Medical Engineering & Physics 21 (1999) 525–532

www.elsevier.com/locate/medengphy

A pharmacokinetic model for multiple sites discontinuous

gastrointestinal absorption
a,b,*
Yves Plusquellec , Constantin Efthymiopoulos c, Pierre Duthil d, Georges Houin b

a
UFR de Mathématiques, Université Paul Sabatier, 118 route de Narbonne, 31062 Toulouse, France
b
Cinétique des Xénobiotiques, Faculté des Sciences Pharmaceutiques, 35 chemin des Maraı̂chers, 31062 Toulouse, France
c
Glaxo-Wellcome Research and Development, Clinical Pharmacology Division, Greenford UBG OHE, UK
d
Service de Biophysique, Faculté de Médecine, C.H.U. Rangueil, avenue du Professeur Jean Poulhès, 31403 Toulouse, Cedex 4, France

Received 15 March 1999; received in revised form 11 May 1999; accepted 5 July 1999

Abstract

Purpose. To build a pharmacokinetic model taking into account a discontinous absorption along the gut, from n successive sites,
a non-absorbing intestinal segment being always in between two successive sites. To solve the mathematical model linked with the
pharmacokinetic model to obtain the concentration and contribution of each site to absorption, area under curve and bioavailability.
Methods. Whatever the number n of sites, we obtained the Laplace transform of amounts of drug in each site, then plasma
concentration, so that concentration or AUC were expressed analytically. When only two absorption sites are present, concentration
is obtained from Heaviside’s theorem; but for nⱖ3, Bromwich’s theorem is necessary, a pole being of the order of more than two.
Results. Simulations performed with data gathered from the literature allow to find, with n=2 sites, the particular case used for
ranitinine and to show the efficacy of each site. For n=3 sites, real data exhibiting three peaks of various magnitude were fitted on
our model.
Conclusion. This general discontinuous oral absorption pharmacokinetic model may be taken as a possible tool to characterize
each site of absorption and to estimate the area under curves or bioavailability.  1999 IPEM. Published by Elsevier Science Ltd.
All rights reserved.

Keywords: Discontinuous absorption; Laplace transform and inverse; Delay; Secondary peaks; Ranitidine

1. Introduction tion window, it is necessary to build a compartmental

Double peaks in plasma concentration time profile
have been encountered following oral administration of
veralipide [1], ranitidine [2], cimetidine [3,4], furosenide
[5] or piretanide [6]. Furthermore, Efthymiopoulos et al.
[7] published data related to a drug (ranitidine) exhibit-
ing multiple peaks in its plasma concentration profile. If
the assumption of enterohepatic recirculation is rejected
by physiological arguments, a possible explanation for
multiple peaks occurrence may be a discontinuous
absorption along the gut.
Many other drugs display two, three or more absorp-
tion peaks following oral administration. To take into
account these peaks, that is the influence of each absorp-
* Corresponding author. Tel.: +33-05-61-55-69-15; fax: +33-05-61-
55-83-26.
model allowing to separate parameter values correspond-
ing to each site: for example, the part of the administered
dose absorbed in a given site may be of interest.
Two kinds of models have been proposed to deal with
discontinuous absorption. We have proposed a model
with two absorption sites used for veralipide after a sin-
gle oral dose [1] or chronic administration [8]. Then we
extended this model to n sites in [9], also with a time
lag in the case of ranitidine [7]. In these models, the
transit time duration for the crossing of a site was taken
into account. Other types of discontinuous absorption
models have been published by Suttle et al. [10] with
two absorption sites and n non-absorbing segments, by
Witcher et al. [11] with two sites and one non-absorbing
segment and by Murata et al. [12] for multifraction
absorptions. The models of [10,11] have been used to
fit ranitidine and cimetidine data.
As we have obtained data exhibiting several second-

1350-4533/99/$ - see front matter  1999 IPEM. Published by Elsevier Science Ltd. All rights reserved.
PII: S 1 3 5 0 - 4 5 3 3 ( 9 9 ) 0 0 0 6 0 - 0
526 Y. Plusquellec et al. / Medical Engineering & Physics 21 (1999) 525–532

