168
I
NTRODUCTION
In recent years, bicarbonate (HCO
3
) has been increasingly used as the main buffer with substitution fluids for bloodpurification. Consequently, acetatebuffered fluids havebeen progressively discarded. However, acetate has not been completely abandoned. In substitution fluids containing both electrolytes and HCO
3
(usually 3234mmol/L), 34 mmol/L of acetate are still added to prevent calcium salt precipitation. Therefore, due to the negligiblelevels of acetate in blood, there is a dialysatetoplasma gradient of 34 mmol/L in patients on bicarbonate hemodial ysis (HD). Concerning HCO
3
, if a theoretical plasma concentrationof 20 mmol/L is considered, then a dialysatetoplasmagradient as high as 1214 mmol/L is expected. Therefore,in bicarbonate HD, about 25% of the net mass balance of buffers can in fact be accounted for by acetate. Acetate infusion can induce some untoward clinical effects, including cardiovascular instability and increased nitric oxide synthesis. A new technique for renal replacement therapy (RRT), acetatefree biofiltration (AFB
®
),aimed at obviating the aforementioned problem, was proposed about 20 yrs ago (1, 2). With AFB
®
, there is no simultaneous mixing of calcium and HCO
3
. Only HCO
3
isinfused into blood (as NaHCO
3
), whereas calcium is supplied only with the electrolytecontaining, HCO
3
freedialysate. Consequently, there is no need for acetate.
Due to the complex interactions between the convectiveand diffusive fluxes of bicarbonate and sodium during AFB
®
, dedicated computer software (Skipper
®
) was necessary to predict their kinetics and enable the physician toselect the most suitable working parameters of each session (3).
Some clinical observations reported that AFB
®
could improve both acidbase control and hemodynamics in patients on regular dialysis treatment for endstage renal disease(4). Consequently, AFB
®
could also be advisable for critically ill patients with acute renal failure (ARF)in intensive care units (ICUs), in particular those treated with continuous renal replacement therapy (CRRT) because of their poor cardiovascular stability. However, AFB
®
JN
EPHROL2006; 19: 168175
O O
RIGINAL RIGINAL
I I
NVESTIGA NVESTIGA TION TION
www.sinitaly.org/jnonline/vol19n2/
Computer program to prescribe acetatefreebiofiltration
®
as a continuous renal replacement therapy: Theoretical descriptionand
in vivo
validation
Corrado Vitale, Cristiana Bagnis, Martino Marangella
Nephrology and Dialysis Unit, A.S.O. Ordine Mauriziano, Turin  Italy
ABSTRACT:We describe the theoretical features and
in vivo
assay of an srcinal computerbased program, which describes Na and HCO
3
kinetics with acetatefree biofiltration
®
performed as a continuous venovenous renal re placement therapy (CVVAFB). In 14 patients, CVVAFB sessions were performed as follows. Machine Hospal Prisma
®
,membrane AN69 0.9 m
2
, blood circuit Hospal M100pre Set, dialysate bags of electrolyte solution with basic composition of: Na 139 mEq/L, K 2 mEq/L, Ca 3.0 mEq/L, Mg 2 mEq/L, Cl 146 mEq/L, and glucose 5.5 mmol/L. Predilutional infusion bags of 167 mEq/L of NaHCO
3
solution (Hospal Biosol). To guide the prescription of the dialysate(Q
D
) and the infusate (Q
Inf
) flow rates we designed a suitable computer program, which calculated Na and HCO
3
kinetics and predicted their plasma profiles at equilibrium. Prescribed flow rates were Q
B
150 ml/min, Q
D
1636 ±42ml/hr, Q
Inf
639 ±74 ml/hr and ultrafiltration (UF) rate 110 ±41 ml/hr. During CVVAFB, plasma profiles of HCO
3
and Na became stable after 24 hr. With HCO
3
