Laboratory of Computational Biology and Risk Analysis, National
Institute of Environmental Health Sciences
Physiologically based pharmacokinetic modeling of the parent
chemical primidone and its two metabolites phenobarbital and phenylethylmalonamide (PEMA) was applied to investigate the differences of primidone metabolism among humans, rats, and mice. The model simulated previously published pharmacokinetic data of the parent chemical and its metabolites in plasma and brain tissues from separate
studies of the three species. Metabolism of primidone and its
metabolites varied widely among a sample of three human subjects from
two separate studies. Estimated primidone metabolism, as expressed by
the maximal velocity Vmax, ranged from 0 to 0.24 mg·min
1·kg1 for
the production of phenobarbital and from 0.003 to 0.02 mg·min1·kg1 for
the production of PEMA among three human subjects. Further model
simulations indicated that rats were more efficient at producing and
clearing phenobarbital and PEMA than mice. However, the overall metabolism profile of primidone and its metabolites in mice indicated that mice were at higher risk of toxicity owing to higher residence of
phenobarbital in their tissues and owing to the carcinogenic potential
of phenobarbital as illustrated in long-term bioassays. This result was
in agreement with a recently finished National Toxicology Program (NTP)
carcinogenicity study of primidone in rats and mice.
 |
Introduction |
Primidone is a desoxybarbiturate
anticonvulsant used for the treatment of partial seizures in humans. It
is one of the major drugs used for the treatment of psychomotor
epilepsy. Fig. 1 is an illustration of
the metabolic pathways of primidone. Oxidation of the second carbon on
primidone converts it to phenobarbital in vivo in both
animals and humans. Alternatively, ring cleavage of the drug at the
second carbon position converts primidone to phenylethylmalonamide
(PEMA)1[fnc] (Baumel et al., 1973
).
Phenobarbital is a well known anticonvulsant, whereas PEMA has been
shown to have anticonvulsant effects in rats. Because of its extensive
use by humans, the pharmacokinetics of primidone was extensively
investigated in humans, rats, and mice. Plasma concentrations of
primidone and its metabolites were measured in two human subjects who
individually received a single gavage dose of 500 mg of primidone
(Baumel et al., 1972). In another study, application of more
sensitive methods was used to investigate the pharmacokinetics of
primidone and its metabolites in a single human subject given a single
primidone gavage dose of 600 mg (Sato et al., 1986).
Pharmacokinetics of primidone was also examined in Swiss-Webster mice
where plasma levels of primidone, phenobarbital, and PEMA were
determined after the animals received a single gavage dose of 50 mg/kg
of the parent drug (Leal et al., 1979). In a different
study, plasma levels of the three chemicals were investigated simultaneously in male albino rats after administering single gavage
doses of 125, 250, and 500 mg/kg of primidone (Baumel et al., 1973).
The toxicity of primidone is compounded by evidence of the carcinogenic
potential of phenobarbital in rodents (McClain, 1995) and the mitogenic
activity of PEMA in Salmonella (Zeiger et al., 1988). These findings, in addition to its high volume use by humans, stimulated the National Toxicology Program (NTP) to conduct a 2-year
chronic carcinogenicity study to fully characterize the chronic
toxicity of primidone in Fischer 334 rats and
B6C3F1 mice (NTP, in press). The NTP study
indicated that primidone was a more potent carcinogen to mice than it
was to rats.
The complexities of primidone metabolism complicates the contribution
of each metabolite to its potential toxic effect because of exposure to
primidone and could impair extrapolations of risk across species.
Mechanistic modeling in the form of physiologically based
pharmacokinetic (PBPK) models offers an approach to analyze the
variability of metabolic profiles among species whenever adequate data
exist. PBPK models characterize the deposition and metabolism of
chemicals in tissues based on established or hypothesized mechanisms. They consist of tissue compartments that are linked by sets of mass
balance equations that realistically describe the disposition and
metabolism of modeled xenobiotics. Although PBPK models are deterministic in nature, the data are not, so the model parameters are
generally estimated using routine statistical methods (e.g. least squares). Hence, whenever distinct kinetic data are available, PBPK models can be used to estimate kinetic constants for individuals of the same species.
