Characterization of impurities formed by interaction of duloxetine HCl with enteric polymers hydroxypropyl methylcellulose acetate succinate and hydroxypropyl methylcellulose phthalate

Article abstract of DOI:10.1021/js970133r

Duloxetine hydrochloride ((S)-N-methyl-3-(1-naphthalenyloxy)-2- thiophenepropanamine hydrochloride) has been found to react with polymer degradation products or residual free acids present in the enteric polymers hydroxypropyl methylcellulose acetate succinate (HPMCAS) and hydroxypropyl methylcellulose phthalate (HPMCP) in dosage formulations to form succinamide and phthalamide impurities, respectively. The rate of formation of the impurities is accelerated by heat and humidity. The structures were deduced using molecular weights obtained from LC-MS experiments and confirmed by comparison of UV spectra, HPLC retention times, and electrospray mass spectra to independently synthesized material. It is proposed that polymer-bound succinic and phthalic substituents can be cleaved from the polymer, resulting in the formulation of either the free acids or the anhydrides. It is postulated that the reaction is enabled by migration of either (1) the free acid or anhydride or (2) the parent drug through the formulation. The formation of these impurities was minimized by increasing the thickness of the physical barrier separating the enteric coating from the drug.

Full text of DOI:10.1021/js970133r

Characterization of Impurities Formed by Interaction of Duloxetine HCl
with Enteric Polymers Hydroxypropyl Methylcellulose Acetate Succinate
and Hydroxypropyl Methylcellulose Phthalate
P
ATRICK J. JANSEN*, PETER L. OREN, CRAIG A. KEMP, STEVEN R. MAPLE
,
AND
STEVEN W. BAERTSCHI*
Contribution from the Lilly Research Laboratories, Pharmaceutical and Analytical Development Division, Eli Lilly and Company,
Indianapolis, Indiana 46285
Received March 31, 1997.
Accepted for publication October 23, 1997X.
Abstract
0 Duloxetine hydrochloride ((S)-N-methyl-3-(1-naphthale-
nyloxy)-2-thiophenepropanamine hydrochloride) has been found to
react with polymer degradation products or residual free acids present
in the enteric polymers hydroxypropyl methylcellulose acetate succinate
(HPMCAS) and hydroxypropyl methylcellulose phthalate (HPMCP) in
dosage formulations to form succinamide and phthalamide impurities,
respectively. The rate of formation of the impurities is accelerated
by heat and humidity. The structures were deduced using molecular
weights obtained from LC−MS experiments and confirmed by
comparison of UV spectra, HPLC retention times, and electrospray
mass spectra to independently synthesized material. It is proposed
that polymer-bound succinic and phthalic substituents can be cleaved
from the polymer, resulting in the formation of either the free acids or
the anhydrides. It is postulated that the reaction is enabled by
migration of either (1) the free acid or anhydride or (2) the parent
drug through the formulation. The formation of these impurities was
minimized by increasing the thickness of the physical barrier separating
the enteric coating from the drug.
Figure 1sRepeating units of HPMCAS and HPMCP.
Introduction
Duloxetine hydrochloride (1, (S)-N-methyl-3-(1-naphtha-
lenyloxy)-2-thiophenepropanamine hydrochloride) is a new
drug currently being developed for the treatment of major
depressive disorder and urinary incontinence. It is a potent
in vitro and in vivo inhibitor of serotonin and norepineph-
rine uptake in nerve terminals of mammalian brains and/
or serotonin uptake in human platelets ex vivo.1 Because
1 is unstable in solution at pH values less than 2.5,2 enteric
polymer-coated formulations have been developed to pre-
vent its acid degradation in the stomach and to provide
for subsequent rapid disintegration and release in the small
intestine. A tablet formulation coated with the enteric
polymer HPMCP was originally developed, but a formula-
tion consisting of pellets coated with the enteric polymer
HPMCAS contained in a capsule is the desired market
formulation. The structures of the enteric polymers are
given in Figure 1.
During the course of development of the HPMCAS-
coated pellet formulation, an unknown impurity which
eluted after 1 (impurity A) was detected by HPLC analysis
of samples stressed at 60 °C for 14 days. This impurity
was also detected in stability samples stored at 30 °C/60%
relative humidity and 40 °C/75% relative humidity.3 Sub-
sequent analysis of stability samples of HPMCP-coated
tablets indicated the presence of a different impurity
(impurity B) that also eluted after 1. The formation of
significant levels of these unknown impurities at pharma-
ceutically relevant storage conditions prompted an inves-
tigation into their identity and mechanism of formation.
