Maillard reaction of lactose and fluoxetine hydrochloride a secondary amine

Article abstract of DOI:10.1021/js9702067

Analysis of commercially available generic formulations of fluoxetine HCl revealed the presence of lactose as the most common excipient. We show that such formulations are inherently less stable than formulations with starch as the diluent due to the Maillard reaction between the drug, a secondary amine hydrochloride, and lactose. The Amadori rearrangement product was isolated and characterized; the characterization was aided by reduction with sodium borohydride and subsequent characterization of this reduced adduct. The lactose-fluoxetine HCl reaction was examined in aqueous ethanol and in the solid state, in which factors such as water content, lubricant concentration, and temperature were found to influence the degradation. N- Formylfluoxetine was identified as a major product of this Maillard reaction and it is proposed that N-formyl compounds be used as markers for this drug- excipient interaction since they are easy to prepare synthetically. Many characteristic volatile products of the Maillard reaction have been identified by GC/MS, including furaldehyde, maltol, and 2,3-dihydro-3,5- dihydroxy-6-methyl-4H-pyran-4-one. Close similarity between the degradation products of simple mixtures and formulated generic products was found; however, at least one product decomposed at a rate nearly 10 times that predicted from the simple models. Maillard products have also been identified in unstressed capsules. The main conclusions is that drugs which are secondary amines (non just primary amines as sometimes reported) undergo the Maillard reaction with lactose under pharmaceutically relevant conditions. This finding should be considered during the selection of excipients and stability protocols for drugs which are secondary amines or their salts, just as it currently is for primary amines.

Full text of DOI:10.1021/js9702067

Maillard Reaction of Lactose and Fluoxetine Hydrochloride, a Secondary
Amine
D
AVID D. WIRTH*, STEVEN W. BAERTSCHI, ROSS A. JOHNSON, STEVEN R. MAPLE, MARYBETH S. MILLER
,
D
IANA K. HALLENBECK AND TEPHEN M. GREGG
,
S
Contribution from the Lilly Research Laboratories, Eli Lilly and Company, Lafayette, Indiana 47905
Received May 21, 1997.
Final revised manuscript received October 20, 1997.
Accepted for
publication October 23, 1997X.
tautomers are in equilibrium with their more reactive
aldehyde forms; nonreducing carbohydrates such as man-
nitol, sucrose, and trehalose are not subject to Maillard
reactions. Although early scientists believed that only
primary aromatic amines were capable of this reaction,
subsequent research has shown that nearly all primary and
secondary amines, aromatic or aliphatic, are capable of this
reaction.9 Amino acids and proteins are also substrates
for the Maillard reaction.10 The impact of these reactions
on the stability of pharmaceuticals has also been known
for some time and was recently reviewed.11
Abstract
0 Analysis of commercially available generic formulations
of fluoxetine HCl revealed the presence of lactose as the most common
excipient. We show that such formulations are inherently less stable
than formulations with starch as the diluent due to the Maillard reaction
between the drug, a secondary amine hydrochloride, and lactose. The
Amadori rearrangement product was isolated and characterized; the
characterization was aided by reduction with sodium borohydride and
subsequent characterization of this reduced adduct. The lactose
−
fluoxetine HCl reaction was examined in aqueous ethanol and in the
solid state, in which factors such as water content, lubricant
concentration, and temperature were found to influence the degrada-
tion. N-Formylfluoxetine was identified as a major product of this
Maillard reaction and it is proposed that N-formyl compounds be used
Prozac, the world’s largest selling antidepressant, is a
selective serotonin reuptake inhibitor;12 it has been used
by over 25 million people in more than 90 countries. Its
active ingredient is fluoxetine hydrochloride, 1, and the
diluent is starch. Within the past few years, many generic
formulations containing fluoxetine HCl have been manu-
factured and sold in countries lacking patent protection or
in which the patents have expired. We have recently
examined several dozen of these products and find that a
majority of them contain lactose as their primary diluent.
Lactose is widely used as a diluent for capsules and tablets
due to its low price, high purity, and excellent compression
and stability characteristics. Although its ability to un-
dergo nonenzymatic browning (Maillard) reactions has long
been known to formulation scientists, some reports have
suggested this interaction to be largely restricted to
primary amines.13 The origin of this view is likely founded
in early literature reports of the slower browning of glucose
mixtures of N-methyl glycine as compared to glycine itself14
and work on the browning of mixtures of lactose and
methylated amphetamines,15 studies which measured opti-
cal density only, not characteristic Maillard products.
Although these reports showed faster rates of browning of
mixtures of reducing carbohydrates with primary amines
as compared to secondary ones, they do not support the
conclusion that secondary amines should generally be
incapable of the Maillard reaction.
