ECOLE

NATIONALE
VETERINAIRE
TOULOUSE

Allometric scaling to predict

pharmacokinetic and pharmacodynamic
parameters in man

PL Toutain
UMR 181 Physiopathologie et Toxicologie Exprimentales
INRA, ENVT

1
 Introduction to allometry

Allometry (a term coined by Huxley

& Tessier 1936) is the study of size
and its consequences

2
 Range of body size in mammals

Shrew 2 g Blue whale: >108 g

Allometry is the study of size and its

consequences
Interspecies allometric scaling is based on the
assumption that there are anatomical, physiological
and biochemical similarities among animals which
can be described by simple mathematical models 3
Range of body size in mammals:
extrapolation within species

Adult to adult Young to adult

4
 Many allometric relationships have been
established between body size and organ weight
as well as body size and physiological process

5
 Simple allometry

Y=aBWb

6
 The power function
Y = aBWb
Where Y is the parameter of interest, BW is the body weight, a & b are
the coefficient and exponent of the allometric equation respectively

The log transformation of this equation is represented as :

log Y = log a + b x logBW

Linear plot: slope=b and intercept=log A

the slope of the line (b) indicates the type of scaling relationship

7
 Simple allometry:
the log-log transformation

logY=log a +b log BW

b=slope

Y=aBWb

log a is the Y-intercept

8
The scaling exponent (b) i.e. the slope
defines the type of scaling relationship

b=1.25
Y increase faster than BW
Positive allometry

b=1.0
Y increase proportionally
with BW (isometry)

b=0.75
Y increase slower than BW
Negative allometry

9
The assumption behind the log-log
transformation

It is assumed that there is a constant %CV

about the value of PK parameter
associated with BW being considered

10
 The log-log transformation

log-log transformation of the data will visually minimize

the deviations from a regression line
A high R2 (e.g. 0.95) do not guarantee that all the data
point will be close to the regression line
The extrapolation of this regression line to obtain a
predicted human value may have a great uncertainty
The regression process does not treat the weight of
each animal species comparably
Direct fitting of power function with incorporation of a
weighting strategy has been shown not to improve the
prediction performance
11
 The log-log transformation

When there is a limited number of species

associated with the regression analysis,
each data point has the greatest impact
on the prediction of Y for animals whose
value of BW are closer to the deviant
observation

12
 How does a the distribution of body weight used
in the regression analysis influence the
prediction of Y
For any species included in the regression
analysis, how does its location on the X-axis (i.e;
its value of BW relative to other observed data
points) influence prediction of Y
Can we anticipate the impact on prediction error
by the goodness of fit (R2) of the regression line

13
 Number of species and the
regression line
When there is a limited number of species associated
with the regression analysis, each data point has the
greatest impact on the prediction of Y for animals whose
value of BW are closest to the deviant observation

When a midpoint species (dog in vet medecine) is the

source of the error, the change is primarily in the
intercept rather the slope; consequently the resulting
magnitude of prediction error is comparable throughout
the range of BW values examined

14
 Influence on the predicted value in man of a 30% decrease
of the clearance value for a given species

species BW (kg) CL CL CL CL

Mouse 0.03 0.72 8 0.72 8 0.72 8 0.5046

Rat 0.2 2.99 2.09 2.99 2.99

Rabbit 4 28.28 28.28 28.28 28.28

monkey 8 47.56 47.56 47.56 47.56

dog 15 76.21 76.21 54.25 76.21

Man 70 242 247 200 212

predicted bias 0% +2% +17% +12%
15
ACCURACY OF ALLOMETRICALLY PREDICTED PHARMACOKINETIC
PARAMETERS IN HUMANS: ROLE OF SPECIES SELECTION

Huadong Tang and Michael Mayersohn

16
Drug Metabolism Disposition, 2005, 33 (9) 1288-1293
 ACCURACY OF ALLOMETRICALLY PREDICTED PHARMACOKINETIC
PARAMETERS IN HUMANS: ROLE OF SPECIES SELECTION

Huadong Tang and Michael Mayersohn

Drug Metabolism Disposition, 2005, 33 (9) 1288-1293

As demonstrated by both theoretical and literature experimentation,

rats had no significance in predicting human PK parameters as long as
the body weight of the rat is not the smallest in the species used in the
allometric relationship.