, the average predicted and observed plasma levels at equilibrium were27.1 ±1.5 and 27.3 ±1.5 mEq/L, respectively (p=ns); with Na, the predicted and observed levels at equilibrium were140.2 ±1.8 and 141.3 ±1.7 mEq/L, respectively (p=ns). There was a good correlation between predicted and observed values for both Na (p<0.01, r=0.78) and HCO
3
(p<0.01, r=0.87). These results confirmed the reliability of ourmathematical model for CVVAFB. A Longterm trial is needed to obtain more information on its clinical effects andcompare CVVAFB with other continuous renal replacement therapy techniques. Key words:
Continuous renal replacement therapy, Acetatefree biofiltration, Hemodiafiltration, Hemodialysis, Intensive care
Vitale et al
169
has never been switched from an intermittent to a continuous technique for blood purification. The solute kinetics with CRRT are different from those occurring with intermittent blood purification techniques. With applied flow rates of 23 l/hr, the dialysate becomesfully saturated with respect to plasma, i.e. the concentrations of diffusible solutes tend to be the same in bothblood and dialysate (5). Similarly to intermittent AFB
®
, if CRRT is made by using HCO
3
free dialysate, a greateramount of HCO
3
moves from the blood to the dialysateand must be carefully compensated by infusing NaHCO
3
.However, just because of the different kinetics of solutes with continuous and intermittent dialysis techniques, Skipper
®
cannot be used to tailor the best dialysate and infusateflow rates with continuous venovenousAFB
®
(CVVAFB).Therefore, a dedicated computer program was necessary for calculating kinetics and mass balances of Na andHCO
3
. The correct management of Na balance is important incritically ill patients, who can both be given Na with parenteral nutrition, and lose it through drainage and urine.This gain and loss participates with the substitution fluidssupplied during CRRT in determining the final electrolytebalance. In this computer program, additional determinants of Na balance are taken into account, thereby helping the nephrologist to accomplish and maintain physiological plasma levels of Na and HCO
3
during CVVAFB. In this paper, we detail the mathematics of our computerprogram for CVVAFB prescription and report the resultsof its
in vivo
assay.
M
ETHODS
The machine
CVVAFB was tailored to the Hospal Prisma
®
machine, set on its program for predilution continuous hemodiafiltration. The choice of the pre

dilution infusion, instead of thepost

dilution used for intermittent standard AFB
®
, wasaimed at reducing the anticoagulant dose.The standard technical performances of Prisma
®
are as follows. Dialysate flow rate (Q
D
), up to 2500 ml/hr; infusional flow rate (Q
Inf
), up to 2000 ml/hr and blood flow rate(Q
B
), up to 180 ml/min. Four rotating pumps control theblood, dialysate, infusate and the effluent (spent dialysate,infusate and net ultrafiltration (UF)) flow rates; a syringepump provides anticoagulant supplementation. The prescribed fluid balance is automatically controlled by threegravimetric scales, which are connected to the dialysate, infusate and the effluent bags.
CVVAFB design
In CVVAFB, Na and HCO
3
are infused predilutionally andmove from the plasma to the dialysate through the filtermembrane. An “equilibrium point” is finally reached, when the gain and loss of these solutes compare and theirplasma concentrations level off over time. Starting from defined values of dialysate and infusate flow rates (the main determinants of solute clearances), it ispossible to predict the “equilibrium points” of Na andHCO
3
concentrations in plasma. Based on these assumptions, we designed a computer program (running on Microsoft Excel
®
) that estimates the concentrations at whichNa and HCO
3
reach their respective “equilibrium points”,based on the prescribed flow rates of blood and fluids(Tab. I).
