Beruflich Dokumente
Kultur Dokumente
1152–1163
Department of Pediatric Nephrology, Polish-American Children’s Hospital; and Faculty of Electrical Engineering, Automatics,
Computer Science, and Electronics, University of Mining and Metallurgy, Krako´w, Poland
Accuracy of hemodialysis modeling. from patient weight are the simplest coefficients for mathemati-cal
Background. One- and two-compartmental models of he- models and have sufficient precision as well. The global value of
modialysis (HD) are well known. These models make it possi-ble to both compartments is slightly greater than the corre-sponding value
analyze the course of treatment and to predict the effect of dialysis for a one-compartmental model. The effective-ness of dialyzers is
procedures. Mathematical modeling helps physicians to match in practice lower than might be expected on the basis of the data
dialysis therapy to the individual needs of the patient; however, the provided by their manufacturers. Urea cellular clearance is two
efficiency of the models depends on the accuracy of the coefficients. times greater than creatinine and uric acid cellular clearances. The
How to select coefficients in the case of one-compartmental models clearance differences are more prominent for the cellular membrane
is known for urea and creatinine. Less information is available for than for artificial semiper-meable membranes.
two-compartmental models. Results on modeling of uric acid
concentrations have not been published.
Methods. The identification of the mathematical model coef-
Mathematical modeling for hemodialysis aids the phy-
ficients was based on the concentration measurements of three
markers of uremic toxicity (urea, creatinine, and uric acid) in both sician in providing an accurate dialysis prescription as well
patients’ blood and dialysate. Blood samples were taken from the as predicting the course of the individual’s renal disease.
arterial line several times throughout the dialysis period. Although the two-compartmental models are more effective
Simultaneously, dialysate samples were taken from a test port in the than one-compartmental models in de-termining the urea,
dialyzer outflow line. The mathematical model parameters were
determined so as to minimize the deviations between the measured
creatinine and uric acid concentra-tions, they are much more
points and the calculated curves. In this way, distribution volumes, complicated to use in the dialysis setting.
cellular clearances, and dialyzer mass transfer coefficients were
estimated. Studies were carried out on six chronically dialyzed
Results. For a one-compartmental model, the median value of patients in the course of three regularly performed he-
distribution volume V _ 0.56 DW was obtained, where DW is the
patient’s dry weight. For a two-compartmental model, modialysis (HD) sessions for each patient. The dialyses were
intercompartment volume Vi _ 0.36 DW and extracompart-ment performed at Dialysis Unit, Department of Pediat-ric
volume Ve _ 0.21 DW. The following median values for cellular Nephrology, Jagiellonian University, Krakow, Po-land.
clearances were established: urea 415 (mL/min), creatinine 207 Braun Secura hemodialysis machines were used, and low-
(mL/min), and uric acid 257 (mL/min). flux polysulfone Fresenius F5 and F6 and Baxter CA-110
Conclusions. One- and two-compartmental models describe the
concentration of the urea, creatinine, and uric acid very effectively, dialyzers were employed. During each HD ses-sion,
in contrast with phosphorus, in which modeling results are not concentrations of four components (urea, creati-nine, uric
satisfactory. Although two-compartmental mod-els are more acid, and phosphorus) were determined on an average of 11
effective, they are much more complicated than one-compartmental times at 20-minute intervals by the use of the automated dry
models, which justifies using the one-com-partmental model for method (Kodak Ectachem 700 XR analyzer). About 1600
hemodialysis modeling. A two-compart-mental model must be used
in the case of rebound phenomenon modeling. The total body water determinations of toxin concen-trations in blood and
values we have obtained are similar to the anthropometrically based dialysate were done. All the pa-tients gave their informed
values for urea and creatinine and to a lesser degree for uric acid. consent for their participation in the studies carried out in the
Distribution volumes for one- and two-compartmental models course of regular HD sessions.
obtained
Computer software was prepared to calculate the opti-mal
Key words: compartment models, distribution volume, clearance, dial-ysis coefficients for the mathematical models. The values of the
modeling, urea clearance. distribution volume for three commonly measured uremic
Received for publication January 21, 1999 and in
toxins (urea, creatinine, and uric acid) were com-puted and
revised form August 26, 1999 Accepted for compared with volumes obtained via five an-
publication September 16, 1999 thropometrically based methods for total body water
2000 by the International Society of Nephrology assignment. The usefulness of these methods is discussed
1152
Zio´łko et al: Hemodialysis modeling 1153
and analyzed according to the results derived from the uted volume V to experimental data, correction for mean
experimental data. ultrafiltration was done automatically.