The purpose of this effort is to apply PBPK models to determine the
metabolic profile of primidone among three human subjects, rats, and
mice partially by simulating published data with optimized pharmacokinetic constants and partially by using existing model structures on the distribution and metabolism of its major metabolites. The derived metabolic profiles of primidone and its metabolites will be
used to assess the role of pharmacokinetics in the difference of
species response to the parent chemical between rats and mice as was
illustrated by the NTP study.
| |
Materials and Methods |
PBPK Model Compartments and Construct.
A central PBPK model of primidone was connected, by way of the liver
compartment, to a PBPK model for each metabolite (fig. 2). This was done so that plasma levels
of primidone and its metabolites can be simulated. Each PBPK model
includes the following tissues: lung, fat, muscle, skin, heart, kidney,
gastrointestinal (GI) tract (represented by small intestine), brain,
and arterial and venous blood compartments. The following detailed
description of the mathematical equations used in the model reveals all
assumptions embedded in the overall model.
View larger version (39K):
[in this window]
[in a new window]
|
Fig. 2.
A schematic of the parent PBPK model for
primidone and the two submodels for phenobarbital and PEMA.
Primidone is supplied into the GI lumen of the model. Once it is
metabolized, phenobarbital and PEMA are introduced to the liver
compartments of each submodel. All compartments are flow limited except
for the brain.
|
|
The usual approach for primidone application is by oral administration,
especially in experimental settings where an initial gavage dose is
introduced into the lumen and is then absorbed through the GI tissue to
be distributed via the portal vein to the liver. To accommodate this
mechanism, a compartment for the GI tract was included that consisted
of a lumen and tissue subcompartments for the primidone model only.
Absorption of the chemical from the GI lumen into the GI tissue is
modeled by a first order rate equation as follows:
where AMTlmn is the amount (mg) of
primidone in the GI lumen, and Kabs is the
first order absorption rate constant (min1) for
transfer from GI lumen into the GI tissue. Using simple, equilibrium
partitioning between arterial blood and GI tissue, the rate of change
in the GI tissue subcompartment is defined as follows:
where AMTgi is the amount of
primidone (mg) in the GI tissue, Qgi is the
portal vein blood flow (ml/min), Pgi is the
GI tissue/blood partition coefficient, Cart
is the concentration (mg/ml) of primidone in arterial blood, and
Cgi is the concentration of primidone
(mg/ml) in GI tissue. For this tissue and all other tissues,
concentration is defined as the ratio of amount (AMT) to tissue volume.
Following absorption through the GI tissue, primidone is distributed to the liver where it is either metabolized (assuming Michaelis-Menten kinetics) or further distributed to other tissues as follows:
where AMTlvr is the
amount (mg) of primidone in the liver, Qlvr
is the hepatic arterial blood flow (ml/min),
Clvr is the concentration (mg/ml) of
primidone in hepatic tissue, Plvr is the
liver tissue/blood partition coefficient of primidone,
Vmaxph and Kmph
are the Michaelis-Menten metabolism constants for primidone oxidation
to phenobarbital, and Vmaxpm and
Kmpm are the metabolism rate constants for
primidone cleavage to PEMA. Both Vmaxph and Vmaxpm are scaled to the 0.7 power of body
weight.
Distribution of primidone and its metabolites to brain tissue is
governed by a diffusion limited process through the blood-brain barrier. For this reason, the brain compartment was divided into separate blood and tissue subcompartments similar to a previous phenobarbital model construct (Igari et al., 1982). Assuming
simple diffusion kinetics, the following equation describes the mass balance of the brain blood capillary subcompartment:
where AMTbrnc is the
amount (mg) of primidone in brain capillary,
Volbrnt is the volume (ml) of brain tissue,
DR is the diffusion rate constant
(min1) of primidone between tissue and blood,
Cbrnt is the concentration (mg/ml) of
chemical in brain tissue, FR is the ratio of free to tissue
concentrations of the chemical, Cbrnc is
the concentration (mg/ml) of primidone in brain capillary,
Bplasma is the binding fraction of
primidone to red blood cells, and Qbrn is
the brain blood flow (ml/min).
Using a similar set of assumptions, the mass balance equation for the
brain tissue compartment is as follows:
where the parameters are defined analogously to those
for the brain capillary.
