Acta Anaesthesiol Scand 2003; 47: 37–45
Printed in Denmark. All rights reserved
New mathematical model for the correct prediction of
the exchangeable blood volume during acute
normovolemic hemodilution
J. MEIER1, M. KLEEN2, O. HABLER2, G. KEMMING1,2 and K. MESSMER1
1
Institute for Surgical Research, 2Clinic of Anesthesiology, Klinikum Großhadern, Ludwig-Maximilians-University Munich, Germany
Background: The blood volume that has to be exchanged for
crystalloids and/or colloids during acute normovolemic hemo-
dilution (ANH) in order to reach a preset target hemoglobin
concentration (hb) is usually predicted by the Bourke and Smith
formula developed in 1974. This formula systematically over-
estimates the ‘true’ exchangeable blood volume (EBV), a fact
that may potentially endanger patients because the target hb
will be missed and the normovolemic anemia might turn out to
be more severe than a priori intended. Our objective was to de-
velop a more accurate mathematical model of hemodilution kin-
etics and to validate this new model in animals and in patients
undergoing ANH.
Methods: Twenty-two anesthetized beagle dogs and 18 patients
under balanced anesthesia underwent isovolemic hemodilution
with hydroxyethyl starch (HAES 6%, 200 000) to a target hb of 7
g dlª1 or 9 g dlª1, respectively. Exchangeable blood volume pre-
dicted by use of the different mathematical models was com-
pared with the blood volume actually exchanged to meet the
preset target hb.
Results: Calculation of EBV by the Bourke and Smith formula
A CUTE normovolemic hemodilution (ANH) is an es-
tablished procedure for the avoidance of allo-
genic blood transfusions in elective surgery. Between
induction of anesthesia and start of surgery, fresh
whole blood is withdrawn from the patient and sim-
ultaneously replaced by the identic volume of a col-
loid solution or the triple volume of a crystalloid solu-
tion (1). Presuming that normovolemia is maintained,
hemodilution entails a reduction of red blood cells
(RBC), and thus of hemoglobin concentration (hb).
However, despite the reduced arterial oxygen content
and systemic oxygen delivery, systemic oxygen con-
sumption remains constant over a large hb range as a
result of different compensatory mechanisms (2),
mainly the increase of cardiac output, and the increase
of tissular oxygen extraction.
From the clinical point of view ANH offers several
advantages: (1) in hemodiluted subjects blood lost
Copyright C Acta Anaesthesiol Scand 2003
ACTA ANAESTHESIOLOGICA SCANDINAVICA
0001-6349
(EBVB π S) systematically overestimated the volume actually ex-
changed (overestimation: dogs 15%, patients 20%), whereas our
new iterative model predicted EBV (EBViterative) more reliably
(overestimation: dogs 1%, patients 8%). In both cases EBVB π S
differed significantly from the EBViterative.
Conclusion: Exchangeable blood volume is predicted more ac-
curately by the new iterative model than by the Bourke and
Smith formula. The iterative model leads to an improvement in
patient safety and provides a physiologically adequate basis for
future studies investigating the efficacy of ANH in reducing
allogenic blood transfusions.
Accepted for publication 17 June 2002
Key words: exchangeable blood volume; hemodilution; math-
ematical model.
c Acta Anaesthesiologica Scandinavica 47 (2003)
during surgery contains fewer RBC per milliliter (3–
5); (2) in case of transfusion need, the patient receives
fresh autologous whole blood, containing coagulation
factors and platelets in physiological concentrations;
and (3) transfusions may be effectively avoided with
application of ANH, presuming that a low target hb
is accepted (6). Hemoglobin concentration declines ex-
ponentially during ANH. This was described for the
first time 1974 by Bourke and Smith (7).
冉 冊
hb0
EBV Ω BV ¿ ln [1]
hbt
This is the original formula of Bourke and Smith (7).
EBV Ω exchangeable blood volume, BV Ω initial (i.e.
pre-ANH) blood volume, hb0 Ω initial hb, hbt Ω target
hb.
Equation 1 is also used in clinically practiced ANH
37
J. Meier et al.
to predict the exchangeable blood volume (EBV)
necessary to reach a target hb (hbt). Simultaneously it
constitutes the mathematical basis for theoretical con-
siderations concerning the efficacy of ANH in avoid-
ing allogenic transfusions.
This formula is a simplified version of the plasma
exchange equations presented by Wallerstein et al. (8)