Experimental data were also used to establish cellular Equation 1 has a simple interpretation. The rate of toxin
clearances for three toxins: urea, creatinine, and uric acid. concentration (left hand side of Equation 1) is a decreasing
Additionally, actual clearances of the dialyzers em-ployed (minus sign in the right hand side of (Equa-tion 1) function
were studied. The results were compared with the data of time t and is proportional to the time variable toxin
supplied by the manufacturers. We found that actual concentration (Ce), proportional to constant clearance (Kd )
clearance values were lower than those reported by the and inversely proportional to constant distribution volume (V
manufacturers. ). The solution of Equa-tion 1 has the form:
The study was undertaken with the following aims: (1 ) to
estimate the mathematical model coefficients by fitting
Ce(t) _ C0e_tKd/V (Eq. 2)
theoretical curves to the measurements of urea, creatinine,
and uric acid concentrations in blood and dialysate; (2 ) to where e < 2.72 is the base of the natural logarithm. Equation
compare one-compartmental curve fits with two- 2 describes the decrease of toxin concentration during the
compartmental curve fits; (3 ) to verify the mathematical hemodialysis treatment. Coefficient C0 is an initial toxin
formulae describing dialyzers; and (4 ) to compare the concentration.
practical effectiveness of dialyzers for three toxins with the Volume distribution (V ) must be known in order to
effectiveness expected on the basis of the manufacturers’ calculate the allowable dialyzer clearance (Kd ) or the
data. optimal dialysis time. From Equation 2:
T_ V ln C0 (Eq. 3)
METHODS Kd CT
Models of hemodialysis therapy where T is the HD duration; CT is the toxin concentration
Wolf and his collaborators were the first who described value that should be arrived at as a result of HD; and ln is the
dialysis kinetics and dialyzer clearance almost 50 years ago natural logarithm.
[1]. Renkin was also a pioneer in the mathematical Equation 3 is used in the calculation of dialysis dura-tion
description of dialysis [2], while Sargent and Gotch suc- (T ) necessary to decrease the toxin concentration level from
cessfully introduced the one-compartmental model to clinical the value C0 to CT if the toxin distribution area in the patient’s
practice in the late seventies [3, 4]. The benefits and risks of body is V, and the dialyzer clearance Kd. Equation 3 could
this model application have been described and discussed in be helpful in clinical practice if the data (V, Kd and C0) are
many clinical and theoretical articles [5–13]. not encumbered with significant errors. The dependence of
relative deviations _T/T of HD duration on the relative
deviations of distribution volume _V/V, dialyzer clearance
Figure 1A shows the flows of any uremic toxin during
dialysis treatment according to the one-compartmental _Kd/Kd and toxin con-centrations _C0/Co and _CT/CT has the
form:
model. All body fluids and plasma water are considered as
T V Kd C C C
one volume of distribution [3, 4]. The mathematical model _ _ _ _ _ _ 1_ 0 _ _ T2 ln 0
for this case has the form of an ordinary differen-tial T V Kd C0 CT CT
equation:
(Eq. 4)
dCe _ _ Kd Ce (Eq. 1) The minus indicates the descending character of the
dt V dependence (for example, between dialysis time and dia-
where Ce is the toxin concentration; t is the time; Kd is the lyzer clearance). In clinical practice ln(C0/CT) < 1, be-cause
dialyzer clearance, and V is the compartment volume. toxin concentration values decrease by approxi-mately 2.7
It was assumed that the change of volume (V ) during HD times as a result of hemodialysis. It follows from Equation 4
had little influence on the modeling efficiency and could be that the deviations of the four argu-ments on the right hand
neglected. Another assumption was that urea generation and side of Equation 3 have a similar influence over the value of
the residual urea removal by the kidneys were very low in deviation _T/T of the dialysis duration. The V volume can be
comparison with the dialyzer clearance (Kd). The established relatively easily and precisely enough on the
mathematical model given by (Eq. 1) took only the most basis of anthropomet-ric data, mainly the patient’s body
important phenomena into consideration. The less important mass. These methods will be presented in later in this article.
phenomena sometimes cancelled each other out, and The clearance Kd can be determined on the basis of the
sometimes their effects could be represented by the information provided by the manufacturer of the dialyzer.
