Journal of Pharmacokinetics and Biopharmaceutics, Vol. 22, No.

2, 1994

The Bateman Function Revisited: A Critical

Reevaluation of the Quantitative Expressions to
Characterize Concentrations in the One Compartment
Body Model as a Function of Time with First-Order
Invasion and First-Order Elimination

Edward R. Garrett ~

Received June 8, 1993

The Bateman function, A" (e-k't _ e-k~t), quantities the time course of a first-order invasion (rate
constant 1%) to, and a first-order elimination (rate constant 1%)from, a one-compartment body
model where A" = ( TDose)k~/ (ka- 1%)V. The rate constants (when 1%> 31%) are frequently deter-
mined by the "'method of residuals" or 'feathering. '" The rate constant 1% is actually the sum of
rate constants for the removal of drug from the invading compartment. "'Flip-flop," the interchange
of the values of the evaluated rate constants, occurs when 1%>3k~. Whether -1% or -1% is
estimable from the terminal In C - t slope can be determined from which apparent volume of
distribution, V, derived from the Bateman function is the most reasonable. The Bateman function
and "feathering'" fail when the rate constants are equal. The time course is then expressed by
C = 7Dtk e -kt. The determination of such equal k values can be obtained by the nonlinearfitting
of such C-t data with random error to the Bateman function. Also, rate constant equality can be
concluded when l/t~,ox and the km~, (value ofke at the minimum value of ek~t~/1%plotted against
variable Ice values) are synonymous or when k,~tm~x approximates unity. Simpler methods exist
to evaluate C-t data. When a drug has 100% bioavailability, regresswn of Dose/V /C on A U C /
C in the nonabsorption phase gives ke no matter what is the ratio of m=1%/1%. Since 1%t,,~x=
In m / (m - 1 ) , m can be determined from the given table relating m and 1%t,~x. When y is
unknown, ke can be estimated from the abscissas of intersections of plots of Cm~ ek't~ and
k ~ U C , both plotted vs. arbitrary values ofke, and ?D/V values are estimable from the ordinate
of the intersection. Also, when 7 is unknown, Ice can be estimated from the abscissas of intersections
(or of closest approaches) of ek't~'~/ke and A U C / C ~ , both plotted vs. arbitrary values of 1%.
The C - t plot of the Modified Bateman function, C = B e -x2t - A e -xtt, does not commence at the
origin (i.e., when tc=o=0 and when a lag time does not exist). However, tc=o=ln(A/B)/
(~.~-,~.2) when A > B . AUC A~ without time lag is the same as AUC Ar and A " = B e - ~ =
A e -aw'. The t , ~ of the C-t plot of the latter is tc=o later than the t,,~ of the C-t plot of the

JWe regretfully announce that Dr. Garrett passed away on October 25, 1993, after an extended
illness.
2Send reprint requests to The Beehive, Department of Pharmaceutics, College of Pharmacy,
University of Florida, Gainesville, Florida 32610-0494.
103
0090-466x/94/04(0-0103507.00/09 1994PlenumPublishingCorporation
104 Garrett

former which commences at t=O. However, (AUMCA~n,,)~o=B/AJ-A/gt2 differs from

(AUMC~o~B) = A " t c = o ( 1 / A 2 - 1 / ~ ) + A"(I/g~ - l/g2). (AUMCA~,) = A " ( 1 / ~ - l/;t 2) when
C-t plots start at t=0. AUMC..~o,r zs not valid. The (MRT~.co,,)~o ts also an invalid MRT
A , # B 9 . A # B 9 9 9

( B / g 2- A/;tt ), but when A > B, C-t curves which start at the origin,
estimate, (B / ;L2 - A / ;~2) / etr176
Ct =o, have MRT values displaced by t~~ o, i.e. . .1.i.A.D. .T Aco~
# B _ _ l t I D T t A ' o r A' = A = B J j _ r
- . . . . . . . . c= o. The t~=
o f the Bateman function is also displaced by tc=o when the A exceeds the B o f its modifiedform.
Dose-dependent pharmaeokineties can be concluded from C-t data generated by various first-
order invading nonintravenous doses i f drug absorption is 100% The k. values can be determined
i f the apparent volume o f distribution o f the one-compartment body model is known. Plots o f
m/AUC~" vs. time t have a slope o f - CLME, (the negative o f the clearance o f the metaboBte)
and an intercept o f the clearance o f the precursor, CLpM, provided that all o f the precursor had
been absorbed. Similar studies could determine the apparent volume o f distribution o f the metabo-
IRe and the clearance (and thus the rate constant, ke~=CLpM/Ve) o f the precursor to the
metabolite.

KEY WORDS: Bateman function; modified Bateman function when A #B; feathering; flip-
flop ka, ke, when k values are equal, simplified pharmacokinetic methods;
A rrt'A'~s - ~ A r r C A " ~ - A r r C ['~=m" (lag time), t. . . . Cmax, AUMC[.~=n o, 4"1.
#B ~B ~ [ A = B o r A']
MRT~.~o~,~# MRTA~,~ - MRT + tc~o.

INTRODUCTION
As a consultant and reviewer (or referee of manuscripts), I have fre-
quently encountered the inflexible and erroneous use of the Bateman expres-
sion to represent first-order invasion to, and first-order elimination from the
one-compartment body model. The term "Bateman Function" (1,2a) is used
in Europe for the expression that characterizes these phenomena, named
after the person (1) who first developed it in the explanation of transitory
radioactive phenomena.
The quantitative Bateman expression for the time (t) course of concen-
tration, C, in an invariant apparent volume of distribution, V, of the avail-
able dose, yD = A0, with a first-order rate constant of invasion, ka, and a
first-order rate constant of elimination, k~

A k , VC ~ " E (1)
is (1, 2a, 3a, 4a)
C = [k~7"D/(ka - k~) Vl(e -k~ e -k'') (2a)
= A'(e -k" - e -k't) (2b)
where, when k . / ( k , - k e ) = l, .4o = Ao/V. Eq. (2) can be obtained by integrat-
ing (l, 3a, 3b, 4a, 5a)

dC= ( k a A / V - k e C ) dt (3)

A in Eq. 1 is the amount of drug (A0) not-yet-absorbed or removed

from the invading compartment at time t.
Bateman FunctionRevisited 105

k, is Actually the Sum of Rate Constants for the Removal of Drug from the
Invading Compartment, A
The invasion rate constant, k~, is actually the sum of all first-order rate
constants for the removal of drug from the invading compartment (4b,
6). For example, when the gastrointestinal contents are both absorbed and
eliminated by parallel first-order rate constants, k~ is the sum of the true
first-order constant of absorption into V and of all other parallel first-order
rate constants of removal from the gastrointestinal tract. To obtain the true
absorption rate constant, one would have to determine yk~ where Y can be
obtained from the ratio of the areas under the plasma concentration-time
curves from oral (po) and intravenous (iv) administration of the same doses
of the drug, when the pharmacokinetics of both are first-order and dose-
independent,
7 = (A UCpo/DOSEpo)/(A UC~/DOSE~) (4)

When kl = k, ~ 3k~ = 3k2 (4c)

If ka > 3kr (4c), the second exponential, e -k't, of Eq. (2) tends to vanish
with increasing time, t, while the first exponential, e-k~ still has significance.
Absorption is insignificant when C-t data shows terminal data attributable
solely to elimination. The velocity of the invasion of the intermediate com-
partment, kaA, is significantly greater than the elimination rate, ke VC, so
that e -k't approaches e-~~ 1/e ~1761/oo ~ 0 faster than e -k't. It follows that
the terminal phase of a semilogarithmic plot of the natural logarithm of the
concentration, In C, against time, t, can be related to the single exponential,
(e-k~ since
In C' = In C = ln{[k~Ao/(k~- ke) V](e - k ' t - ~ 0)} (5a)
t--~ O0

=In{Aok~-ke) V} -k~t (5b)

= In A ~ - k j (5c)
When ka> > >ke, ka/(k~-k,)~-l, and A~=dok~/(k~-ke)V and C ' = C
when the drug's absorption has been completed. Since the terminal C-t data
plotted semilogarithmically (natural) has a slope of - k e
C' = k,Ao/(ka - k~) V e -kct '~--( h o / V ) e -kct (6)
This is shown clearly by the semilogarithmic plot of Fig. 1 where the
terminal elimination rate constant, k~, is determined from the negative of
the slope (-k2) of the terminal data when there is no longer any significant
invasion (or absorption).
106 Garrett

10"

% . ' . . . ..... El
,.," ,' - , . . ,
v a
m

" C'-C
t-

%.

