0090-9556/93/2

DRUG METABOLISM

10I-0125S02.00/O
AND DisrosITloN

Vol.21,No.

Copyright

1993 by The American

Society for Pharmacology

and Experimental

Therapeutics

Priniedi,,

U.S.A.

A PHARMACOKINETIC

STUDY An Analog
SCHMITZ,

OF

PHENYLCYCLOHEXYLDIETHYLAMINE of Phencyclidine
E. M. LANDAW, A. S. CHANG, S. RAMAMURTHY, AND

A.

K.

CHO,

M.

HIRAMATSU,

DEBRA

A.

D. J. JENDEN Departments of Pharmacology (A.K.C., M.H., D.A.S., A.S.C., SR., D.J.J.) School (Received May 27, 1 992; and Biomathematics of Medicine accepted September 1 1992) (E.M.L., D.J.J.), University of California-Los Angeles,

ABSTRACT: The pharmacokinetics of three phencyclidine analogs-phenylcycompartment model (two pools for each analyte) that included sat-

clohexyl-diethylamine (PCDE), phenylcyclobuxylethylamine and phenylcyclohexylamine (PCA)-were determined

(PCE),
in rats after

urable components for the conversion PCA gave the best fit to the combined
ties

of PCDE to PCE and PCE to data. Despite large uncertain-

intravenous administration of each drug. Because PCE and PCA are major metabolltes of PCDE, their plasma levels were also measured administration of PCDE. termined after administration
after

for some

clearances, veiled
estimated

Similarly, PCA concentration was deof PCE. The data were combined and procedures using
compartmental

microparameters, useful estimates were obtained for distribution volumes, and fraction of PCDE or PCE conto PCE and PCA in vivo under nonsaturating conditions. The

analyzed

by nonlinear

regression

value

fraction of PCDE converted to PCA and the apparent Km for the conversion of PCDE to PCE were comparable to values in vftro with microsomal
preparations, suggesting that met-

and noncompartmental of PCDE metabolism. slants for the

models to determine the kinetic parameters The object was to estimate the kinetic consequence, PCDE to PCE to PCA. A 6-

obtained

metabolic

abolic studies In vitro provide reasonable predictors of the biotransformation process in vivo for this class of compounds.

In a previous study of PCP analogs, PCDE (fig. 1), a ringopened analog, was found to have very little affinity for the MK 801 binding site on the NMDA receptor, the purported site of
action for these compounds in vivo were comparable (1). In contrast, the actions of PCDE to those of PCP itself (2, 3). One

Materials

and Methods

explanation for the discrepancy between actions of the compound in vitro and in vivo was the conversion of PCDE to an active metabolite (3). The proposed active metabolite is PCE, a compound slightly more potent than PCP itself (2, 3). PCE is converted to PCA, a compound with very low affinity for the NMDA receptor binding site and a correspondingly low activity in vivo (2, 3). The conversion of PCDE to PCE and PCA is therefore an important process in the pharmacology of PCDE. We describe herein results of a pharmacokinetic study of the conversion of PCDE to PCE then to PCA in the rat, the most common behavioral subject. In the study, equimolar doses of PCDE, PCE, or PCA were administered separately to three groups ofrats, and plasma concentrations ofall three compounds
were measured in each rat at specified times after administration.

The

plasma

regression
compartmental

concentration data were analyzed techniques using various pharmacokinetic models to assess the relationships
conversion and alternate pathways, and

by

nonlinear and nonbetween the

to examine

sequential

different

approaches

for the analysis

of such

data.

This study was supported by U.S. Public Health Serve Grants CA-16042 (E.M.L.), DA 02431 (A.K.C.), and MH 17691 (D.J.J.). I Abbreviations used are: PCP, phencyde; PCDE, phenylcyclohexyldiethyamine; PCE, phenylcyc$ohexylethylamine; PCA, phenylcyclohexylamine; NMDA,
N-

from the Research Technology Branch of the National Institute on Drug Abuse (Rockville, MD). RCA was synthesized by published methods (4). Deutenum-labeled internal standard for PCDE was synthesized by reduction of N-acetyl-phenylcyclohexylethylamine with lithium aluminum deuteride, and the internal standard for PCA was obtained from 2H5-bromobenzene carried through the procedure of Maddox et a!. (5). The purity of each compound was confirmed by elemental analysis, OC, and MS. Animals. Male Sprague-Dawley rats weighing 260 to 330 g (Charles River Breeding Laboratories, Wilmington, MA) were used in the experiments. Vascular-Access-Ports, model SLA, from Norfolk Medical Products (Skokie, IL) were implanted 24 hr before the experiment by procedures described previously (6). These ports allow sequential collection of blood from the jugular vein of a conscious and moving animal. In the experiment, the ports were flushed with heparinized saline, the drug injected, and the port again flushed with heparinized saline. Serial blood samples were taken at 5, 10, 20, 30, 60, 120, 180, 300, and 420 mm after administration. In each collection, 0.6 ml of blood was collected and set aside before collection of the actual sample (0.5 ml). The 0.6 ml was then reinjected and the port washed by an additional 0.5 ml of saline. The blood samples were centrifuged at 15,000 rpm for 3 mm and the plasma stored at -80#{176}C until assayed. The three compounds were administered to separate groups of animals as the hydrochloride salts in saline at doses of 35.7 Mmol kg. Plasma Analysis. Internal standards in 60% perchloric acid (0. 1 ml) were added to 0.2 ml ofplasma, mixed, and centrifuged. The supernatant

