median postfilter [Mg],
0.01mmol/L greater than prefilter [Mg] [P=0.001; V=
17847]) and probably results from bloodstream concentra-
tion after ultrafiltration to remove the predilution volume.
In contrast, the flux–time integral is universally greater
prefilter compared with postfilter (heparin: median prefilter
minus postfilter difference, 27.4 mmol [IQR, 19.6–
36.4mmol]; citrate: median prefilter minus postfilter differ-
ence, 23.2mmol [IQR, 16.8–28.5mmol; P v heparin=
0.174; W=148.5]), confirming a net loss of magnesium
from the blood path. Further, in citrate where no magne-
Table 3. Arterial magnesium concentrations and parameters of exponential decay curves of arterial magnesium
concentration in venovenous haemodiafiltration
Data are organised by anticoagulation type then by magnesium infusion amount. Curve Name=patient number. filter number. curve letter.
Mg infusion= total mmol of magnesium infused. Mg via TPN=total amount given over the time period of that curve (hourly rate may vary).
CV=coefficient of variation. NRMSE=normalised root mean square error. QB=volumetric blood flow rate. QD=volumetric flow of dialysate.
TPN=total parenteral nutrition *Curves marked with an asterix contain a single extreme outlier. Ignoring these outliers significantly improves the
numerical parameters of curve fit as follows: Curve ID/NRMSE/CV(RMSE)/R^2; ID 11.1.A/4.09%/0.59%/0.852; ID 4.4.F/7.93%/0.77%/0.620;
ID 4.1.A/2.98%/0.36%/0.900. †RMSE linear equation: 0.013, RMSE exponential equation: 0.088. F-ratio negative. ††RMSE linear equation: 0.002;
RMSE exponential equation: 0.009. F-ratio negative. †††RMSE linear equation: 0.005; RMSE exponential equation: 0.029. F-ratio negative.
‡RMSE linear equation: 0.013; RMSE exponential equation: 0.055. F-ratio negative.
Critical Care and Resuscitation Volume 14 Number 4 December 2012280
sium is added in predilution, the flux–time integrals are not
significantly different between the arterial sampling site
(surrogate for access line) (median, 168.7mmol) and the
prefilter flux (median, 166.0mmol; V=135; P=0.1134).
To summarise, with magnesium-free predilution, there is
a concentration drop between the access line and the filter
due to dilution; after the filter, the concentration rises
slightly. This rise in concentration occurs despite magnesium
flux across the filter and is due to haemoconcentration from
ultrafiltration. This is shown in the example in Figure 1.
In nine curves in the citrate group (magnesium-free
predilution), the effluent [Mg] fell below the dialysate
concentration of 0.5mmol/L for part of the time (Table 4
and Figure 1). This suggests that sufficiently low [Mg]
occurred within the filter to drive net flux of magnesium
from the dialysate to the blood path.
Attention to magnesium homoeostasis is recognised as an
important component of critical care.
Using typical fluid flow rates for adult patients receiving
CVVHDF, our findings suggest that responding to once-
daily measurement of serum magnesium results in substan-
tial time periods near or below the lower reference range
value. This effect was greatest in those receiving citrate and
is likely due both to loss of magnesium chelated to citrate in
the effluent18 and the lack of magnesium in the predilution
used with this method.
We determined a half-life for magnesium to decay to a
new baseline after starting CVVHDF or receiving a magne-
sium bolus. It should be noted that this baseline only implies
a steady concentration (slope negligible) and does not
necessarily imply a state where net intake equals loss. As
magnesium exists in multiple compartments, mobilisation
from bone3 as well as enteric and intravenous magnesium
administration will contribute to maintenance of a steady
plasma concentration. Magnesium supplementation prac-
tices vary between units and non-CRRT losses will vary
between patients making approximating net balance in a
Table 4. Non-compartmental analysis of magnesium (Mg) concentration and flux
AUC=change in the area under the concentration time curve. AUMC=change in the area under the first moment curve. C=citrate. L
DH=low-dose heparin. WBH=weight-based heparin. FTI=flux-time integral. [Mg]=magnesium concentration. Mg from Blood=the amount in mmol
that left the haemofilter blood path over curve duration and is calculated from Mg-FTI Prefilter Mg-FTI Postfilter. Postdilution flux used in Patient Loss is
not shown but can be calculated as 0.2L/h 0.5mmol/L curve duration. Mg To Effluent=the amount of Mg entering the effluent above the dialysate
concentration. Effluent: Prefilter Flux Ratio=(Mg To Effluent)/(Mg-FTI Prefilter).