The remaining model compartments are all assumed to follow first order,
flow-limited tissue distribution leading to equations of the following
form:
where AMTt is the amount
(mg) of primidone in tissue t, Qt is the
blood flow (ml/min) through tissue t, and
Ct and Pt are the tissue concentration (mg/ml) and tissue/blood partition coefficient of the chemical, respectively.
Additional compartments for arterial and venous blood were also
included in the model as follows (respectively):
where AMTart is the amount (mg) in
arterial blood, Qtotal is the total cardiac
output (ml/min), and Clung and
Plung are the lung tissue concentration
(mg/ml) and tissue/blood partition coefficient, respectively.
where AMTvns and
Cvns are the amount (mg) and concentration
of (mg/ml) of the chemical in venous blood, respectively. The remaining
parameters are as defined above.
The PBPK submodels for phenobarbital and PEMA (see fig. 2) are
constructed in a similar fashion to the one for primidone except for
the GI and liver compartments. In the GI tissue compartment, the lumen
subcompartment is eliminated, as these two metabolites are not
administered orally for this analysis, and the liver compartment is
modeled as follows:
where AMTlvrph or lvrpm is the
amount (mg) of phenobarbital or PEMA in liver as they are produced from
primidone, Cartph or artpm is the arterial
concentration (mg/ml) of phenobarbital or PEMA, Clvrph
or lvrpm is the concentration (mg/ml) of phenobarbital or
PEMA in liver, Cgiph or gipm is the GI
tissue concentration (mg/ml) of phenobarbital or PEMA,
Plvrph or lvrpm is the liver tissue/blood
partition coefficient for phenobarbital or PEMA, Pgiph
or gipm is the GI tissue/blood partition coefficient of phenobarbital or PEMA, and where dProd/dt
represents the rate of production of either metabolite from primidone
and dmetab/dt is the rate of elimination of
either phenobarbital or PEMA.
For phenobarbital, the rate of production is given as follows:
whereas for PEMA it is as follows:
The dmetab/dt rate of
metabolism for phenobarbital is given as follows:
and for PEMA it is as follows:
where Kmet is the first order rate
constant (min1) of phenobarbital metabolism and
Vpema (mg/min) and
Kpema (mg/ml) are the Michaelis-Menten
constants for PEMA metabolism. Vpema is
also scaled to the 0.7 power of body weight. Documented slow metabolic rate of phenobarbital was better assumed by a first order kinetic rate,
which is a modification of the saturable Michaelis-Menten equation when
affinity is low (Igari et al., 1982).
Additionally, the model accounts for enzyme induction resulting from
the accumulation of phenobarbital in the liver. This induction only
affects the metabolism rates of primidone and PEMA and no other portion
of the model. The induction of phenobarbital metabolism by
phenobarbital itself has been shown to be insignificant (Maniara
et al., 1988). Therefore, when applicable, the induced rates
were multiplied by a factor (1 + Kind)
where Kind is calculated as follows:
Finally, the value of Kind was
lagged for 2 hr in all cases to allow for the induction process to take
course.
Model Parameters.
Body weights were averaged as 70, 0.133, and 0.035 kg for humans, rats,
and mice, respectively. The remaining physiological parameters were
obtained from a report prepared by the international Life Sciences
Institute as shown in table 1 (Risk
Science Institute, 1994).
The concentrations of all chemicals in the brain tissues of rats and
mice were sensitive to the values for FR and DR.
These parameters were optimized to the available brain tissues data, as
shown under Results. The value of
Bplasma was set at 0.438 in all model
simulations of brain data (Igari et al., 1982).
Volume of blood (Vblood) was calculated
based on plasma volume (Vplasma) as
follows:
where hematocrit was set at 0.45 for all species
(Igari et al., 1982).
