Journalof AnalyticalToxicology,Vol.
27, September2003
A Comparative Solid-PhaseExtractionStudyfor
the SimultaneousDetermination of Fiuoxetine,
Amitriptyline, Nortriptyline, Trimipramine, Maprotiline,
Clomipramine and Trazodone in Whole Blood by
Capillary Gas-LiquKJChromatographywith
'~ 9 9
Nitrogen-PhosphorusDetection
M.A. Martinez*, C. S~nchez de la Torre, and E. Almarza
Department of Chemistry, National Institute of Toxicology, Ministry of Justice, C/ Luis Cabrera 9, 28002 Madrid, Spain
Abstract 1
This paper reports the simultaneous detection of the seven
antidepressantsfluoxetine, amitriptyline, nortriptyline,
trimipramine, maprotiline, clomipramine, and trazodone in
whole blood at concentration levels of 100-2000 ng/mL by
gas chromatography with a nitrogen-phosphorus detector
(GC-NPD). A comparative and validation study using two
solid-phaseextraction (SPE)columns, Chem Elut and Bond
Elut Certify, were developed regarding their recovery,
precision, sensitivity, and matrix purification efficiency. The
Chem Elut columns, a diatomaceous earth, are closely related
to conventional liquid-liquid extraction. The Bond Elut
Certify columns, more recently developed in the market,
are a mixed SPE:reversed-phaseand cation exchange
sorbent. Recoveriesof the compounds using Chem Elut columns
at 500 ng/mL were in the range 30-50%, with intra- and
interassayprecisions of less than 9% and 17%, respectively.
Limits of detection (LODs) and quantitation (LOQs) ranged
from 13 to 146 ng/mL and from 44 to 485 ng/mL, respectively.
Recoveriesof the compounds using Bond Elut Certify columns
at 500 ng/mL were in the range 59-84% with intra- and
interassayprecislons of less than 8% and 11%, respectively.
LODs and LOQs ranged from 8 to 67 ng/mL and from 25
to 223 ng/mL, respectively. An excellent linearity was observed
with both extraction procedures from the LOQs up to 2000
ng/mL Higher recoveries, cleaner extracts, better sensitivity,
better precision, and less solvent consumption and disposal
were achieved for the screening of these antidepressantswith
the use of the mixed SPE Bond Elut Certify compared with
Chem Elut columns.
* Author to whom correspondenceshould be addressed.E-mail:mariantmart@lerra.es.
Reproduction(photocopying)of editorialcontentof thisjournalis prohibitedwithoutpublisher'spermission.
II
Introduction
Antidepressants are widely used for the treatment of a variety
of depressive states and other psychiatric disorders. An increase
in antidepressants intoxication led to the development of reli-
able analytical methods for their analysis. In systematic toxi-
cological analysis (STA),one of the main purposes is screening
analysis, and antidepressants are an important class of drugs in
forensic and clinical cases.
The extraction and concentration of the compounds of
interest from complex biological matrices play a great role
in STA. Solid-phase extraction (SPE) has been established
over recent years as a very effective method for sample pre-
teatment and clean-up for the pretreatment of biological
samples for clinical and toxicological drug analysis. SPE offers
several advantages over traditional liquid-liquid extraction
(LLE) such as higher selectivity, cleaner extracts, more repro-
ducible results, and absence of emulsion formation (1). How-
ever, the major applications of SPE have been limited to a
relatively clean biological sample such as plasma, serum, or
urine (2-7). Until now, only a few publications have described
SPE methods for whole blood (8--16). However, in practice,
whole blood is the sample encountered most frequently in
forensic cases.
The Chem Elut column is an SPE procedure closely related
to conventional LLE. It involves the absorption of the aqueous
phase on the diatomaceous earth, a porous material that acts as
a support for the aqueous phase. This provides a large surface
area for partition into an eluting solvent, which flows through
the immobilized specimen under gravity, eluting the analytes
of interest (17,18). However, large volumes of hazardous or-
353

ganic solvents are required. For STApurposes, in which acidic,
neutral, and basic substances may be present, this type of SPE
must be carried out with at least two columns: one for the
acidic and neutral substances and one for basic and neutral sub-
stances (1). The Bond Elut Certify column, a mixed-mode
bonded-silica column, can retain at a suitable pH acidic and
neutral substances by hydrophobic interactions with the alquil
chains (octylsilane,n-Ca) and the basic substances by interac-
tion with the cation exchange groups (benzenesulphonyl-
propylsilane) using one column. This is a great advantage in the
field of toxicology,in which blood specimens availablefor anal-
ysis are sometimes very small (15).