Critical Care and Resuscitation Volume 14 Number 4 December 2012281
Our results highlight that the driving concentration gradi-
ent for diffusive clearance will be the mean of the pre- and
postfilter diffusible magnesium after the addition of predilu-
tion to the bloodstream. Assuming a bound fraction
approximating 30% and an approximate 15% dilution
effect of the predilution, it can be appreciated that the
bloodstream diffusible [Mg] will be close to (and sometimes
below) the 0.5mmol/L dialysate concentration. We demon-
strated that ultrafiltration of water must offset the mass
transfer of magnesium to the effluent to maintain or slightly
raise the blood compartment concentration over the filter.
The results are subject to several limitations due to the
observational nature of the study and the small number of
patients. A limitation of our sampling frequency is that we
may have underestimated peak concentrations; however,
administration over 2 hours would have decreased the
magnitude of this error. Where magnesium is given at faster
rates, higher peak levels than we sampled may be achieved
and may have therapeutic19 or adverse effects depending
on the clinical scenario. Decay curve analysis is subject to a
risk of overfitting; however, visual inspection of the data
and derived curves, and the parameters of curve fit suggest
this was not the case. In four cases where a magnesium
bolus was not given, we found that a linear model best
described the data. We do not suggest alternative kinetics
in these cases but rather sampling limitations obscuring
detection of exponential decay.
A multicompartment model with two to three extra
plasma pools has been described in healthy subjects20 using
radiolabelled magnesium; however, to our knowledge mag-
nesium compartments in CRRT have not been described. It
is possible that the model we have used is not the best
approximation and that more frequent sampling under
controlled conditions would reveal kinetics consistent with a
multicompartment model. As our data were generally well
described by monoexponential decay and gave similar
results with a non-compartmental analysis, an alternative is
that the effect of CRRT on the system masks detection of
relatively smaller shifts between compartments.
Imprecision in concentration measurement and recorded
sample timing may have compounded errors in construct-
ing the true decay curve. These sources of error may
account for us not detecting a half-life difference between
lower and higher effluent flow rates. We have studied total
[Mg]; however, newer generation blood gas analysers can
report ionised magnesium. Ionised hypomagnesaemia is
less frequent than total hypomagnesaemia and may be
more clinically relevant.21,22 Access to the ionised magne-
sium result on blood gas analysis may have improved
detection and response to low concentrations in the citrate
group, where blood gas analysis was frequently performed
as part of protocol-driven ionised calcium monitoring.
Future studies incorporating ionised magnesium may better
define the effects of citrate and protein binding and with
more frequent sampling could describe other factors influ-
encing magnesium kinetics in CRRT such as body weight,
gastrointestinal losses, filter age and effluent dose.
Although many of our findings can be predicted from an
understanding of CRRT, these data quantify the time for
magnesium to fall with two widely used CRRT fluids and
reinforce the need to monitor [Mg] frequently or increase
the frequency of supplementation if a high–normal arterial
concentration is to be maintained consistently. The demon-
stration of significant exponential decay makes determining
the timing of magnesium sampling in relation to dosing
important when interpreting results. We would suggest at
least twice-daily magnesium supplementation for patients
receiving CRRT, with measurement of trough concentra-
tions to guide dosage or further increases in frequency. It
may potentially be advantageous to administer magnesium
supplementation over longer time periods to attenuate
peaks and promote steady concentrations.
Our findings should hopefully encourage a shift towards
the supply of CRRT fluids with a higher [Mg], just as a trend
towards avoiding hypophosphataemia has led to the pro-
duction of CRRT fluids with higher phosphate ion concen-
tration. We suggest that dialysate [Mg] of 0.8mmol/L may
be more suited to the critically ill population. In those
receiving magnesium free predilution fluids with concomi-
tant citrate anticoagulation, a dialysate concentration of
1.0mmol/L may be required and should be the subject of
We thank Dr David Pilcher for assisting with statistical analysis, Isaac
Brain for vector art, and the staff of the Launceston General Hospital
intensive care unit and pathology service for (unfunded) sample
collection and analysis.
No relevant disclosures.
Matthew J Brain, Post Fellowship Registrar in Intensive Care1,2
Mike Anderson, Senior Intensive Care Specialist3
Scott Parkes, Intensive Care and Respiratory Physician1
Peter Fowler, Clinical Pharmacist1
1 Launceston General Hospital, Launceston, TAS, Australia.
2 The Alfred Hospital, Melbourne, VIC, Australia.
3 Royal Adelaide Hospital, Adelaide, SA, Australia.
1 Reinhart RA, Desbiens NA. Hypomagnesemia in patients entering
the ICU. Crit Care Med 1985; 13: 506-7.