10" I I I 9
0 25 50 75 100
t

Fig. 1. Feathering [the method of residuals (3b, 4d, 7) applied to a semilogarithmic (natural)
plot of plasma concentration, C in ng/ml vs. time (t)] on the presumption of the validity of
the Bateman function [Eq. (2)] with first-order invasion (/ca) and first-order elimination (k,).
The example given here is: V=volume=20 L, Ao (bioavailable dose) = 100 mg, ke=0.01 (the
negative of the slope of the terminal phase). The extrapolated intercept of C is the natural
logarithm of kado/[(ka - ke) V l -~Ao/V if ka >>k~. If the antiiogarithmic values of C (solid line)
are subtracted from the antilogarithmic values of C' (dotted line), the negative slope of the
natural logarithm of the difference ( C ' - C, dashed line) is/ca = 0.05. If ke >>k~, flip-flop occurs
and the respective slopes of the natural logarithms of C' and C ' - C are reversed and are k,
and k~, respectively.

The process of deducing the invasion rate constant is by the method

of residuals (stripping or feathering) (3b,4d,7), since the nonlogarithmic
formulation [Eq. (2)] of the orginal data, C, can be subtracted at any time
(t) from the nonlogarithmic formulation of C', the extrapolated terminal
phase [Eq.(5) or (6)] to yield
C' - C = {A6 e -k~ } - {Ao k~/(k~ - k . ) V} (e -k~ - e -k~') (7a)
= [k.Ao/(ke-k~) V] e -k't =.4~ e -k't (7b)

Thus,
l n ( C ' - C) =In A 6 - k ~ t (8)
so that a semilogarithmic plot of the natural logarithm of the difference,
C ' - C , (Fig. 1) against time has a slope, (-kl), that is equivalent to the
negative invasion rate constant, -k~, when k~>>ke.
The apparent volume of distribution, V, of the one-compartment body
model considered herein is estimable from
V=Ao/Co=D~/Co (9)
Bateman FunctionRevisited 107

where the iv bolus dose, Div, to a one-compartment body model (the drug-
equilibrated body fluids) has,a presumed or extrapolated zero time concen-
tration of Co. A rougher estimate of V can be obtained from the intercept
of the In C' line vs. time (Fig. 1), i.e., In Int, from which the volume can be
estimated as
V= k.Ao/(k, - ke) Int (lO)
where
Int = k , A o / ( k , - k~) V (11)
When the rate constant of invasion,/ca, is very much greater than that of
elimination, k~, in the function [Eq. (2)], i.e., k , > > >k~, k , / ( k , - k e ) ~ - I
and
V_~ Ao/Int (12)
It is obvious that the intercept, Int, to permit the estimation of the
apparent volume of distribution o f the invaded compartment, V, [Eq. (12)]
may frequently overestimate V since k , / ( k , - k e ) in Eq. (10) is less than
unity when k, is not sufficiently larger than k..
When the invasion rate, k,A, does not greatly exceed the elimination
rate, k, VC (or/ca does not greatly exceed k~) invasion continues while a
great deal of elimination is in progress. Linear approximation of the terminal
data of semilogarithmic plots (natural) of concentration in the invaded com-
partment against time tends to underestimate the k, and overestimate any
corresponding half-life, tl/2 = 0.693/k~.

Flip-Flop, when ke ~ 3k. (2a,4c)

Although invasion rate is usually larger than elimination rate (4e), an
unusual phenomenon occurs when k,<<3ke (4c). It is/ca and not Ice that is
determinable from the terminal slope (~2) of semilogarithmic plots of the
concentration of the invaded compartment against time. A particular
example would be when slowly releasing intramuscular formulations of pen-
icillin (a drug actively tubularly secreted and of short terminal half-life on
iv administration) are administered. In these cases, e -kct vanishes with time
before e -k't does when ke>>k,. The natural logarithm of terminal data
becomes
In C' = In C = = ln[k,Ao/(k~ - k,t) V'] e-k" (13a)
t --~ O0

~_ln(a [ / V) - k,t (13b)

where
A~=k,Ao/(ke-k,) (14)
108 Garrett

when ke>>ka. After feathering, i.e., subtracting the values of C from the
values of the extrapolated C' at the same time
l n ( C ' - C ) = I n A~/V-ket (15)
and the elimination rate constant, ko, can be determined from the negative
of the slope (k0 of the natural logarithm of the feathered values plotted
against time (Fig. 1).
There are alternatives (8) to the method of residuals (3b,4d,7). Ao/V
values (quotients of the bioavailablr dose, A0 = ?Dose, and the apparent
volume of distribution, V) can be calculated from Cmaxek~ 9Excellent first-
order elimination rate constants, kr can be obtained for the one-compart-
ment body model with first-order rate constants of invasion (k~) and elimina-
tion (k~) from the regression of (Ao/V)/C on A UC/C no matter the ratio
of m=k~/k~ when absorption is negligible. V/y can be calculated from
CL(clearance)(MRT-1/k~) where CL=Ao/AUC. The values of k~
(abscissa) and V(ordinate) can be determined from the intersection of CL/ke
and C L ( M R T - I/k~), both plotted against arbitrary kr values. The ka values
can be estimated from mko or from 1/MATwhere M A T = M R T - 1/kr (8).
When ? is unknown, kc can be estimated from the abscissa of the intersection
of plots of Cmaxe~~ and k~AUC (where the ordinate estimates Ao/V)
(Method A) or from intersections of ek~ and A UC/Cmax (Method B),
all plotted against arbitrary ko values. The relation between kr and m is
kdm~=ln m / ( m - 1 ) and m can be estimated from a table based on this
equation (8).
The curves frequently do not intersect at low m values because errors
introduce inconsistencies. Under such conditions, the abscissa of the closest
approach of one curve to another (or the minimum of the difference) pro-
vides the best estimates of kr particularly with Method B.
When the terminal rate constant, ;t2, is assigned tothe elimination rate
constant, ko, and k~ > 3ko, then
V=~.,Ao/(21 - E2) Int (16)
whereas, when ko > 3ka
V' = ~.2Ao/( 3.~- )~2) Int (17)
The terminal rate constant rate constant, ;t2, as either k~ or k~, could
be based on the reasonability of the estimated apparent volume of distribu-
tion [Eq. (16) or Eq. (17)]. If V [Eq. (16)] were anomalously less than the
plasma volume, a value expected if there were total plasma protein binding
of the drug [e.g., plasma volume in the human is ca. 0.6 (plasmacrit) x 75
(blood volume, ml/kg) = 45 (37-59) ml/kg (9) the hypothesis ofke > k~ could
be rejected. The expected apparent volume of distribution of the total body
Bateman Function Revisited 109

fluids in humans would be 240 (210=290) ml per kg (10) if there were no

plasma protein binding. Plasma volume in the dog is 55 (44-73) ml per kg
(9) and its expected apparent volume of distribution of total water is 300
(240-350) ml/kg (10)1.
In the example of Fig. 1 it is presumed that ka = 5ke and the rate constant
ke= 0.01, as can be estimated from the negative of the slope of the terminal
(natural) logarithmic concentrations vs. time. The - k a = - 0 . 0 5 from the
slope of the feathered curve. The estimated apparent volume of distribution
(I0 of the one-compartment body model is 20 L [Eq. (16)], close to the
19 L of the total of equilibrated body water expected in an 80 kg human
(10), the expected V when the drug is distributed among body fluids and is
not bound to plasma protein. If the negative of the slope of the terminal
phase were due to flip-flop and k~ = 0.01, the apparent volume of distribution
would be V ' = 4 L [Eq. (17)]. This is approximately equivalent to the true
plasma volume and would be improbable if the drug were not completely
(100%) plasma protein-bound. The volume, V, could be greater if the
unbound drug were sequestered by body tissues or solubilized in body fat.
These facts could deny or affirm the possibility of flip-flop.