PCDE

and PCE were obtained

was transferred

to a 20 ml culture

tube containing

1 ml of 1.5 M sodium

methyl-o-aspartate; CV, coefficient SC, Schwarz criterion.

of variation;

AIC, Akaike information

criterion;

ic
PCDE 125

9J
PCE FIG. l.Structures. PCA

Send reprint r.qu.sts

tkiversity 1735. of California-Los

to: Arthur K. Cho, Ph.D., Department of Pharmacology, Angeles School of Medicine, Los Angeles, CA 90024-

126
carbonate (pH 9.5) and 5 and Jackson, Muskegon, centrifuged at 2000 rpm transferred to a 12 ml

CHO ml ofglass distilled dichloromethane (Burdick MI). The mixture was shaken for 10 mm and to separate the layers. The organic layer was conical tube, and most of the solvent was

ET

AL.

evaporated
transferred

under a stream
to an autosampler

of nitrogen.

The residue,
Glass

about 0. 15 ml, was

Co., Milville, NJ) for

vial (Wheaton

GC/MS

analysis.

A Hewlett-Packard 597lA GC/MS system, operating in the selected ion mode at 70 eV, was used to analyze the samples. The GC contained a methyl silicone (0.33 Mm film thickness) capillary column that was 12.5 m in length and 0.2 mm in internal diameter. The chromatograph was heated with a temperature ramp of80 to 2l0C with a rate of 3YC/ mm. The temperatures of the injection port, transfer line, and mass analyzer were l90, 280, and 180C, respectively (3) The m/z values for the internal standards used in the selected ion monitoring mode were mhz = 190 for PCDE (PCDEd2) and I 37 for PCA (PCAd5). PCDEd2 was used as the internal standard for PCE. The m/z values used to

FIG. 2. Compartmental This

ineq.

mode/for is described

the serial toPCA.

conversion

ofPCDE

to PCE

block
I.

diagram

by the system ofdifferential

equations

monitor

analytes

were

188, 160, and

132 for PCDE,

PCE,

and PCA,

respectively. Concentrations were determined from standard curves that were run with each analysis. Recoveries of amines from spiked plasma samples were more than 90% and varied by less than 10%. Data Analysis. Plasma levels of the three analytes varied over a range of more than three orders of magnitude, and it was necessary to assess the error structure of the data prior to evaluation of specific pharmacokinetic models. Analysis of replicate assay data demonstrated that meas-

urement

error variability

had a nearly

constant

CV (approximately

6%

for all three analytes). Interindividual variability was assessed for each compound at each sample time by the variance of measurements across rats. Log-log plots of the variance against the mean yielded slopes not markedly different from 2.0 for each of the three compounds, and there were no significant differences among the three straight line fits. This indicated that the population standard deviation was also proportional to the mean (CV approximately 30%). Because a logarithmic transformation yields a more homogeneous variance than the raw data when CVs are constant (7), all subsequent least squares analyses used constant unit weights (i.e. were unweighted), with the logarithm ofthe concentration as the dependent variable. Because the three analytes did not appear to differ in their error structure, identical constant weights were used for each compound for those regressions involving simultaneous fitting of more than one analyte. Compartmental Analysis ofPooled Data. Because of the known irreversible conversion sequence among the three compounds and an apparent biexponential decay curve for each compound alone, the basic model likely to provide a reasonable fit to the data consists of three sequential central compartments (PCDE, PCE, and PCA), each with reversible entry into a peripheral compartment with first-order kinetics (fig. 2). The system ofdifferential equations for this model is (fig. 2): -(k10 -(km
+ k12 + k14)q + k23 + k25)q2 + k..1q4 + k,2q1 + k52q5

compartmentj. Note that k1q1 represents the metabolic flux for conversion of PCDE to centrally appearing PCE and k23q2 represents the flux for conversion of PCE to centrally appearing PCA. Also the leaks k10, k, and k represent all remaining irreversible metabolic and excretory pathways for clearance of the analytes. Thus we assume that all irreversible formation and elimination steps occur exclusively in those tissues kinetically associated with the central pool. We note that the lumped constant k10 includes any portion oftotal PCDE elimination kinetics due to the formation of PCE that undergoes further irreversible elimination
without

ever having

appeared

in the systemic

circulation.