In this report we describe the identification, synthesis, and
mechanism of formation of these two impurities identified
as duloxetine succinamide (2) and duloxetine phthalamide
(3).
Experimental Section
An a lytica l HP LC a n d P h otod iod e Ar r a y UV Detection s
HPLC analyses were carried out on a Waters photodiode array
system consisting of a 600E controller and pump, a 715 Ultra Wisp,
and a 996 photodiode array detector. The UV spectra were
obtained on the compounds as they eluted from the chromato-
graphic column. The column, a 4.6 × 250 mm Zorbax SB-CN 5
* To whom correspondence should be addressed. Telephone: (317)
276-1388. Fax: (317) 277-2154. E-Mail: baertschi@lilly.com.
jansen_patrick_j@lilly.com.
X Abstract published in Advance ACS Abstracts, December 15, 1997.
© 1998, American Chemical Society and
American Pharmaceutical Association
S0022-3549(97)00133-0 CCC: $15.00
Published on Web 01/02/1998
Journal of Pharmaceutical Sciences / 81
Vol. 87, No. 1, January 1998

Table 1
s
NMR Chemical Shift Assignments for 1
−3
1
2
3
site
δ
H
δ
C
δ
H
δ
C
δH
δC
1
3.07
44.99
3.43, 3.56a
44.30, 45.62
(3.14, 3.30)
(3.56, 3.62)
2.25, 2.44
5.71, 5.99
43.50, 47.01
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
2.40, 2.58
6.17
34.37
72.58
2.14, 2.32a
5.88, 5.97
36.07, 36.96
72.75, 73.24
35.32, 36.32
73.38
143.56
126.07
126.75
125.83
152.14
107.50
125.81
120.48
134.06
127.42
125.40b
126.40b
121.60
125.37
144.18, 144.59
125.88, 125.95
126.60, 126.71
125.46, 125.49
152.30, 152.40
107.27, 107.30
125.62, 125.79
120.22, 120.38
134.10, 134.11
127.43, 127.45
125.37, 125.41
126.37, 126.42
121.54, 121.59
125.46
144.19, 144.84
125.50, 125.77
126.68, 126.69
125.37, 125.44
152.10, 152.55
106.78, 107.31
125.80, 125.98
120.14, 120.20
133.96, 134.11
127.32, 127.44
125.10, 125.32b
126.34, 126.38b
121.28, 121.65
125.37
7.28
6.99
7.46
7.23, 7.27
6.97
7.42
7.10, 7.31
6.90, 6.99
7.44
7.08
7.34
7.46
7.02, 7.05
7.33
6.86, 7.08
7.26, 7.34
7.59
7.41, 7.45a
7.85
7.52
7.52
8.26
7.84
7.51
7.51
8.26
7.85
7.52
7.52
8.31
170.52, 170.91
27.04, 27.76b
28.94, 29.01b
173.90, 174.01
169.95
127.89
2.47
2.47
12.01
138.76, 139.10
129.97, 130.06
132.29, 132.63
126.64, 126.68
128.52, 128.32
166.79, 166.93
31.91, 36.32
7.81, 7.93
7.50
7.62
7.26
N
−CH3
2.55
32.37
2.83, 2.95
33.00, 35.09
2.62, 2.91
a Proton assignments from one-bond heteronuclear multiple quantum correlation (HMQC) experiment. b Carbon-13 assignments may be reversed within the
column.
µm (MACMOD Analytical), was operated at ambient temperature.
The flow rate was maintained at 1.0 mL/min. Mobile phases A
and B consisted of 80/20 and 25/75 (v/v) 25 mM KH2PO4 (pH 3.0)/
acetonitrile, respectively. Mobile phase B was increased linearly
from 0% to 30% at 2%/min for 15 min and then from 30% to 100%
at 5.83%/min for 12 min.
MSsNominal mass pseudo-molecular ions were determined
using a Fisons Instruments VG Quattro triple quadrupole mass
spectrometer equipped with a pneumatically assisted electrospray
liquid chromatography/mass spectrometry interface. For the on-
line LC-MS work, the flow rate was maintained at 1.0 mL/min.
Mobile phases A and B consisted of 80/20/0.05 and 25/75/0.05 (v/
v) water/acetonitrile/trifluoroacetic acid, respectively. Mobile
phase B was increased linearly from 0% to 30% at 2%/min for 15
min and then from 30% to 100% at 5.83%/min for 12 min. Positive
ion electron impact (EI) MS data was obtained with a VG 7070E
double focusing magnetic sector mass spectrometer using a direct-
insertion heated probe for sample introduction.