The generally accepted mechanism of glycosylation and
the Amadori rearrangement are summarized in Scheme 1
for a secondary amine.16 The imminium ion intermediate
can be formed from either primary or secondary amines
(but not tertiary ones) and is the key intermediate in both
closure to the glycosylamine and deprotonation to the enol
version of the Amadori rearrangement product, ARP. The
multitude of individual steps in this mechanism allows for
considerable variation in the kinetics. For example, the
hydrolysis of the glycosylamine of benzylamine and glucose
is faster at pH 5 than hydrolysis of the corresponding
glycosylamine with piperidine, but the reverse order of
reactivity is observed at pH 9.17 Kinetic studies of the
Maillard reaction have been reviewed and will not be
explored in depth in this paper.18 In view of the well-
as markers for this drug−excipient interaction since they are easy to
prepare synthetically. Many characteristic volatile products of the
Maillard reaction have been identified by GC/MS, including furaldehyde,
maltol, and 2,3-dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one. Close
similarity between the degradation products of simple mixtures and
formulated generic products was found; however, at least one product
decomposed at a rate nearly 10 times that predicted from the simple
models. Maillard products have also been identified in unstressed
capsules. The main conclusion is that drugs which are secondary
amines (not just primary amines as sometimes reported) undergo the
Maillard reaction with lactose under pharmaceutically relevant condi-
tions. This finding should be considered during the selection of
excipients and stability protocols for drugs which are secondary amines
or their salts, just as it currently is for primary amines.
Introduction
The Maillard reaction is named after Louis Maillard,
who reported over 80 years ago that some amines and
reducing carbohydrates react to produce brown pigments.1
It has been extensively studied and reviewed, especially
in the food and nutrition literature.2 More recent reviews
cover not only the chemistry of foods but human physiology
as well.3 The first product of this reaction is a simple
glycosylamine,4 which readily undergoes the Amadori
rearrangement to produce 1-amino-1-deoxy-2-ketoses.5 The
large body of literature on these reactions is due to the
multitude of possible reaction pathways and products,
including fragmentations of the carbohydrates and forma-
tion of aromatic compounds from cyclization/dehydration
processes.6 Reducing disaccharides also undergo this reac-
tion,7 and it is a well-documented process for the degrada-
tion of lactose during the heating of milk.8 Reducing
carbohydrates such as glucose, maltose, and lactose are
substrates for the Maillard reaction since their cyclic
X Abstract published in Advance ACS Abstracts, December 15, 1997.
© 1998, American Chemical Society and
American Pharmaceutical Association
S0022-3549(97)00206-2 CCC: $15.00
Published on Web 01/02/1998
Journal of Pharmaceutical Sciences / 31
Vol. 87, No. 1, January 1998

Teflon-lined rubber septum. The vials were heated at 100 °C for
20 h. A stainless steel GLT short path thermal desorption tube
packed with 100 mg of Tenax-TA which had been preconditioned
with helium purge at 250 °C and fitted with a needle was inserted
into the vial. A second needle connected to a helium source was
inserted into the vial and the flow rate set at approximately 7 mL/
min. The vial headspace gases were purged onto the Tennax-TA
tube for either 5 or 60 min. The GLT tube was then transferred
to the short path thermal desorption unit for desorption onto the
gas chromatographic column. The following conditions were used
for the transfer and chromatographic analysis. Sample transfer:
gas purge time, 0.5 min; desorption time, 5 min; GLT desorption
temperature, 100 °C initial then 40 °C/min to 200 °C; column
cryotrap at -100 °C. GC conditions: injector, 150 °C; transfer
line to mass spectrometer, 300 °C; split flow, 7 mL/min or 50 mL/
min; the cryotrap was heated rapidly to 250 °C to desorb material
onto the GC column; oven temperature program, 40 °C for 3 min
then 10 °C/min to 300 °C and hold 3 min; chromatographic column,
DB-1, 30 m × 0.25 mm i.d. and 1.0 mm film.
HP LCsHigh performance liquid chromatography was per-
formed with a Spectra Physics SP8700XR pump, a Spectroflow
757 detector (Anspec) set at 260 nm, a ChromJ et integrator by
Spectra Physics, and an Alcott 728 autosampler incorporating a
Valco injection valve with a 10 µL fixed loop. The column was a
Zorbax RX-C8, 25 cm by 4.6 mm, 5 µm particles (Mac-Mod
Analytical, Inc., Chadds Ford, PA) maintained at ambient tem-
perature. The flow rate was 1.00 mL min-1 throughout. The two
eluting solutions used were A ) 0.07% trifluoroacetic acid (TFA)
in water, v/v; B ) 0.07% TFA in acetonitrile. The elution program
started at 80:20 A:B for 5 min, ramped linearly to 15:85 A:B at 30
min, held there until 35 min, and returned to 80:20 A:B at 40 min.
Injections were made no sooner than every 50 min. The ap-
proximate elution times are 1 at 21 min, 2 at 18.4 min, 3 at 27
min, and 4 at 17.9 min. For MS detection, samples were analyzed
on MicroMass (formerly Fisons) Quattro II and Platform II LC/
MS systems in the positive ion electrospray mode, scanning from
100 to 1000 amu, generating centroid data. Further details and
example chromatograms have been published elsewhere.20 This
method allows for the detection and quantitation of impurities
which span a wide range of polarities, including nonpolar com-
pounds which are not eluted using the isocratic US Pharmacopeia
method.21
Scheme 1sMechanism of glycosylation and Amadori rearrangement with
secondary amines.
known interaction of reducing carbohydrates and secondary
amines, as reported in the carbohydrate and food science
literature,19 we have investigated the Maillard/Amadori
reactions of the lactose-fluoxetine system and report our
findings here. Detailed mechanistic work such as identi-
fication of the rate-determining step and full kinetic
analysis as a function of the amine, carbohydrate, and
reaction conditions are left for future studies.