and (40
Two fundamental characteristics of the Bateman function are the maxi-
mum concentration in the invaded phase, Cmax, and the time tmax, that it
occurs. If the Bateman function [Eq. (2)] is differentiated with respect to
time, it can be shown, after proper rearrangements of the results that (4e)

t ~ x = [ln( ka/ke) l / ( k a - k~) (18a)

= In m / [ k e ( m - 1)] (18b)

Since the initial concentration, Co = [~,D (or Ao)]/V, decays monoexponenti-

ally to (?maxat tmax

C,,~x = ([TO(or Ao)]/V) e -k~ (19)

and since, when Eq. (18a) is recognized

In e -k~ . . . . . ketmax = - k e l n ( k a / k e ) / ( k a - ke) (20)

it follows from Eq. (19) that

(?max= = [TO(or do)]/V(k~/ke) -k~/~k'-k~ (21a)

110 Garrett

and, since rn = ka/ks and -kctmax = - k , In m/(k~-Ice) = - k , In m~

( k , ( m - 1) =In m/(1 - m )
Cm~x= ([yD(or .40)]/V)m 1/~ -m) (21b)
where
-kjm~x = ( k a / k s ) k , / ( k a - k~) = In Cma~- In A o / V (22)
Thus, Cm~ [Eqs. (19), (21a), and (21b)] can be calculated for the one-
compartment body model with first-order invasion and elimination rate con-
stants from knowledge of the bioavailable dose, (A0), the apparent volume
of distribution, (V), and the known ratio, m, of k~ to ks.
Plots of In Cma~of the derived linear logarithmic expression
In (?max= ln[(TD(or `40)/V] - k,t,,~ (23)
against tm~ [Eq. (23)] pass through the In Cm~, values for all/ca values
including when k~ =k, when the k, value of the Bateman function [Eq. (2)]
is held constant (10). 3
If a straight line is drawn through In Cma~at tm~ with a -ko slope, it
has an intercept on the In C vs. time plot of In Ao/V, the natural logarithm
of the quotient of the bioavailable dose (`40= 7D) and the one-compartment
body model apparent volume of distribution (V) (8). Thus, if V and the
negative slope (-k,) of the appropriate logarithmic (natural) phase are
known from a n iv dose, Di~, the bioavailability of the first-order invasion
can be estimated as 7 = (`4o/Div).
The rearranged natural logarithmic form of Eq. (21b) when solved for
the natural logarithm of m is (8)
In m = ( m - 1)(In . 4 o / V - l n Cm~) (24a)
Realizing Eq. (22)
In m = (m - 1)kjm~ (24b)
or

kjma~ =In m / ( m - 1) (24c)

so that the ratio, m, can be estimated from parametric equations [(Eqs.
(24a) or (24b)] (see Fig. 2). From the Cm,x of the invaded compartment
(e.g., plasma) and the ratio of the bioavailable dose, .40, to the apparent

~l'his section and many of the others following that are unreferenced were based, in part, on
an unpublished manuscript "Basic Pharmacokinetics" by Edward R. Garrett that was used
as a text for many pharmacokinetic workshops given by E.R.G. in the United States, England,
Egypt, France, and Germany.
Bateman Function Revisited 111

1.9-
1.7
1.5
1.3- A)
1.1
0.9
0.7
0.5- B)
0.3-
0.1
--0.1 I . . . .
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 #.5 5.0 5.5 6.0
M
Fig. 2. Determination (8) of m=ka/k== 5 and k~=0.05 from parametric Eqs. (24a) or (24b).
Curve A is a plot of In m vs. arbitrary m values and curve B can be calculated from (m - l)kjma,
or (m- 1)(ln(Ao/V-ln Cma~)and plotted against the same arbitrary m values. The specific
example given is for data supporting Fig. 1 where tma~=40.24, k~=0.01, Cmax=3.344and
Ao/V= 100 rag/20 L=5 ng/ml.

volume of distribution, V [Eq. (21)], the value of m can be estimated. When

the calculated values [Eq. (24a) or (24b)] are plotted against arbitrary m
values (line B in Fig. 2) the line intersects the values of In m plotted (line A
in Fig. 2) against the same m value to permit the proper estimate of m from
the abscissa of the intersection. Thus ka = mk~ is calculable from this k~. The
calculated values [Eqs. (24a) and (24b)] (line A in Fig. 2) plotted against
the various m values, also intersect the value of In m at the same abscissa
value (Fig. 2). An alternate procedure is to determine m by interpolation in
a table that relates m and ketmaxvalues (10) based on Eq. (24c). Even though
insufficient numbers of lower concentration values to determine a terminal
phase ke were unobtainable from the invaded compartment, k~ could be
estimated from the terminal slopes of ln(A U/At) or ln(Uoo - U) against time
when the terminal values of U (amount of drug excreted in the urine or the
amount metabolized at time, t, by first-order processes) were of sufficient
magnitude to permit the proper estimation of k~.
The plotted data of Fig. 2 were based on the parameters of Fig. 1. It
is interesting to note that the A and B curves always also intersect when
m = 1 in Eqs. (24a) and (24b), since In m is 0 when m = 1 or ka = k~.

W h e n Rate Constants of First-Order Invasion and First-Order Elimination

are Equal, ka = k= = k, m = kalk~ = 1

Feathering of the Bateman function [Eq. (2)] is inoperative when ka =

k~ = k (2b); semilogarithmic plots of concentration, C, against time are not
112 Garrett

terminally linear (Fig. 2). The equation that then best describes the time
course of concentration in the invaded compartment is (2b); semilogarithmic
plots of concentration, C, against time are not terminally linear (Fig. 2).
The equation that then best describes the time course of concentration in
the invaded compartment is (2b)
C =.4okt e - k ' / V (25)