coefficients that characterize domi-nant phenomena. For Usually this is the theoretical value resulting from in vitro
example, in adjusting the distrib- experi-
1154 Zio´łko et al: Hemodialysis modeling
dCe _ _Kc(Ci _ Cd) _ KdCe When HD starts (t _ 0), toxin levels are equal within all the
(Eq. 6)
body compartments:
dt Ve Ve
Ce(0) _ Ci(0) _ C0 (Eq. 7)
Zio´łko et al: Hemodialysis modeling 1155
and can be measured in the patient’s blood. This assump-tion Assuming that KA is known, the estimation of toxin
is justified from a practical viewpoint and it consti-tutes the concentration in blood Ce can be calculated from the
initial conditions for Equation 6. For the Ce variable, the measured dialysate toxin concentration Cd:
solution of Equation 6 with the initial condi-tions given by
Cd(1 _ _Qd/Qb)
Equation 7 is a time dependent function: if Qb ? Qd
Ce(t) _ 0.5C0 e _ta
1 [(1 _ a3)e _ta
2
ta
_ (1 _ a3)e 2]
(Eq. 8)
Ce _ 5 1__
Cd(1 _ Qd/KA)if Qb _ Qd
(Eq. 11)
where
where KA(Qd_Qb)
_ _ e QbQd
a1 _ Kc _ Kd _ Kc
2Ve 2Vi
Identification of model parameters
KdKc
a2 _ !a 21 _ The method of model parameter identification pre-sented
ViVe in this article is based on toxin concentrations
˜
in dialyzer outflow line. Both concentrations, C and Cd, Vmale _ 0.58DW and Vfemale _ 0.55DW
must be measured at the same time. (Eq. 14)
1156 Zio´łko et al: Hemodialysis modeling
According to Bock, the compartment volume V de-pends related to its determination; the intercompartmental mass
only upon height and weight: transfer coefficient has a contractual character. Such Kc
ought to be fitted so that the toxin concentration described by
Vmale _ _14.249 _ 0.19678H _ 0.29571DW
Equation 8 meets the measured toxin con-centration in the
and Vfemale _ _9.926 _ 0.17003H _ 0.21371DW (Eq. 15) best way. If the amount of measurement data is sufficient,
The approach of Hume-Weyers et al [14] is similar to then it allows for a precise Kc calcula-tion.
Bock’s, but the coefficients are different:
Vmale _ _14.249 _ 0.19678H _ 0.296785DW
RESULTS
and Vfemale _ _35.270121 _ 0.344547H _ 0.183809DW
An example of measurements and modeling
(Eq. 16)
The results of a dialysis session modeling for a 23-year-
The equation old male patient (R.S.), whose postdialysis body weight was
V _ _0.07493713 · A _ 1.01767992 · G 58 kg is presented. The patient was dialyzed thrice a week
employing a polysulfone F-6 dialyzer (Fre-senius), and each
_ 0.12703384 · H _ 0.04012056 · DW
session lasted 3 hours and 30 minutes. The blood flow rate
_ 0.57894981 · D _ 0.00067247 · DW 2 through the dialyzer was 250 mL/ min, and the dialyzer flow
rate 500 mL/min. Blood sam-ples were collected from an
_ 0.03486146 · A · G _ 0.11262857 · DW · G
arterial line before entering the dialyzer, simultaneously with
_ 0.00104135 · A · DW _ 0.00186104 · H · DW (Eq. dialysate samples that were collected from a test port in the
17) dialyzer outflow line. The samples were collected 11 times
presented by Chertow et al is the most complicated but it according to the schedule presented in column 2, Table 1.
was obtained from data recorded for dialysate patients Determina-tions of urea, creatinine, uric acid and phosphorus
[16]. G in Equation 17 is equal to 1 for male and to 0 for in blood and in dialysate were made. On the basis of Equa-
female; D is equal to 1 for diabetes and equal to 0 for tion 10, dialyzer mass permeability coefficients KA were
nondiabetes. calculated for urea, creatinine and uric acid. By minimiz-ing
Few data are available in the literature on the inter- the criterion (Eq. 12), the distribution volumes (V, Ve, Vi)
compartmental mass transfer coefficient Kc, although this were determined for particular toxins in one- and two-
parameter is an integral part of the two-compartmen-tal compartmental models, as well as cellular clearances Kc. The
model (Equation 6). Most likely there are two reasons results are presented in the Tables and Figures. Only one
responsible for this situation. The first results from a much graph is presented for phosphorus, since the modeling results
less frequent use of the two-compartmental model than the were unsatisfactory (Fig. 11). The spe-cific kinetics of
one-compartmental model. The second reason for the lack of phosphate also have been observed by Maasrami et al [11].
information about Kc lies in difficulties
Zio´łko et al: Hemodialysis modeling 1157
Fig. 5. Blood creatinine concentration. Symbols are: (r) measured; (j) one-
compartmental model; (m) two-compartmental model.