Experimental Section
La ctose-F lu oxetin e Am a d or i Rea r r a n gem en t P r od u ct
(ARP , 2)sIn a 500 mL round-bottom flask equipped with an
overhead stirrer, thermometer, and condenser were combined 20
g of R-D-lactose monohydrate (55.5 mmol), 10 g of fluoxetine HCl
(28.9 mmol), 250 mL of ethanol, 120 mL of dimethylacetamide,
and 4 mL of triethylamine (28.7 mmol). The nearly homogeneous
mixture was stirred at reflux for 24 h. The reaction mixture was
filtered at ambient temperature and the wetcake was washed with
50 mL of ethanol. The filtrate was evaporated in vacuo to a thick
brown solution (112 g), to which were added 300 mL of ethyl
acetate, 300 mL of 20% aqueous sodium chloride, and 30 mL of
ether. The pH was adjusted to 11 with 50% sodium hydroxide.
White solids which formed during the pH adjustment were filtered
and the layers were separated. The bottom layer was discarded
and the upper layer was set aside. To the milky middle layer were
added 300 mL of ethyl acetate and 30 mL of ether. The layers
were separated and the combined organic layers were evaporated
in vacuo to a thick brown oil (42.8 g), to which were added 175
mL of 20% aqueous sodium chloride, 400 mL water, and 550 mL
chloroform. The pH was adjusted to 1.5 with concentrated HCl.
The layers were separated and to the aqueous layer were added
60 g of NaCl, 500 mL of ethyl acetate, and 50 mL of ether. The
pH was adjusted to 10.6 with 50% NaOH, and the layers were
separated. The ethyl acetate layer was evaporated in vacuo to a
residue (4.7 g) which was 85% pure by HPLC analysis at 260 nm.
Standard 1H and 13C NMR spectra were acquired on a sample of
2 prepared by dissolving 50-60 mg of the material in DMSO-d6.
A partial list of assignments has been made based on comparison
to fluoxetine and is shown in Table 1. Flow-injection mass
spectrometry indicated the molecular weight to be 633 (M + H+,
m/z 634). Accurate mass FAB-MS measurements of the proto-
nated molecular ion (M + H+) indicated the elemental composition
to be C29H39NO11F3 (calcd 634.2475, measured 634.2487).
Ma ter ia lssFluoxetine HCl and Prozac were obtained from Eli
Lilly and Co. generics A, N, O, and Z were obtained in pharmacies
outside the United States and were unexpired formulated 20 mg
products. Lactose monohydrate was obtained from Aldrich Chemi-
cal Co, Milwaukee, WI, and spray-dried lactose was from Foremost
Ingredients Group, Baraboo, WI. The anhydrous lactose forms
were obtained from Sheffield Products, Norwich, NY. The aceto-
nitrile was spectroscopic grade, obtained from EM Science, Gibbs-
town, NJ . Reagent grade trifluoroacetic acid was obtained from
Aldrich Chemical Co. and was distilled prior to use. Other
chemicals and the authentic samples of compounds listed in Table
6 were purchased from Aldrich Chemical Co. and were used as
received. Ethanol was anhydrous and denatured with 5% metha-
nol.
NMRsNuclear magnetic resonance (NMR) spectroscopy for
compounds 1, 2, 3, and 6 was carried out on a Bruker AC300
spectrometer with 1H and 13C operating frequencies of 300.13 and
75.46 MHz, respectively. Additional experiments on compound 6
were carried out on a Bruker AMX500 with 1H and 13C operating
frequencies of 500.14 and 125.77 MHz, respectively.
GC/MSsGas chromatography/mass spectrometry was per-
formed on a MicroMass TRIO II quadrapole mass spectrometer
controlled using the Masslynx operating system. The instrument
was equipped with a Hewlett-Packard Model 5890 gas chromato-
graph which was fitted with a Scientific Instruments Services
short path thermal desorption Model TD 4 system and CryoTrap
unit. The mass spectrometer was operated in the EI+ mode and
calibrated daily against FC-43. Mass spectral data were obtained
in the centroid mode by scanning the mass range from 33 to 600
amu at a rate of 1260 amu/s and using an interscan time delay of
0.05 s.