and Veall et al. (9) for replacement transfusion as a
form of treatment for hemolytic disease of the new-
born. Although developed for another therapeutical
concept these equations are related to the kinetics of
the decline of hb during hemodilution.
Calculation of EBV with the Bourke and Smith for-
mula presents three major weaknesses:
1. The formula has been developed on the basis of
an integration process based on the assumption of
continuous and simultaneous substitution of
blood with a colloid. This is not always the case in
the clinical setting.
2. The formula has been developed on condition that
the RBCs are distributed homogeneously through-
out the body, a fact that can not be assumed.
3. One prerequisite for validity of the Bourke and
Smith formula is constancy of blood volume
throughout the entire hemodilution process.
In the original publication the formula was validated
in patients using calculated instead of measured blood
volumes. Therefore validation of equation 1 with
these data is difficult.
It has been repeatedly reported that the Bourke and
Smith formula systematically overestimates EBV (11–
13). This implicates that exchange of the calculated
blood volume leads to a lower hb than targeted, a fact
that might endanger the patient.
The reasons for overestimation of EBV by the Bour-
ke and Smith formula are not completely clear; poss-
ible explanations are a systematic error arising from
estimation of blood volume; incorrect description of
the decline of hb during ANH by this simplified for-
mula itself; or changes of the actual total blood vol-
ume during the hemodilution process.
Therefore the first aim of the present study was to
find out whether a more extensive mathematical
model can describe ANH kinetics more accurately
than the original integrative model presented by
Bourke and Smith and whether the new model en-
ables a more precise prediction of EBV. It is undis-
puted that the formula of Bourke and Smith is math-
ematically correct. However it is supposed that it does
not sufficiently take into account all physiologic fun-
damentals involved in clinically practiced ANH.
The second aim of this study was to validate our
38
new iterative model in an animal model and in our
clinical setting.
Methods
Theoretical dilution model
The equation provided by Bourke and Smith corre-
sponds to an isovolemic, continuous and simultaneous
substitution of homogeneously distributed red blood cells
with a colloid solution. This is not necessarily the way
a clinical hemodilution procedure takes place. We de-
veloped a model that corresponds to an intermittent
hemodilution procedure considering inhomogeneous red
blood cell distribution throughout the body.
Clinical performance of ANH requires the following
preconditions:
P
Knowledge of patients prehemodilution hb (hb0);
P
Definition of the posthemodilution target-hb (hbt);
P
Measurement or estimation of a patient’s actual
blood volume (BV);
P
Determination of the modality of the hemodilution
procedure (continuous vs. intermittent substitution,
sequence of blood removal and substitution).
The following terms are used in the derivation of our
new iterative mathematical model: BV Ω total blood
volume, E Ω volume exchanged during each single di-
lution step, EBV Ω total exchangeable blood volume,
i.e. blood volume to reach the target post-ANH hb.
It has to be noted that RBC are not homogeneously
distributed within the vasculature, a fact that has not
been taken into account for development of equation
1. Hemoglobin concentration within large vessels
(large vessel hb Ω hblv) is higher than hb within small
vessels (small vessel hb Ω hbsv) (14–16). The whole
body hb (hbwb) depends on hblv and hbsv, and can
only be measured by dilution kinetics. It has to be
noticed that hb values are usually measured by taking
blood samples from an arterial or venous line placed
in a large vessel. The hb concentration therefore corre-
sponds more likely to hblv than hbwb.
For the development of the new iterative model the
following terms are used:
P
hblv n Ω large vessel hb before each single dilution
step n (can be measured from arterial or venous
blood samples);
P
hbwb n Ω whole body hb before each single dilution
step n (can not be measured easily, but corresponds
to the number of RBC in the whole body).
Therefore hb0 and hbt of the original formula corre-
spond to hblv 0 and hblv t.