Nomenclature
AUC area under concentration curve in compartment 1
Cl
冉
clearance Cl=
Qabs
AUC 冊
D dose of drug
F bioavailability
fi (i=1 to n) the fraction of the dose D absorbed in site i
g fraction of D which is absorbed: g= f1 +... + fn
G gut (site of administration)
GAi (i=1 to n): the absorbing segments
Gai(t) the amount of drug at time t in the absorbing segment GAi
GTi (i=1 to n⫺1): the non-absorbing segments; GTj is between GAj and GAj+1
Gti(t) the amount of drug at time t in the non-absorbing segment GTi
kai (i=1 to n): rate of absorption from GAi to the central compartment 1 (in time⫺1)
k1 rate of absorption from G to GAi (in time⫺1)
k2 rate of absorption from GAi or elimination from GTi (in time⫺1)
k10 rate of elimination from compartment 1 (in time⫺1)
k20 rate of elimination from compartment 2 (in time⫺1)
k12, k21 rates of exchange between compartments 1 and 2 (in time⫺1)
Ki=k2+kai for i=1 to n
Qabs the total amount of drug which is absorbed: Qabs=Q1+..... + Qn
Qelim the amount of drug eliminated via faeces
Qi Qi
(i=1 to n): the amount of drug which is absorbed in site i (fi= )
D
ti (i=2 to n): time at which the drug reaches the absorbing segment GAi
V distribution volume of compartment 1 (in l)
Xi(t) for i=1,2: the amount of drug at time t in compartment i
zai(s) the Laplace transform of Gai(t) for i=1 to n

ary peaks [7], we have built a more general model than
[10,11] including any number, n, of absorption sites
along the gastrointestinal tract. Furthermore, we have
added delays in drug moving, taken into account various
coefficients for absorption related to each site, included
a non-absorbing segment in between two successive sites
and assumed that the drug can be absorbed from several
sites simultaneously. Thus, for any drug undergoing dis-
continuous gastrointestinal absorption, there is no limi-
tation in terms of the types which it is possible to analyse
by this model.
From the analytical expression of concentration,
obtained by Laplace transformation, we demonstrated
the theoretical contribution of each site to the total
absorption. Furthermore, the area under concentration
curve (AUC), clearance and bioavailability were obtain-
ed.
Using data gathered from the literature and our model
with two sites, we were able to describe the usual profile
for ranitidine.
Furthermore, we have fitted raw data provided by
Glaxo-Wellcome laboratories on our three-site model,
and we succeded in describing secondary peaks.
2. Theoretical part: method
2.1. The pharmacokinetic model
The dose D of the drug is orally administered, reaches
the gut G and goes to the compartment GA1 which is
the first site of absorption along the gut (see Fig. 1), in
a first order process (coefficient k1): we assume a first
order process because absorption depends on the rate at
which the drug leaves the site (analogous to a gastric
emptying rate). Absorption from this site is described by
a first order rate constant ka1. The non-absorbed amount
of drug begins its transit through the non-absorbing com-
partment GT1 of the gut in a first order process
(coefficient k2) and reaches the second site of absorption
GA2 at time t2 where again absorption occurs
(coefficient ka2): t2 is the time at which the first compo-
nent of the drug reaches the second site.
The non-absorbed amount of the drug travels through
the non-absorbing part GT2 (coefficient k2) and reaches
the third site, GA3, at time t3 where absorption again
occurs (coefficient ka3).
This process is repeated until the (n⫺1)th non-absorb-
 Y. Plusquellec et al. / Medical Engineering & Physics 21 (1999) 525–532 527

Fig. 1. Discontinuous oral absorption model with n absorbing windows Gai along the gastrointestinal tract.