Capillary gas-liquid chromatography with nitrogen-phos-
phorus detection (GC-NPD)has proved to be a powerful tool in
the area of underivatized drug analysis. NPD is the most con-
venient in screening toxicological analysis because of its out-
standing sensitivity for the detection of traces of drugs and
negligible interference from non-nitrogenous compounds, both
endogenous and exogenous (19).
In our study, antidepressants were submitted in parallel to
SPE with Chem Elut and Bond Elut Certify columns. Mepiva-
caine and prazepam were used as extraction and chromato-
graphic standards, respectively.
We report a simple and reliable GC method with NPDwithout
derivatization as part of STAthrough a comparative validation
study of two SPE procedures. This method allows the determi-
nation of blood therapeutic and toxic concentrations of seven
antidepressants belonging to the main classes of these drugs:
tricyclic and tetracyclic antidepressants and selectiveserotonin
reuptake inhibitors.
The chemical structures of the seven antidepressants
fluoxetine, amitriptyline, nortriptyline, trimipramine, mapro-
tiline, clomipramine, and trazodone studied in this work
and the compounds mepivacaine and prazepam, used to
F~Ir: v
~ 3
I system for the manual mixed-mode bonded-
I~'L~
IVlgl'lt.~
0 cross-linked methylsilicone (0.11-pro film
W
Figure 1. Chemical structures of the seven antidepressants studied in this work.
354
Journal of Analytical Toxicology, Vol. 27, September 2003
control the whole analytical procedure, are presented in Figures
I and2.
Experimental
Materials
All chemicals (Merck, Darmstadt, Germany) and solvents
(Scharlau, Barcelona,Spain) wereof analytical grade. The tested
drugs were obtained from commercial suppliers and were of
pharmaceutical quality. Borate buffer (pH 9.0) and
dichloromethane/isopropanol (85:15) were used for the di-
atomaceous earth (Chem Elut) procedure. Phosphate buffer
(0.1M, pH 6.0), 0.01M acetic acid, acetone/dichloromethane
(1:1), and dichloromethane/isopropanol/ammonia (78:14:8)
were used for the mixed-mode bonded-silica (Bond Elut
Certify) procedure. Individual stock solutions (1 mg/mL) were
prepared by dissolving the appropriate amount of each drug
in methanol. These stock solutions were stored in glass tubes
at 4~ The extraction standard solution was prepared by di-
luting the stock solution of mepivacainewith deionized water
to 16 pg/mL. The chromatographic standard solution was
prepared by diluting the stock solution of prazepam with
methanol to 5 pg/mL.
Chem Elut CE 1010 columns (10 mL of column reservoir
volume) and Bond Elut Certify columns (130 mg of sorbent
mass, 3 mL of column reservoir volume) were both supplied by
Varian Sample Preparation Products (Harbor City, CA).
A pool of citrated human whole blood samples was obtained
from Hospital 12 de Octubre (Madrid, Spain) and verifiedto be
drug free. No interferences were found for the studied com-
pounds, and the samples were kept frozen at -20~ until used.
The blood was spiked with appropriate drugs. The concentra-
tion of each drug in spiked samples was 0.1,
0.5, and 2 pg/mL, respectively.
Instrumentation
A VAC-ELUT SPS 24 vacuum manifold
N(HI.11'~ silica SPE was purchased from Varian. A P-
Selecta sonication bath and a P-Selecta Cen-
tronic S centrifuge were both obtained from
Selecta (Barcelona, Spain).
The chromatographic analysis of the ex-
tracts was performed on a Hewlett-Packard
(HP) (Avondale, PA) model 5890 series II
GC equipped with an NPD and linked to an
HP 3396A integrator. A 25-m x 0.20-mm i.d.
fused-silica capillary column coated with
thickness) (HP) was employed. The carrier
gas was helium (Air Liquid, Madrid, Spain)
delivered at a column head pressure
of 195 kPa. Injection port and detector tem-
peratures were 280~ and 300~ respectively.