Iterative model
The first step towards the derivation of a new iterative
model is to determine that hb arising after a given
blood volume is replaced in one single dilution step.
‘One step’ in this context is defined as the sequential
procedure of withdrawing a given amount of whole
blood and infusing a suitable volume of a colloidal or
crystalloid solution. This is one possible way ANH is
performed in clinical practice. It corresponds to the
‘intermittent substitution formulas’ presented by Wall-
erstein et al. (8) for blood substitution in the treatment
of erythroblastosis fetalis. For homogeneous distri-
bution of RBC throughout the body the following for-
mula for this intermittent procedure has been de-
scribed (17):
冉 冊
Y Vªx n
Ω [2]
Y0 V each single hemodilution step.
This is the original dilutional equation from McCul-
lough et al. (17) referring to the observations of Wall-
erstein et al. (8): Y Ω final concentration; Y0 Ω initial
concentration; V Ω plasma volume, ¿ Ω size of aliquot
removed; n Ω number of aliquots removed.
It has been shown that hb declines faster when
blood removal is performed before replacement of the
volume removed takes place (17). Even though equa-
tion 2 simulates a discontinuous procedure it does not
consider the fact that the RBCs are not distributed
equally throughout the whole body, a fact that has
been considered for replacement transfusion as a form
for treatment for hemolytic disease in the newborn by
Veall et al. (9) and for hemodilution kinetics by Hahn
(10). We propose a combination of these two concepts:
The volume of RBC that is withdrawn in any of the
dilution steps n of a discontinuous hemodilution pro-
cedure is hblv n ¿ E. The volume of RBC remaining is:
(hbwb n ¿ BV) – (hblv n ¿ E). The new resulting whole
body hb (hbwbn π 1) reads:
(hbwb n ¿ BV) ª (hb lv n ¿ E
hbwb n π 1 Ω [3]
BV
Using this equation it is possible to obtain each fol-
lowing hbwb by using the actual hbwb and hblv. Dur-
ing clinical ANH it is impossible to measure hbwb for
each dilution step. Therefore hbwb has to be calcu-
lated from hblv. The relationship between hbwb and
hblv is known to vary within small limits (15). Al-
though there might be a dependency on the actual
hb level (16), it can be assumed that the relationship
between hbwb and hblv is a fairly constant quotient
(dogs ⬇0.9, humans ⬇0.9), which only slightly varies
ANH: prediction of exchangeable blood volume
with changes of hb, temperature, arterial pressure, etc.
(14).
hbwb
⬇ 0.9 ; hbwb ⬇ 0.9 ¿ hblv [4]
hblv
Represents the relationship of hbwb to hblv in regard
to Gibson et al. Similar values for dogs and humans
have been reported (14, 15).
Using Equation 4 it is possible to obtain the new
large vessel hblvn π 1.
hbwb n π 1
hblv n π 1 Ω [5]
0.9
Thus an algorithm has been found enabling calcu-
lation of the new hblv (hblvn π 1) after each single
hemodilution step. The exchangeable blood volume
can be obtained as the sum of volumes exchanged in
a
EBV Ω 兺 Ex Ω E ¿ a [6]
xΩ1
The exchangeable blood volume can be obtained as
the sum of volumes exchanged in each single hemo-
dilution step. a Ω number of dilution steps
To determine EBV it is necessary to choose E and to
repeat the previously described steps until hbt is
reached (Fig. 1). This can be automated using a com-
puter program. Therefore an algorithm has been
coded in Visual Basic (Visual Basic 6, Microsoft Cor-
poration, Redmond, WA), enabling calculation of EBV
based on hblv 1, BV and E. This program is freeware
and may be obtained from the authors via e-mail
(jens.meier/firemail.de). Although the new iterative
approach differs from the integration process de-
veloped by Bourke and Smith, it can be demonstrated
Fig. 1. The two steps of the iterative model. The volume of blood to be
exchanged reads as the sum of the blood volumes replaced in the differ-
ent hemodilution steps. This algorithm can easily be implemented
using Microsoft Excel, or can be coded in any programming language.
A version for Microsoft Windows and PalmOS may be obtained from
the authors. Hb, hemoglobin concentration; BV, blood volume; wb,
whole body; lv, large vessels; E, volume exchanged in one dilution step.
39
J. Meier et al.
Fig. 2. The relation of exchangeable blood volume (EBV) calculated
with the original formula of Bourke and Smith (EBVB π S) and EBV
using the computer program (EBViterative). An initial hemoglobin con-
centration (hb) of 15 g/dl has been assumed. X-axis: different target
hb, Y-axis: EBV calculated as a fraction of blood volume. BViterative is
always lower than EBVB π S, indicating that using the original formula
results in an overestimation of EBV. Hb, hemoglobin concentration;
BV, blood volume.
that the two formulas correspond to each other, pre-
suming that three conditions are fulfilled:
1. Homogenous distribution of RBC within the vas-
culature.
2. Infinitely small E (i.e. removal of blood and in-
fusion of the substitute take place simultaneously
and slowly, otherwise homogeneous distribution of
RBC can not be warranted).
3. Constant BV.
The complete confirmation of equivalency and the ex-
planation of the three assumptions is given in the Ap-
pendix section. However the above-mentioned re-
quirements are not fulfilled during clinical ANH.
Therefore the original formula of Bourke and Smith,
although mathematically correct, systematically over-
estimates EBV. This is because of the fact that the
underlying mathematical preconditions are physiolo-
gically not realizeable.
Comparison of EBV obtained by the Bourke and
Smith formula or by the iterative model
In contrast to the Bourke and Smith formula, EBV cal-
culated with the new iterative model accounts for the
inhomogeneity of the distribution of RBC within the
vasculature and for the fact that the removal of blood
and the infusion of the substitute do not take place
slowly and simultaneously. No known mathematical
40
ANH model as yet takes these two facts into consider-
ation at the same time.
In order to investigate whether EBV calculated with
the new iterative model deviates from EBV calculated
with the Bourke and Smith formula and whether
using the iterative model enables a better prediction
of EBV, EBV has been calculated proportionally to
blood volume either with the Bourke and Smith for-
mula or with the iterative model for different target
hb values. For this purpose E has been chosen very
small (this corresponds to a slow and simultaneous
exchange of blood with the substitute). An overview
of this comparison is given in Fig. 2.
Then the influence of E on EBV was investigated.
The results are shown in Fig. 3. An overview of the
derivation of this figure is given in Appendix B.
Validation of the iterative hemodilution model in
dogs
In a large experimental protocol, basically investigat-
ing the effects of hyperoxemia on oxygen transport
and tissue oxygenation during extreme hemodilution
(18,19), 22 healthy beagle dogs of either sex, weighing
15.4 ∫ 1.9 kg were diluted with hydroxyethylstarch
(HAES 6%, MW 200 000/0.5%, Braun, Melsungen,
Germany) in one step from their initial hb of 12.2 ∫
2.7 g dlª1 to hbt Ω 7 g dlª1. The dogs had been splenec-
tomized at least 8 weeks before entering the study to
exclude changes of the red cell mass during ANH in-
duced by splenic contraction (20).
Fig. 3. The relation of exchangeable blood volume (EBV)iterative and
EBVB π S depending on the aliquot removed in a single hemodilution
step. The aliquot removed is given as a part of the total blood volume
(E/BV). If the aliquot removed in a single hemodilution step is not
larger than 10% of BV then the difference of EBViterative and
EBVB π S is not more than 5% (see Appendix B). Hb, hemoglobin con-
centration; BV, blood volume; E, volume exchanged in one dilution
step.