The tissue/blood partition coefficients for the primidone and PEMA
submodels were calculated from their respective
n-octanol/water partition coefficients
KOW for all species according to a
previously published procedure (Poulin and Krishnan, 1995a). Their
procedure depends on profiles for water and lipid contents (neutral and phospholipids) of all tissues in each species. Hence, tissue
composition data were collected from available literature for humans,
rats, and mice. For human tissues, water and lipid contents were
collected from literature for blood, liver, lung, muscle, kidney,
brain, and adipose tissues (Poulin and Krishnan, 1995b). Water content of human GI tissue was assumed to be 0.7 of tissue weight, whereas total lipid and phospholipids contents were calculated to be 0.012 of
total tissue weight and 0.39 of total lipid contents, respectively (Nakazawa et al., 1977). Water content of human skin tissue
was assumed to be 0.7, whereas total and phospholipid contents were obtained as 0.1 of total tissue weight and 0.53 of total lipids contents (Gray and Yardley, 1975). Human heart tissue contents and
profile were assumed to be similar to muscle. Tables
2 and 3 are
listings of the water and lipid contents in rats and mice tissues,
respectively. Table 4 lists the partition coefficient used in the model
for all three chemicals in human and rodent tissues.
The partition coefficients for phenobarbital in rat tissues were
available in the literature (Igari et al., 1982). However, for verification purposes, phenobarbital partition coefficients were
also calculated based on its KOW value and
compared with corresponding values for rats (table
4). The parameters agreed in all cases
except for the fat tissue; the calculated partition coefficient was 10 times higher than the experimentally observed one. For this reason, all
calculated fat tissue/blood partition coefficients were divided by 10 for all species. In this manner, order of lipophilicity of the
chemicals (PEMA < primidone < phenobarbital) is also
maintained. The discrepancy between experimental partition coefficients
and those predicted in the adipose tissue may be attributed to the
physical and chemical differences between octanol and lipids, which
constitute most of the adipose tissue. The deviation is less obvious in
other tissues where lipids is not a major component.
Simulation Software.
The model was constructed using SCoP Simulation Control Program version
3.51 (Simulation Resources, Inc.) on a Silicon Graphics workstation.
The SCoP program uses an optimization algorithm called PRAXIS, for
"Principal axis method." The optimization method is included in the
simulation package as a SCoPfit program.
| |
Results |
Estimation of Metabolic Parameters.
The purpose of this work was to estimate the metabolic constants for
primidone and its two metabolites by applying PBPK modeling. Derived
metabolic constants are related to the metabolism of primidone to
phenobarbital and PEMA, metabolism of phenobarbital and PEMA themselves, and phenobarbital enzyme induction constants
(Vind and
Kmind) for each species. The
interdependencies of these constants is a reflection of the
relationships between the parent chemical and its metabolites.
Therefore, estimation of these metabolic constants can only be achieved
for data where the levels of the three chemicals are measured
simultaneously. Once data were available, the metabolic constants were
estimated simultaneously by least square optimizing methods of the
model's simulations to data.
Metabolic Constants of Primidone and Its Metabolites in Humans.
Primidone and its metabolite plasma levels were determined in two
studies. In the first study, a human subject (subject A) received a
single oral dose of 600 mg (Sato et al., 1986). Two other
human subjects (subject B and subject C) were given a single primidone
oral dose of 500 mg in the second study (Baumel et al., 1972). Analysis of subject A's plasma showed the presence of the three
chemicals in contrast to subjects B and C, who only produced levels of
primidone and PEMA in their plasma. The model-derived metabolic
constant for all three human subjects is given in table 5. Fig. 3
is an illustration of the model simulation with optimized metabolic
constants for subject A's data. Fig. 4
depicts the model simulations for subjects B and C. Comparing all human
parameter values indicated that all subjects differed in their
abilities to absorb primidone. Although absorption rates changed among
subjects, extended comparisons of their metabolic profiles are still
adequate because these profiles are controlled by the descending parts of the pertinent plasma concentration curves. Hence, further analysis of the parameters in table 5 implies that subject B was not able to
produce phenobarbital in contrast to subjects A and C. Although subject
C did not show plasma levels of phenobarbital, the model best fit to
the plasma levels of both primidone and PEMA in this case had to yield
a value for the metabolism of primidone to phenobarbital. This best fit
resulted in a least square value of 0.327 compared with 0.513 when the
production of phenobarbital was set to zero in this case. The fact that
phenobarbital was not detected in the plasma of subjects B and C may be
attributed to the deficiency of subject B to produce it
(Vmaxph = 0) and the efficiency of subject
C to clear it (high Kmet). Furthermore, as
shown in table 5, the apparent maximum velocities of primidone
metabolism to PEMA were 0.0026, 0.02, and 0.01 mg·min1·kg1 for
subjects A, B, and C, respectively. Once more, these values indicate a
wide range of variability among these three individuals.