ing segment GTn⫺1 of the gut and the nth site GAn dGt2
reached at time tn, where absorption again occurs, 쐌 ⫽k2Ga2⫺k2Gt2; Gt2(0)⫽0 (5)
dt
always in a first order process (coefficient kan).
After the nth absorption site, the remaining drug exits dGa3
∗ ⫽k2Gt2⫺K3Ga3 if tⱖt3; Ga3(t)⫽0 if 0ⱕt⬍t3 (6)
from this site (coefficient k2) and is eliminated via the dt
faeces.
dGtn−1
Drug absorption into the central compartment 1 occurs 쐌 ⫽k2Gan−1⫺k2Gtn−1; Gtn−1(0)⫽0 (7)
as the result of n delayed absorption phases. dt
The drug is then distributed to a peripherial compart- dGan
ment 2 with first order coefficients k12 and k21. Elimin- ∗ ⫽k2Gtn−1⫺KnGan if tⱖtn; Gan(t)⫽0 if 0ⱕt⬍tn (8)
ation may occur via k10 from compartment 1, possibly dt
via k20 from compartment 2 and via the faeces. The dis- dX1
tribution volume of compartment 1 is denoted by V when ∗ ⫽ka1Ga1⫹ (9)
dt
concentrations are used instead of amounts.

2.2. The mathematical model 再 if tⱖt2

再
ka2Ga2⫹%⫹ kanGan⫹k21X2−(k10⫹k12)X1; X1(0)⫽0
if tⱖtn
Let G(t), Gai(t), Gti(t), X1(t) and X2(t) denote the dX2
amounts of drug, at time t, in compartments G, GAi, ∗ ⫽k12X1⫺(k20⫹k21)X2; X2(0)⫽0 (10)
dt
GTi, 1 and 2 respectively. Time t is 0 when the drug
reaches compartment G (it would be easy to add a delay Thus, Eqs. (1)–(8) give the input function in compart-
t1 to take into account the time t1 necessary for the drug ment 1 to get the concentration in this compartment,
to be released from the pharmaceutical formulation). while Eq. (10) allows to deal with elimination from a
The differential system connected with this model is, peripheral compartment.
with Kj =k2+kaj for j=1 to n:
2.3. The Laplace transforms
쐌G(t)⫽De−k1t (1)
dGa1 Let zai(s), z1(s) and z2(s) denote the Laplace trans-
∗ ⫽Dk1e−k1t⫺K1Ga1; Ga1(0)⫽0 (2)
dt forms of Gai(t), X1(t) and X2(t) respectively.
We give, in the Appendix, a calculation of zai(s):
dGt1
쐌 ⫽k2Ga1⫺k2Gt1; Gt1(0)⫽0 (3) Dk1
dt
za1(s)⫽
(s+k1)(s+K1)
dGa2
∗ ⫽k2Gt1⫺K2Ga2 if tⱖt2; Ga2(t)⫽0 if 0ⱕt⬍t2 (4)
dt and for i=2 to n
528 Y. Plusquellec et al. / Medical Engineering & Physics 21 (1999) 525–532

Dk1k 2i−2
2 e−sti
zai(s)⫽ i
(11) The contribution of each absorbing segment to the
(s+k1)(s+k2)i−1 ⌸ (s+Kj ) total AUC appears in this summation.
j⫽1

Then 3.2. The non-absorbed amount Qelim

A part of the amount reaching the compartment GAn

冘
is absorbed into the central compartment but a part Qelim

冦
n

sz1= kaizai+k21z2−(k10+k12)z1 is not absorbed and travels along the gut to be eliminated
i⫽1
(12) via faeces:
sz2=k12z1−(k20+k21)z2
Qelim⫽k2 冕 ⬁

tn
Gan(t)dt⫽k2zan(0) because Gan(t)⫽0 if 0ⱕt
With
⬍tn
s+k10+k12 −k21
⌬⫽
|
−k12 s+k20+k21 | ⫽(s⫹k10⫹k12)(s⫹k20⫹k21) Thus
Dk n2
⫺k12k21 Qelim⫽ (15)
K1K2%Kn
we denote by a and b the two roots of ⌬=0, that is
⌬=(s⫺a)(s⫺b)
then 3.3. The fraction g of D which is absorbed

冘
n

冘
n
(s+k20+k21) kaizai

⌬
0 |
1 kaizai −k21
z1⫽ i⫽1
s+k20+k21
⫽
i⫽1
(s−a)(s−b)
| (13)
g⫽
D−Qelim
D
⫽1⫺
k n2
K1K2%Kn
(16)