Measurement of hb and BV
Hemoglobin concentration was measured immedi-
ately before (hb0) and after ANH (hbt) in blood
samples withdrawn from the aorta using a CO-Oxy-
meter (CO-Oximeter 682, Instrumentation Laboratory,
Lexington, MA).
Before and after ANH the animals’ BV was meas-
ured using indocyanine green (ICG) dilution kinetics
(21,22). The BV measured before ANH was used for
the calculation of EBVB π S and EBViterative. After injec-
tion of 0.25 mg kgª1 BW indocyanine green, the mono-
exponential decay of the blood ICG concentration was
determined intermittently for at least 5 min every 30
s. Extrapolation of the decay to the time point of injec-
tion enabled calculation of the ICG dilution and thus
blood volume:
ICG doseinjected
BV Ω [7]
ICG concentrationt Ω 0
Represents BV calculated by indocyanine green di-
lution kinetics.
Deviation from the Bourke and Smith formula
In order to quantify the difference of the quantity of
whole blood exchanged to reach hbt (Vexchanged), and
the exchangeable blood volume calculated with the
Bourke and Smith formula (EBVB π S) the following
ratio was defined:
Vexchanged
ratio1 Ω [8]
EBVB π S
Here, the volume actually exchanged is compared
with the volume calculated by the Bourke and Smith
formula.
Ratio1 equals 1 if EBV and Vexchanged are identical. If
the ratio is less than 1, a smaller volume of blood than
predicted by the original formula of Bourke and
Smith could be finally exchanged. If the calculated
blood volume would have been entirely withdrawn,
hb would have exceeded the targeted value. A similar
ratio is used to determine the quantity of whole blood
to be exchanged to reach hbt (Vexchanged), and the ex-
changeable blood volume was calculated with our
iterative computer model (EBViterative).
Vexchanged
ratio2 Ω [9]
EBViterative
Here, the volume actually exchanged is compared
with the volume calculated by the computer model.
ANH: prediction of exchangeable blood volume
Validation of the iterative hemodilution model in
patients
Presuming that EBV can be accurately predicted using
our new iterative model with blood volume having
been measured, it still has to be demonstrated that EBV
can also be accurately predicted when the blood vol-
ume is estimated on the basis of a mathematical model.
In the latter case any calculation of EBV depends on
correct estimation of BV. Therefore our iterative model
was also validated in patients undergoing ANH, in
whom blood volume was estimated.
For that purpose specific data obtained in a clinical
protocol including normovolemic hemodilution were
retrospectively analyzed (23). In the present analysis
we refer to 18 patients who were contributed to the
multicenter study by our own center. After induction
of anesthesia, the patients (weight 72.94 ∫ 15.7 kg)
were diluted with hydroxyethylstarch (HAES, 6%,
MW 200 000/0.5%, Braun, Melsungen, Germany) from
hb0 Ω 11.2 ∫ 1.0 g dlª1 to target hbt Ω 9 g dlª1.
Estimation of BV and measurement of hb
Blood volume was estimated using the following for-
mulas:
Men: BV Ω 0.417 ¿ T3 π 0.045 ¿ BW ª 0.03Women:
BV Ω 0.414 ¿ T3 π 0.0328 ¿ BW ª 0.03[10]Estimation
of blood volume for men and women. T Ω height in
m, BW Ω body weight in kg (24).
Hemoglobin concentration in arterial blood was
measured using a CO-Oxymeter (HemoCue, AB Leo
Diagnostics, Helsinborg, Sweden) before and after
each hemodilution step.
The ratios presented were the same as described
earlier for the animal model.
Statistical analysis
A Wilcoxon signed-rank test was used to assess statis-
tically significant differences between ratio1 and ratio2
in both studies. P ⬍ 0.05 was considered statistically
significant.
Results
Figure 2 depicts the comparison of the data obtained
by the Bourke and Smith formula and by the new iter-
ative model for very small E (E»0). Exchangeable
blood volume/BV calculated by the new iterative
model is approximately 10% lower than the values ob-
tained using the Bourke and Smith formula. Therefore
the new iterative model suggests systematically lower
values for EBV than the Bourke and Smith formula,
and an overestimation of EBV is less likely to occur
using the new iterative model.
41
J. Meier et al.