Similarly,

any

portion of PCE elimination due to conversion to PCA that does not appear in the systemic circulation is included in the lumped constant k. Plasma concentrations (the dependent variables, assumed to represent samples from the central pool) at time t are given by c,(t) = where V1, V2, and V3 are the central volumes of distribution for PCDE, PCE, and PCA. With the observed concentrations in nM units and V1 in units ofliter. the initial (Ja& all zero for each injection except q1(0) = 35,700 nmol.kg for the PCDE injection data, or q2(0) = 35,700 nmol.kg for the PCE injection, or q3(0) = 35,700 nmoI.kg for the PCA injection. Because of the large number of parameters in the compartmental model, the limited number ofrats studied for each injection, and the fact that each rat could be injected with only one compound, we concluded that a population modeling approach would be problematic and elected to use a naive pooled-data method (8, 9). That is, we fitted a single compartmental model simultaneously to the data from all three sets of injection experiments to estimate mean population parameters. This is in contrast to the noncompartmental estimates, described subsequently, based on fits of simpler models to individual data for each rat. The nonlinear regression program BMDPAR (10) was used with the DIFEQ

dt d2

feature

to fit a numerical

solution

of the differential

equation

model

to

-(k k4q

+ k)q3

+ k23q2

( 1)

the data from all experiments simultaneously. The basic system described in eq. 1 was fitted to the data, and F tests for nested models were used to compare this to fits of simpler models (i.e. with peripheral compartments removed) or to fits of more complex models (i.e. saturable kinetics as described in Results for one or more metabolic conversions). To compare fits among several candidate models we also used the AIC [AIC = N.ln(RSS) + 21] and SC [SC = N.ln(RSS) + ln(N).Pj, where Nis the number of data points, RSS is the residual sum of squares, and P is the number of estimable model parameters (1 1, 12). Each criterion may be considered a measure of parsimony, and the model having the smallest criterion value is considered best (13). Because eq. 1 can be solved analytically, we compared the BMDPAR fits of the basic model with those using the analytic solution and program BMDP3R (10). The

,
dj6

k25q2 kq3

/c63q6,

where

q1, q2, and q3 are the quantities of PCDE, PCE, and PCA, respectively, in the central compartments; q4, q3, and q6 are their respective quantities in the peripheral compartments; and the k,1 are the fractional transfer or conversion coefficients from compartment i to

parameter estimates and standard errors were very similar comparisons, and only BMDPAR fits are reported.
Functions ofparameters and their standard errors

for these

were also estimated.

PHARMACOKINETICS With subscripts total clearances 1, 2, and are CL1 CL3 The
2)

OF

A PCP

ANALOG

127

3 denoting

PCDE,

PCE,

and PCA

respectively,

V1(k, V3k.

+ k,,+),

1, 2

(2)

pool. As V, = CL. MRT, bounds on the steady-state volume of distribution can be computed to test the sensitivity of V, estimates to assumptions about elimination sites. These lower and upper bounds on the true distribution volume are: =CL3.MRD = CL3 (11)
.

fraction PCE

ofcentral in concept

PCDE

converted circulation k2/(k0 to centrally k23/(k

to centrally oftotal (14) and

+ k2).

appearing is given by

PCE

(f... that

/eT

MRTT

is equivalent

to the fraction

clearance

of PCDE

furnishes

to the systemic
J-.2

(3) PCA is (4)

The fraction

of PCE converted
f2-.3

appearing + k23).

If we assume that the appearance of PCA from administered PCDE arises only from the metabolism of centrally appearing PCE, then the fraction ofPCDE eventually converted to centrally appearing PCA is f... 3 = fI_.2 . f2..3. Finally, the apparent steady-state distribution volume ( Vi,)
is defined as the steady-state mass of analyte in the whole system (i.e. central plus peripheral compartments) divided by the plasma concentration ofanalyte. With steady-state maintained by a constant flux of analyte entering only the central pool, the predicted ratio ofperipheral compartment mass to central compartment mass is k14/k1 for PCDE, k25/k52 for PCE, and k/k3 for PCA. It follows then that

v,_r3=v1(l

+)

(5)

Clearances, conversion fractions, and distribution volumes were also computed for fits to subsequent models that included saturable kinetics for the conversion of PCDE to PCE and/or PCE to PCA (see eqs. 13-16 and accompanying discussion). For these more complex models, the k12 and k23 terms in eqs. 2-4 were replaced by the corresponding ratio of the metabolic conversion flux to the amount of parent compound in the limit of vanishingly small concentrations; these limiting ratios are denoted ki2 and k3. Therefore estimated clearances and metabolic conversion fractions for saturable models are meant to apply when concentrations of analytes remain well below their Michaelis (half-saturation) constants. IndividualMultiexponential Fits and NoncompartmentalAnalysis. To complement the pooled-data compartmental analyses, we used alternate noncompartmental methods for linear systems to analyze individual rat data. For PCA injection experiments, a multiexponential model c3(t) = Aexp(-X,t) was fitted to the PCA concentration data for each rat, and the area under the curve (AUC), area under the moment curve