Fig. 9. Mass transfer coefficient (KA mL/min) calculated for uric acid.
Fig. 12. Comparison of average errors for Equations 13 through 17, and
both one- and two-compartmental models. Symbols are: (m) Wat-son; (j)
Compartment volume identification
Hume; (h) simplified; (X) Chertow, and (_) Bock. The above presented methods were employed to cal-culate
the compartment volumes V, Ve and Vi in six chronically
dialyzed patients. All patient data are listed in Table 5. The
(5 ) mean value deviation (Eq. 12) between measured volumes calculated according to anthro-pometric
concentrations (column 7 in Table 1) and the val-ues measurements are listed in Table 6. In each column the first
of the two-compartmental model is Q _ 0.33 mmol/L. value was calculated when the patient’s post-HD weight was
introduced to the equation, and the second for his/her pre-HD
The results of the creatinine concentration modeling for weight. The differences between the two values are small (up
patient R.S. are: to 6.3%), but they are noticeable in statistical analysis. If the
patient’s predi-alysis weight was introduced into
(1 ) one-compartmental distribution volume V _ 31385 anthropometric equa-tions, then the resulting distribution
mL, volume was always
Zio´łko et al: Hemodialysis modeling 1159
Table 4. F6 Fresenius dialyzer clearance (blood flow rate 250 mL/min and tions of each toxin concentrations were made. The medi-ans
dialysate flow rate 500 mL/min)
from the three assessed sessions for each patient and each
In vitro In vivo Identified compartment are separately listed in Table 7.
Urea 208 188 169 _ 19 The anthropometrically derived volume results (Table
Creatinine 184 169 124 _ 12 6) were compared with the results (Table 7) obtained from
Uric acid No manufacturer data available 132 _ 9
the measurements and computer minimization of criterion
(Eq. 12). For each method and each toxin, the relative error
was calculated as:
greater than the values arrived at through kinetic identi- 1 6 3 V(formulan) _ V(identificationn,i)
fication. E_ oo (n,i) (Eq. 18)
18 V
In Table 7 the results of one- and two-compartmental n_1 i_1 identification
model volume identification for four commonly mea-sured along with standard deviation
uremic toxins (urea, creatinine, uric acid and phos-phorus) 1 6 3 V(formulan) _ V(identificationn,i) 2
Table 7. Median values of compartment volumes V and Ve _ Vi identified from one- and two-compartmental models
Volume L
Urea Creatinine Uric acid Phosphorus
Patient one-comp. two-comp. one-comp. two-comp. one-comp. two-comp. one-comp. two-comp.
M.P. 14.4 5.0 _ 9.9 13.6 6.4 _ 10.8 14.4 4.8 _ 10.8 3.2 2.5 _ 30.7
M.D. 29.0 11.5 _ 17.7 27.3 14.0 _ 17.7 19.9 6.6 _ 14.6 33.6 11.6 _ 66.4
Ma.K. 36.0 3.6 _ 29.1 30.1 10.8 _ 23.9 27.0 7.0 _ 20.3 75.8 4.0 _ 78.3
M.S. 32.7 18.1 _ 21.1 33.8 25.8 _ 19.9 25.2 11.2 _ 15.7 19.5 8.6 _ 114.5
Mi.K. 16.9 5.2 _ 11.6 19.5 15.7 _ 3.9 17.2 5.0 _ 11.6 16.2 1.9 _ 24.2
R.S. 32.9 10.6 _ 22.6 30.4 22.7 _ 52.3 25.5 11.7 _ 16.3 20.1 7.5 _ 39.7
Table 8. Average errors (Eq. 18) for anthropometrically based methods of volume determination
Urea Creatinine Uric acid
one-comp. two-comp. one-comp. two-comp. one-comp. two-comp.
Watson 0.15 0.12 0.17 _0.02 0.39 0.32
Equation 14 0.01 0.02 0.03 _0.14 0.24 0.18
Bock 0.08 0.05 0.10 _0.08 0.31 0.25
Hume 0.07 0.04 0.08 _0.09 0.30 0.24
Chertow 0.21 0.17 0.23 0.03 0.46 0.39
Table 9. Standard deviations (Eq. 19) for anthropometrically based methods of volume determination
Urea Creatinine Uric acid
one-comp. two-comp. one-comp. two-comp. one-comp. two-comp.