NMRsProton and carbon-13 NMR spectra were recorded at
299.957 and 75.432 MHz, respectively, using a Varian Unity
spectrometer equipped with a 5-mm Nalorac indirect detection
probe and a Matrix shim system. Samples consisted of 20 mg/
mL 1, 27 mg/mL 2, and 20 mg/mL 3 dissolved in dimethyl
sulfoxide-d6. Chemical shifts are reported in parts per million
relative to dimethyl sulfoxide-d6 (2.50 ppm for 1H, 39.50 for 13C).
All spectra were recorded in a nonspinning mode.
IRsFourier transform IR spectra were collected on a Nicolet
60SXB spectrometer. 1024 scans were coadded at 4 cm-1 resolu-
tion from 3775 to 625 cm-1 using Happ-Geutzal apodization.
Syn th esis of 2 a n d 3sThe synthesis of 2 was accomplished
by allowing duloxetine HCl to react with a slight molar excess of
succinic anhydride in chloroform containing a trace of triethy-
lamine. HPLC analysis of the reaction mixture after 15 min at
40 °C confirmed nearly complete conversion of duloxetine to 2.
Approximately 200 mg of crude 2 was purified by preparative
reverse phase HPLC for characterization. The synthesis of 3 was
accomplished in the same manner as 2 using phthalic anhydride
in place of succinic anhydride. It was also purified by preparative
reverse phase HPLC.
Figure 2
s
HPLC chromatogram of duloxetine hydrochloride pellet formulation
stressed at 60
°
C for 14 days.
NMR assignments. 2: EI MS m/z (fragment) 397 ([M]+), 297 ([M]+
of 1), 254 (C12H16NO3S), 152 (C8H10NS), 123 (C7H5S), 44 (CO2);
IR 1730 (CdO), 1644 (amide I), 1599 (aryl-CH), 1595 (aryl-CH),
1396 (aryl-CH), 1237 (aryl-O), 1095 (O-CH), 1079 (CdC), 837 cm-1
(HCdCH). 3: EI MS m/z (fragment) 297 ([M]+ of 1), 266 (C17H14
-
OS), 237 (C15H9OS), 144 (C10H8O), 76 (C6H5); IR 1713 (CdO), 1610
(amide I), 1596 (aryl-CH), 1398 (aryl-CH), 1237 (aryl-O), 1095 (O-
CH), 1066 (CdC), 834 cm-1 (HCdCH).
Results and Discussion
Id en tifica tion of Im p u r ities A a n d BsPellets of 1
coated with the enteric polymer HPMCAS were stressed
at 60 °C for 14 days. HPLC analysis of the stressed sample
(Figure 2) revealed a late-eluting impurity (impurity A) at
a level of approximately 0.5%. The UV spectra of 1 and
impurity A (Figure 3), obtained by PDA detection, were
similar. Electrospray LC-MS analysis of the impurity
peak showed a [M + H]+ ion at m/z 398, suggesting a
Ch a r a cter iza tion of Syn th esized Im p u r itiessBoth impuri-
ties were characterized using NMR, MS, and IR. See Table 1 for
82 / Journal of Pharmaceutical Sciences
Vol. 87, No. 1, January 1998

electron impact (EI) MS, IR, and NMR data supported the
presence of the duloxetine moiety; however, the proton and
carbon-13 NMR spectra exhibited the spectral complexity
typically observed for amides in solution.6 Amides in
solution can exist in two interconverting conformers, cis
and trans, of unequal population leading to complex proton
and carbon-13 NMR spectra because each magnetically
inequivalent site may show two resonances. To establish
that the spectral complexity observed was attributable to
changes in molecular configuration, the proton NMR
spectra of 2 were recorded as the temperature was raised
from 20 to 100 °C (Figure 6). As the temperature in-
creased, the separation between resonance pairs decreased
and the line widths broadened until at 90 °C the proton
resonances of position 3 collapsed and the line width began
to sharpen. These observations are consistent with the
proposed amide cis-trans isomerism.
Complete NMR chemical shift assignments (Table 1) for
2 and 3 were based upon comparison with the chemical
shifts of 1 and the use of single and multiple bond
heteronuclear correlation experiments optimized for three-
bond C-H coupling constants. No efforts were made to
make the assignments cis or trans specific, although in
solution the trans conformer is preferred.6 The connectivity
of the succinoyl and phthaloyl moieties to the N-methyl
nitrogen was established by the long-range coupling ob-
served between the N-methyl protons and the carbonyl
carbon at position 18 for both 2 and 3. Starting with the
N-methyl protons and carbon, which can be readily as-
signed on the basis of their corresponding chemical shifts,
the carbons at positions 2 and 3 can be assigned by their
long-range couplings to the N-methyl protons. Assignment
of the resonances attributable to the thiophene ring can
also be readily sorted out. Only one carbon from the
thiophene ring shows a long-range coupling to the proton
at position 3 and it is assigned to position 5. The assign-
ments of the carbon resonances to positions 6 and 7 were
based upon the observed long-range couplings to positions
5 and 7 and positions 5 and 6, respectively. The naphtha-
lenyloxy resonances are similarly assigned using observed
long-range couplings. Of particular interest are the as-
signments of positions 13 and 16 which were made based
upon the coupling observed between positions 11 and 13
and positions 12 and 16.