Computer program: theoretical basis
To use the computer program for CVVAFB, some input variables are required, as follows. Blood flow rate (Q
B
ml/min), hematocrit (Hct, %), dialysate flow rate (Q
D
ml/hr), dialysate Na concentration (Na
D
mEq/L), infusate flow rate (Q
inf
ml/hr), Na and HCO
3
concentrations in infusate (Na
inf
and HCO
3inf
mEq/L), prescribedUF rate (Q
uf
ml/hr), Na intake due to total parenteral nutrition and therapies (Na
TPN
mEq/24 hr), Na losses withdrainage and/or urine (Na
loss
mEq/24 hr), estimated endogenous acid production (H
+
mEq/kg bw) and patient body weight (kg) (Tab. I). The input variables depend on both the individual characteristics of each patient and the CVVAFB program prescribed. Starting from defined values of input variables,the program simulates Na and HCO
3
kinetics during CV VAFB and predicts the plasma levels at equilibrium, as detailed below. In other words, input variables are
independent variables
, whereas plasma Na and HCO
3
levels are
dependent variables
and are the main output of the program. Dialysate and infusate flow rates of 1.6 and 0.6 L/hr, respectively, were routinely prescribed to provide a wholeurea clearance of at least 2 L/hr (6). The equilibriumpoints were set and changes in working parameters weremade if predicted plasma Na and HCO
3
levels at equilibrium differed from desired physiological values (Tab. I). Concerning HCO
3
balance, a daily HCO
3
consumptionroughly equivalent to the daily ashacid production was assumed. While under normal conditions a daily acid production of about 1 mEq/kg body weight occurs (7), inthese patients increased metabolic production occurs, fixing the daily metabolic consumption of HCO
3
at 1.2mEq/kg bw, to be used for calculations. As far as blood analyses are concerned, Na was reported astotal plasma concentration (NaP) and HCO
3
as totalblood concentration (HCO
3
). However, because diffusiveand convective kinetics involve only the diffusible or ultrafilterable electrolyte fraction, it was necessary to introducesome correcting factors in the calculations, as follows. The ionic concentration of Na in plasma water (Na
+
P) wascalculated as NaP divided by 0.935 (correction for protein) and multiplied by 0.96 (8). Then, Na
+
P was consid
TABLE I 
CVVAFB COMPUTER PROGRAM: A WORKSHEET (4 EXAMPLES)BLOODExample # 122a2bPrescribed flow rate (ml/min)input150150150150Hematocrit (%)input25252525DIALYSATE Prescribed flow rate (ml/hr)input1600200016001600[Na] concentration (mEq/L)input144139144144INFUSATEPrescribed flow rate (ml/hr)input600500500600[Na] concentration (mEq/L)input167167167167[HCO
3
] concentration (mEq/L)input167167167167Prescribed ultrafiltration (ml/hr)input120150150150Net ultrafiltration (ml/hr)output100505050OTHER METABOLIC PARAMETERSIntravenous infused Na (mEq/24 hr)input75757575Intravenous infusion, flow rate (ml/hr)input20100100100Na output (urine, drainage) (mEq/24 hr)input0000Endogenous fixed acid (mEq/kg bw)input1.21.21.21.2Body weight (kg)input65656565Plasma HCO
3
at equilibrium (mEq/L)output26.416.622.025.6Plasma sodium at equilibrium (mEq/L)output 141135138139
Continuous acetatefree biofiltration
170
ered as a reliable estimate of ultrafilterable Na (ufNa) (8).Diffusible plasma Na (dNa
+
P) was calculated as Na
+
P multiplied by 0.97 (Donnan factor) (8). In bicarbonatefreedialysate, Na (Na
D