Even though one rate constant is not more than three times another, their
values can frequently be estimated by curve fitting to the Bateman function
[Eq. (2)] with computer-based nonlinear regression techniques using least-
squares criteria curve fitting (11,12) such as Metzler et al.'s NONLIN (13)
or Yamaoka et aL's nonlinear regression fitting (14), provided that small
differences between estimated k~ and ko values were contributed by r a n d o m
assay error.
Bialer (15) proposed a simple method to determine whether first-order
absorption and elimination rate constants were equal in the one-compart-
ment open model with first-order processes so that the proper equation
[either the Bateman function of Eq. (2) or Eq. (25)] could be chosen to
characterize the relevant pharmacokinetic parameters (4c).
The determination of a relation for t,,~ can be effected by maximizing
Eq. (25) [differentiating Eq. (25) with respect to time, t, equating to zero
and solving for t]. Thus
tm~ = 1/k = 1/k, = 1/k2 (26)

and solving Eq. (25) for Cmaxwhen t = tm~x= 1/k results in

C ~ = AolVe (27)

where e is the base of natural logarithms. Thus

Cmax/max = .40/].s V e = Ao/eCL = `4 UC/e (28)

where CL (clearance)=k V (4i) and .4 UC= .4o/CL. Thus, when the product
of two experimentally observable values, Cmaxand tmax, can be shown to be
equal to .4 UC/e, the rate constants should be equal and the Bateman func-
tion [Eq. (2)] should be indeterminate. Eq. (25) must be used to calculate
plasma concentration-time values.
Barzegar-Jalali and Toomianian (16) have criticized the Bialer proposal
(15) by demonstrating that rate constant ratios ($,/~,2) as widely divergent
as 1.6 and 1.0 generated relatively equivalent respective products, Cmad. . . .
which were 99.3% and 100% of the corresponding .4 UC/e values for error-
free data so that the use of Eq. (28) "may not be a suitable criterion for the
Bateman Function Revisited 113

determination of the equivalence" (of the rate constants). They stated that
their theoretical mathematical derivations supported their assertion. [How-
ever, their definition of C,~ax in their Eq. (6) does not make sense; substitu-
tion of tmax= t into their Eq. (1) for Cmax does not lead to their Eq. (6)
which merely defines the relation between the C,,ax and tm~ values when
values are maintained constant and ~.1 values are varied, our Eq. (19)].
Recent criticisms (17) of the Bialer premise have also been made.
Our experience with nonlinear fitting of the Bateman function with
formulated slight differences in the rate constants led to estimates of their
equivalence by Yamaoka et al. (14) nonlinear fitting. Chan and Miller (11)
and Patel (12) concluded similarly when the NONLIN program (13) was
used.
Patel (12) proposed the "Concentration Ratio Method" to determine
the rate constant k. The reciprocal of the ratios of two consecutive plasma
concentrations, defined by Eq. (25), are solved for k values and several such
sequential values can be averaged. If the variation among these values exceed
random error, it cannot be concluded that k~ = k2 ~_k.
Patel (12) had stated that pharmacokinetic analysis of the data by
NONLIN, nonlinear fitting to the Bateman function, Eq. (2), was the "only
method k n o w n . . , to identify the equality" of the rate constants and to
obtain such a k. Nevertheless, identification of the equality of rate constants
of invasion and elimination can be effected by observing that k = 1/tm~ and
that k~in (value o f k at the minimum value ofe~rm'x/k plotted against variable
k values) and tm~x are synonymous [or the value of kmi~tm~ approximates
unity (8).

tin; W h e n Co = 0 After the Origin, t = 0

Curves generated by the Bateman function [Eq. (2b)] pass through the
C-t origin. Frequently, the time course of concentration in the invaded
compartment [Eq. (1)] is best characterized by the difference between two
exponentials in the Modified Bateman function:

C = B e - x 2 ' - A e -~'' (29)

and only when A = B = A' can we have the true Bateman function [Eq. (2b)].
and

(30)
114 Garrett

1 50~r /
c -Z--3
. uucA B _ _ -

lOOt / I /#..~..~.7-- unr

x'
9 " . , ,

..- " Auo~A;; or A=I1]

opu c oo
-251-/!/' __

-75 __ 75
-100 -5O
C . . . . . .

-125- ~ " " ~ -25

L C/i '
I I 0
- 150 0"~la0 25-~iag 50 75 100 125 50
Minutes
Fig. 3. Plots against time of invaded compartment concentrations (C in #g/ml, solid lines
constructed from Eq. (29) upper panel when A > B ; lower panel, when A = B = A ' or when A ' =
A", with a first-order invasion and a first-order elimination. There is a tl,s of 7.7 min in the
upper panel and none in the lower [Eq. (31)]. The t,,,x = 33 min in the upper panel and Cmax=
131 in both panels. The product of concentration and time, [C x t, (mg/ml) x hr], are represen-
ted by the dotted lines9 The cumulative" areas A U C uA~n A'=a~Bor,~'=.l" , under the C
n c o r r o r A UCco~r
vs, t curve [Eq. (34) or Eq. (35)] are, respectively, plotted against time (dashed line). The
uncorrected area under the C x t p r o d u c t - t i m e first moment curve [AUMC~:~r~, (gg/
ml) x hr 2, Eq. (44a) or (44b)] are given (x 10-3) against time as the dot-dashed curves and for
a brief 7.7 min lag time A U M C 2 ~ , is virtuall~ the same as A U M C ~ B at the scaling shown
and not too discrepant from the A U M C A=B=A o~=#, in Fig. 5A. The curves of the lower panel
(use the right ordinate values) are for C-t data transposed to the origin (without lag time, i.e,,
tc=o=t~,g=O) where an area (9168) equivalent to the A UC under the C-t curve and above
C = 0 is calculable from Eqs. (36); A" can be calculated from Eq. (37). Note that the t~x
values of the C-t plots of the two panels differ by h,g=tc=o=7.7 rain. The dashed curve in
the lower panel is for 10- 3 x A UMC A '= B = A orA '~ .4". Note that the calculated concentration (C)
in the upper panel before tc=o when A > B is negative [Eq. (34)]. The calculated uncorrected
cumulative A UC~7~B is also negative [Eqs. (44)], even beyond tc=o = h g = 7 . 7 rain.

When A # B in Eq. (29), a plot of invaded compartment concentration,

C, against time does not pass through the origin (Fig. 3, upper). When
B>>A, the plot has finite concentration (C,=0) at time zero. This implies an
immediate introduction of a bolus of drug into the invaded compartment
(e.g., an amount of drug is instantaneously released from an oral, intramus-
cular, parenteral, or prodrug dosage form and relatively rapidly absorbed
into the system circulation) to give a finite value, C,=0. Subsequently there
can be a first-order rate of absorption.
When A>>B, the plot of Eq. (29) (as in the example of Fig. 3, upper
as well as in Figs. 4 and 5b) would have a negative concentration value
Bateman Function Revisited 115

350
1.00~

300

_J
250 g ~
E
200 +\.. ~- I ",~

I
150 100

100

50

0 I I, , I ' I ' T , I ,
0.0 0.5 1.0uou
rs,1 1.5 2.0 2.5

Fig. 4. Plots against time of the average ibuprofen plasma concentrations (closed circles) from
10 rabbits (C in #g/ml, fitted by the appropriate parameters (solid line) of Eq, (29), i.e.,
Ai = 2.64 hr-i, A2= 2.06 hr-l and A = 2006/,g/ml > B = 1863 g g/ml) from an ibuprofen lysinate
suppository containing the totally bioavailable 146.3 mg of ibuprofen acid equivalents. There
is a t~,g of 7.7 rain [calculated from Eq. (31)and see graph] as calculated for the mean values.
The average calculated tla8 of the individual rabbits was 7.5+0.5 vain. Also plotted are the
mean plasma concentrations after intravenous administration of the same dose to 5 of the
same 10 rabbits. The cross bars and line extent above and below each plotted value give the
ranges of the standard errors and the standard deviations, respectively. The lack of cross bars
indicates that symbol size exceeds the standard error. The inset presents a semilogarithmic
plot against time of the Wagner-Nelson (12, 4i) derived unabsorbed drug fractions from the
suppository. The dashed lines are based on the determined iv elimination rate constant, k, =
2.23 hr -t which is approximately the k~=2.06 hr -t from the semilogarithmic fitted terminal
plasma concentrations.