P r op osed Mech a n ism of F or m a tion sIn both the
pellet and tablet formulations the duloxetine-containing
layers are separated from the enteric coating by a subcoat
layer consisting primarily of hydroxypropyl methylcellulose
and other ingredients. This raises the question about how
the impurities form since the duloxetine layer is physically
separated from the enteric polymers. Two possibilities can
be envisioned: (1) the impurities are not actually present
in the formulation but rather form during sample prepara-
tion or (2) migration of formulation components is occurring
that allows 1 to come in contact and react with either the
enteric polymers or an active succinoyl derivative present
in the polymer.
Figure 3sUV spectra of 1 and impurities A and B.
molecular weight of 397. From these results the source of
the impurity was postulated to result from a reaction of
the secondary amine of 1 with a UV-transparent compo-
nent of the formulation.
The list of ingredients used in the manufacture of the
pellets was examined for compounds that could possibly
react with 1. The enteric coating used for the pellets,
HPMCAS, is a hydroxypropyl methylcellulose polymer
substituted with acetate and succinate groups, which is
known to contain small amounts of residual free succinic
acid.4 The secondary amine of 1 could potentially react
with the succinic acid (or an active succinoyl derivative)
to form a duloxetine succinamide (2) that would have a
molecular weight of 397.
This hypothesis was tested by reacting 1 with succinic
anhydride in chloroform (see the Experimental Section).
HPLC analysis of the reaction product revealed that 1 had
been consumed and a new peak had appeared that matched
the retention time, UV spectrum, and electrospray mass
spectrum (Figure 4) of impurity A. The structure of the
synthesized material was shown to be that of 2 (see below).
The discovery of 2 in duloxetine pellets coated with the
enteric polymer HPMCAS prompted an investigation of
duloxetine tablets coated with the enteric polymer HPMCP.
HPMCP is a hydroxypropyl methylcellulose polymer sub-
stituted with phthalate groups and is known to contain
small amounts of residual free phthalic acid.4 One would
expect that the same type of condensation reaction leading
to 2 in duloxetine pellets might also occur in duloxetine
tablets, resulting in the formation of a duloxetine phthala-
mide (3), which would have a molecular weight of 445.
HPLC analysis of a sample of enteric-coated duloxetine
tablets stressed at 40 °C/75% relative humidity for 2
months (Figure 5) detected a late-eluting impurity (impu-
rity B). Comparison of the UV spectrum of impurity B to
that of 1 (Figure 3) again suggested that the impurity was
related to 1. Electrospray LC-MS results on impurity
peak B showed a [M + H]+ ion at m/z 446, suggesting a
molecular weight of 445. From this information, the
structure of impurity B was proposed to be 3.
To test whether 2 was actually present in the formula-
tion or if it was formed during the extraction procedure,
two experiments were conducted. In the first experiment,
pellets containing 2 were extracted with both methanol and
methanol containing approximately 4% (v/v) methylamine.7
If 2 were formed during the extraction procedure, methy-
lamine, which is a primary amine and a better nucleophile
than 1, should compete with 1 for the active succinoyl
derivative and diminish the amount of 2 formed. This,
however, was not the case as the same levels of 2 were
measured in both samples.
Structure 3 was therefore prepared by reaction of ph-
thalic anhydride with 1 in chloroform (see the Experimen-
tal Section). Impurity B was shown to be 3 by comparison
of the HPLC retention time, UV spectrum, and electrospray
mass spectrum (Figure 4) to that of synthetically prepared
3.
Str u ctu r e Elu cid a tion of 2 a n d 3sHigh-resolution
FAB MS yielded [M + H]+ ions of m/z 398 and 446,
corresponding to molecular formulas of C22H23NO4S and
C26H24NO4S for 2 and 3, respectively.5 Accurate mass
In the second experiment, placebo pellets were stressed
in the same manner as pellets containing 1. Both the
placebo and the duloxetine pellets were extracted in the
Journal of Pharmaceutical Sciences / 83
Vol. 87, No. 1, January 1998

Figure 4sPositive ion electrospray mass spectra of (a) impurity A, (b) synthesized 2, (c) impurity B, and (d) synthesized 3. The spectra of impurities A and B
were obtained from LC
synthesized products.