17
 Historical developments:
the direct extrapolation of doses
from animals to man

18
The Use of Body Surface Area as a Criterion
of Drug Dosage in Cancer Chemotherapy
Donald Pinkel
(Department of Pediatrics, Ronwell Park Memorial Institute
and
University of Buffalo School of Medicine, Buffalo, N.Y.)

Cancer Res 1958 28 853-856

19
 The use of body surface area as a criterion of
dosage regimen in cancer chemotherapy
(From D Pinkel :Cancer Res 1958 28 853-856)

Mouse=0.018

Body weight in Kg

Infant=8
Rat=0.25 Child=20
20
Adult=70
 Body surface area in man

The DuBois and DuBois formula

BSA (m) = 0.20247 x Height(m)0.725 x Weight(kg)0.425
The Haycock formula
BSA (m) = 0.024265 x Height(cm)0.3964 x Weight(kg)0.5378
The Gehan and George formula
BSA (m) = 0.0235 x Height(cm)0.42246 x Weight(kg)0.51456
The Boyd formula
BSA (m2) = 0.0003207 x Height(cm)0.3 x Weight(grams)(0.7285 -
( 0.0188 x LOG(grams) )

21
Comparison of toxicity data acquired during clinical studies of
18 anticancer agents with those obtained in mice, rats, dogs,
and rhesus monkeys uncovered close interspecies
correlations when doses were related to body surface, much
closer than when doses were related to mass. This finding
has guided numerous trials of anticancer and other agents.
22
Comparison of toxicity data on anticancer agents for
the Swiss mouse and man (on a mg per m 2 basis)
From Freireich et al 1966

1000
Maximum tolerated dose (mg per m2)

100

Antimetabolites
10
Alkylating agents
Others

1.0

0.1
10 1000
Mouse LD10 mg per m2 23
Observed and predicted dosage (mg per m 2) in
man using animal system (Freireich & al 1966)

24
 Interspecies scaling of maximum
tolerated dose of anticancer drugs

In general, small animal require larger dose than

human to reach the MTD.
Wanatabe et al used the LD10 mice data from
25 anticancer drugs and concluded that the MTD
in human can be predicted from mice LD1 using
a scaling power of 0.75
Actually the use of a fixed exponent cannot be
justified

25
 Slope actually from 0.60 to 0.84

Data from Freireich & al 1966

26
Body weight or body surface area?
BSA is not directly measured but
estimated with allometric equations
For a given species, it may exist several
equations predicting BSA
There is no advantage using BSA over
BW

27
28
What is exactly a Dose?

29
 The determination of an ED50 or any ED%

PD

Clearance x target EC50

ED50 =
Bioavailability

PK
ED50 - is a hybrid parameter (PK and PD)
- is not a genuine PD drug parameter
30
 What is a dose?

Cardiac _ output (L / day ) 321 BW (kg )0.75

clearance plasma Cardiac _ Output ER

clearance plasma ECtherapeutical

Dose
Bioavailability

31
 Cardiac output in mammals

Cardiac _ output 223 BW 0.75

In mL per minute Body Weight in 32

kg
 Interpretation of body clearance

Interpretation of body clearance consists of

calculating an extraction ratio

Body clearance (blood)

Ebody =
Cardiac output

33
 What is a dose?

Cardiac output (L per day)

g/L

321 BW 0.75 ER ECtherapeutical

Dose
Bioavailability

g per day

34
Dose (IV) for an hepatic cleared drug with a low or
a high hepatic extraction ratio (ER)

V max
Low ER Dose fu ECtherapeutical
Km
The plasma protein binding and metabolism activity are the major
determinants for the elimination of low hepatic clearance drugs;
therefore it is not expected to have a good allometric relationship
with BW across species for this kind of drug

High ER Dose 68BW 0.76 ECtherapeutical

Because hepatic blood flow is shown to have an allometric
relationship with BW, it is expected that the elimination of high
hepatic clearance drug can show an allometric relationship with BW 35
 Interspecies scaling of
pharmacodynamic parameters