For PCE injection data, a multiexponential model for PCE and a second multiexponential model for metabolite PCA (constrained to equal zero at time 0) were fitted to the data for each rat. Similarly for PCDE injection data, models were fitted to the parent PCDE and metabolite PCE data for each rat; PCA concentrations were not used in these noncompartmental analyses. Constrained biexponential, triexponential, and polynomial exponential models were tried for the metabolite curves, and no model fitted better than a difference of two exponentials; no attempt was made to constrain the time constants for the fit for the metabolite curve to match the time constants for the parent compound. Central volumes, clearances, bounds for MRT, and bounds for steadystate distribution volumes were estimated by computations analogous to those used for PCA. As the standard errors for each noncompartmental parameter were roughly constant across rats, population means were estimated by unweighted averages of the point estimates, and precision ofeach such mean was measured by the usual SEM. Additionally, given that the dose in moles/kg was identical for all three analytes, a bioavailability calculation was used to estimate the population mean fraction of PCDE converted to centrally appearing PCE and mean fraction of PCE converted to centrally appearing PCA:
fI-.2

AUC12/AUC22 AUC23/AUC33,

(12)

f2-.3

(AUMC),

central

volume,

and total clearance AUC

=

were estimated:
(6)

A,/X,

AUMC

A,/X

(7)

where AUC denote the average across rats of the AUC for analyte j given analyte i injected. The validity of eqs. 12 depends on the usual linearity and stationarity assumptions plus an equivalent source constraint (15) between analyte injected intravenously and that produced metabolically and appearing centrally. Finally, eqs. 12 and the modelspecific counterparts in eqs. 3 and 4 underestimate the fractional formation of a metabolite if any portion of formed metabolite undergoes irreversible elimination (e.g. further metabolism or direct secretion into bile) prior to first appearance in the systemic circulation (14). A biexponential model (N = 2) was used in all cases for fining the decay curve for each parent compound, because higher order models did not improve the fit. Therefore, assuming the compartmental model of eq. 1 is correct, for each rat receiving a PCDE injection k14, k., and (k10 + k2) were estimated from the fit to the parent compound; similarly k25, k52, and (k20 + k23) were estimated for each rat receiving a PCE injection, and k, /(63, and k for each rat receiving PCA. Again, population mean estimates were made by simple unweighted averages across rats. Finally, population mean estimates for k0, k12, k, and k23 were made using the relationships in eqs. 3 and 4 and the population mean estimates for (k10 + k2), (k20 + k23), f1.2, and f2...3. Standard errors for those population means computed derived as functions method. of other population mean estimates were by the delta

V3 CL3

35,700/s 35,700/AUC.

A,

(8) (9)

Results
Compartmental described

Letting MRT#{176}and MRT denote the lower and upper bounds the mean residence time for PCA in the body given that it starts plasma, their estimates are (15, 16):

of
in

Analysis.
and

The
Methods

linear

in Materials and the

from MRT#{176} =
MR1
= AUMC/AUC (10)

six experiments were

for each

in which Groups
=

PCDE, of 5-6
data

6-compartment model was fitted to pooled data PCE, or PCA was adminand parent com-

istered pound time

concentrations

of metabolites

measured.
combination

rats
injected

were
and

studied
analyte Parameter

across
measesti-

ofanalyte 277 usable standard

We note
exclusively

that
from model

the true
the central with

MRT

MRT#{176}

pool (as assumed

= MRT

model
catenary

ofeq.

1), and true MRT

elimination

exclusively

when elimination occurs in the basic compartmental for the opposite extreme-a from the most peripheral

ured,
mates

resulting
and their

in N

points. errors

asymptotic

centage ofCV) for this basic model of table 1, and derived clearances,

as perare listed in the first column conversion fractions, and

(expressed

128

CHO

ET AL. 1 estimates Model 2 PCE - PCA

Saturation

TABLE
Summary Model No
saturation

ofmicroparameter I

(% CV)

PCDE

Model 3 PCE PCE - PCA Saturation

-.

Derived from Fits to mdividual Rats

-

RSS
No. of parameters AIC SC k0 (min) k4 (min) k41 (min) k2 V1(nmol.min.kg) KMI(nM)

49.77
14 1110.3 1161.1 0.0142 (27.9%) 0.0733 (24.9%) 0.0184 (17.8%) 0.0309 (16.0%)
-

41.17
15 1059.8 1114.2 0.00905 (22.7%) 0.0445 (31.6%) 0.0198 (22.2%) 0.0167 (15.7%)
-

40.56
16 1057.6 1115.6 0.00964 (24.3%) 0.0552 (28.1%) 0.0176 (19.0%) 0.0258#{176} (26.9%) 780 (45.5%) 13172 (57.8%) 0.00660 (22.9%) 0.0315 (28.1%) 0.0137 (21.1%)

0.0106 (36.6%) 0.0486 (7.2%) 0.0191 (12.2%) 0.0214 (18.7%)

-

(min)

k25 (min) k52 (min) k23

Vmax2

0.00413 (24.3%) 0.00968 (46.4%) 0.01045 (47.9%) 0.00668 (13.7%)

-

0.00543 (22.4%) 0.0234 (30.8%) 0.0129 (24.7%)

0.00529 (65.1%) 0.05 18 (19.8%) 0.0227 (10.3%)