After inserting Eq. (11) into Eq. (13), we are able to 3.4. The amounts of drug absorbed in each site
get the analytical expression of concentration in com-
partment 1 by inverse Laplace transform, whatever the The amount of drug absorbed in site 1 is
number n of absorbing sites. Q1=ka1兰 ⬁0Ga1(t)dt
If n=2, all poles of z1 are of order one, so that Heavisi- Thus
de’s theorem allows to obtain the concentration. This Dka1
case easily gives the results of [10,11]. Q1⫽ka1za1(0)⫽ (17)
K1
If nⱖ3, as the pole ⫺k2 of z1 is of the order of more
than one, we have to use the Bromwich’s theorem to get Thus, the fraction f1 of D which is absorbed in site 1 is
the expression of concentration. An example of such a ka1
f1= .
study is given in the practical section. K1
This gives an explanation for the assumption of Eq.
(12) in [10].
3. Results The amount of drug which is absorbed in site 2 is

冕
⬁
3.1. Area under concentration curve in compartment 1 Dka2k2
Q2⫽ka2 Ga2(t)dt⫽
K1K2
t2

冘
n
1 k20+k21 and the fraction f2 of D which is absorbed in site 2 is
AUC⫽ z1(0)⫽ ka za (0)
V Vab i⫽1 i i Q2 k2ka2 ka2
f2⫽ ⫽ ⫽ (1⫺f1) (18)
Or: D K1K2 K2

冘
n
(k20+k21)D kaik 2i−1 Similarly, the amount Qi of drug which is absorbed
AUC⫽ (14)
V[k10(k20+k21)+k12k20]i⫽1K1K2…Ki in site i is:
 Y. Plusquellec et al. / Medical Engineering & Physics 21 (1999) 525–532 529

冕
⬁
Dkaik 2i−1 ⫽f1⫹%⫹fn⫽g
for i:1,2,%n Qi⫽kai Gai(t)dt⫽ i (19)
⌸ Kj which is in agreement with Eq. (9) of ref. [11], but with
ti
j⫽1 n sites instead of 2 sites.
Qi
and the fraction of D absorbed in site i is fi= .
D
The total amount of drug which is absorbed is: 4. Practical section

冘
n

Qabs⫽Q1⫹Q2⫹%Qn⫽ ka1zai(0) (20) As this model includes n absorption sites with non-
i⫽1
absorbing parts of the gut, the special case n=2 allows
to deal with the study of ranitidine or cimetidine pro-
posed in [10,11] but with analytical expressions instead
Thus Eq. (14) may be rearranged as: of numerical process. Furthermore, data were fitted on
(k20+k21)Qabs the model with n=3 sites. These drugs have been chosen
AUC⫽ (21) due to well pronounced secondary peaks. Thus, no doubt
V[k10(k20+k21)+k12k20]
is possible between a peak and a small variation of con-
centration resulting from the experiment.
In the special but commonly used situation where
k20=0: 4.1. The case n=2
Qabs
AUC⫽
k10V
which is a classical result. 4.1.1. Concentration: with K1=k2+ka1 and K2=k2+ka2

3.4.1. Remarks
z1⫽
Dk1(s+k20+k21)
冋
ka1
(s−a)(s−b) (s+k1)(s+K1)

册
1. Obviously, the fraction g of D which is absorbed is ka2k 22e−st2
equal to: ⫹
(s+k1)(s+K1)(s+k2)(s+K2)
g⫽f1⫹f2⫹...⫹fn.
Thus, from Heaviside’s theorem we get X1(t)=L−1z1
2. The total clearance may be expressed as: X1(t)
then the concentration C(t)= , used in the follow-
Q1+Q2+%+Qn Qabs V
Cl⫽ ⫽ (22) ing example.
AUC AUC
V([k10(k20+k21)+k12k20] 4.1.2. Simulation for ranitidine
⫽
k20+k21 Using a set of parameter values gathered from the
which becomes Cl=V.k10 when k20=0. literature about ranitidine [10,11], we have got the
plasma concentration profile C(t) given in Fig. 2 with:
k1=0.47 h⫺1; k2=1.85 h⫺1; ka1=3.5 h⫺1; ka2=4 h⫺1;
Thus, Cl does not depend on delays or number of sites.
k12=1.42 h⫺1; k21 =0.96 h⫺1; k10 =0.8 h⫺1; k20=0.01 h⫺1;
V=63 l; t2=3.7 h and D=120 mg.
3.5. Bioavailability The results for f1, f2, AUC and clearance are included
in Fig. 2 (clearance in l/h and AUC in µg/l/h). Thus, we
If the same dose D of the drug is intravenously admin- are able to find a secondary peak, as mentioned in the
istered in compartment 1, then: clinical report about ranitidine and to add results about
D(k20+k21) absorption of the drug in each site.
AUCIV⫽ (23)
V[k10(k20+k21)+k12k20]
4.2. The case n=3