PD

Clearance x target EC50

ED50 =
Bioavailability

36
 Interspecies scaling of
pharmacodynamic parameters

Very little information is available for the

prediction of pharmacodynamic (PD)
parameters from animal to man
It is conceptually difficult to accept that the
efficacy and potency of a drug will relate
with body weight of the species

37
 Allometry of pharmacokinetics and
pharmacodynamics of the muscle relaxant
metocurine in mammals

38
 Interspecies scaling of
pharmacodynamic parameters:
The case of Ketoprofen (sKTP)

Cat, goat, sheep, calf, horse

Endpoints: inhibition of the synthesis of
thromboxan (TXB2) and prostaglandinE2
(PGE2)
No relationship between IC50 (or other PD
parameters) with BW
39
Modeling and allometric scaling of s(+)-ketoprofen
pharmacokinetics and pharmacodynamics: a
retrospective analysis
E.-I. LEPIST & W.J. JUSKO, J. Vet. Pharmacol. Therap. 27, 211-218, 2004
ANTIINFLAMMATORY DRUG

40
41
Interspecies scaling of pharmacodynamic
parameters:
the case of anaesthetic potency minimum
alveolar concentration (MAC)

Poor correlation between BW and

MAC for several inhalation anesthetics

Travis & Bowers 1991in: Toxicol Ind Health 1991 7 249-260

42
 In vitro data: Drug affinity & drug potency

Drug potency
from in vitro:
MIC for
antibiotics

Benzodiazepine dose and benzodiazepine affinity

43
 Interspecies scaling of
pharmacokinetic parameters

Clearance x target EC50

ED50 =
Bioavailability

44
 Absorption
Volume of
Clearance
distribution
bioavailability

Half-life Systemic
exposure

Dosing regimen Dosage regimen

How often? How much 45
 Acute toxicity of anticancer drugs
human versus mouse
Dose Ratio AUC Ratio
External dose Internal dose
Frequency

46
Interspecies scaling of clearance

47
Simple allometry: Diazepam

48
Scaling of antipyrine intrinsic clearance in 15
mammalian species

49
Boxenbaum & Fertig Europ J Drug Metab Pharmacokinet 1984 9 177-183
 The concept of neoteny
Retention of juvenile
characteristics in the
adults of species
The modern man
retained its juvenile
characteristics of its
ancestors (apes)
through the
retardation of somatic
development for
selected organs
50
Exemple of Neoteny

51
 Interspecies scaling of clearance
1. Simple allometry
2. Allometry with various biological
correction factors
1. Product of maximum life-span (MLP) and
clearance
2. Product of brain weight and clearance
3. Ratio of clearance and GFR
4. Two-term power equation
5. Incorporation of molecular structure parameters
6. incorporation of in-vitro data in in-vivo clearance
7. Correction for protein binding
52
Simple allometry & allometry with standard
correction factors (MLP and Brain weight)

Clearance or Clearance multiplied by MLP or Brain

weight of several species are plotted against BW on
a log-log plot

Clearance aBW b
Clearance MLP aBW b

Clearance BrainWeight aBW b

53
 Product of maximum life-span (MLP) and
clearance
The clearance of different species are multiplied by
their respective MLP and are plotted against a
function of BW on a log-log scale

a(MLP Clearance ) b
Clearance man
8.18 10 5

MLP ( years ) 185.4 * Brain _ weight 0.636 * BW 0.225

54
 Prediction of Cefazolin Clearance in man:
standard vs. corrected allometry (MLP)

Simple allometry Allometry with MLP as a correcting factor

Predicted: 141 mL/min Predicted: 50.55mL/min
Actual: 61 mL/min Actual: 61mL/min
Error: 131% Error:17.1%

55
Selection of a standard correction factor
and the so-called rule of the exponent

The random use of the different correction factors is of

no practical value
Mahmood & Balian 1996 investigated 40 drugs and
found that the exponent of the simple allometry ranged
from 0.35 to 1.39
Based on these exponents ,it was found that there are
conditions under which only one of the three methods
can be used preferentially for reasonably accurate
prediction of clearance