0.0163
(17.5%) 171.4 (17.5%)

0.02 1 1
(20.7%) 156.2 (16.7%)

0.0216
(22.8%)
-

(nmol.min.kg)

K,

(nM)

2886
(23.7%)

2445
(23.3%) 0.0187 (15.9%) 0.0422 (34.6%) 0.0164 (12.9%) 0.0617 (23.8%)

k k
/(63

(min) (min)

0.0128 (16.9%) 0.0286 (52.5%)

0.0181 (15.7%) 0.0415 (35.1%)

(min)

0.0234
(40.1%) 1.49 (20.9%) 5.47 (1 1.2%) 4.67 (17.8%)

0.0215
(27.4%) 2.74 (17.9%) 3.65 (13.1%) 3.34 (16.9%)

0.0209
(26.4%) 2.29 (21.0%) 3.02 (15.6%) 3.25 (17.0%)

0.0398
(15.0%) 2.29 (5.8%) 2.54 (18.6%) 3.82 (10.4%)

V1 (liter.kg) V2(liter.kg) V3 (liter.kg)

a k.

distribution volumes are listed in the first column of table 2. The full 6-compartment model was distinctly better fitting than any simpler model lacking one or more peripheral cornpartments (p < l06 for all F tests). However, for a subset of the data corresponding to early times after PCE administration, there were systematic positive deviations of the observed PCE data and negative deviations of observed PCA data from their fitted values. This suggested that conversion of PCE to PCA may be saturable, and an expanded model was fitted in which all flux terms k23q2 in eq. 1 for conversion of PCE to PCA were replaced by steady-state k3q2 1 +
q2/ V2 -i---

where
I .4 K23

Vmax2
,

(14)

The two estimable parameters replacing k23 are V,,,2 and K,,,, respectively, the maximum flux (mass/time/kg body weight) and Michaelis constant for the conversion of PCE to PCA. This resulted in a significant improvement in the fit (F1262 = 54.7; p
< 10-6), and estimates are presented in the second column of

tables 1 and 2. The Michaelis constant for PCE to PCA conversion in vivo was 2886 683 (SE) nM. Assuming PCE concentrations in the linearizing range (i.e. well below the Michaelis
constant), the plasma
-

clearance

for all three min

.

analytes

under

this

(13)

than

model were 0.060 clearance for PCE

0.079

liter.

that

based

on model

(0.0792

0.0059

1 (0.059

kg, and the estimated liter. min . kg) fractionlarger of was 1 0.0040). The

PHARMACOKINETICS

OF

PCP

ANALOG

129
substrates
to PCE

TABLE

2
(%CV)

might
that the

be competitive
flux for PCDE

for the same

is
Vmazi q

enzyme

system,

so

_________________________________________ Summary ofderived parameter estimates Model 1 No saturation CL1 (liter.min CL2 (liter. min .kg)
.

Model 3 Model 2 PCDE -0 Derived from PCE - PCA PCE - PCA Fits to Individual Rats Saturation Saturation 0.0706 (6.1%) 0.0792 (7.2%) (6.3%) 0.0604 0.648 (6.4%) 0.750 (6.7%) 8.9 1 (10.8%) 10.26 (10.6%) 9.78 (1 1.6%)
-

(17)

K51 v(i

+ KM:VI
is Vma,2q1

KV2)

kf)

0.0673 (7.0%) 0.0591 (6.8%) (7.0%) 0.0599 0.685 (7.4%) 0.618 (10.2%) 7.46 (12.6%) 10.54 (14.6%) 10.37 (12.5%)
-

0.0813 (9.5%) 0.0839 (8.0%) (6.2%) 0.0609 0.728 (7.4%) 0.762 (6.3%) 9.47 (1 1.3%) 9.96 (10.2%) 9.8 1 (1 1.6%)
-

0.0703 (5.1%) 0.0557 (14.9%) (5.4%) 0.0598 0.657 (1 8.9%) 0.769 (19.4%) 8.25 (5.7%) 7.55 (13.5%) 9.38 (3.6%) 12.14 (7 0%)
.

and

the flux for PCE

to PCA

CL3 (liter.min
J-.2

.kg) K

v2(
are displayed

q1 + KMI V1

(18) q2

_____

_____

V2J

f2-.3

This modification was not significantly better than model 2 (F261 = 0.84, p = 0.4) and is not displayed. The plotted final fits for model 3 are nearly indistinguishable from those for model

v,:7

iter.kg (liter.kg) (liter.kg) (liter.kg) (liter.kg) (liter.kg)

2,

and the former

with

means

SE ofthe

observed

Vcr

ver

datain fig. 3 forinjectionsofPCDE, PCE, and PCA, respectively. A formal goodness-of-fit test indicated that the lack of fit mean squared error is about 1.9 times larger than the mean squared replicate error (F38.223 = 1 .89, p = 0.002). There was no gross systematic misfitting in the residuals, and this moderate lack of

I/UPPCf Lu

interindividual However,
significantly artifacts

fit may

be a result

of the usual problem associated with variability in the naive pooled-data interpretation
fit.