) was considered freely diffusible (8). Due to the diffusibility of CO
2
in whole blood, the measureof its concentration in whole blood was considered as a reliable estimate of its concentration in plasma water. Consequently, the derived value of whole blood HCO
3
concentration (HCO
3
) was also used as ultrafilterable HCO
3
(ufHCO
3
). Diffusible HCO
3
(dHCO
3
) was calculated asHCO
3
x 1.15 (9). Other derived variables were obtainedfrom input variables, according to the procedures detailedbelow. Due to the predilution infusion, Na is slightly higher inthe plasma in the filter capillaries than in the plasma that enters the arterial line of the circuit. Therefore, plasma Nain the filter (NaP
f
) was calculated as: [(Na
inf
* Q
inf
+ NaP *Q
P
) / (Q
inf
+ Q
P
)], where Q
P
stands for plasma flow rate,calculated as Q
B
* (100Hct)/100. The same procedure was made to calculate plasma HCO
3
in the filter (HCO
3f
). In CRRT, dialysate is fully saturated with diffusible plasmasolutes. Therefore, diffusive balances of Na (NaDB) andHCO
3
(HCO
3
DB) were calculated as the plasmatodialysate gradient of their diffusible fractions, times thedialysate flow rate [NaDB =
(NaD  dNa
+
P
f
) * Q
D
; HCO
3
DB=
dHCO
f3
* Q
D
]. Considering the convective balances of Na (NaCB) andHCO
3
(HCO
3
CB), they were calculated as the ultrafilterable fractions of NaP
f
(ufNa
f
) and HCO
3f
(ufHCO
3f
) times the whole convective flow (Conv), which is thesum of Q
inf
and Q
uf
. The amount of Na infused in CRRT was calculated as(Na
inf
* Q
inf
); the whole reinfusive balance of Na (NaRB) was calculated as the sum of (Na
inf
* Q
inf
) plus Na
TPN
. Theamount of HCO
3
reinfused in CRRT was calculated as(HCO
3inf
* Q
inf
). The estimated metabolic consumptionof HCO
3
was considered equal to the estimated daily generation of endogenous fixed acid, namely, H
+
x kg of body weight (GenH
+
). Finally, the calculations of the whole balance of Na and HCO
3
(NaWB and HCO
3
WB) were madeas follows. 1) NaWB: NaDB + NaCB + NaRB Na
loss
2) HCO
3
WB: HCO
3
DB + HCO
3
CB + HCO
3
RB  GenH
+
Input variables were inserted in the MSExcel
®
electronicspreadsheet, where the derived variables were calculated,according to the criteria described above. Algebric values(positive or negative) of each balance were taken into account. Finally, equation [1] was solved for NaP whenNaWB = 0 and equation [2] was solved for HCO
3
WB = 0,thereby obtaining an estimate of NaP and HCO
3
at equilibrium. Table II summarizes the mathematical procedures
.
Patients
After preliminary assays
in vitro
, the reliability of our computer program for CVVAFB was also tested
in vivo
in 20critically ill patients with ARF, admitted to the ICU after
Vitale et al
171
cardiovascular surgery. Of the patients, 14 had been treated continuously for at least 48 hr and were selected for thestudy. All patients were treated using the same machine. The ethical committee of our institution approved the study.
CVVAFB prescription
The target “equilibrium points” of Na and HCO
3
had beenpredicted from the prescribed dialysate and infusate flow rates. Other water and electrolyte intakes or losses were
TABLE II 
VARIABLES USED FOR NaAND HCO
3
MODELING IN CVVAFB
a) Input variablesUnit
HctHematocrit% Weight Body weightkgQ
B
Blood flow rateml/minQ
D
Dialysate flow rateml/hrNa
D
Dialysate sodium concentration mEq/LQ
Inf
Infusate flow rateml/hrNa
Inf
Infusate sodium concentrationmEq/LHCO