(C,=0) when extrapolated to below zero time. This would imply a delay (lag
time) to a time, tc=o, in release and/or absorption of the drug. When Eq.
(30) is equated to zero, it can be readily shown that the lag
time, tc=o, (when the extrapolated concentration is zero), can be calculated
from

tt~g = t c = o = [In(A/B)]/(AI - Aa) (30

Some pharmacokineticists have preferred forcing data to fit the true
Bateman function, assuming A = B = A' as in Eq. (2), after a lag time before
the function commences (4g).
116 Garrett

12. /

10. I A)
I C x t-" IAUMC A=B
8. I" . II
I ." / #
6-
.~//// / #
f / A"
--
4- / AUCcorr
f
2.
9y . I / ~ / t tlag=tC=o=O

0 , ,,--, ~ ~ , ,~ ~ ~ , w
10.0
Cxt" //~.~:-
7.5 B• 9" ,,,c'Z'~-~,'2/ /
5.0
2.5
tc=o ~ rJr
0.0

--2.5"
--5.0-
\ ,,
--7.5- C~ I ~..=t,,_~=7.7,
flog = ~=o = 7.~'
--10.0 I I ,, h : : : : , ,
1.5

1.0-
c) G,i
9 ,,u~,,. I
i ~z I
0.5.
: d I .
0.0-
e

,, ",, :1 I
-0.5-
".,, 9I /, .-3,,.,,.,,r
"- :1 / ....... ""=""
-1.0. \.. : i j/
":~[" ~og - ~-o - ~-~'
-1.5 I I I '1 I I I I I F
0 2 4 6 8 10 12 14 16 18 20 22
( flog----7.7 ) Minutes
Bateman Function Revisited 117

tmax and Cm~ when A ~ B in the Modified Bateman Function, Eq. (29)
Peak invaded compartmental concentration (Cm~x) can be calculated
from Eq. (29) using the calculated t = t,,~:
tAm~n = [ln(AZ,/B2~)]/(Z, - Z~) (32a)
= [ln(A/B)]/(Z, - Z2) + [ln(Zt/g2)/(Z, - ~.2)] (32b)
= tc~0+ tm~x~=n (32c)
These equations are readily derived by differentiating Eq. (29) with
respect to time, equating to zero, solving for time and realizing Eqs. (18a)
and (31). When A = B=/1' and the plot of concentration against time (Fig.
3, lower) starts from the origin, -m~x
t a ' s ~:-the same as t ~ since t c ~ o = 0 which
is solely a function of the ~ values [Eq. (18)].
Cm~x of the Modified Bateman can be calculated from Eq. (29) when
t ~- lAmB
-m~ [Eq. (32)]. The Cm,~ of the curve transposed to the origin ( t c ~ o =
0) has exactly the same value when calculated [Eq. (31)] from t~m~x n=A' [Eq.
(18)].

Correctly Calculated Area Under the Curve (A UC) when A ~ B in the

Modified Bateman Function, [Eq. (29)]
The area under the concentration-time curve useful in bioavailability
estimates is above the C = 0 ordinate. It can be estimated by the linear
trapezoidal rule (4h) which, when the curve is approximated by successive
short straight lines, is
A UCtrap= [E (Ci~- C i - l ) ( t i - t,-i)/2] + C , / ~ 2 (33a)
However, when the curve best-fitted by Eq. (29) is best approximated
by successive exponential decays, i.e., when the curve has sharply ascending
or descending exponential changes with time, pronounced changes of curva-
ture with time, and/or the time intervals, t ; - ti- ~, are too long to adequately
approximate the curve by linear segments, the true mean concentration of

Fig. 5. Plots against time at early periods of plasma concentrations [C, as solid lines in panels
A and B ) / l g / m l , Eq. (29)] and of the calculated products of these concentrations and time
(C x t,/~g-hr/ml, dotted lines) for the suppository with the Modified Bateman A > B parameters
considered in Figs. 3 and 4. The calculated uncorrected areas in panels B and C under the
plasma concentration-time curve IAUC~f~,~ in (/~g/ml) x hr are from Eqs. (34) and (35)] and
under the first moment-time curve [AUMC~o~ in ( / l g / m l ) x h r 2 are from Eq. (44)]. The
dashed plots are the corrected calculations of A UMC[A U M C ~ B, Eq. (47)] and start from
C = 0 at t = tcffio= tlag. The A UMC~s and A UMC~rSvalues are practically indistinguishable
at the scales of panel B (Fig. 3) and that of Fig. 5. The plots in panel A are for C-t data
transposed to the origin (without lag time, i.e., tc=o = tmag-0).
118 Girt'ell

an interval differs widely from the average of the initial and final C values o f
the intervals and the area under the curve is best estimated by the logarithmic
trapezoidal approximation (4h, 7b):
A U C ~ . p = [ ~ ( C ~ - C ~ _ O ( t ~ - t , _ O / ( I n C~-ln C~_~)]+C./2, (33b)
The A UC must be zero at tc=o. Terminal areas can be estimated from
C./,12 where (7. is the terminal concentration in the invaded compartment
at time t. and ;% = 22 is the final elimination rate constant.
However, the calculated area, based on integration of Eq. (29), ifA #B,
can lead to erroneous A UC estimations since [as is apparent by inspection
of Figs. 3 (upper), 5B, and 5C] the uncorrected A UC is

and
A UC~,, =
;o C dt = [ B / ~ - A/~.] (34)

A trC..co,, = A trC' 3 - IAtrc,, j (35)

The latter term could be referenced as +A UC, c=o since A UC, c~ is negative
/IUCun~o,, may differ widely from A UCtrap [Eqs. (33a) or Eq. (33b)]. A
significant lagtime (tc=0), before the commencement of first-order absorp-
tion, leads to a calculated area [Eq. (34)] that is less by I A U C t c J (the
negative of the negative area from time zero to time tc=o) than the area
true
A UC~I, above the C = 0 ordinate (see Figs. 3 [upper], 5B and 5C). The
correct calculation of A U
~,C ~lc
true should be

A ~ ~ ~a~ -
"r'~true-
fo fo"
Cdt- Cdt=AUCu.o~,,-AUGc~~

= B/~a - A/~,, - [B(1 - e - k ~ ' ~ - A(1 - e-~'~c'~

(36a)

(36b)
= ( a e-*~'c=~ - (,4 e-*"~-~ (36e)
_~,,(l_•
\22 2,1 (36d)

where tc=o is calculated from Eq. (31) and tc=o=0 ifA = B (Fig. 5A) and
B e-X2tc'o=A e-~"c~~ " (37)
Thus Eqs. 36 and 37 show that Eq. (29) can be transformed to [Eq. (36d)]
with the same A UCt"fg as a true Bateman function with tc=o and A = B = A " .
In the ibuprofen lysinate suppository (18) where plasma pharmaco-
kinetics were fitted (Fig. 4) to the difference between two exponentials, the
Modified Bateman function [Eq. (29)], had an A UCu~o,, value of 145 (/zg/
Bateman Function Revisited 119

ml) x hr [Eq. (35)] which underestimated AUCealc t,~o= 153 (pg/ml) x hr [Eq.
(38)], or the similarly valued A UCt~o of the experimental mean values [Eqs.
(33a) or (33b)]. The A UCtr~p values would be greater than the calculated
true
A UC~Ic values if the areas estimated by the trapezoidal rule when A > B
were extrapolated to the origin of the concentration-time curve rather than
to C = 0 at tc=o.
T h e plasma concentration-time curve of Eq. (29) (Fig. 3) can be shifted
to the origin and transformed to plots without lag times (compare upper
and lower panels of Fig. 3 and panel A with panels B and C of Fig. 5)
on the premises of A' = A = B or A' = A" (Eq. 37). The relations developed
previously for the Bateman function [Eq. (2)] are then applicable.
It is apparent from Eq. (36d) that the explicit value of A" or A' is

A' or A" =A UCA"/[(1/~) - (1/A,)] (38)

title
where the A UC A" value is either A UCtrap [Eqs. (33)] or A UC~j~ [Eqs. (36)].
The tm~ values of such transformed plots are shifted by tc=0 = t~"
toward the origin [Eq. (32c) and Figs. 3 and 5].