The bioavailability F of the oral form may be 4.2.1. Concentration

expressed as: The contribution of the third site is shown in z1:

冘 冋
n
AUCoral Qabs 1 1 Dk1(s+k20+k21) ka1
F⫽ ⫽ ⫽ kaizai(0)⫽ (Q1⫹Q2⫹%⫹Qn) z1 ⫽
AUCIV D Di⫽1 D (s−a)(s−b) (s+k1)(s+K1)
530 Y. Plusquellec et al. / Medical Engineering & Physics 21 (1999) 525–532

Fig. 2. Simulation for n=2 sites with parameter values corresponding to ranitidine (see these values in the text).

Table 1
ka2k 22e−st2 Experimental data after administration of 150 mg of ranitidine
⫹ (24)
(s+k1)(s+k2)(s+K1)(s+K2)
t time (h) Concentration (ng/ml)

⫹
ka3k 42e−st3
(s+k1)(s+k2)2(s+K1)(s+K2)(s+K3) 册 0.25
Whenever we add a site in the model, we add a new
term in z1 thus in concentration obtained by inverse
Laplace transform using Bromwich’s theorem because
the pole (⫺k2) of z1 is of order two.
4.2.2. Data fitting
In the study reported in [7], ranitidine (150 mg) was
administered to 24 healthy male subjects as a chewable
tablet and as a standard tablet, according to a randomized
crossover design. Blood samples were collected up to 12
h post-dosing. Ranitidine was assayed in plasma by a
validated radioimmunoassay with a lower quantification
limit of 1 ng/ml.
We used data from one subject (subject #5) from this
study to illustrate the fitting of the model when three
peaks were observed in the plasma profile; the observed
plasma concentrations in subject #5 are reported in Table
1. The corresponding fitting using the model with 3 sites
is shown in Fig. 3 and the results are reported in Tables
2 and 3.
Here again, we found a satisfactory fitting of data
allowing to get optimized values for parameters. For
example, it may be interesting to know the delay neces-
sary for the drug to reach the absorbing window in order
to show in what part of the gut this site is located. Fur-
thermore, the model, separating each absorption part,
indicates what part of the drug is absorbed in the first
site (here almost half).
0.5
116.9
210.3
1 190.4
1.5 167.5
2 178
2.5 167.4
3 160.2
3.5 179.1
4 139
4.5 119
5 93.6
6 74.6
8 52.1
10 29.8
12 15.3
5. Conclusion
By adding several changes to previous models, such
as the number of absorption windows along the gastro-
intestinal tract, we have built up a general pharmacoki-
netic compartmental model which may be a possible tool
for discontinuous absorption study.
As we have obtained the analytical expression of
plasma concentration, we are able to deal with areas
under curve, bioavailability, clearance and furthermore
to point out the contribution of each site to the total
absorption of the drug.
From the two-site model, with parameter values
related to ranitidine, we have shown the occurrence of
secondary peaks. From the three-site model, we have fit-
ted raw data and obtained the contribution of each site
 Y. Plusquellec et al. / Medical Engineering & Physics 21 (1999) 525–532 531

Fig. 3. Fitting with the triple absorption site model of ranitidine plasma concentrations obtained in subject #5 after administration of a chew-
able tablet.