3inf
Infusate bicarbonate concentrationmEq/LQ
uf
Prescribed ultrafiltration rateml/hrNa
TPN
Na intake (parenteral nutrition)mEq/24 hrH
+
Endogenous acid generationmEq/kgNa
loss
Sodium losses (urine, drainages)mEq/24 hr
b) Basic settingsCalculations
NaPSodium concentration in plasmamEq/LNa
+
PIonized Na in plasma water[NaP * 0.96 / 0.935]mEq/LufNaUltrafilterable plasma Na [Na
+
P]mEq/LdNa
+
PDiffusable plasma Na[Na
+
P * 0.97]mEq/LHCO
3
Plasma HCO
3
[blood HCO
3
]mEq/LdHCO
3
Diffusable plasma HCO
3
[HCO
3
* 1.15]mEq/L
c) Derived variablesCalculations
Q
P
Plasma flow rate [Q
B
* (100Hct)/100]ml/minNaP
f
NaP into the filter[(Na
inf
* Q
inf
+ NaP * Q
P
)/(Q
inf
+Q
P
)]mEq/LNa
+
P
f
Na+P into the filter[NaP
f
* 0.96 / 0.935]mEq/LufNa
f
ufNa into the filter[Na
+
P
f
]mEq/LdNa
+
P
f
dNa+P into the filter[Na
+
P
f
* 0.97]mEq/LHCO
3f
HCO
3
into the filter[(HCO
3inf
* Q
inf
+ HCO
3
* Q
P
)/(Q
inf
+Q
P
)]mEq/LdHCO
3f
dHCO
3
into the filter[HCO
3f
* 1.15]mEq/LConvWhole convection[Q
inf
+ Q
uf
]ml/hrNaDBNa diffusive balance [(Q
D
/1000) * (NaD  dNa
+
P
f
) * 24] mEq/24 hrNaCBNa convective balance [ufNa
f
* effUF * 24 / 1000]mEq/24 hrNaRBNa reinfusive balance [(Na
inf
* Q
inf
* 24 / 1000) + Na
intake
] mEq/24 hrNaWBWhole sodium balance[NaDB + NaCB + NaRB  Na
loss
] mEq/24 hrHCO
3
DBBicarbonate diffusive balance [(Q
D
/1000) * ( dHCO
3f
) * 24]mEq/24 hrHCO
3
CBBicarb. convective balance [  dHCO
3f
* effUF * 24 / 1000] mEq/24 hrHCO
3
InfBBicarbonate infusive balance [HCO
3inf
* Q
inf
* 24 / 1000] mEq/24 hrGenH
+
Endogenous acid generation [H
+
* Weight]mEq/24 hrHCO
3
WBWhole bicarbonate balance [HCO
3
DB + HCO
3
CB + HCO
3
InfB  GenH
+
]mEq/24 hrNaPat equilibrium point =(((Q
D
* Na
D
+ Na
inf
* Q
inf
+ (Na
intake
 Na
loss
) * 1000/24) * 0.935 * (Q
P
* 60 + Q
inf
)) / / ((Q
D
* 0.97 + Conv) * 0.96 * Q
P
* 60))( Na
inf
* Q
inf
) / (Q
P
* 60)mEq/LHCO
3
at equilibrium point =(((HCO
3inf
* Q
inf
 H
+
* Weight * 1000/24) / (Conv + Q
D
* 1.15)  (HCO
3inf
* Q
inf
) // (Q
P
* 60 + Q
inf
)) * (Q
P
* 60 + Q
inf
) / (Q
P
* 60))mEq/L
Continuous acetatefree biofiltration
172
considered as well. Values were entered into the calculator worksheet (Tab. I).The correspondence between the
prescribed
flow rates of dialysate and infusate and the average
actual
flow rates supplied by the Prisma
®
machine was evaluated by randomized hourly sampling over CVVAFB sessions. With blood,the correspondence between
prescribed
and
actual
flow rates was determined
in vitro
.
Blood circuit
Standard Prisma M100pre Set was used. This device is alocked circuit, including a 0.9 m
2
AN69 filter previously connected with lines for blood, dialysate, predilutional infusion, effluent and anticoagulant. The circuit is not equipped with a bubble chamber; the upper part of the filter works as an airtrapping system.
Fluids
Fiveliter bags of sterile electrolyte solution were used as acetatefree dialysate, the basic composition was Na 139mEq/L, K 2 mEq/L, Ca 3.0 mEq/L, Mg 2 mEq/L, Cl 146mEq/L and glucose 5.5 mmol/L. Sterile solutions of concentrated NaCl, KCl and CaCl
2
were used to modulate the basic dialysate composition according to the physician’s prescriptions.Predilution infusion was performed with bicarbonatecontaining bags currently used for standard intermittent AFB,ie Hospal Biosol
®
(NaHCO
3
concentration 167 mmol/L). Anticoagulation was performed with unfractionated heparin.