Estimation of Absorption Rate Constant on Assumption of

Dose-Independent, First-Order Elimination: Wagner-Nelson Procedures
and Statistical Moment Theory and Calculated A UMC Values of
the Modified Bateman Function, Eq. (29)
One method to estimate the valid k, value can be based on the assumed
validity of the intravenously obtained ke value on the premise that the elimin-
ation pharmacokinetics is not dose-dependent. The Wagner-Nelson pro-
cedure (19, 4i) [Fig. (4)] permits estimation of the fraction of drug
unabsorbed

(39a)

= 1 - ([c,+ ~eAuc,]/~eA uc~) (39b)

where Ct is the concentration in the invaded compartment at time (t), A UC,

is the area under the concentration curve up to that time (t), A UCoo is the
total area under the curve up to infinite time and ke is the intravenously
determined elimination rate constant. The apparent first-order rate constant,
k,, of absorption can be estimated from the slope of the linear semilogar-
ithmic plot (inset, Fig. 4) of the fraction of drug unabsorbed, fu [Eq. (39)],
against time or by nonlinear regression (9) offu on time, t, in accordance
120 Garrett

with

A = ( A ) t ~ o e -~' (40)

This expression for the monoexponential loss of unabsorbed drug, f~,

permits the estimation of the hypothetical or apparent fraction, (f.),=0, of
drug that is unabsorbed at time zero, i.e., the start of drug administration.
When (f,),=o is greater than unity, there is lag time before monoexponential
decay begins. When it is less, a fraction [ 1 - (fu),=0] must be instantaneously
absorbed at t = 0. The time, tA = 1, when the fraction of drug unabsorbed
is unity can be obtained by solving the logarithmic forms of Eq. (40) for
the time when f~=

tA= I =In (f~)t=o/k. (41)

A statistically significant positive if.= ~, when (f~)t=0 > 1, is an estimate

of the time lag before the commencement of monoexponentially decreasing
unabsorbed drug. Obviously, no absorption occurs until fu decreases from
unity.
A statistically significant negative tf.~ j, when (f~)t=0< l, implies that
a drug fraction, l - (f~),=o, was instantaneously available to the invaded
compartment.

Calculated A UMC Values of the Modified Bateman Function Eq. (29)

An alternative method for estimating a first-order absorption rate con-
stant (k.) is by noncompartmental analysis based on statistical moment
theory (4j) where the mean residence time (MRT) of the drug in the invaded
compartment can be estimated from the quotient of the first moment
(A UMC) to the zero moment (A UC)

MRT- AUC Jo tCdt Cdt (42)

and both values of the quotient can be evaluated by the trapezoidal rule
(4h, 6b). The terminal value of the A UMC is calculable from (4j)

A UMC,_~~ = (t.)C./~z + C./$~ (43)

Integration of Ct with respect to time [C is given in Eq. (29)], since

S t e-~t dt=e-a~t( - ~ ' t - 1)/$~, yields
AUMCuncorr=A[e (Xd+ 1) 1]/$~-B[e-aet(~zt+ 1 ) - 1]$~ (44a)
Bateman Function Revisited 121

so that
A~B ~ 2
(A U M C , ~ , , , , ) ~ - B / ~ - A/~,~ (44b)
By analogy to the Modified Bateman A UMC~u~oB,~ [Eqs. (34) and (35)], Eqs.
A~s since the value of
(44) should urlderestimate the true A U M C , A UMC~o,~
A U M C , c.o, when A > B in Eq. (29), is a negative {~c-o Ct dt} at the time,
tc=o, (when the concentration C is zero).
There is a real disparity between the two A U M C values
A U 'M c'~ "~n = A U M C UlIr
~COI'I" A,, s _ A U
L.IM ~" ~" n
~IC~O (45)
where
f tc~oCt f tC~O
A UMC,A
c .~ os = dt = ( B t e -~'2t- A t e -x'~) dt (46a)
~'0 vO

= A[e-X"c=o(~ltc=o + 1 ) - 1]/Z~
- - B[e-X~tc'~ + 1) - 1] / ~ (46b)
The corrected or true A U M C from tc=o to infinity conforms to Eq. (45)
and should be equated to zero at times between zero and tc=o (i.e., tc=o
should be the variable t in Eqs. (45) and (47) at times less than tc=0) and
d~ B) ~ = B[e-Z~'c=~
(/i UgCcorr o+ 1)]/~,~ - d[e -xt tc=~ tc= 0 + 1)]/~,~
(47a)
and
(A U M C ~ n ) ~ = B e-X2'c-~ + B e-X2tr176 A e-X'tc'~
-A e-~"c'~ (47b)
When Eq. (37) is recognized
(A U M C ~ B ) ~ = A"tc= o/~a + A"/L~ - A"tc= o/~,1 - A"/~,~ (47c)
= A " t c = o ( 1 / A , 2 - 1/;h) + A " ( 1 / Z ~ - 1/&~) (47d)
-- t c ~ oArrctru9
~ o ~,c [Eq. ( 3 6 d ) ] + A " ( 1 / ~ - l / ~ , ~ ) (47e)
The A U M C A ' = A U M C ~ c T r n=~'~ (see A" in Eq. (37) when tc=o=0
and e - ~ c ' ~ = 1 and the Modified Bateman function shifts to the origin, so that
A U M C ~ T f l =A, = A U M C A"

= A " ( e - ~ ' t ( Z d + 1 ) - l)/~,~-A'[e-X~'(~at+ 1) - 1 ] / ~ (48a)

where
(A U M C A") ~ = A"( 1 / ~ - 1/A,~) (48b)
122 Garrett

In the suppository example (18) considered in Fig. 4 with a short tc~o =

hag of 7.7 min, A U M C ~ r [Eq. (44)]/A UMC~c~,~n [Eq. (47)] =0,9978 (see
Figs. 3 and 5) whereas A u n c t i o n , [Eq. (44)]/A UMC~7~n=A' [Eq. (48)] =
1.148 and AUMCAco~n [Eq. (47)]/AUMC ~" [Eq. (48)]=1.1475 where
A U M C ~ n [Eq.(47)] =545,529.6. However, larger h~g values when [A is
much greater than B in the modified Bateman function of Eq. (29)] result
in greater discrepancies among A U ! f f C ~ [Eq. (44)] and A UMC~,rr [Eq.
(48)]. If A = 5000 and B = 2000 with the same ;t~ values so that tt~g[Eq. (31)]
is 94.8 min, A UMCu~r A,,n [Eq. (44)]/AUMCco~, A~B [Eq. (47)] and
AUMC~, [Eq. (44)]/AUMC A" [Eq. (48)], respectively, -12.23 and
-34.53 are impossibly negative. However, A u g c c o~~~ s [Eq. (47)]/A U M C A"
[Eq. (48)] =2.82 where A UMC A~ [Eq. (48)] = 25,656. It is apparent that the
proper estimate of A U M C when A # B is from Eq. (47) and that a similar
value would be given by the trapezoidal rule for the positive area under the
Ct-t curve. In contrast to/1 UC determination from C-t data, transposition
of Ct-t data to the origin gives significantly different A UMC values whether
or not the area (,4 UMC A=n=A') under the Ct-t plots were calculated [Eq.
(48)] or were determined by a trapezoidal rule.