For the analysis of products from solid-state samples, the
contents of one 20 mg capsule of fluoxetine HCl product or
mixtures of 20 mg of bulk fluoxetine HCl and 200 mg of lactose
were added to a 10 mL glass vial which was crimp sealed with a
Red u ced Ad d u ct (6)sIn a 50 mL round-bottom flask equipped
with an overhead stirrer, a thermometer, and a condenser were
combined 1.5 g of ARP, 2 (2.37 mmol), 15 mL of ethanol, and 1.79
32 / Journal of Pharmaceutical Sciences
Vol. 87, No. 1, January 1998

g of NaBH4 (47.4 mmol, 20 equiv). The mixture was stirred at
reflux for 1 h and at 70 °C for 1 h. The mixture was cooled to
ambient temperature and 10 mL of ethanol was added. Excess
NaBH4 was quenched by the dropwise addition of 5 mL of acetone
at 5 °C. The mixture was evaporated to a wet residue under
vacuum at 35 °C. To the residue were added 200 mL of 20%
aqueous sodium chloride and 200 mL of CHCl3, and the pH was
adjusted to 2.2 with concentrated HCl. The layers were separated
(after settling overnight). To the CHCl3 layer were added 100 mL
of CHCl3 and 200 mL of water, and the pH was adjusted to 2.2
with concentrated HCl. The layers were separated, to the water
layer were added 300 mL of ethyl acetate and 30 mL of ether,
and the pH was adjusted to 10.6 with 50% NaOH. The layers
were separated, and the ethyl acetate layer was concentrated
under vacuum, producing 0.4 g of residue which contained 80% of
the desired reduced fluoxetine-lactose adduct, 6, by HPLC
analysis at 260 nm. Extensive NMR experiments to confirm the
structure of 6 were carried out on a sample prepared by dissolving
∼60 mg in DMSO-d6. The experiments performed were 13C DEPT
(distortionless enhancement by polarization transfer), 2D 1H-1H
COSY (correlation spectroscopy),22 2D 1H-13C HMQC (hetero-
nuclear multiple quantum coherence),23 and 2D 1H-13C HMBC
(heteronuclear multiple bond correlation).24 The NMR assign-
ments are shown in Table 1. Mass spectrometry (infusion,
electrospray) verified the molecular weight by the presence of a
large M + 1 ion at 636 amu.
N-F or m ylflu oxetin e (3)sAcetic formic anhydride was pre-
pared by stirring acetic anhydride (4.87 mL, 51.6 mmol) and formic
acid (7.78 mL, 206 mmol) in a beaker for 5 min. Fluoxetine HCl
(4.5 g, 12.9 mmol) and methylene chloride (30 mL) were placed in
a 100 mL round-bottomed flask. Sodium hydroxide (1 M, 30 mL,
30 mmol) was added. The solution was stirred at room temper-
ature for 5-10 min. The layers were separated, and the organic
layer was dried with sodium sulfate and filtered. The filtrate was
placed in a 100 mL round-bottomed flask, and the acetic formic
anhydride was added. The reaction was stirred at room temper-
ature for 5 h and extracted with a 10% sodium bicarbonate solution
(3 × 50 mL). The organic layer was dried with sodium sulfate
and filtered. The filtrate was concentrated to an oil (1.5 g) which
solidified upon standing (mp 65-67 °C). The purity by the
gradient HPLC method was 98.9%. The major and minor rotomers
about the amide bond are identified in the following NMR spectra.
1H NMR (300 MHz, CD2Cl2) 8.03-7.96 (dd, 1 H), 7.52-7.43 (d,
2H), 7.41-7.23 (m, 5H), 6.98-6.88 (dd, 2H), 5.30-5.22 (m, 1H,
minor), 5.22-5.16 (m, 1H, major), 3.62-3.48 (m, 2H, major), 3.44-
3.16 (m, 2H, minor), 2.94 (s, 3H, minor), 2.88 (s, 3H, major), 2.32-
1.96 (m, 2H); 13C NMR (CD2Cl2, major rotomer) 162.79, 160.56,
140.55, 129.28, 128.51,127.15, 126.14, 122.99, 116.16, 77.44, 46.22,
37.23, 29.56; (minor rotomer) 162.75, 160.84, 141.01, 129.15,
128.34, 127.05, 126.24, 123.42, 116.25, 78.51, 41.52, 36.11, 34.89;
19F NMR (282 MHz, CD2Cl2) -62.14, -62.2; IR (KBr) 2934, 2875,
1689, 1334, 1249, 1117, 847 cm-1; MS (CI+) m/e 338 (25), 176 (100),
143 (43).
Rea ction s in Aqu eou s Eth a n olsFluoxetine HCl (0.50 g, 1.45
mmol) and lactose monohydrate (5.0 g, 13.9 mmol) were combined
with 5.0 mL of ethanol and 7.5 mL of water in each of four separate
flasks. The mixtures were heated to 60 °C and vigorously stirred
with a mechanical agitator. Potassium hydroxide (85% assay,
amounts of 0, 7, 47, or 100 mg) was added, and the reactions were
monitored by the above HPLC method. The equivalents of KOH
based on the fluoxetine and the results are given in Table 2. The
apparent pH (called pH*) was measured by direct insertion of a
calibrated glass electrode into the warm reaction mixture. Samples
were diluted 1:5 with eluent for HPLC measurement. The
fluoxetine concentration was measured versus an external stan-
dard and the mole percent concentrations of 2 and 3 were taken
as equal to their area percent values since their molar response
factors were measured on authentic material and found to be
within 5% of that of fluoxetine itself.
Scheme 2sMaillard reaction of lactose and fluoxetine HCl.
headed Mg stearate, were repeated with 0.04 g of magnesium
stearate added before the blending or grinding operation. Four
commonly available types of lactose were screened: crystalline
monohydrate; a mixture of crystalline and spray-dried monohy-
drate; and two particle size grades of anhydrous lactose, granular
and finely milled. After mixing, the samples were heated in an
oven at 98 °C for 24 h in glass vials with loose-fitting plastic caps.