from ratio2 in the animals (0.85 ∫ 0.18 vs. 0.99 ∫ 0.15;

P ⬍ 0.05), as well as in the patients (0.8 ∫ 0.33 vs. 0.92
∫ 0.33; P ⬍ 0.05).
The dogs’ blood volume (Fig. 6) was not altered sig-
nificantly by the hemodilution process (1330 ∫ 198 ml
vs. 1348 ∫ 128 ml, NS).
The Bourke and Smith formula systematically over-
estimates the exchangeable blood volume in animals
as well as in patients (ratio1 ⬍ 1.0). The application of
our new iterative model enabled a significantly better
prediction of EBV in both settings.
The new iterative model correctly predicts EBV if

Fig. 4. Values of equation 7 and equation 8 obtained in the animal

model. ANH, acute normovolemic hemodilution.

The influence of E on EBV can be seen in Fig. 3. If

blood is withdrawn quickly (i.e. E is not infinitely
small) homogeneous distribution of RBC is not
achieved during the blood exchange and Hb will de-
cline more rapidly. The bigger the E that is chosen,
the lower the EBV will become as calculated by the
computer program (EBViterative). If 10% of the total
blood volume of an animal is withdrawn and subse-
quently (not simultaneously) replaced (E/BV Ω 0.1),
then the ratio of EBViterative/EBVB π S will be more
Fig. 5. Values of equation 7 and equation 8 obtained in the patients.
than 0.95. Hence, even if in the clinical setting single BV, blood volume; ANH, acute normovolemic hemodilution.
dilution steps amount to 10% of the total blood vol-
ume (E Ω 0.1 ¿ BV), the ratio EBViterative/EBVB π S re-
mains still greater than 0.95. Therefore the error im-
posed by large exchange volumes may explain ap-
proximately 5% of the deviation of the Bourke and
Smith equation. In a clinical setting blood withdrawal
and infusion of the substitute usually take place sim-
ultaneously. Therefore the time necessary for distri-
bution of the substitute is less than is needed for in-
fusion of 10% of the blood volume, so this error might
even be smaller. A smaller E corresponds to a better
prognosis of EBV by the Bourke and Smith formula
as the influence of E on EBV will diminish. Neglecting
the influence of E on EBV as done by the Bourke and
Smith formula leads to a further overestimation of
‘real’ EBV.
The ratios obtained from applying equations 8 and
9 are depicted in Fig. 4 for the experimental ANH and Fig. 6. The dogs’ blood volume before and at the end of the hemodilu-
in Fig. 5 for the patients. Ratio1 differs significantly tion procedure. BV, blood volume.