Correctly Calculated Statistical Moments (4i) (e.g., MRT) when A ~ B in

the Modified Bateman Function, Eq. (29)
M R T ~ " ~ r is calculated [Eq. (42)] without correction, i.e., when nega-
tive A UMC contributions are not ignored, from
A=PB A~B
(MRT..~o.)oo = (A UMC..~o.)oo [Eq. (44b)]/A w , ~ r [Eq. (36c)] (49a)
""'t~
= (B/A 2 - A/g~)/etC=~ - A/AI) (49b)

Similarly

' 4 " n-- A U M C corr [Eqs. ( 4 7 ) ] / ( A U C t ~ ) [Eq. (36c)]

. M R Tco, (50a)

and
( M R T ~ B ) ~ = B[e-~2tc=o(~,2tc=o+ 1)]/~L22
--A[e-;~'tc=~ + 1)]/A~ [Eq. (47b)]/(B e-e'2tc~~
- ( A e-X~tc~~ [Eq. (36c)] (50b)

If A = B = A " or is converted to/1" [Eq. (37)], the M R T value tc=o = 0 and

e -a'tc'~ 1 is

MRTA'~ a'[Eq. ( 4 8a )]/ A UC~,c

true [Eq. (36d)] (51a)
Bateman Function Revisited 123

and from the quotient of Eq. (48b) and Eq. (36d)

(MRT'~")oo = A"(11~.~- 1/~2)/A"(1/~" 1/~.,) (51b)
= l/X2+ 1/~) (51c)
,-.~1c [Eqs. (30) or (36d)] when
or from the quotient of [Eq. (47e)] and A U''tru~
Eq. (51c) is recognized
(MRTA~B)~o = tc=o+ (MRTtA"~ (51d)
M R~-~un~o.
T A ~ B [Eq. (49)] is virtually the same as MRT~o~ [Eq. (50)] for
the suppository example considered in Fig. 4 with a brief fC=O=tlag of
7.7 min. In this example M R T ~ = 5 9 . 3 7 (see Figs. 3 and 5) whereas
M O T A ~ B - 59.50 and M R T A"-- 51.85 so that the latter was displaced by the
7.7 min, of tc=o (see Eq. 51d). Larger flag values, when A is much greater
than B in the Modified Bateman function of Eq. (29) result in greater discrep-
ancies. If A = 5000 and B = 2000 with the same ~. values as the previous
example, the thg [Eq. (31)] is 94.8 rain, M R T A ~ is the impossibly negative
-1791 value whereas MRTrA,,H= 146.4 and M R T A'= 51.9 differ by the lag
time [Eq. (51d)], tc=o [Eq. (31)]. The proper estimate of M R T A ' s for the
Modified Bateman function [Eq. (29)] is from Eq. (50). A similar value
would be given by the trapezoidal rule for the positive area under the Ct-t
curve. Transposition of Ct-t data to the origin gives a significantly different
M R T "~ value whether or not the area (A UMC A") under the Ct-t data were
calculated [Eq. (50)] or determined by a trapezoidal rule. This is in contrast
to the invariance of the positive A UC, above the C= 0 ordinate, on the
transposition of a C vs. t plot along the time axis.
The mean absorption time (M/IT, 4i) of an orally or parenterally
administered drug invading by a first-order process can be calculated from
the difference between its resulting mean residence time MRTp, and that of
an iv bolus administration
M A T = M R Tp - MR ~,, (52)
where MRTi,, for a one-compartment body model is 1/~.2 [Eq. (51c) when
~,1 = oo and 1/~.l = 0].
Dose-adjusted A UC values, i.e., A UC/ Dose = l / CLtot, that show lower
(or different) bioavailabilities at the lower doses, e.g., (A UC/Dose)lower dose
is greater than (A UC/Dose)highr dose, are indicative of dose-dependent elimi-
nation processes with larger apparent kr values at the higher oral dose when
the absorption rate constant, ka, can be considered invariant. One possible
explanation would be saturable plasma protein binding. A greater unbound
drug fraction could permit faster elimination rates (-kr at the higher
free drug concentrations of higher doses. Saturable metabolism, as the major
124 Garrett

elimination process, could also give relatively higher oral or parenteral dose-
adjusted bioavailabilities at the higher dose since the apparent k~ is lessened
and the resultant systemic concentrations are heightened. Examples are
diphenylhydantoin (20) and amobarbital (21) which had increased ,4 UC/
Dose ratios with lessened elimination rates. Alternatively, there are possibili-
ties of saturable tissue uptake, substrate, and product (metabolite) inhibition
of a major metabolic process (22). Such phenomena could still exhibit appar-
ent dose-dependent first-order elimination with smaller ko rate constants at
higher doses. They have been observed with dicoumarol in humans (22) and
with phenylbutazone, probenecid, and biscoumacetate in dogs (23).

Elimination Rate Constants (ke) and their Ratios at Different Oral and
Parenteral Doses with Dose-Dependent Pharmacokinetics
The first-order elimination rate constant, k~, for an iv dose into the
one-compartment body model where A UCiv = Co/ke is

ko = (c0)i~/`4 uci~ (53)

The (C~)i~ of an hypothetical iv dose that gave the same apparent dose-
adjusted bioavailability, ,4 UC/Dose = 1/CLtot a s a completely absorbed oral
or parenteral dose D' with dose-independent pharmacokinetics can be esti-
mated from
(C~)iv = D ' / V (54)
where the apparent volume of distribution, V, did not vary with dose or the
mode of drug administration. V could be experimentally determined from
one intravenous study with a known dose, Div, and an evaluated A UCiv
[from (Co)iv/(ke)iv or by the trapezoidal rule]. If an intravenous study were
not available or possible, 100% bioavailability could be concluded at low
doses (Dp) if all orally or parenterally administered drug were renally
excreted either unchanged or as a metabolite and Dp would be a value
proportional to Co with the proportionality constant of an invariant appar-
ent volume of distribution, V [Eq. (54)]. Thus, relative ke values from com-
pletely bioavailable different oral or parenteral doses could be estimated
from the ratio of the doses since
( kr (kr = (C0)p,/`4 UCpJ ( Co)p2/`4 UCp2 (55)
The (k')iv elimination rate constant, of an intravenous dose equivalent
to that of the oral dose could be estimated by an equation similar to Eq.
(53) on substitution of (Cb)iv for (C0)iv and A UClv for A UCIv| A difference
between (ke)i, and (k')iv would imply dose-dependent pharmacokineties. It
Bateman FunctionRevisited 125

is indicative of dose-dependent pharmacokinetics when the ratio differs sig-

nificantly from 1.0.
Whereas the A UC values of low-dosed ibuprofen suppositories in rab-
bits are the same as that of low-dosed intravenously administered ibuprofen
(18) to indicate complete absorption from these suppositories at these dose
levels, dose-dependent pharmacokinetics of ibuprofen were concluded from
higher dosed suppositories by the above methods.

Absorption Rate Constants of Completely Absorbed Drugs on the

Assumption of Dose-Dependent First-Order Elimination
The first moment when 1/~j = 1 / ~ = 0 in Eq. (44b) or (48b) for postula-
ted intravenous conditions is

A UMC~,, = (C~)iv/(k'e)2 (56)

so that the appropriate mean residence time, M R T , [Eq. (51c)], is

M R T'~ = 1/ k'~= A UMCj~/A UC~ (57a)

-~"A M
U' C i v / ( C o )'i v / k'r (57b)

and plasma concentration-time data, when transposed to the orgin of such

plots, show mean absorption time values, M A Table (and derived values of
ka = 1 / M A T ) that are displaced by the magnitude of the lag time, a function
of the magnitude of the difference between A and B [Eq. (30) in the Modified
Bateman Eq. (29)].