The samples were analyzed with the gradient HPLC method and
the results are reported using fluoxetine as the standard. Since
compounds 2 and 3 have the same molar response as fluoxetine,
the levels given for these impurities are molar yields. The total
impurity level is estimated by assuming that the molar responses
of all impurities are equal to fluoxetine’s and thus can be compared
on a relative basis only.
Solid -Sta te Kin etic Stu d iessLactose (5.0 g, in separate
experiments, the monohydrate, anhydrous lactose, and monohy-
drate containing an additional 0.10 g of magnesium stearate) and
fluoxetine HCl (0.50 g) were mixed thoroughly with a mortar and
pestle and 1.8 g of the mixtures was added to each of three 10 mL
glass vials. The vials were lightly capped and placed in equili-
brated ovens at 75, 85, or 95 °C. Samples of the solids were
removed at 23, 47, 95, 175, and 287 h for the 75 °C experiment;
at 6, 23, 47, 99, and 122 h for the 85 °C experiment; and at 2, 6,
13, 21, and 28 h for the 95 °C experiment. The samples were
assayed as described for the screening studies.
Results and Discussion
P r ep a r a tion a n d Ch a r a cter iza tion of Am a d or i Re-
a r r a n gem en t P r od u ct (ARP ), 2sThe Maillard reaction
between fluoxetine and lactose to produce the ARP, 2, is
shown in Scheme 2. An authentic sample of 2 was obtained
by reaction of lactose and fluoxetine HCl in N,N-dimeth-
ylacetamide and ethanol as solvent with triethylamine as
the base. It was purified by extractions but was not
crystallized.
Solid -Sta te Scr een in g Stu d iessFor the screening studies,
5.0 g of lactose and 0.50 g of fluoxetine HCl were combined and
mixed one of three ways, described here according to their
designation in Table 3. Those tumbled were placed in capped glass
bottles and strapped to a horizontal rotating rod for 40 min. Those
that were ground were mixed thoroughly in a standard glass
mortar and pestle for several minutes. To those labeled “wet
grind" was added 0.5 g water before the mortar and pestle grinding
operation. Some experiments, identified with a Y in the column
Although the structure of 2 is shown as acyclic, it exists
in solution as a mixture of pyranose and furanose forms
both of which can be diastereomeric.25 For this reason,
Journal of Pharmaceutical Sciences / 33
Vol. 87, No. 1, January 1998

Table 1sNMR Data for Compounds 1, 2, and 6
Table 2
sApparent Rates of Product Formation for the Reaction of
Fluoxetine HCl and Lactose in Aqueous Ethanol at 60
°
C
percent per hour
N-formyl-
pH*
equiv KOH
ARP, 2
fluoxetine, 3
5.0
7.0
8.0
8.4
0
0.0014
0.084
0.31
0
0
0.07
0.50
1.05
0.040
0.22
fluoxetine, 1
13C 1H
2.99
ARP, 2
1H
reduced adduct, 6a
0.16
positionb
1
13C
13C
1H
44.99
53.88
51.30
31.88
77.00
160.10
116.36
126.93
121.56
139.87
126.07
128.88
128.22
41.45
39.21
124.47
58.95
58.06
64.48
71.41
71.08
75.32
68.34
61.06
103.97
70.65
73.44
70.01
77.56
62.20
3.29
3.26
2.36, 2.19
5.53
s
7.07
7.56
s
s
7.41
7.37
7.29
2.84
2.78
2
3
4
5, 9
6, 8
7
34.01
76.38
2.31, 2.23
5.74
s
7.08
7.55
s
s
7.42
7.36
7.27
2.51
2.35, 2.28
5.58
s
7.06
7.55
s
s
7.46
7.36
7.28
160.09
116.28
126.77
121.36
139.97
125.94
128.74
128.01
32.25
160.66
116.20
126.82
120.98
141.21
126.19
128.58
127.64
10
11, 15
12, 14
13
N−CH3
CF3
124.38
s
124.53
s
s
1′
3.44, 3.04
3.33, 3.15
4.16
3.58
3.52
3.39
3.58
3.54
4.24
3.33
3.30
3.65
3.75
Figure 1
s
HPLC data for the reaction of fluoxetine HCl, lactose, and KOH at
2
3
′
′
pH* 8.0, 60
°
C in aqueous ethanol.
weight for compound 2, consistent with the reduction of
the ketone with sodium borohydride.
4
5
6
1
2
3
4
5
6
′
′
′
′′
′′
′′
′′
′′
′′
Rea ction s in Aqu eou s Eth a n olsFluoxetine HCl and
lactose monohydrate were heated to 60 °C in aqueous
ethanol and the concentrations of 2 and 3 were measured
as a function of the amount of added base and time.
Apparent rates for the appearance of the two products are
given in Table 2. They were obtained from a linear fit to
their concentrations obtained by HPLC using at least five
data points but excluding the origin, since a slight induc-
tion period was evident in the reactions. The linear fits
had coefficients of determination (R2) of 0.97 or greater.
Data for only the first few percent conversion were used,
where the order is effectively zero with all reactants so that
apparent rates in (mole) percent per hour are reported
rather than rate constants to add relevance.