ignoring method. that the

VIr
I/UPPCT 3,3

10.25 (14.4%) 1 1.07 (4.9%)

this
of the

last

raises

the

possibility model may

improved pooled-data

fits for the saturation

also be

PCE

converted

to
below

centrally
saturation)

appearing
was 0.750

PCA

(again

assuming
using model

Comparisons with sults. Biexponential and point estimates

Materials

concentrations 2, in contrast

0.050

to 0.618 0.063 for model 1; the fraction of PCDE converted to centrally appearing PCE was similar for both models (0.648 0.042 for model 2 vs. 0.685 0.05 1 for model 1). Finally, the estimated central volumes for PCDE, PCE, and PCA (2.7 0.5, 3.7 0.5, and 3.3 0.6 liter. kg) were more similar to each other than those for model 1 (1.5 0.3, 5.5 0.6, and
4.7 0.8 liter. kg).

clearance, A and
the last systems

Individual Fits and Noncompartmental Remodels were fitted to individual rat data, were averaged across rats, as discussed in and Methods, for noncompartmental estimates of conversion fractions of PCDE to PCE and PCE to steady-state distribution volumes. These are listed in
of table tend 2 and, to agree assuming despite being based on a linear

column analysis,

mates
and

from

the

best-fitting

more saturable
that mean

closely with compartmental

linear

derived estimodels 2
compartmen-

3. Additionally,

the basic population

tal model

of fig. 2 is correct,

microparameters

A similar deviation of the data from the fitted model at early times following the injection of PCDE led to evaluation of a third model in which additionally the flux from PCDE to PCE is saturable, with k12q1 terms being replaced by kq1
1 + q1/V1

and central volumes were estimated column of table 1 . The most striking volume estimates tend to be in much saturable models than with the model

and are listed in the last result is that the central better agreement with the 1 estimates. on each individual to have the lowest
ofthe sample

( 1 5)

Of the parameters that could be estimated rat, the central volumes and clearances tended
estimation variability, nitude of population permitting variability.

simple assessments For example, the

magstand-

where
k4
12K

ard deviation
respectively,
Vm,i MI V1

(16)

liter. ard
0.0034,

min errors

of the point estimates of CL1, CL2, and CL3 were, 1 (N = 5 rats), 0.0204 (N = 6), and 0.0072 . kg (N = 5), whereas the median asymptotic stand0.008

resulted in a borderline improvement in fit over the model in which only PCE to PCA is saturable (F,26 = 3.95; p = 0.048). The estimates for this third model are presented in the third column of tables 1 and 2. The Michaelis constant for PCE to PCA in vivo was 13170 7610 nM, and the remaining estimates are generally close to those for model 2. AIC and SC in table 1 are measures of parsimony for model discrimination, each a function of the residual sum of squares plus a penalty term increasing with the number of parameters. The
both

This second

for individual clearance estimates were, respectively, and 0.0040 liter#{149}min .kg. Thus of the total variance among the point estimates, only about 18%, 4%, and 1 % could be accounted for as being due to estimation variability, suggesting that the majority of the observed clearance variability represents true population heterogeneity. In contrast, the sample standard deviations for central volumes V1, V2, and V3 were 0.30 (N = 5), 1.16 (N = 6), and 0.90 liter. kg (N = 5), with the median ofthe estimation variances being approximately 100%, 39%, and 190% of the variance of the point estimates.
0.0043, Discussion

AIC picked model 3 and the SC picked model by a small margin, in agreement with the marginal ment of model 3 over model 2. Finally the possibility was considered that PCDE

2 as best, improveand PCE

The sequential conversion of PCDE to PCE and then to PCA is a major metabolic pathway. In a study with microsomes in vitro, the sequence accounted for 57% ofthe metabolic products

130

CHO

ET

AL.

Measured

Injected

:
I I

PCDE

PCE
4.
C

PCA

0 S
3,

PCDE

8
-S

2
1

I!
I I U
U I U U I

100

200 urn.

300

400

100

200 Tim.

300

400

100

200

300

400

lime

4.
C

.2
S .

3.

PCE

I
I

I 100

I 200

I 300

I 400

I 100

I 200

I 300

I 400

lime
4.
C

Time

.2
S C

3.

PCA if
U U I

100

200
Tim.

300

400

FIG. 3. Model

3 (saturable

conversion

ofPCDE

to PCE

and PCE

to PCA). predicted

For each compound injected (left; dose = 35,700 nmol. kg) and each compound measured log1(concentration) from the model. Concentration is in nM, time is in mm. Circles represent all cases, the standard error bars for the means are covered by the plot symbols.

after that injection (top), solidlines indicate means of the observed log concentrations;

in almost

of PCDE appearance
f-.2f2-.3

(1).