−MS analysis of stressed duloxetine formulations while the spectra of 1 and 2 were obtained from LC−MS analysis of solutions of the
Figure 5
s
HPLC chromatogram of duloxetine hydrochloride tablet formulation
stressed at 40
°
C/75% relative humidity for 2 months.
same manner except that 1 was added to the extraction
solvent used for the placebo sample. If 2 were formed
during the extraction procedure it would be detected in the
placebo extract. Analysis of the two extracts revealed
significant quantities of 2 present in the duloxetine pellet
sample but not in the placebo extract.
The results of these two experiments suggest that 2 is
formed in the pellet formulation and not during the
extraction procedure. For this to occur 1 must come in
contact with the active succinoyl derivative either by
migration of one or both of the reactants through the
subcoat layer or by the breakdown of the subcoat layer.
Thus, to minimize the formation of 2, the thickness of the
subcoat layer was increased in the proposed commercial
formulation. Stability tests conducted on this formulation
Figure 6sVariable temperature NMR spectra of the proton at position 3 of 2.
have shown that the levels of 2 are negligible (<0.1%)
following storage at 25 °C/60% relative humidity for 18
months.
In summary, we propose that 2 and 3 are formed from
reaction of succinic and phthalic anhydrides with 1,
respectively. Intimate contact required for reaction could
84 / Journal of Pharmaceutical Sciences
Vol. 87, No. 1, January 1998

be facilitated by either migration of the anhydrides or of 1
between the physically separated layers. The anhydrides
could form as a result of either dehydration of the residual
free succinic and phthalic acids present in the enteric
polymer or an intramolecular attack of the free acid on the
ester linkage to the polymer backbone, cleaving to form free
anhydride. It does not appear likely that 1 is reacting
directly with the succinic and phthalic ester moieties on
the intact polymer, since one would also expect the forma-
tion of a duloxetine acetamide degradation product, which
has not been observed. We cannot rule out the possibility
that the free acids are reacting directly with 1; however,
the reaction of carboxylic acids with amines is not a favored
reaction (in contrast to the reaction of amines with the
corresponding anhydrides). Further investigation would
be required to conclusively define the active reactants
leading to 2 and 3.
either 1 or the phthaloyl or succinoyl moieties are migrat-
ing through the subcoating to enable physical contact and
reaction. The amount of impurity 2 formed in the com-
mercial formulation of 1 has been minimized by increasing
the thickness of the physical barrier separating the reacting
species.
References and Notes
1. Wong, D. T.; Bymaster, F. P.; Mayle, D. A.; Reid, L. R.;
Krushinski, J . H.; Robertson, D. W. Neuropsychopharacology
1993, 8, 23-33.
2. Bopp, R.J .; Breau, A. P.; Faulkinbury, T. J .; Heath, P. C.;
Miller, C.; Stephan, E. A.; Weigel, L. O.; Wong, D. T. The
Rearrangement of Duloxetine Under Mineral Acid Condi-
tions. Presented at the 206th National American Chemical
Society Meeting, San Diego, CA; March 13, 1993; ORGN 111.
3. The level of the impurity was measured to be 0.3% in a
sample stored at 30 °C/60% relative humidity for 12 months
and 3.6% in a sample stored at 40 °C/75% relative humidity
for 6 months.
4. Information obtained from product literature supplied by the
manufacturer of HPMCAS. According to the manufacturer’s
certificate of analysis, the HPMCAS contained 0.06% free
succinic acid and the HPMCP contained 0.57% free phthalic
acid.
Concluding Remarks
This study suggests that new impurities in formulations
can result from reactions between enteric polymer substit-
uents and drugs containing nucleophilic functional groups
even when they are physically separated in the formula-
tion. Two impurities detected in formulations of 1 upon
aging were determined to be the result of the reaction of 1
with phthaloyl and succinoyl moieties present in the enteric
polymers HPMCP and HPMCAS, respectively. Because
the enteric polymers are physically separated from 1 by a
subcoating, the formation of these impurities indicates that
5. 2 (calculated 398.1426, found 398.1458); 3 (calculated 446.1426,
found 446.1416)
6. Dorman, D. E.; Bovey, F. A. J . Org. Chem. 1973, 38, 1719-
1721.
7. This corresponded to a molar ratio of methylamine to
duloxetine of ∼450:1.
J S970133R
Journal of Pharmaceutical Sciences / 85
Vol. 87, No. 1, January 1998