Without addition of KOH, the apparent pH (pH*) was
5.0 and the reaction was very slow. However, addition of
only 0.07 equiv of KOH (pH* ) 7.0) increased the rate of
appearance of 2 by a factor of 60. At higher pH values,
the rate of formation of 2 increased further and the
production of N-formylfluoxetine, 3, was observed. Rep-
resentative data from pH* ) 8.0 is shown in Figure 1. The
peak assigned to the glycosylamine, 4, eluted just before 2
and was assumed to be this compound based on the fact
that it was the first visible product of the reaction, was
quickly converted to 2, and had a molecular weight of 633,
as determined by LC/MS analysis. It is clear from this
solution-phase study that the Maillard and Amadori reac-
tions of fluoxetine HCl and lactose are viable processes and
that they are readily catalyzed by base.
4.28
3.66
a Proton and carbon chemical shifts are taken from the HMQC data.
b Numbering for 1 and 2 are similar to that shown for 6.
complete characterization of 2 was not accomplished.
Rather, sufficient mass spectral and NMR characterization
was performed to be reasonably confident of the structure
and this was supplemented with more complete charac-
terization on the alcohol 6, prepared by reduction of 2 with
sodium borohydride. The NMR assignments can be found
in Table 1. Several key elements were needed to confirm
the structure of 6. Specifically, the combination of proton-
proton coupling (COSY) and proton-carbon long-range
coupling (HMBC), along with basic proton and carbon
chemical shift information, was used to show the following
(refer to Table 1). First, the fluoxetine moiety in the
molecule is intact. Second, the N-methyl protons of flu-
oxetine show coupling to the adjacent methylene C1 of
fluoxetine, and more importantly to the methylene C1′ of
the lactose moiety, thereby establishing the site of union
between the fluoxetine and the lactose. Finally, the
galactose unit of the lactose has been retained. The
anomeric carbon at 1′′ maintains the downfield chemical
shift of 103.9 ppm seen in lactose.26 The HMBC experi-
ment also shows coupling from the proton at 3.39 ppm of
4′ to the 1′′ carbon. Similarly, the 1′′ proton at 4.24 ppm
shows coupling to the 4′ carbon at 75.3 ppm.
Formamides and acetamides are well-known products
of the decomposition of Amadori rearrangement products,
both for primary27 and secondary amines.28 Although both
formylation and acetylation of fluoxetine have been ob-
served in the studies reported here, the former has been
observed throughout, often as the primary path of decom-
position, whereas only small amounts of acetylation have
been observed. Two paths to formamides in the Maillard
The NMR data is supported by the mass spectral work,
which identified the molecular weight of the sample to be
635 amu, consistent with the proposed structure of 6. This
is also exactly two mass units higher than the molecular
34 / Journal of Pharmaceutical Sciences
Vol. 87, No. 1, January 1998

Table 3
−Solid Phase Screening Experiments for Mixtures Heated at
98 C for 24 h
°
% Impurities
N-formyl-
lactose
type
mixing
mode
Mg
stearate
total
ARP, 2
fluoxetine, 3
monohydrate
spray-dried
tumble
grind
N
Y
N
Y
N
Y
N
Y
N
Y
N
Y
N
N
N
N
N
N
0.48
0.46
0.48
0.6
6.13
17.3
1.2
0.92
1.48
1.44
22.8
24.2
0.46
0.5
3.11
0.44
0.53
2.06
0.073
0.16
0.089
0.21
3.23
8.56
0.26
0.4
0.023
0.039
0.02
0.4
wet grind
tumble
grind
0.036
0.02
0.12
0.078
0.48
0.45
0.41
5.1
Figure 2sLactose monohydrate and fluoxetine HCl at 85 °C.
wet grind
13
under normal storage conditions. The gradient HPLC
method was used to measure both the total amount of
impurities as well as amounts of 2 and 3.
anhydr gran
tumble
grind
wet grind
tumble
grind
0.072
0.079
1.51
0.079
0.107
0.86
0.014
0.011
anhydr milled
The level of total impurities by HPLC in the bulk
fluoxetine HCl used for this experiment was 0.23%; the
level did not increase by heating fluoxetine HCl by itself
and no significant impurities were observed by heating a
lactose control individually. Analysis of replicates indicates
the relative error in the impurity measurements to be
somewhat high, about 20%. However, the wide range of
results still allows for several statistically valid conclusions.
The Maillard reaction was observed in all experiments and
its rate was slightly faster with spray-dried lactose than
with lactose monohydrate. Many products were detected
by HPLC analysis, most of which were early-eluting
carbohydrate-containing materials based on LC/MS analy-
sis. Water accelerates the process. The rate of ARP
formation with anhydrous and monohydrate lactose forms
were similar, but the anhydrous form was less sensitive
to added water, possibly due to conversion to lactose
monohydrate. The rate was not significantly different in
blended mixtures versus ground mixtures. The addition
of magnesium stearate catalyzed the formylation process.
On the basis of the pH-dependent data generated in
solution, this effect may be due to localized changes in pH
rather than changes in physical mobility within the solids
due to the lubricant. This set of experiments clearly
indicated that the Maillard reaction of fluoxetine HCl and
lactose were likely to occur, even in the solid phase, and
that factors such as moisture content, type of lactose, and
the presence of magnesium stearate could have an influ-
ence on the rates and the identities of subsequent decom-
position products.