The

corresponding

in

vivo

estimate

for

central

of PCA

from

PCDE

from

our

study

is the product

(i.e. 55.5 5.7 (SE) % for model 3 or 5 1 14% from the linear noncompartmental analysis). The good agreement among these estimates suggests that elimination in vivo of unchanged analyte is a negligible fraction of total clearance of PCDE and PCE, and that nearly all of the PCE and PCA produced metabolically is available to the systemic circulation. Other studies report that only 10% of administered PCE is excreted unchanged (17). PCDE is much less polar than PCE and may resemble PCP, of which only 2% is excreted as unchanged drug. Thus, these compounds are extensively metabolized. The metabolic sequence studied (i.e. the oxidative metabolism of the amine portion of the multiring compound) is a quantitatively important pathway. In the case of PCP, the reaction leads to the carboxylic acid (18-20), the dominant excretion product in many species. The analogous reaction of PCDE, however, is a metabolic activation. The metabolite, PCE, has an affinity for the NMDA receptor that is comparable to that of PCP, and produces a response in vivo that is 3 to 5 times greater

than that of PCP (21). Separate studies with PCDE and other analogs indicate that PCE must contribute to PCDE actions (.3). The kinetic parameters of PCDE metabolism are estimated here from the combined kinetics of the parent compound and its two metabolites. Although some microparameter estimates had large standard errors, the values for individual clearances,
conversion

fractions,

and

volumes

of distribution

were

more

precise. Chakrabarthi and Law (1 7) have compared the pharmacokinetics of PCE and PCP after intravenous doses of radiolabeled drugs. These workers reported a rate constant for total elimination of 0.020 min for PCE, in fair agreement with our estimate of M3 + k20 (0.0277 0.0050 min) using the present data. Using model 3, estimated steady-state distribution volumes were ofsimilar size among the three analytes (9.5 - 10 liter.kg) and at least 3-fold greater than central volumes. This similarity among the distribution volumes suggests that all three compounds have ready and comparable access to body compartments. One assumption for these model-based estimates of distribution volume is that eliminating tissues are associated exclu-

PHARMACOKINETICS

OF

A PCP

ANALOG

131
than the linearity
approach,

sively with the central pool. We have no direct evidence to either refute or support this standard assumption, but comparison of the noncompartmental upper and lower bounds for the true distribution volume (last column, bottom oftable 2) suggests the estimates are not very sensitive to violations of this assumption. In particular, the ratio of the upper bound to lower bound was only 1.5, 1.4, and 1.2 for PCDE, PCE, and PCA, with Vf corresponding to the worst-case extreme of elimination exclusively non-central and j,lowcr corresponding to the standard central elimination assumption. The saturation models gave apparent Michaelis constants for PCDE to PCE of 1 3 iM and for PCE to PCA, 2.4 tiM. The corresponding value for microsomal metabolism of PCDE was 38 M (1), a value that is in reasonable agreement when the 58% CV for the in vivo constant is considered. The large volumes of distribution, associated with ratios k14/k1 and k25/k52 exceeding unity, suggest that high tissue levels ofthe drug might be achieved relative to plasma. PCP has been shown to have brain/plasma ratios that range from four (22) to six (23), depending on the
dosage accounting plasma would schedule. The concentrations of the and compounds in the

may
implicit

be less important
in

assumptions
particularly

generally
if the

a noncompartmental

nonlinearity our present

significantly experiments.

affects

a larger

data

domain

than

in

References 1. M. Stefek, R. W. Ransom, E. W. DiStefano, and A. K. Cho: The alpha carbon oxidation of some phencyclidine analogues by rat tissue and its pharmacological implications. Xenobiotica 20, 591-

600 (1990).
2. H. E. Shannon: Evaluation of phencyclidine analogs on the basis of the discriminative stimulus properties in the rat. J. Pharm. Exp. Ther. 216, 543-551 (1981). 3. A. K. Cho, M. Hiramatsu, D. A. Schmitz, T. Nabeshima, and T.

Kameyama: Pharmacokinetic and pharmacodynamic of some phencyclidine analogs in rats. Pharmacol.

39, 947-953 (1991).

properties
Biol. Noveile (1975). Behav. voie

liver at early

time
for the

points

could than

therefore
saturation,

substantially
because KM would

exceed
levels

KM, in

4. P. Geneste, P. Hermann, J. M. Kamenka, and A. Pons: dacces aux isomeres des phenyl-2 cyclohexylamines au cyclohexane. Bull. Soc.Chem. France 1619-1626 5. H. V. Maddox, E. F. Godefroi, and R. F. Parcell: The phencyclidine and other 1-arylcyclohexylamines. J.
8, 230-235 (1965).

substituees synthesis
Med. and Chem. D. J.

of

apparent

be lower

in liver,

be underesti-

6. A. K. Cho,

M. Hiramatsu,

E. W. DiStefano,

A. S. Chang,

mated. It is also possible that the kinetics in each rat are truly linear but the apparent saturation is due to our data-analytic approach
in the presence of significant variation of kinetic parameters rats. Because it was not possible to collect complete data sets following administration of all three compounds to each individual rat, we used a pooled-data analysis that lumps interacross individual variation with measurement error. Nevertheless, there

Jenden:

Stereochemical

differences

in the metabolism

of 3,4-

7. 8.