The splitting ratio was 1:20.The column tern-
Journal of Analytical Toxicology, Vol. 27, September 2003
perature was initially held at 180~ for 1 rain and then in-
creased to 300~ at 10~ The final temperature was held
for 3 rain. The total chromatographic time, including 2 rain of
equilibration time, was 18 rain. Insert liners silanized with
dimethyldichlorosilane/toluene (5:100) and packed with Su-
pelco silanized glass wool (Bellefonte, PA) were used. Special
care must be taken in silanizing the glass wool and insert liner
to obtain good chromatographic behavior. Furthermore, no
more than 100 injections should be performed without
replacing the insert liner, to prevent deterioration of the chro-
matographic system.
Samples
Human whole blood (100 rnL) spiked with the appropriate
drugs at three concentration levels 0.1, 0.5, and 2 1Jg/mL,was
sonicated in a sonic bath for 15 rnin at room temperature. The
spiked blood was submitted in parallel to SPE with Chem Elut
and Bond Elut Certify columns.
Extraction procedure using Chem Elut columns
The extractions were performed using a Chem Elut extraction
procedure used in our laboratory and described in our pre-
vious work (15). Briefly, aliquots (2 mL) of spiked blood, along
with the extraction standard (100 ~L of mepivacaine aqueous
solution of 16 pg/mL) were added with 7 mL of borate buffer
CH3 H~
MI~IVAC,A~I8
PI~AM
Figure2. Chemicalstructuresof the compoundsusedfor controlling the
analytical procedure.
..
I matographic standard) was fixed at 5 pg/mL.
ia
4,5
3 8
7 =
'ti
Figure3. RepresentativeGC-NPD chromatogramsof blank human whole blood (A) and blood
spikedwith 0.5 pg/mL(B) after extractingwith ChemElutcolumns. Peakidentification: I, caffeine:
1, fluoxetine; 2, mepivacaine (extraction standard);3, amitriptyline; 4, nortriptyline; 5, trim-
ipramine; 6, maprotiline; 7, clomipramine; 8, prazepam (chromatographicstandard);and 9,
trazodone.
(pH 9), vortex mixed, and loaded onto the Chem Elut columns.
Elution was carried out with 20 mL of dichloromethane/iso-
propanol (85:15). Eluates were evaporated under a nitrogen
stream. The extraction residues of blood were reconstituted
with 200 pL of methanolic solution of chromatographic stan-
dard (5 IJg/mL),and 2 lJL was injected for GC analysis.
Extraction procedure using Bond Elut Certify columns
The extraction procedure with Bond Elut Certify columns
used in our laboratory was described in our previous work (15).
It was performed on a VAC-ELUTSPS 24 vacuum manifold
system. Briefly, aliquots (2.5 mL) of spiked blood, along with
the extraction standard (125 IJL of mepivacaine aqueous solu-
tion of 16 IJg/mL),were added with 7.5 mL of phosphate buffer
(pH 6.0), sonicated, and centrifuged; the supernatants, equiv-
alent to 2 mL of whole blood, were loaded onto the previously
conditioned columns. After washing the columns with deion-
ized water, they were acidified with 0.01M acetic acid. Then
the columns were dried. Methanol (60 IJL) was added, and
the columns were dried under full vacuum. The elution was
performed first with 3.5 mL of acetone/dichloromethane (1:1)
and after with 3 mL of dichloromethane/isopropanol/ammonia
(78:14:8). The combined eluates were evaporated under a
nitrogen stream. The extraction residues of blood were
reconstituted with 200 ~L of methanolic solution of chro-
matographic standard (5 ~g/mL), and 2 pL was injected for
GC analysis.
Validation of the methods
Analytical method validation is the process used to establish
that a quantitative analytical method is suitable for a purpose,
in our case STA.
Antidepressants were added to human whole blood (100 mL)
to achieve concentrations of 0.1, 0.5, and 2 IJg/mL.The blood
was sonicated for 15 rain at room temperature. The spiked
blood was submitted in parallel to SPE with Chem Elut and
Bond Elut Certify columns.
Calibration curves were performed using antidepressant
methanolic standard solutions of five concen-
tration levels of 0.5, 1, 5, 15, and 30 lJg/mL,
B
and the concentration of prazepam (chro-
Antidepressant-to-prazepamarea ratios were
O measured, and the calibration curves were
| generated from least-squares linear regres-
'"
sion. The regression lines were used to calcu-
late the absolute recoveries (n = 6) of
individual antidepressants from spiked blood
at three concentration levels.