Watson 0.26 0.27 0.20 0.29 0.23 0.24
Equation 14 0.16 0.19 0.16 0.23 0.24 0.25
Bock 0.16 0.18 0.12 0.23 0.15 0.16
Hume 0.17 0.19 0.11 0.24 0.21 0.22
Chertow 0.20 0.21 0.15 0.27 0.16 0.17
Table 10. Identified cellular clearances Table 11. Estimates of cellular clearances and distribution volumes
The parameters Ve, Vi and Kc identified on the basis of Replacement of Renal Function by Dialysis, edited by MAHER JF, New
York, Kluwer Academic Publishers, 1989, pp 87–143
experimental data have considerable standard devia-tions as 5. BURGELMAN M, VANHOLDER R, FOSTIER H, BEIGAIR S: Estimation
a result of the relatively small sensitivity of the quantity of parameters in a two-pool urea kinetic model for hemodialysis. Med
criterion (Eq. 12) near the optimal solution. The two- Eng Phys 19:69–76, 1997
6. DEPNER TA: Prescribing Hemodialysis. New York, Kluwer Aca-demic
compartmental model is mathematically more complicated Publishers, 1991
than the corresponding one-compartmental model. The 7. GOTCH FA: Kinetic modeling in hemodialysis, in Clinical Dialysis,
edited by NISSENSON AR, FINE RN, GENTILE D, Norwalk, Appleton and
effectiveness in hemodialysis modeling is similar for both Lange, 1990, pp 118–146
models, thus justifying the use of a one-compartmental 8. HAKIM RM, DEPNER TA, PARKER TF: Adequacy of hemodialysis. Am J
model. The two-compartmental descrip-tion of toxin kinetics Kidney Dis 20:108–122, 1992
9. KOPPLE JD, JONES MR, KESHAVIAH PR, BERGSTROM J, LINDSAY RM,
ought to be used when explaining a steep increase of toxin MORAN J, NOLPH KD: A proposed glossary for dialysis kinetics. Am J
concentration immediately after the end of the dialysis (the Kidney Dis 26:963–981, 1995
so-called rebound phenom-enon). 10. LOPOT F (editor): Urea kinetic modelling, in EDTA/ERCA Series
(vol 4), Belgium, D. Verlinde, 1990
11. MAASRANI M, JAFFRIN MY, FISCHBACH M, BOUDAILLIEZ B: Urea,
creatinine and phosphate kinetic modeling during dialysis: Applica-tion
ACKNOWLEDGMENTS to pediatric hemodialysis. Artif Kidney Dial 18:122–129, 1995
12. PASTAN S, COLTON C: Transcellular urea gradients cause minimal
This research was supported in part by Scientific Research Commit-tee depletion of extracellular, during hemodialysis. ASAIO Trans 35:247–
grant KBN 0553/P4/93/05. 250, 1989
13. SCHNEDITZ D, FARIYIKE B, OSHEROFF R, LEVIN MW: Is intercompar-
Reprint requests to Dr. Jacek Pietrzyk, Department of Pediatric Ne- tmental urea clearance during hemodialysis a perfusion term? A
phrology, Polish-American Children’s Hospital, ul.Wielicka 265, 30–663 comparison of two pool urea kinetic models. J Am Soc Nephrol 6:1360–
Krako´w, Poland. 1370, 1995
14. WATSON PE, WATSON ID, BATT RD: Total body water volumes for adult
males and females estimated from simple anthropometric
REFERENCES measurements. Am J Clin Nutr 33:27–39, 1980
15. HUME R, WEYERS E: Relationship between total body water and surface
1. WOLF AV, REMP DG, KILLEY JE, CURRIE GD: Artificial kidney area in normal and obese subjects. J Clin Pathol 24:234–238, 1971
function: Kinetics of HD. J Clin Invest 30:1062–1070, 1951
2. RENKIN EW: The relationship between dialysance, membrane area, 16. CHERTOW GM, LAZARUS JM, LEW NL, MA L, LOWRIE EG: Develop-
permeability, blood flow in the artificial kidney. ASAIO Trans 2:102– ment of a population-specific regression equation to estimate total body
105, 1956 water in hemodialysis patients. Kidney Int 51:1578–1582, 1997
3. SARGENT JA, GOTCH FA: Mathematical modeling of dialysis ther-apy. 17. BESARAB A, ROSS R, FRINAK S, ZASUWA G: Increased urea kinetic
Kidney Int 18(Suppl 10):2–10, 1980 modeling, possible mechanisms and its significance. ASAIO J 43:303–
4. SARGENT JA, GOTCH FA: Dialysis principles and biophysics, in 310, 1997