9.

is moderately good agreement between the derived pooled-data estimates of volumes, clearances, and conversion fractions and the corresponding linear noncompartmental estimates using fits to individual rats. Even if saturation does occur, most of the observed concentrations (beyond early times) were probably low enough to be in the linearizing range, accounting for the agreement between our two approaches and providing additional confidence in the validity ofthese derived estimates. Quantifying population variability was not a primary goal of this study, but we note that the variance in individual clearance estimates across rats most probably represents true population heterogeneity, because estimation variability is a small fraction ofthis variance. On the other hand, the observed variance among central volume estimates may be due primarily to estimation error. It is worth noting that the compartmental analyses in this work utilized simultaneously all ofthe available data, specifically including plasma concentrations of PCDE, PCE, and PCA following administration of PCDE, and plasma concentrations of

10. 1 1. 12.

methylenedioxymethamphetamine in vivo and in vitro: A pharmacokinetic analysis. Drug Metab. Dispos. 18, 686-691 (1991). J. A. Ziven, and J. J. Bartko: Statistics for disinterested scientists. LifeSci. 18, 15-26(1976). L. B. Sheiner The population approach to pharmacokinetic data analysis: rationale and standard data analysis methods. Drug Metab. Rev. 15, 153-171 (1984). A. Racine-Pcon, and A. F. M. Smith: Population models. In Statistical Methodology in the Pharmaceutical Sciences (D. A. Berry, ed.), pp. 139-1 62. Marcel Dekker, Inc., New York, 1990. W. J. Dixon, ed.: BMDP Statistical Software Manual. University ofCalifornia Press, Los Angeles, 1990, pp. 395-423, 92 1-958. H. Akaike: A new look at the statistical model identification. IEEE Trans. Autom. ControlAC-19, 716-723 (1974). G. Schwarz: Estimating the dimension of a model. Ann. Statist. 6,
461-464 (1978).

13. E. M. Landaw,

and

J. J. DiStefano:

Multiexponential,

Part 2. Data sis and statistical considerations. Am. J. Physiol. 246 (Regul. Comp. Physiol.) 15, R665-R677 (1984). 14. K. S. Pang: A review of metabolite kinetics. J. Pharmacokin.

partmental

and noncompartmental

modeling.

multicomanalyInteg.

Biopharm. 13, 633-662 (1985). 15. J. J. DiStefano, and E. M. Landaw: Multiexponential, multicompartmental and noncompartmental modeling. Part 1. Methodological limitations and physiologic interpretations. Am. J. Physiol. 246 (Regul. Integ. Comp. Physiol.) 15, R665-677 (1984). 16. J. J. DiStefano, B. C. Chen, and E. M. Landaw: Pool size and mass

flux bounds

and quasi-identifiability

of the metabolites to fit similar models

both

of each. Attempts obtained after administration of only one of the analytes were less successful, frequently failed to converge, and always yielded larger variances of the parameter estimates when they did converge. These findings highlight the desirability of fitting pharmacokinetic models to a comprehensive data set rather than using a piecemeal fitting procedure to individual data sets, at least in catenary models of the kind considered here. Finally, the generally good agreement between the results of the two data-analytic approaches used in this investigation suggests that unless the assessment of population heterogeneity is a major objective of a study, the errors involved in lumping population variability with estimation error

after administration to individual data sets

The dispositional kinetics of analogue in rats. Eur. J. Drug (1985). A. K. Cho: Phencyclidine Noxide. Synthesis, decomposition and in vitro metabolism studies. Acta Pharm. Suec. 20, 181-192 (1983). 19. J. K. Baker, J. G. Wohlford, B. J. Bradbury and P. W. With: Mammalian metabolism of phencyclidine. J. Med. Chem. 24, 666-669 (1981). 20. Ward, D., Kalir, A., Trevor, A., Adams, J., Baillie, T. and Castagnoli, N: Metabolic formation ofiminiium species: metabolism of phencycidine. J. Med. Chem. 25, 492-494 (1982). 21. H. E. Shannon, and C. M. DeGregorio: Effects of N-substituted

Mathematical Biosciences 88, 1-14 17. S. Chakrabarthi, and F. C. P. Law: phencyclidine and its N-ethylamine Metab. Pharmacokinet. 8, 383-388 18. G. Hallstrom, C. H. Nguyen, and

relations (1987).

for catenary

models.

132

CHO analogs and metabolites of phencyclidine on avoidance behavior in the rat.J. Pharmacol. Exp. Ther. 218, 55-62 (1981). R. N. Pechnick, R. E. Poland, E. W. DiStefano, and A. K. Cho: High neonatal brain levels of phencyclidine following exposure during gestation. Substance Abuse 10, 185-1 88 (1989).

ET AL.
23. A. K. Cho, J. F. Brady, J. F. and J. N. Burstyn: Nitrogen oxidation and toxicity. In Toxicokinetics and New Drug Devdopment(A. Yacobi, J. P. Skelly, and V. K. Batra, eds.), pp. 149-159. Pergamon Press, New York, 1989.

22.