The analysis of several lactose-containing generic fluox-
etine HCl products by X-ray powder diffraction (XRPD)
indicated that only lactose monohydrate (or spray-dried
lactose) was being used; therefore, additional work with
the anhydrous forms was not pursued. X-ray analysis was
also used to qualitatively verify that even after heating for
several days at 100 °C and in mixtures with substantial
amounts of impurities, crystalline lactose monohydrate
remained; anhydrous lactose was not detected, although
some amorphous material could have been present. To
examine the rates of this drug-excipient interaction in
more detail and at a lower temperature, a series of kinetic
experiments was performed as described in the experimen-
tal section. These conditions were chosen to be more
representative of stress stability conditions typically used
during drug development. The impurity content of the
starting fluoxetine HCl was 0.20%. Representative results
for 85 °C are displayed in Figures 2-4, which use the same
vertical scale for ease of comparison.
wet grind
reaction are generally considered: the oxidative cleavage
of the glycosylamine (or a smaller fragment of it)3 or the
acylation of free amine with a 1,2-dicarbonyl compound,
especially glyoxal.29 In the present case, addition of 0.1
equiv of glyoxal to a 0.1 M solution of fluoxetine HCl
containing 0.5 equiv of KOH in aqueous ethanol at 60 °C
did produce N-formylfluoxetine, 3, as well as a few other
unidentified products. This is consistent with but does not
prove that glyoxal is the precursor to N-formylfluoxetine
in the solution or solid-phase experiments. Similar spiking
experiments with formic acid or glyoxalic acid produced no
new products.
Solid Sta te Rea ction s, Mon itor ed by HP LCsHaving
shown that the Maillard reaction of lactose and fluoxetine
occurs in solution and that 2 and 3 are the expected
products, attention turned to the more pharmaceutically
relevant case of solid-phase reactions. The simplest ex-
periment to identify such a drug-excipient interaction is
visual observation of a heated mixture.30 Such an approach
was used over 30 years ago to warn formulation scientists
of the browning of lactose formulations of drugs which
contain primary and secondary amine functionality.31
Indeed, when lactose and fluoxetine HCl as a solid mixture
were heated overnight at 100 °C, both browning and a
characteristic malt-like odor were obvious. More recent
studies designed to detect drug-excipient interactions
generally rely upon the use of DSC32 or diffuse reflectance
FT-IR.33 We chose to use the gradient HPLC method to
directly analyze products from solid-state reactions.
To identify the important factors which would influence
the solid state Maillard reaction of lactose and fluoxetine
HCl, a series of screening experiments with various lactose
types, mixing methods, water content, and lubricant con-
centrations were performed. The experiments are de-
scribed in the experimental section and the results are
summarized in Table 3. Four commonly available types
of lactose were screened. Three mixing modes were chosen
to mimic typical drug product unit operations including
blending, milling, and wet granulation. Experiments with
lactose monohydrate were also repeated with magnesium
stearate added since this lubricant was commonly found
in the generic lactose formulations. The relatively high
temperature of 98 °C was chosen for expediency; the results
are for comparison within the experiment, not for predic-
tion of overall rates of decomposition of drug mixtures
Journal of Pharmaceutical Sciences / 35
Vol. 87, No. 1, January 1998

Figure 4
sLactose monohydrate, fluoxetine HCl, and magnesium stearate at
Figure 3sSpray-dried lactose and fluoxetine HCl at 85 °C.
85 C.
°
It is evident that in this study, substantially more ARP
(2) was formed with spray-dried lactose (Figure 3) than
when lactose monohydrate was used (Figure 2), consistent
with previous reports.13 In the former case, the total level
of impurities exceeded 1% after just 1 day at 85 °C.
Consistent with the screening study, the presence of
magnesium stearate increased the amount of N-formylflu-
oxetine, 3; compare Figures 4 and 2. Control experiments
with isolated fluoxetine HCl and lactose were performed;
both are stable under the conditions studied.
Temperature-dependent analysis of the results was
performed using the simplest model for the production of
an unstable product (eq 1). Product X refers to decomposi-
tion products of 2, not any specific compound. The rates
of formation of 2 and X were derived from eqs 2 and 3 in
which Tr, the reference temperature, is 348 K. This
modification was used to make the preexponential factor
more independent of the activation energy, as suggested
by Blau and co-workers in their description of Simusolv,34
the simulation software in use. Only data for the concen-
tration of 2 were used in the model since the identities of
its degradation products, and thus their response factors,
are largely unknown. Arrhenius parameters are sum-
marized in Table 4.