View larger version (34K):
[in this window]
[in a new window]
|
Fig. 3.
Model simulations (solid lines) against
plasma concentrations of primidone (a), phenobarbital (b), and PEMA (c)
for one human subject (subject A).
Data were obtained from Sato et al., 1986.
|
|
View larger version (20K):
[in this window]
[in a new window]
|
Fig. 4.
Model simulations of primidone (solid lines)
and PEMA (dashed lines) for human subject B (a) and human subject C
(b).
Phenobarbital was not detected in either case. Data were obtained from
Baumel et al., 1972.
|
|
Metabolic Constants of Primidone and Its Metabolites in Rats and
Mice.
Primidone and its metabolite plasma concentrations were determined in
groups of albino rats at single gavage doses of 125, 250, and 500 mg/kg
(Baumel et al., 1973). Fig. 5
depicts the model simulations in comparison with data when optimization
of the metabolic constant is exercised. Additional optimization of FR
and DR resulted in the model simulations to the brain tissue data as
shown in fig. 5d. The overall optimized constants in rats
are given in table 6.
View larger version (55K):
[in this window]
[in a new window]
|
Fig. 5.
Model simulations of plasma concentrations
of primidone (a), plasma concentrations phenobarbital (b), plasma
concentrations PEMA for Wistar rats (c), and brain tissue
concentrations of primidone (d) of three different single primidone
gavage doses.
Solid, dashed, and dotted
lines are model simulations of the 500, 250, and 125 mg/kg primidone gavage doses, respectively. Triangles,
squares, and circles depict data
corresponding to initial gavage doses of 500, 250, and 125 mg/kg of
primidone, respectively. Data were obtained from Baumel et
al., 1973.
|
|
Simultaneous measurements of plasma levels of the three chemicals were
also determined in Swiss-Webster mice (Leal et al., 1979).
The animals in these experiments received a single gavage dose of 50 mg/kg. Fig. 6 represents the model
simulations of the plasma levels of the three chemicals in comparison
with the published data for mice. Additional simulations of the brain
tissue of primidone levels from the same study are also provided in
fig. 6d. Optimized parameters derived from these simulations
are given in table 7.
View larger version (41K):
[in this window]
[in a new window]
|
Fig. 6.
Model simulations (solid lines) of plasma
concentrations of primidone (a), plasma concentrations of phenobarbital
(b), plasma concentrations of PEMA (c), and brain tissue concentrations
of primidone (d) of a single gavage dose experiment in Swiss-Webster
mice.
Data were obtained from Leal et al., 1979.
|
|
Parameter Sensitivity Analysis.
Sensitivity analysis reflected the regions along the simulated curves
where the parameter variability is of significant effect. Hence,
absorption first order constant (Kabs) and
the GI partition coefficient are most sensitive at the beginning of
each simulation. As simulations proceeded, metabolic constants
increased and absorption-related ones decreased in sensitivity.
| |
Discussion |
Metabolism of primidone varied widely among three human subjects,
rats, and mice as predicted by PBPK modeling. In all cases presented in
this study, rats were the most capable species of producing
phenobarbital from primidone. This is illustrated by computing their
apparent metabolic rate
(Vmax/Km) so
that the mathematical interdependence of
Vmax and Km is
avoided. Rats had an apparent metabolic rate of 2.5 compared with 0.94 ml·min1·kg1 for
mice and close to the highest one computed for humans (subject C at 2.4 ml·min1·kg1).
Moreover, the model simulations suggested that mice had the lower rate
of phenobarbital clearance among all the investigated cases. This lower
rate of phenobarbital clearance results in a steady state level of
phenobarbital in plasma at a higher value in mice than rats when
primidone is administered chronically. Fig.
7 shows the simulations of the steady
state levels of primidone and its metabolites in rats and mice when a
daily gavage dose of 50 mg/kg of primidone is administered for a week.
The figure shows about a 2-fold increase of the plasma levels of
phenobarbital for mice than the predicted levels in rats. The
continuous presence of this level of phenobarbital in mice in a chronic
carcinogenicity study of primidone may partly explain the higher
potency of the drug to mice than rats. This conclusion is in agreement
with a recent NTP study on the carcinogenicity of primidone, which
found clear evidence that the chemical was a hepatic carcinogen in
B6C3F1 mice (at dietary levels of 30, 65, or 150 mg/kg) but not in Fischer 344 rats (at dietary levels of 25, 50, or 100 mg/kg) (NTP, In Press).