Table 2 could be of great help to characterize the steps involved

Optimised values of the three-site model parameters for ranitidine and then to optimize formulations.
data fitting

Notation Optimized Notation Optimized

values (h⫺1) values Acknowledgements
k1 2.589 ka1 4.058 h⫺1
k2 5.271 ka2 2.421 h⫺1 We Thank Mme Aragon M. Christine (service de
k10 10.587 ka3 3.234 h⫺1 Biophysique) for her efficient secretarial assistance. The
k12 10.874 t2 1.621 h authors gratefully thank the reveiwer for his relevant
k20 10.067 t3 2.185 h
suggestions about the model structure.
k21 10.625 V 145 l

Table 3
Appendix. Calculation of the Laplace transform
Parameters deduced from optimisation in data fitting zai(s) of Gai(t) with Kj =k2+kaj

fi is the fraction of D Clearance AUC 앫

absorbed in site i
DK1 Dk1
f1=43.3% 97.38l/ h 1169 µg/l/h sza1⫽⫺K1za1⫹ , thus za1(s)⫽
f2=17.7%
s+k1 (s+k1)(s+K1)
f3=14.7%
앫 if zGti denotes the Laplace transform of Gti(t)
Dk1k2 Dk1k2
to absorption: half of the administered dose is absorbed szGt1⫽⫺k2zGt1⫹ , zGt1⫽
in the first site for almost all the subjects participating (s+k1)(s+K1) (s+k1)(s+k2)(s+K1)
in the study. When we wish to add another site of
absorption in the model, we have only to add a term in 앫 sza2=k2zGt1⫺K2za2,
the concentration expression. Dk1k 22
At this time, extensive work is being carried out to thus za2(s)=
(s+k1)(s+K1)(s+k2)(s+K2)
optimize drug absorption through the gastrointestinal
앫 szGt1=k2zGt2⫺
tract in order to increase drug delivery to the body, after
Dk1k 32
oral administration. Among the mechanisms involved in k2za2(s), zGt2(s)=
this process, the absorption of drugs through the intesti- (s+k1)(s+k2)2(s+K1)(s+K2)
앫 Dk1k 42
nal membrane becomes a key, especially for drugs of sza3=k2zGt2⫺K3za3, thus za3(s)= 3
protein origin. In vitro, tools were developed to study (s+k1)(s+k2)2 ⌸ (s+Kj )
such mechanisms. In vivo, the sites of absorption are j⫽1
hardly reached directly, even in animals. Pharmacoki-
netic models, such as the one described in this paper, Then
532 Y. Plusquellec et al. / Medical Engineering & Physics 21 (1999) 525–532

Dk1k 62 References
za4(s)⫽ 4
(s+k1)(s+k2)3 ⌸ (s+Kj )
j⫽1 [1] Plusquellec Y, Campistron G, Staveris S, Barre J, Jung L, Tille-
men JP, Houin GJ. Pharmacokin Biopharm 1987;15:225–39.
and, by recurrent process [2] Garg DC, Weidler DJ, Eshelman FN. Chir Pharmacol Ther
1983;33:445–52.
Dk1k 2i−2
2
[3] Walkenstein SS, Dubb JW, Randolph WC, Westlake WJ, Stote
zai(s)⫽ i RM, Intoccia AP. Gastroenterology 1978;74:360–4.
(s+k1)(s+k2)i−1 ⌸ (s+Kj ) [4] Oberle RL, Amidon GL. J Pharmacokin Biopharm
j⫽1 1987;15:529–45.
[5] Hammarlund MM, Paalzow LK, Odlind B. Eur J Clin Pharmacol
but as the absorption phase in site i is delayed until time 1984;26:197–207.
ti to begin only at this time: [6] Brockmeier D, Grigoleit HG, Leonhardt H. Eur J Clin Pharmacol
1986;30:79–82.
Dk1k 22i−2e−sti [7] Efthymiopoulos C. 9th Annual Meeting of AAPS 6 Nov. 1994,
for i⫽2, 3,..., n zai(s)⫽ San Diego.
i
[8] Staveris S, Plusquellec Y, Campistron G, Barre J, Rochas MA,
(s+k1)(s+k2)i−1 ⌸ (s+Kj ) Lung L, Tillement JP, Koffel JC, Gouin GJ. Pharm Sci
j⫽1
1988;77:64–70.
Dk 2i−2 [9] Plusquellec Y, Houin G. Arzneimitt-forsch 1994;44:679–82.
Especially zai(0)= i for i=1 to n [10] Suttle AB, Pollack GM, Brouwer KLR. Pharm Res 1992;9:350–6.
⌸ Kj [11] Wright Witcher J, Boudinot FD. Pharm Res 1996;13:1720–4.
j⫽1 [12] Murata K, Noda K. Pharm Res 1993;10:757–62.