Mahmood & Balian 1996 xenobiotica 26 887-895 56

 The rule of exponents
to predict clearance in man
Mahmood & Balian 1996

1. 0.55 b <0.71 : no correction factor is

necessary

2. 0.71 b <1.00 MLP should be incorporated

into scaling method

3. B>1.00 Brain weight should be incorporated

into the scaling method
57
 The rule of exponents
to predict clearance in man for 50 drugs

Methods % Mean absolute error

(MAE)
Simple allometry 106

CL x MLP 40

CL x brain Weight 49

Rule of exponents 25

58
Mahmood In interspecies pharmacokinetic scaling 2005 pp49
 A Comprehensive Analysis of the Role of
Correction Factors in the Allometric
Predictivity of Clearance from Rat, Dog, and
Monkey to Humans

RAKESH NAGILLA, KEITH W. WARD

103 compounds investigated

Standard allometry and allometry including various correction factor

(MLP, brain weight, GFR) were performed

Scaling were performed on all compounds universally and on

segregated subset based on allometric exponent, clearance,
physicochemical properties etc

776 allometric combinations with 27913 outcomes were preformed

A predicted-to-observed clearance ratio of 0.5 to twofold was

preselected as the criterion for predictive success
59
Nagilla & Ward JPS 2004 60
 No correction MLP

Brain weight Rule of the exponents

61
Nagilla & Ward 2004
 A Comprehensive Analysis of the Role of Correction
Factors in the Allometric Predictivity of Clearance from
Rat, Dog, and Monkey to Humans

When all three species were utilized in scaling

using simple allometry, 48 of 103 compounds
yielded a ratio (predicted/observed) that was not
within twofold of the observed value
Incorporation of the empirical correction factor
MLP or brain weight, either universally or
judiciously according to the rule of exponents,
failed to improve the predictive performance of
the method.
62
 A Comprehensive Analysis of the Role of Correction
Factors in the Allometric Predictivity of Clearance from
Rat, Dog, and Monkey to Humans

The success rate of allometric scaling

ranged from 18 to 53%
None of the correction factor resulted in
substantially improved predictivity
None of the methods attempted in this
study achieved a success rate greater
than that observed by simply estimating
human clearance based on monkey
hepatic extraction
63
Influence of species, routes of elimination and correction factors
% outliers

Nagilla & Ward 2004 0.5-to twofold window 64

 Value of the allometric approach
Conclusion: the prospective allometric
scaling , with or without correction factors,
represent a suboptimal technique for
estimating human clearance based on in
vivo preclinical data
Nagilla & Ward J Pharmac Sci 2004 1 2522-
2534

66
See also Obach & al for the value of
allometry as a predictive tool

67
Correction factors for renally and biliary
excreted drugs
Renally excreted drugs
Clearance / GFR aBW b

Biliary excreted drugs

Cl Bile _ flow aBW B

Cl UDPGT aBW b
UDPGT=UDP-glucuronyltransferase activity 68
 Interspecies scaling of clearance
1. Simple allometry
2. Allometry with various biological correction factors
1. Product of maximum life-span (MLP) and clearance
2. Product of brain weight and clearance
3. Ratio of clearance and GFR
4. Two-term power equation
5. Incorporation of molecular structure
parameters
6. incorporation of in-vitro data in in-vivo clearance
7. Correction for protein binding

69
 Incorporation of molecular structure
parameters
Wajima et al. 2002 suggested to use descriptors of
drugs related to clearance to predict clearance in man
e.g.:
Molecular Weight ,Calculated partition coefficient (c log P;
Number of hydrogen bound acceptors (Ha)).
Then using some types of regression (multiple linear
regression analysis, partial least square analysis or
artificial neuronal network), a regression equation can
be derived to predict clearance in man:

Log (CLman ) Log (CLrat ) Log (Cldog ) MW Hydrogen _ bounding ....