ORIGINAL ARTICLES Download full-text
Critical Care and Resuscitation Volume 14 Number 4 December 2012282
2 Morimatsu H, Uchino S, Bellomo R, Ronco C. Continuous veno-
venous hemodiafiltration or hemofiltration: impact on calcium,
phosphate and magnesium concentrations. Int J Artif Organs 2002;
3 Vaporean ML, Van Stone JC. Dialysate magnesium. Semin Dial
1993; 6: 46-51.
4 Ru GJ, Thi M, Hardie D, Carlson RW. Association of hypomag-
nesemia and mortality in acutely ill medical patients. Crit Care Med
1993; 21: 203-9.
5 Soliman HM, Mercan D, Lobo SSM, et al. Development of ionized
hypomagnesemia is associated with higher mortality rates. Crit
Care Med 2003; 31: 1082-7.
6 Safavi M, Honarmand A. Admission hypomagnesemia — impact on
mortality or morbidity in critically ill patients. Middle East J Anesthe-
siol 2007; 19: 645-60.
7 RENAL Replacement Therapy Study Investigators. Intensity of con-
tinuous renal-replacement therapy in critically Ill patients. N Engl J
Med 2009; 361: 1627-38.
8 Najafi M, Hamidian R, Haghighat B, et al. Magnesium infusion and
postoperative atrial fibrillation: a randomized clinical trial. Acta
Anaesthesiol Taiwan 2007; 45: 89-94.
9 Toraman F, Karabulut EH, Alhan HC, et al. Magnesium infusion
dramatically decreases the incidence of atrial fibrillation after
coronary artery bypass grafting. Ann Thorac Surg 2001; 72: 1256-
61; discussion 1261-2.
10 Tiryakioglu O, Demirtas S, Ari H, et al. Magnesium sulphate and
amiodarone prophylaxis for prevention of postoperative arrhythmia
in coronary by-pass operations. J Cardiothorac Surg 2009; 4: 8.
11 Burgess DC, Kilborn MJ, Keech AC. Interventions for prevention of
post-operative atrial fibrillation and its complications after cardiac
surgery: a meta-analysis. Eur Heart J 2006; 27: 2846.
12 Blum DJ, Ko YH, Pedersen PL. Mitochondrial ATP synthase catalytic
mechanism: a novel visual comparative structural approach empha-
sizes pivotal roles for Mg2+ and P-loop residues in making ATP.
Biochemistry 2012; 51: 1532-46.
13 Llach F. Renal osteodystrophy. In: Jacobs C, Kjellstrand C, Koch K,
Winchester J, editors. Replacement of renal function by dialysis.
Dordrecht: Kluwer Academic Publishers, 1996.
14 Brain M, Parkes S, Fowler P, et al. Calcium flux in continuous
venovenous haemodiafiltration with heparin and citrate anticoagu-
lation. Crit Care Resusc 2011; 13: 72-81.
15 White J, Massey L, Gales SK, et al. Blood and urinary magnesium kinetics
after oral magnesium supplements. Clin Ther 1992; 14: 678-87.
16 Bourne D. Basic pharmacokinetics. Chapter 20. http://www.
boomer.org/c/p4/c20/c2001.html (accessed Jun 2012).
17 Gunaratna C. Drug metabolism and pharmacokinetics in drug
discovery: a primer for bioanalytical chemists, part I. Curr Sep 2000;
18 Kozik-Jaromin J, Nier V, et al. Citrate pharmacokinetics and calcium
levels during high-flux dialysis with regional citrate anticoagulation.
Nephrol Dial Transplant 2009; 24: 2244-51.
19 Davey MJ, Teubner D. A randomized controlled trial of magnesium
sulfate, in addition to usual care, for rate control in atrial fibrillation.
Ann Emerg Med 2005; 45: 347-53.
20 Sabatier M, Pont F, Arnaud MJ, Turnlund JR. A compartmental
model of magnesium metabolism in healthy men based on two
stable isotope tracers. Am J Physiol Regul Integr Comp Physiol
2003; 285: R656-63.
21 Huijgen HJ, Soesan M, Sanders R, et al. Magnesium levels in
critically ill patients. Am J Clin Pathol 2000; 114: 688-95.
22 Mazzaferro S, Barberi S, Scarda A, et al. Ionised and total serum
magnesium in renal transplant patients. J Nephrol 2002; 15: 275-80.❏