The intra-assay precision was assessed at
! three concentration levels by the extraction
and analysis on the same day of six spiked
blood samples for each level. The interassay
precision was assessed by analyzing, on two
different days, a set of nine spiked blood sam-
pies at a concentration of 0.5 pg/mL.
The limits of detection (LODs) and quanti-
tation (LOQs) were determined as the lowest
355
 Journal of Analytical Toxicology, Vol. 27, September 2003

concentration giving a response of 3 and 10 times, respectively, Results and Discussion

the average of the baseline noise defined from six control sam-
ples. Extraction and chromatography
The linearity of the method for each compound was checked The extraction procedures used in this study were the same
by preparing six replicates of the calibration curves at three as described previously (15). After developing our comparative
different concentrations, ranging from 100 to 2000 ng/mL, validation study, good chromatographic results were obtained
by addition of known amounts of each drug to human whole from both SPE procedures. Figures 3 and 4 show representative
blood. GC-NPD chromatograms of extracted human blood for each
method. Reliable separations of the seven an-
tidepressants and the chromatographic stan-
A B dard were obtained using the chromatographic
conditions described in short chromatographic
o times (16 rain). An excellent chromatographic
5 7
behavior with good peak shapes without
4" "
derivatization was obtained. Both procedures
provided extracts free of chromatographic in-
terferences in the areas corresponding to the
retention time of the studied compounds and
the control standards. The use of mepivacaine
!
as extraction standard and prazepam as chro-
i ' i matographic standard allowed us to control
the whole analytical procedure during the rou-
Figure4. RepresentativeGC-NPDchromatograms• blankhumanwholeblood(A)and blood tine performance of an STA.The nitrogen-con-
spikedwith0.5 pg/mL(B)afterextractingwithBondElutCertifycolumns.Peakidentification: taining structures of the antidepressants
I, caffeine:1, fluoxetine;2, mepivacaine(extractionstandard);3, amitriptyline;4, nort@Nline; fluoxetine, amitriptyline, nortriptyline, trim-
5, trimipramine;6, maprotiline;7, domipramine;8, prazepam(chromatographicstandard);and ipramine, maprotiline, clomipramine, and tra-
9, trazodone.
zodone, together with the use of SPE columns
are responsible for the high sensitivity ob-
tained using an NPD, as well as for the mini-
Table I. Extraction Recovery (%), Intra- and Interassay Precision (RSD %), mization of interferences.
Linearity (r2), LOD, and LOQ for the Seven Antidepressants in Whole Blood The chromatographic techniques may be
Using Chem Elut and Bond Elut Certify Columns and Analyzed by GC-NPD combined to obtain the maximum benefit from
Intra-assay Interassay them in the performance of STA. In accor-
Recovery(%) precision, precision, dance with the routine strategy for STAused in
Mean RSD (%) RSD (%) LOD LOQ our laboratory, in our study with the Bond
(0.5 pg/mL) (0.5 pg/mL) (0.5 pg/mL) Linearity, (ng/mL) (ng/mt) Elut Certify columns, we collected the com-
Compound (n = 6) (n = 6) (n = 9) F (n = 6) (n = 6) (n = 6) bined eluates, acidic-neutral and basic, in order
to complete a unique sample extract the
Fluoxetine GC-NPD screening/GC-MS confirmation and,
Chem Elut 38 6 9 0.999 123 411 in case it would be necessary,with HPLC-scan-
Bond ElutCertify 59 8 11 0.999 66 220 ning IN detection because some drugs cannot
Amitriptyline be detected in the GC-NPD screening, do not
Chem Elut 34 6 11 0.999 146
485 show good chromatographic behavior, or are
Bond ElutCertify 69 6 7 0.999 8 25
Nortriptyline partially decomposed.