Biochemistries
During CVVAFB, plasma urea and creatinine (Cr) were determined at the start and every 24 hr; electrolytes, bloodpH and HCO
3
were determined at the start and every 6 hr. In plasma, urea and Cr were assayed by routine methods;Na was assayed by potentiometry and, after the suitablecorrection for coefficients and protein activity, reported astotal plasma Na (NaP, equivalent to photometry assay); pHand ionized Ca, by ionselective electrode. In whole blood,HCO
3
was calculated from CO
2
tension.
Statistical analysis
Data are expressed as mean ±standard deviation (SD). Asfar as the relationships between predicted and observed values of plasma Na and HCO
3
at equilibrium were concerned, two regression models were performed: predictedNa = b
0
+ b
1*
observed Na (equation 1); predicted HCO
3
= b
0
+ b
1*
observed HCO
3
(equation 2). Correlations between estimated and measured plasma Naand HCO
3
levels were performed with Pearson’s correlation coefficients (r). P<0.05 was considered statistically significant. With equations 1 and 2, the Passing and Bablok linear regression procedure (method conversion) (10) was used tocalculate the slope b
1
and the intercept b
0
with their 95%confidence interval to test that b
0
is no different from 0and b
1
is no different from 1.
R
ESULTS
Table II details the equations used by our computer program to calculate plasma of Na and HCO
3
levels at equilibrium. Table I reports the worksheet used for CVVAFBprescription (running on MSExcel
®
). Table III reportsthe main features of CVVAFB sessions (average duration51 ±5 hr).The average
actual
flow rates were lower than the corresponding
prescribed
flow rates: 1560 ±37 vs. 1636 ±42ml/min for dialysate (p<0.01), 617 ±63 vs. 639 ±74ml/min for infusate (p<0.01), and 141 ±4.0 vs. 150ml/min for blood (
in vitro
assay, p<0.01). On average, theprescribed flow rates tended to overestimate the actualflow rates by about 5%. For each CVVAFB session, dialysate composition was tailored to patient requirements by adding concentrated NaCl, KCl and CaCl
2
to the basic formula. Conversely, thesame infusate solution, i.e. hypertonic NaHCO
3
(167mEq/L), was used for all patients. Intravenous therapies were taken into account in calculating Na and fluid balances. On average 49 ±15 ml/min of fluids (saline or glucosecontaining solutions) were infused over CVVAFB sessions, and 96 ±37 mEq/24 hr of Na as well. Table IVreports plasma profiles of urea, Cr, pH and electrolytes during CVVAFB. As expected, with an administered urea clearance of about 38 ml/min, plasma urea andCr levels decreased progressively during CVVAFB. Na, pHand HCO
3
reached a plateau within 24 hr; thereafter, they leveled off until the 48th hr (p=ns). As concerns the plasma K and ionized Ca levels, wider fluctuations than Na andHCO
3
levels were observed, possibly due to the interference of several factors, including K content of transfusedblood and plasma, transcellular shift of K, and alteredbone turnover.In Table V the mean
predicted
and
observed
plasma levels of Na and HCO
3
at equilibrium are compared. The averagesof the values at 2448 hr of CVVAFB were taken as the
ob served values
. The correlation between predicted (140.2 ±1.8 mEq/L)and observed (141.3 ±1.7 mEq/L) plasma Na levels wassignificant (y = x  0.55; r=0.61, p=0.021). From the analysis of their confidence intervals, it emerged that the intercept of the correlation (b
0
) was not significantly different from 0 (which suggests that no
systematic
over orunderestimation of predicted Na occurred) and theslope (b
1
) was not significantly different from 1 (whichmeans that there was not a
percent
over or underestima