70
 Interspecies scaling of clearance
1. Simple allometry
2. Allometry with various biological correction
factors
1. Product of maximum life-span (MLP) and clearance
2. Product of brain weight and clearance
3. Ratio of clearance and GFR
4. Two-term power equation
5. Incorporation of molecular structure
parameters
6. Correction for protein binding
7. incorporation of in-vitro data in in-vivo
clearance 71
 Correction for protein binding
Protein binding varies considerably among
animal species which in turn can influence the
distribution and elimination of drugs
Theoretically unbound clearance should be
predicted with more accuracy than the total
clearance but in practical terms this is not the
case (Mahmood, 2005)
Actually, the correction for binding simply adds
more variability to the unbound clearance of the
species

72
 Interspecies scaling of clearance
1. Simple allometry
2. Allometry with various biological correction factors
1. Product of maximum life-span (MLP) and clearance
2. Product of brain weight and clearance
3. Ratio of clearance and GFR
4. Two-term power equation
5. Incorporation of molecular structure parameters
6. Correction for protein binding
7. incorporation of in-vitro data in in-
vivo clearance
73
 Dose for an hepatic cleared drug with a low
hepatic ER and a total absorption

V max
Dose fu ECtherapeutical
Km

The plasma protein binding and metabolism activity are the major
determinants for the elimination of low hepatic clearance drugs;
therefore it is not expected to have a good allometric relationship with
BW across species for this kind of drug as it is the case for antipyrine
( the Clint of antipyrine in man is only one-seventh of that which would
be predicted from other species)

74
 Incorporation of in vitro data in in vivo
clearance (Lav et al. 1997)
Clearances are normalized with in vitro data
providing a more rational (mechanistic) approach for
predicting metabolic clearance in man

CLhuman ( hepatocytes )
Cl animal a BW b
Cl animal ( hepatocytes )

For 10 extensively metabolized compounds, adjusting the in vivo

clearance in the different animal species for the relative rates of
metabolism in vitro dramatically improved the prediction of human
clearance compared to the approach in which clearance is directly
extrapolated using BW
Lave et a., J Pham Sci., 1997, 86: 584-590 75
 Interspecies Scaling of Bosentan, A New Endothelin
Receptor Antagonist and Integration of in vitro Data into
Allometric Scaling
Thierry Lave, Philippe Coassolo, Genevive Ubeaud, Roger Brandt, Christophe Schmitt, Sylvie Dupin,
Daniel Jaeck ane Ruby C. Chou - Pharmaceutical Research, 13(1), 1996

Cl human _ hepatocytes
Cl animal _ in vivo
Cl a BW b
Cl animal _ hepatocytes

R2=0.525 R2=0.976
Predicted human clearance=196ml/min Predicted human clearance=100mL/min
77
 Hepatocytes vs microsomes

Absence of phase II metabolism on liver

microsomes, which could result in enzyme
inhibition due to the accumulation of the
oxidative metabolites

78
Incorporation of in-vitro data in in-
vivo clearance
Methods %MAE

Simple allometry 164

CL x Brain Weight 61
In-vitro method 40
Rule of exponent 38

Data of Lave al (J Pham Sci 1997 86 584-590) on 10 extensively

metabolised drugs reanalysd by Mahmood 2005
81
Extrapolation of bioavailability

82
 Bioavailability in man:
prediction from rodents, primates & dogs ED%

Clearance x target EC50

ED50 =
Bioavailability

83
Absorption & Bioavailability
(F)

%F fabs (1 fg ) (1 ERH )

where
fabs = fraction absorbed from GI
lumen
fg = fraction metabolized by GI tissue
ERH = hepatic extraction ratio,
equivalent to hepatic first pass
84
effect
 Bioavailability in man:
prediction from rodents, primates & dogs

85
From Grass ADDR 2002
Extrapolation of Vss

87
 Interspecies scaling of volumes of
distribution (Vd)
fup
Vss Vp Vt
fut

Where Vp, is the volume of plasma; Vt is tissue

volume and fup and fut are the fraction of unbound
drug in plasma and tissues respectively
Usually a change in fut has a greater effect than
fup on Vss

88
 The minimal volume of distribution is
7.5 L (0.1 L/kg)
VD = 7.5 + 7.5 x fu + 27L x fu
p

fuT
Drug highly
Volume of
bound to No partitioning
distribution of
plasma protein No tissue binding
albumin
fu=very smal
V = 7.5 L (not 3 L) which is the VD of albumin

Note: plasma volume = 3 L but plasma protein (and

drug) diffuse out of vascular space and thus protein
(and drug) will return through the lymphatic system
89
 Interspecies scaling of volumes of
distribution (Vd)
Because there is no allometric relationship
between protein binding and BW, it will be
difficult to project the Vd of drug in humans from
data in animals
When a drug has a low binding to plasma and
tissue proteins or when a drug only distribute
extracellularly, the Vd of the drug reflect total
body water or extracellular water
In these cases, the Vd in human can be predicted
from data in animals because both the total body
water and extracellular water decrease as animal size
increases in an allometric manner.