42

blood volume is measured (ratio Ω 0.99). If blood vol-
ume is calculated also, the new iterative model will
overestimate EBV (ratio Ω 0.92). Nevertheless, in this
clinically relevant setting calculation of EBV with the
iterative model is significantly better than using the
Bourke and Smith formula (P ⬍ 0.05). The overestima-
tion of EBV by the iterative model can be neglected in
comparison with that by the Bourke and Smith for-
mula.
Discussion
The main findings of the present study are:
1. The formula of Bourke and Smith systematically
and significantly overestimates EBV independent
of the method for blood volume determination
(measurement or estimation).
2. Using our new iterative model a more accurate pre-
diction of EBV was achieved as compared in a stan-
dardized laboratory situation as well as in a clinical
hemodilution procedure.
3. The accuracy of our new iterative model is suffi-
ciently high to justify its preferred application in
clinically performed ANH.
The efficacy of hemodilution in reducing perioperat-
ive allogenic blood transfusions is mainly related to
the fact that in a case of surgical bleeding less RBC
will be lost per milliliter in a hemodiluted patient than
in a patient with normal hb. The lower the hb after
ANH the higher becomes the probability to avoid
allogenic transfusion (6). However the correct predic-
tion of EBV, particularly in extreme hemodilution, is
important because the margin of safety to guarantee
adequate tissue oxygenation narrows with higher de-
grees of dilutional anemia. Exceeding a targeted low
hb due to overestimation of EBV might expose the
patient to the risk of tissue hypoxia. Therefore in ex-
treme hemodilution the new iterative model is su-
perior to the formula of Bourke and Smith concerning
a patient’s safety.
Moreover in theoretical and mathematical studies
on ANH, the use of the Bourke and Smith formula
may also result in an incorrect prediction of EBV. Most
of the studies investigating the clinical efficacy of
ANH use one of two different approaches. Either the
number of allogeneic transfused RBC units is com-
pared between patient groups undergoing ANH or
not (4, 25), or mathematical models are developed to
predict the amount of blood to be saved by ANH (3,
5, 6, 26–28). In most of these studies the original for-
mula of Bourke and Smith is applied, although better
models have already been found (10). As a conse-
ANH: prediction of exchangeable blood volume
quence these results have to be interpreted carefully,
as they usually do not account for the fact that the
original Bourke and Smith formula will systemically
overestimate the exchangeable blood volume.
EBViterative is approximately 10% lower than
EBVB π S if E is infinitely small (Fig. 2). Therefore an
overestimation of ‘true’ EBV is less likely to occur
using our new iterative model. This is confirmed by
the ratios obtained (equations 8 and 9) in the stan-
dardized laboratory situation as well as in the clinical
setting. Irrespective of the fact that blood volume is
measured or estimated the new iterative model al-
ways provides a more accurate description of the de-
cline of hb during a hemodilution process than ob-
tained by using the Bourke and Smith formula.
If blood volume is measured our new iterative
model enables a very reliable calculation of EBV
(ratio Ω 0.99). However, even if blood volume can only
be estimated like in the usual clinical setting the new
iterative model also provides a more accurate esti-
mation of EBV than do the common methods (ratio Ω
0.92). Nevertheless, optimal, e.g. correct results,
would also be obtained if the patient’s blood volume
is measured.
The standard deviation of ratio 1 and ratio 2 in our
animal model (equations 8 and 9) is as high as the
standard deviation of the respective ratios found in
patients (Figs 4 and 5). Therefore prediction of EBV
using our computer model supplies correct but not
highly precise values of EBV. Several reasons might
be responsible for this. Volume shifts from or to the
interstitium might contribute to a fall or increase of
hemoglobin concentration in absence of actual blood
or fluid loss. Correct measurement or estimation of
BV is an indispensable precondition for correct and
precise calculation of EBV, but can usually not be per-
formed in a clinical setting. Therefore any variations
of BV during the observation period will lead to an
incorrect prediction of EBV. No changes of the blood
volume of the dogs were observed before and after
the hemodilution procedure. Therefore a systematic
overestimation of EBV by the Bourke and Smith for-
mula can not be explained by systematic changes of
blood volume throughout the hemodilution process
in this setting. As blood volume was not measured in
patients it can not be excluded that variations in blood
volume can explain part of the systematic overestima-
tion of EBV by the Bourke and Smith formula in pa-
tients.
The ratio obtained by use of equation 5 might not
be precise enough to provide a correct relation of hbwb
and hblv. Because hblv is only one of the many factors
contributing to hbwb, it is not surprising that this cor-
43
J. Meier et al.
relation can not be complete. Obtaining a more precise
prediction of hbwb is difficult as many variables have
to be taken into account (species, body weight, height,
sex, splenic contraction, blood volume, etc.).