Chem Elut 38 5 9 0.999 122 408
Bond ElutCertify 62 5 5 0.999 15 51 Validation
Trimipramine Table I shows a summary of the most repre-
Chem Elut 33 9 10 0.999 13 44 sentative analytical parameters studied in
Bond ElutCertify 65 8 8 0.999 11 37 this research. Recoveries of the compounds
Maprotiline using Chem Elut columns at 500 ng/mL were
Chem Elut 36 6 7 0.999 61 203 in the range 30-50%, with intra- and
Bond ElutCertify 59 6 6 0.999 37 125
interassay precision values (RSD) less than
Clomipramine
Chem Elut 30 7 7 0.999 84 281
9% and 17%, respectively. Recoveries of the
Bond ElutCertify 63 8 10 0.999 67 223 compounds using Bond Elut Certify columns
Trazodone at 500 ng/mL were in the range 59-84%,
Chem Elut 50 5 17 0.999 72 240 with intra- and interassay precision values
Bond ElutCertify 84 2 5 0.999 17 56 less than 8% and 11%, respectively. The
respective LODs and LOQs using Chem

356
Journal of Analytical Toxicology,Vol. 27, September2003
Elut columns ranged from 13 to 146 ng/mL and from 44 to
485 ng/mL. The respective LODs and LOQs using Bond Eiut
Certify columns ranged from 8 to 67 ng/mL and 25 to 223
ng/mL. The use of Bond Elut Certify reversed-phase and ion
exchange sorbent resulted in better LODsand LOQscompared
with the Chem Elut columns extraction procedure. The
linearity of the methods was satisfactory over the range
100-2000 ng/mL The calibration curves obtained using
quadratic regression gave the best correlation coefficients
r 2 > 0.999, (n = 6) for all tested drugs.
A comparative SPE validation study in the general
context of STA
One of the crucial points in STA is the extraction step
that must be able to extract a very wide variety of drugs,
ranging from very lipophilic to moderatelypolar, and exhibiting
acidic, neutral, basic, or zwitterion properties, preferablyusing
one column. The extraction method must be rapid, repro-
ducible, and yield good recoveries and clean extracts. In
addition, STA always requires a validated extraction--concen-
tration procedure, given the low concentration in which drugs
are present in biofluids and the fact that the latter contain
many interfering compounds. In our comparative SPE valida-
tion study, we have demonstrated that the use of the reversed-
phase and cation exchange sorbent Bond Elut Certifyrepresents
an improvement with respect to Chem Elut columns for
the screening of these antidepressants such as higher recov-
eries, cleaner extracts, better precision, and less solvent
consumption and disposal.
Furthermore, our assay encompasses the range of concen-
trations found in the majority of therapeutic or toxic situa-
tions published in the literature (20,21); in this way,the utility
of the proposed method is fully demonstrated.
Conclusions
The present study has shown that the use of Bond Elut Cer-
tify reversed-phaseand ion exchange sorbent resulted in a lot of
advantages, besides these seven antidepressants detection such
as better and more easily reproducible recoveries, cleaner ex-
tracts, and less solvent consumption and disposal, compared
with the Chem Elut columns extraction procedure.
The assays described herein allowed the determination of
seven antidepressants in whole blood with satisfactory recov-
eries and LOQs. The proposed method has good specificityfor
both toxicologicalscreening and therapeutic drug monitoring
of several antidepressants available commercially. These re-
sults complete our previous studies about antidepressant de-
tection during the performance of STA (15).
Acknowledgements
The authors wish to thank Hospital 12 de Octubre (Madrid,
Spain) for providingthem with the blood samples.
References
1. J.P.Franke and R.A. de Zeeuw. Solid-phase extraction procedures
in systematic toxicological analysis. J. Chromatogr. B 713:51-59
(1998).
2. T.J. Good and J.S. Andrews. The use of bonded-phase extraction
columns for rapid sample preparation of benzodiazepines and
metabolites from serum for HPLC analysis.J. Chromatogr. Sci. 19:
562-566 (1981 ).
3. G. Musch and D.L. Massart. Isolation of basic drugs from plasma
using solid phase extraction with cyanopropyl-bonded phase.
]. Chrornatogr. 432:209-222 (1988).
4. B.K. Logan, D.T. Stafford, I.R. Tebbett, and C.M. Moore. Rapid
screening for 100 basic drugs and metabolites in urine using cation
exchange solid-phase extraction and high-performance liquid
chromatography with diode array detection. J. Anal. Toxicol. 14:
154-159 (1990).
5. Bond Elut Certify Instruction Manual, Varian Inc., Harbor City,
CA 1992.