Determination by Integral Methods (24) of the Rate Constants for

Metabolic Formation and Elimination (Even when they Approach
Each Other) when a Drug and its Systemically Formed
Metabolite are Both Assayed in Plasma
The amount of metabolite formed, M, can be related by metabolic
clearance, CLpM, to the area under the precursor drug's plasma
concentration-time curve, A UCp,, (24)

M = CLpMA UCpt (58)

and

U~ = CLM~A U C ~ (59)
126 Garrett

If the processes

Drug in GI(G) kG%Drug in Plasma (P) k~,

Metabolite in Plasma (M) k,~, Elimination(E') (60)

are first order, then the plasma concentration (m) of metabolite (M) is
m=M/Vm
= (total amount of M formed from P
- amount of M eliminated as E')/I'm (61 a)
and
m = CLpMA UCPt- CLME,.,4 U C M (61b)
where d UC P is the area under the plasma C-t curve of the precursor and
A UC~ is the area under the plasma C-t curve of the metabolite where the
CLs are the respective clearances.
Thus
m ,4 v c ?
A UCPt
= CLpM - CLME, A UC~ (62)
and
m -CL AucP
AucM- PM~ CLME (63)

so that plots of m/AUC~ against A UCtM/AUC P, at any time, t or

m/AUCt~ against A UCPt/AUC~ have intercepts of CLpM and --CLME,
respectively, with slopes of - CLUE,= --kME'Vp, and CLpM= kpM VM, respec-
tively. Alternatively, these values can be determined by multiple linear regres-
sion of Eq. (61b).
The amount of metabolite formed (M) could be constructed as the
difference between the amount formed and the amount excreted in the urine.
Thus, the metabolite amount in the plasma
M= m Vm= CLpM A U C P - UM (64)
permits
UM VMm (65)
A UC~t"= CLpM A UCet
or
UM/m = CLpM A uCP / m - Vm (66)
Bateman Function Revisited 127

so that plots o f the q u o t i e n t of e x p e r i m e n t a l values, Um/A UC P' against

m / A U C v' at a n y time, t, have a n intercept a n d slope, respectively, o f CLpM =
k p u V p a n d = - V m ; plots o f Um/m a g a i n s t A UC~ have the slope a n d
intercept reversed. Alternatively, these values can be d e t e r m i n e d b y multiple
linear regression [Eq. (64)] o f m e t a b o l i t e p l a s m a c o n c e n t r a t i o n o n the
A U C p, values o f d r u g p l a s m a c o n c e n t r a t i o n s a n d u r i n a r y a m o u n t s (U,,) of
m e t a b o l i t e at a time t. T h u s kp~ -- C L p M / V p .

REFERENCES
1. H. Bateman. The solution of a system of differential equations occurring in the theory
of radioactive transformations. Proc. Cambridge Phil. Soc. 15:423~27 (1910).
2. F. H. Dost. Grundlagen der Pharmakokinetik, George Thieme Verlag, Stuttgart, 1968,
(a) pp. 44-47, 241-242; (b) p. 143.
3. M. Gibaldi and D. Perrier. Pharmacokinetics, Marcel Dekker, New York, 1975, (a) p.
18, pp. 33-34; (b) Appendix 3, pp. 281-292.
4. M. Gibaldi and D. Perrier. Pharmacokinetics, 2nd ed. Marcel Dekker, New York, 1982,
(a) chap. 4, pp. 145-198, (b) pp. 147-148, (c) pp. 34-39, (d) Appendix C, pp. 433-
436 method of residuals, (e) pp.--37--39, (f) p. 82, (g) pr 146, (h) pp. 445-449, Appendix
D, (i) pp. 149-155, (j) pp. 409-414, chap. 11, (k) pp. 319-353, chap. 8.
5. S. Glasstone. Textbook of Physical Chemistry, 2nd ed., D. Van Nostrand, New York,
1946, pp. 1075-1077.
6. R. E. Notari, J. L. DeYoung, and R.H. Reuning. Effect of parallel first order drug
loss from site of administration on calculated values for absorption rate constants. J.
Pharm. Sci. 61:135-138 (1972).
7. D. S. Riggs. The Mathematical Approach to Physiological Problems, A Critical Primer,
Williams and Wilkins, Baltimore, MD, 1963, (a) pp. 120-161, (b) p. 163.
8. E. R. Garrett. Simplified methods for the evaluation of the parameters of the time-
course of plasma concentration in the one-compartment body model with first-order
invasion and first-order drug elimination including methods for ascertaining when such
rate constants are equal. J. Pharmacokin. Biopharm. 21:689-734 (1993).
9. P. L. Aitman and D. S. Dittmer (eds.). Biology Data Book, Federation of American
Societies for Experimental Biology, Washington, DC, 1964, p. 264.
10. P. L. Altman and D. S. Dittmer (eds.). Blood and Other Body Fluids, Federation of
American Societies for Experimental Biology, Washington DC, 1961, pp. 358-359.
11. K. K. H. Chan and K. W. Miller. Nonlinear regression approach for determining
whether absorption and elimination rate constants are equal in the one-compartment
body model with first-order processes. J. Pharm. Sci. 72:574-576 (1983).
12. I. H. Patel. Concentration ratio method to determine the rate constants for the special
case when ka=ke. J. Pharm. Sci. 73:859-861 (1984).
13. C.M. Metzler, G. L. Elfring, and A. J. McEwen. A package of computer programs
for pharmacokinetic modelling. Biometrics 30:562-563 (1974).
14. K. Yamaoka, Y. Tanigawara, T. Nakagawa, and T. Uno. A pharmaeokinetic analysis
program (multi) for microcomputer. J. Pharm. Dyn. 41:879-885 (1981).
15. M. Bialer. A simple method for determining whether absorption and elimination rate
constants are equal in the one.compartment open model with first-order processes, J.
Pharmacokin. Biopharm. 8:111- 113 (1980).
16. M. Barzegar-Jalali and M. Toomanian. Theoretical consideration for the cases where
absorption rate constant approaches elimination rate constant in the linear one-compart-
ment open model. Int. J. Pharm. 12:351-354 (1982).
17. P. Maeheris, M. Symillides, and C. Reppas. On the assessment of the relative magnitude
of rate constants in the linear open one-compartment model. J. Pharm. Sci. 81:1231-
1233 (1992).
128 Garrett

18. T. W. Hermann and F. K, Glowka. Bioavailability of racemic ibuprofen and its lysinate
from suppositories in rabbits. Unpublished data, given with permission, 1990.
19. J. G. Wagner and E. Nelson. (a) Per cent absorbed time plots derived from blood
levels and/or urinary excretion data. ,I. Pharm. Sci. 52:610-611 (1963). (b) Kinetic
analysis of blood levels and urinary excretion in the absorptive phase after single doses
of drugs J. Pharm. Sci. 53:1392-1403 (1964).
20. J. J. Ashley and G. Levy. Kinetics of diphenylhydantoin elimination in rats. J. Pharmaco-
kin. Biopharm. 1:99-102 (1974).
21. E. R. Garrett, J. Br6s, K. Schnelle, and L. L. Rolf Jr. Pharmacokinetics of saturably
metabolized amobarbital. J. Pharmacokin. Biopharm. 2:43-104 (1974).
22. R. A. O'Reilly, P. M. Aggeler, and L. S. Leong. Studies on the coumarin anticoagulant
drugs. A comparison of the pharmacodynamics of dicumarol and warfarin in man.
Thromb. Diath. Haemorrh. 11:1-22 (1964).
23. P. G. Dayton, S. A. Cucinell, M. Weiss, and J. M. Perel. Dose-dependence of drug
plasma level decline in dogs. J. Pharmacol. Exp. Ther. 158:305-316 (1967).
24. E. R. Garrett. In L. Z. Benet, G. Levy, and B. L. Ferraiolo (eds.), Pharmacokinetics
A Modern View, Plenum Press, New York, 1984, chap. 20.