In conclusion our results demonstrate that by using
our new iterative model, EBV can be predicted more
accurately than by the Bourke and Smith formula be-
cause physiological conditions are more adequately
taken into account. Former formulas regarding di-
lutional kinetics do not take into consideration the
specific problems arising with the use of ANH. Our
new iterative model is easy to apply, its use enhances
patient safety and it provides a correct physiological
basis for further studies addressing the efficacy of
ANH.
References
1. Messmer K, Sunder-Plassmann L, Klövekorn W, Holper K.
Circulatory significance of hemodilution: rheological
changes and limitations. Adv Microcirc 1972: 4: 1–77.
2. Doss D, Estafanous F, Ferrario C, Brum J, Murray P. Mechan-
ism of systemic vasodilation during normovolemic hemo-
dilution. Anesth Analg 1995: 81: 30–34.
3. Feldman J, Roth J, Bjoraker D. Maximum blood savings by
acute normovolemic hemodilution. Anesth Analg 1995: 80:
108–113.
4. Herregods L, Foubert L, Moerman A, Francois K, Rolly G.
Comparative study of limited intentional normovolaemic
haemodilution in patients with left main coronary artery ste-
nosis. Anesthesiology 1995: 50: 950–953.
5. Weiskopf R. Mathematical analysis of isovolemic hemodilu-
tion indicates that it can decrease the need for allogenic
blood transfusion. Transfusion 1995: 35: 37–41.
6. Weiskopf R. Efficacy of acute normovolemic hemodilution
assessed as a function of fraction of blood Volume lost. Anes-
thesiology 2001: 94: 439–446.
7. Bourke D, Smith T. Estimating allowable hemodilution.
Anesthesiology 1974: 41: 609–612.
8. Wallerstein H, Brodie S. The efficiency of blood substitution.
Am J Clin Pathol 1948: 18: 857–866.
9. Veall N, Mollison P. The rate of red-cell exchange in replace-
ment transfusions. Lancet 1950: 2: 792–797.
10. Hahn R. Estimating allowable blood loss with correction for
variations in blood Volume. Acta Anaesthesiol Scand 1989: 33:
508–512.
11. Klövekorn W, Messmer K. Warum entspricht der berechnete
‘in vitro’ Effekt der präoperativen Hämodilution nicht den
klinischen Tatsachen? Anaesthesist 1976: 25: 193–194.
12. Kreimeier U, Messmer K. Hemodilution in clinical surgery:
state of the art 1996. World J Surg 1996: 20: 1208–1217.
13. Manel J, Garric J, Lefevre J, Laxenaire M. Règle à calcul du
Volume sanguin à prélever pour réaliser une hémodilution
normovolémique intentionnelle. Ann Fr Anesth Reanim 1988:
7: 427–432.
14. Gibson J, Peacock W, Seligman A, Sack T. Circulating red
cell volume measured simultaneously by the radioactive
iron and dye methods. J Clin Invest 1946: 25: 838–847.
15. Gibson J, Seligman A, Peacock W, Aub J, Fine J, Evans R.
The distribution of red cells and plasma in large and minute
44
vessels of the normal dog, determined by radioactive iso-
topes of iron and iodine. J Clin Invest 1946: 25: 848–857.
16. Pries A, Fritzsche A, Ley K, Gaehtgens P. redistribution of
red blood cell flow in microcirculatory networks by hemo-
dilution. Circ Res 1992: 70: 1113–1121.
17. McCullough J, Chopek M. Therapeutic plasma exchange.
Laboratory Med 1981: 12: 745–753.
18. Habler O, Kleen M, Hutter J et al. Effects of hyperoxic venti-
lation on hemodilutioninduced changes in anesthetized
dogs. Transfusion 1998: 38: 134–144.
19. Habler O, Kleen M, Hutter J et al. Hemodilution and intra-
venous perflubron emulsion as an alternative to blood trans-
fusion: effects on tissue oxygenation during profound hemo-
dilution in anesthetized dogs. Transfusion 1998: 38: 145–155.
20. Sato N, Shen Y, Kiuchi K, Shannon R, Vatner S. Splenic con-
traction-induced increases in arterial O2 reduce requirement
for CBF in conscious dogs. Am J Phys 1995: 268: 491–503.
21. Haller M, Akbulut C, Brechtelsbauer H, Fett W, Briegel J,
Peter K. Determination of plasma volume with indocyanine
green in man. Life Sci 1993: 53: 1597–1604.
22. Henschen S, Busse M, Zisowsky S, Panning B. Determi-
nation of Plasma Volume and Total Blood Volume using
Indocyanine Green: a short review. J Med 1993: 24: 10–27.
23. Spahn D, Brempt R, Theilmeier G et al. Perflubron emulsion
delays blood transfusions in orthopedic surgery. Anesthesi-
ology 1999: 91: 1195–1208.
24. Allen T, Peng M, Chen K, Huang T, Chang C, Fang H. Pre-
diction of blood Volume and adiposity in man from body
weight and cube of height. Metabolism 1956: 5: 328–345.
25. Weisel R, Charlesworth D, Mickleborough L et al. Limi-
tations of blood conservation. J Thorac Cardiovasc Surg 1984:
88: 26–38.
26. Brecher M, Rosenfeld M. Mathematical and computer
modeling of acute normovolemic hemodilution. Transfusion
1994: 34: 176–179.
27. Kick O, Daniel E. Mathematical considerations in the prac-
tice of acute normovolemic hemodilution. Transfusion 1997:
37: 141–143.
28. Smetannikov Y, Hopkins D. Intraoperative bleeding: a math-
ematical model for minimizing hemoglobin loss. Transfusion
1996: 36: 832–835.
Address
Prof. Dr h. c. mult. Konrad Messmer
Institute for Surgical Research
Ludwig-Maximilians-Universität München
Klinikum Großhadern
Marchioninistraße 15
81366 München
Germany
E-mail: messmer/icf.med.uni-muenchen.de
Appendix A
If red blood cells (RBC) are distributed homogeneously within
the vasculature (hb1), hb after one hemodilution step (hb2)
reads:
hb1 ¿ (BV ª E)
hb2 Ω [11]
BV
And for any further hemodilution step:
冉 冊
hbn ¿ (BV ª E) E
hbn π 1 Ω Ω hbn ¿ 1 ª [12]
BV BV
 ANH: prediction of exchangeable blood volume