6. X.-H. Chen, J. Wijsbeek, J.-R Franke, and R.A. de Zeeuw. A single-
column procedure on Bond Elut Certify for systematic toxicolog-
ical analysis of drugs in plasma and urine. J. Forensic 5ci. 37:
61-71 (1992).
7. R. de la Torre, J. Ortufio, J.A. Pascual, S. Gonz~lez, and J. Ballesta.
Quantitative determination of tricyclic antidepressants and their
metabolites in plasma by solid phase extraction (Bond-Elut TCA)
and separation by capillary gas chromatography with nitrogen-
phosphorous detection. Ther. Drug Monit. 20(3): 340-346 (1998).
8. I.R. Tebbett. Rapid extraction of codeine and morphine in whole
blood for HPLC. Chromatographia 23:377-378 (1987).
9. C.M. Moore and I.R.Tebbett. Rapid extraction of anti-inflammatory
drugs in whole blood for HPLC. Forensic ScL Int. 34:155-158
(1987).
10. R.J. Hughes and M.D. Osselton. Comparison of methods for the
analysis of tricyiclic antidepressantsin small whole blood samples.
]. Anal. Toxicol. 13:163-168 (1989).
11. X.-H. Chen, J.-P.Franke, J. Wijsbeek, and R.A. de Zeeuw. Isolation
of acidic, neutral, and basic drugs from whole blood using a single
mixed-mode solid-phase extraction column. J. Anal Toxicol. 16:
351-355 (1992).
12. X.-H. Chen, J.-R Franke, J. Wijsbeek, and R.A. de Zeeuw. Pitfalls
and solutions in the development of a fully automated solid-phase
extraction method for drug screening purposes in plasma and
whole blood. J. Anal. ToxicoL 17:421-426 (1993).
13. P.G.M. Zweipfenning, A.H.C.M. Wilderink, P. Horsthuis, J.-R
Franke, and R.A. de Zeeuw. Toxicological analysis of whole blood
samples by means of Bond-Elut Certify columns by gas chro-
matography with nitrogen-phosphorus detection. J. Chromatogr. A
674:87-95 (1994).
14. X.-H. Chen, J.-R Franke, J. Wijsbeek, and R.A. de Zeeuw. Deter-
mination of basic drugs extracted from biological matrices by
means of solid-phase extraction and wide-bore capillary gas chro-
matography with nitrogen-phosphorus detection. J. Anal Toxicol.
18:150-153 (1994).
15. M.A. Mart[nez, C. S~nchez de la Torre, and E. Almarza. Simulta-
neous determination of viloxazine, venlafaxine, imipramine, de-
sipramine, sertraline and amoxapine in whole blood: comparison
of two extraction/clean-up procedures for capillary gas-liquid
chromatography with nitrogen-phosphorus detection. J. Anal Tox-
icoL 26(5): 296-302 (2002).
16. J. Yawney, S. Treacy, K.W. Hindmarsh, and F.J.Burczynski. A gen-
eral screening method for acidic, neutral and basic drugs in whole
blood using the Oasis MCX column. J. Anal. ToxicoL 26(6):
325-332 (2002).
17. J. 8reiter, R. Helger, and H. Lang. Evaluation of column extraction:
a new procedure for the analysis of drugs in body fluids. Forensic
$ci. 7:131-140 (1976).
18. B.K. Logan and D.T. Stafford. Liquid/solid extraction on di-
atomaceus earth for drug analysis in postmortem blood. J. Forensic
357

Sci. 34(3): 553-564 (1989).
19. C. Lora-Tamayo, M.A. Rams, and J.M.R. Chacon. Gas chromato-
graphic data for 187 nitrogen- or phosphorus containing drugs and
metabolites of toxicological interest analysed on methyl silicone
capillary columns. J. Chromatogr. B 374:73-85 (1986).
20. R.C. Baselt and R.H. Cravey. Disposition of Toxic Drugs and
Chemicals in Man, 6th ed. Chemical Toxicology Institute, Foster
358
Journal of Analytical Toxicology,Vol. 27, September2003
City, CA, 2002, pp 453-455 and 1052-1054.
21. A.H. Stead and A. Moffat. A collection of therapeutic, toxic and
fatal blood drug concentration in man. Hum. Toxicol. 3:437-464
(1983).
Manuscript received November 11, 2002;
revision received March 18, 2003.