90
Volume of distribution of propranolol

Vtotal
Vfree (Unbound)

For propranonol, Vf should be similar in humans and other species

However this is not a general rule (e.g. large difference for Vf between 91
species for Beta-lactam antibiotics)
 Interspecies scaling of volumes of
distribution (Vd)
Vc is the most important volume parameter
which can be predicted with much more
accuracy than Vss or V
The exponent of all three volume revolve around
1.0 indicating that there exist a direct
relationship between BW and volume
Correction for protein binding is not much help in
improving the prediction of vomume in man

92
Extrapolation of half-life

93
 Interspecies scaling of elimination
half-life
Application of HL to the first time dosing to
man is limited
HL is an hybrid parameter (clearance and
Vd)
Conceptually, it is difficult to establish a
relationship between HL and BW
Unlike clearance and Vd , the correlation
of HL with BW has been found to be poor
94
 R2=0.14
Allometric analysis of
HL
ciprofloxacin half-life,
clearance and volume
of distribution across
CL
R2=0.90 mammals
Poor correlation for
HL while correlation
for CL and Vss are
R2=0.94
VD
good

95
96
 Prediction of drug clearance in
children from adults

Origin of the difference between children

and adults
Variation in body composition
Difference in liver and kidney function

97
Age-related changes clearance

Morphine Fentanyl

98
 Prediction of drug clearance in
children from adults

41 drugs considered
124 observations in children of different
age groups
Infant, children, adolescent (from 1 day to
17 years)

Mahmood BJCP 2006

99
 Tested models
1. Classical allometric equation with different exponents
0.75 _ or _ 0.80 _ or _ 1.0
BWchild
CLin _ child Cl adult
BW adult

2. Correction of adult clearance by the estimated liver and

kidney weight in children

3. The clearance were estimated using a specific method for a

given age (decision tree)
Child<1year: exponent=1
Child >1 years but <5 years: correction by liver and kidney weight
Child >5 years : allometric exponent of 0.75, 0.80 or 0.85

Mahmood BJCP 2006 100

 Results
1. No single method was suitable for all
drugs or for all age groups
2. The %RMSE i.e. (MSE)0.5 was almost
similar for exponent 0.75, 0.80 and 0.85
as well as the approach based on the
liver and kidney weights
3. The lowest RMSE was seen with the
mixed approach
Mahmood BJCP 2006 101
Percent root mean square (RMSE) and percent error in the
prediction of clearance in children by several methods

Number of predictions in error

(>100%) for 124 predictions

Tested Exponents: 0.75, 0.89, 0.85 and 1.0

L+K: liver and kidney weights correction Mahmood BJCP 2006
102
Mixed : decision tree based upon age
 Children <1 year old

The exponent 0.75 overpredicted the

clearance by several folds
When exponent 1.0 (no exponent) was
used on the BW the prediction of
clearance was fairly reasonable and far
less erratic than 0.75

Mahmood BJCP 2006

103
 Children from 1 to 5 years old

The best approach appears to be the liver

and kidney weights corrections

Mahmood BJCP 2006

104
 Children >5 years old

One can use any exponent:

(0.75, 0.80 or 0.85)

Mahmood BJCP 2006 105

Allometry in veterinary medicine

106
107
108
 Conclusions:
Advantages of interspecies PK scaling

Simple and easy to use

Require plasma concentration-time data from
which PK parameters are calculated
Knowledge of elimination pathways, and plasma
protein binding may be helpful but not necessary
Data analysis is short
80% success rate if incorporation of hepatocytes
information's

109
Limits of allometic scaling

110
111
Limits of allometric scaling

112
For more information, consult the
Mahmood book

113