This is the recursive form of a geometric series if E and BV algorithm (implemented in Maple V5, Brooks/Cole Publishing
remain constant. If this can be ensured equation 12 can be trans- Company, Pacific Grove, CA) the equation can be solved:

冉 冊 冉 冊
formed into a recursion-free presentation:
hbn hb1

冉 冊
lim (EBV) Ω ª BV ¿ ln Ω BV ¿ ln [18]
E nª1 E»0 hb1 hbn
hbn Ω hb1 ¿ 1 ª [13]
BV Hence, it is demonstrated that the formula of Bourke and Smith
can be derived exactly and that it is a special case of the iterative
Exchangeable blood volume can be calculated by multiplying
procedure described above. It therefore describes the process of
the number of dilution steps with the volume replaced in one
ANH and its kinetics in a correct way presuming that three pre-
step: EBV Ω (n-1) ¿ E. In order to obtain the resulting hb not from
conditions are met:
the number of dilution steps, but from the volume replaced in
one step (E), one can substitute: 1. Homogeneous distribution of RBC within the vasculature.
2. Constant and infinitely small E.
EBV EBV 3. Constant BV.
EΩ ; nΩ π1 [14]
nª1 E

This expression has to be set in the recursion-free form of the

equation. Several rearrangements give this result: Appendix B
冉 冊冉 冊 Ω hb
EBV
E E n The error that is entailed by the fact that E is not infinitely small
1ª [15]
BV hb1 can be quantified by equation 17. It yields a term to calculate
EBV depending on the amount of E. In the following EBV calcu-
And further: lated by equation 17 is called EBViterative. A ratio of the value for

冉 冊
冢 冉 冊冣
EBV calculated by the limitation process (EBVlim) and the orig-
hbn
ln inal formula of the Bourke and Smith precise value for EBV
hb1 (EBVprecise) can be derived and written as follows:
EBV Ω E ¿ [16]

冉 冊 冉 冊
E
ln 1 ª hct1 E
BV BV ¿ ln ¿ ln 1 ª
EBVB π S hctt BV

冉 冊
Ω [19]
Thus EBV can be calculated depending on the volume replaced EBViterative hctt
by a single dilution step (E). During ANH, withdrawal of blood E ¿ ln
hct1
and infusion of the diluents are performed simultaneously. This
implies that the single dilution step is nearly infinitely small. It is possible to substitute the ratio E/BV with ¿ Ω E/BV. Several
In other words: E»0. One can rearrange the last expression as transformations produce:
follows: EBViterative x
Ωª [20]
E
冉 冊
hbn EBVB π S ln(1 ª x)

冉 冊
EBV Ω ¿ ln [17]
E hb1 This term is plotted in Fig. 3. It depicts the relation of EBViterative
ln 1 ª and EBVB π S, depending on the aliquot removed in a single
BV
hemodilution step. The aliquot removed is given as a part of the
The limit for this expression can be calculated. Using a standard total blood volume (E/BV).

45