View larger version (26K):
[in this window]
[in a new window]
|
Fig. 7.
Model simulations of primidone (dotted line)
and phenobarbital and PEMA (solid lines) for 1 week.
(a) A daily gavage dose of 50 mg/kg of primidone in
mice; (b) a daily gavage dose of 50 mg/kg of primidone
in rats.
|
|
The metabolic profile of primidone varies widely among humans, which
may indicate the presence of sensitive human populations that may
produce greater amounts of primidone metabolites. To illustrate this
point, the levels of the chemicals in human plasma are depicted in fig.
8 in the case if subjects A, B, and C
received a daily primidone gavage dose of 500 mg, respectively. The
simulated levels of primidone and phenobarbital in subject A follows a
sustained increase in blood. This finding results from the slower rates of primidone and phenobarbital metabolic clearances from the blood as
indicated by their model-derived values (see table 5). In the other
subjects, primidone levels sustain a constant level, and phenobarbital
was not detected. Therefore, extrapolating the NTP carcinogenicity
study of primidone, fig. 8 simulations imply that subject A may be at a
higher risk of phenobarbital-induced cancer than the other two
subjects.
View larger version (35K):
[in this window]
[in a new window]
|
Fig. 8.
Multidose model simulations of (a) predicted
plasma levels of primidone in subject A, (b) predicted plasma levels of
phenobarbital (solid line) and PEMA (dotted line) for subject A, (c)
predicted plasma levels of primidone (solid line) and PEMA (dotted
line) for subject B, and (d) predicted plasma levels of primidone
(solid line) and PEMA (dotted line) for subject C.
In all cases, primidone was introduced into the model as a daily gavage
dose of 500 mg.
|
|
In this example, PBPK modeling provided a tool to derive the metabolic
constants of primidone and its metabolites when all chemicals were
measured in plasma. The model parameterization was partly computed
(partition coefficients) and partly optimized to statistically best fit
published data. As in any modeling effort, the issue of identifiability
is of concern. Is there enough data to allow the accurate estimation of
parameters? In general, different parameters are sensitive to specific
portions of the data curves. For example, parameters related to
absorption (e.g. Kabs) has to
fit the rising portion of primidone plasma levels. Metabolism parameters of primidone to phenobarbital (e.g.
Vmaxph, Kmph)
has to fit the descending portions of primidone plasma levels in
addition to the rising portion of the phenobarbital ones. Similarly,
the metabolism parameters for PEMA (e.g.
Vmaxpm, Kmpm)
has to fit, along with the phenobarbital ones, the declining portions
of primidone levels and their rising ones for PEMA. Once the metabolic
constants for primidone to phenobarbital and PEMA are fixed, the
metabolic constants for the later chemicals (e.g.
Kmet, Vpema,
and Kpema) have to fit their
plasma-declining portions of each chemical individually. The necessary
condition that the model had to fit the plasma levels of all three
chemicals simultaneously alleviated some of the concerns regarding this
question. Additional simulations at different levels of the parent
chemical in rats and humans, and in brain tissues of rats and mice,
added more confidence to the model-derived parameter estimations. To
further examine the model-derived metabolic estimates, a search of
available literature values yielded a range of 3 to 3.8 ml·hr1kg1 for the
metabolic rates of phenobarbital in humans (Yukawa et al.,
1992). This range is close to the model-derived metabolic rate in
humans (3.14 ml·min1 for subject A), which
corresponds to a value of 2.69 ml·hr1·kg1 assuming
subject A weighs 70 kg.
Abbreviations used are:
PEMA, phenylethylmalonamide;
PBPK, physiologically based pharmacokinetic;
GI, gastrointestinal.
![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
Top
Abstract
Introduction
![V<SUB><UP>plasma</UP></SUB>=44[<UP>Body weight</UP> (<UP>kg</UP>)]<SUP>0.99</SUP>](/content/vol26/issue6/fulltext/585/img016.gif)
DDT on the absorption of nutrients in vitro and on brush border enzymes in rat intestine.
Arch Toxicol
49:
131-138[Medline].