Table 4
s
Arrhenius Parameters for Fluoxetine HCl and Lactose
3a
Thermolysis Based on Eqs 1
−
monohydrate
Mg stearate
+
parameter
monohydrate
spray-dried
k1
k2
×
×
105, 1/h
102, 1/h
4.8(0.3)
2.4(0.2)
24.9(1.1)
13.4(1.8)
0.97
6.9(0.9)
2.3(0.1)
34.9(1.3)
18.5(2.0)
0.96
11.9(0.8)
4.0(0.3)
26.8(1.3)
16.6(1.5)
0.98
E1, kcal/mol
E2, kcal/mol
R2 for model
months to 2
0.1% (30
+
X
)
5.9
34
3.5
°
C)
a The standard deviation is in parentheses.
k1
k2
1
9
8 2
9
8 X
(1)
[2]
E1 1
E2 1
d[2]
dt
1
1
) k1 exp
-
[1] - k2 exp
-
( (
T))
( (
T))
R Tr
R Tr
(2)
(3)
E2 1
Figure
5sFormation of ARP, 2, from solid-phase mixtures of lactose
d[X]
dt
1
) k2 exp
-
[2]
monohydrate, fluoxetine HCl, and magnesium stearate.. Points are experi-
mental; lines are from eq 2.
( (
T ))
R Tr
The last line in Table 4, an extrapolation to the time
required for the sum of 2 and its degradants, X, to total
0.1% at 30 °C, according to eqs 2 and 3, is included as an
indication of the possible real-time magnitude of this drug-
excipient interaction. The prediction that this process
should be observable at ambient temperature is confirmed
by the detection of ARP, 2, in unstressed, unexpired lactose
formulations; refer to Table 5. Figure 5 shows the experi-
mental concentration of ARP, 2, versus time for the mixture
of lactose monohydrate, fluoxetine HCl, and magnesium
stearate at all three temperatures and the curves generated
from eq 2 using the parameter estimates given in Table 4.
Due to the competitive decomposition of 2, its concentration
rises to a moderate level and remains fairly constant. That
is, it is produced and consumed at approximately the same
apparent rates. That it continues to be produced is
indicated by continued growth in other impurity peaks,
including N-formylfluoxetine, 3.
Due to severe limitations on mobility, the kinetics of solid
state systems, especially bimolecular ones, are not straight-
forward. Cohn utilized a function similar to the Arrhenius
equation but which includes a parameter for the thickness
of the reacting solid phases.35 Carstensen has reviewed
the problem as applied to the stability of solids and solid
dosage forms and uses both the Arrhenius expression as
well as Cohn’s equation.36 In the present study, the use
of data from only the first few percent of the conversion
minimizes the impact of these homogeneity issues, and the
goodness of the fit to the model indicates that valid
assumptions have been made.
The relevance of the above analysis to actual marketed
generic fluoxetine HCl products was examined by replica-
36 / Journal of Pharmaceutical Sciences
Vol. 87, No. 1, January 1998

Table 5
s
Stability of Fluoxetine HCl Products at 40
°C, 75% RH
Table 5. Blanks indicate a level less than the detection
limit of 0.01%.
% impurity by HPLC
Several points are of interest. First, the formulation
with starch showed very limited impurity formation while
the two formulations with lactose more than doubled in
their total impurity content. Second, generic A (and other
formulations examined) contain detectable amounts of 2
even before stressing, proving that the Maillard reaction
occurs even at ambient temperature in unexpired products.
Third, significant amounts of N-formylfluoxetine, 3, was
produced by stressing both lactose formulations and its
growth rate in generic A was faster than in generic Z, which
may be due to the higher concentration of magnesium
stearate in the former. Finally, although formation of 2 is
proof of the Maillard reaction, its concentration may not
rise to a very high level due to its conversion to other
products, including 3. Thus, these formulations exhibited
the same general behavior as noted in the solution phase
experiments and the excipient interaction studies.
Solid -Sta te Rea ction s Mon itor ed by Hea d sp a ce GC/
MSsAlthough many of the products of the Maillard
reaction are nonvolatile, such as the ARPs and melano-
idins, many volatile products are produced as well. These
have typically been identified by gas chromatography,
especially by detection with a mass spectrometer. These
studies frequently involve thermolysis of the ARP itself
rather than the sugar-amine mixture. For example, Mills
and Hodge studied the volatile products from pyrolysis of
1-deoxy-1-L-prolino-D-fructose,37 Hayase and Kato studied
the glucose-butylamine system,27 Birch et al. examined
1-amino-1-deoxyfructoses derived from alkylamines,38 and
Shigematsu et al. has identified volatile products from
pyrolysis from several ARPs derived from glucose and
amino acids.39 Although a diverse array of products has
been identified, some compounds or classes of compounds
are almost always observed. These include carboxylic acids
(especially formic and acetic acid), acyclic carbonyl com-
pounds such as aldehydes and ketones, dicarbonyl com-
pounds such as 2,3-pentanedione, furans such as furan
itself but especially 2-furaldehyde and furfuryl alcohol,
other oxygen-containing heterocycles such as pyranones
and furanones and their dihydro counterparts, amides of
the amines (especially formamides and acetamides), and,
in cases where a primary amine was used, pyrroles and
other nitrogen-containing heterocycles.
product
Prozac
impurity
initial
0.17
1 mo
3 mo
0.21
6 mo
9 mo
0.23
total
2
3
total
2
3
total
2
3
0.19
0.23
(starch)
0.01
0.47
0.01
0.02
0.35
0.01
0.01
0.01
0.60
0.01
0.03
0.45
0.02
0.04
0.01
0.90
0.05
0.13
0.63
0.03
0.08
0.02
1.10
0.05
0.22
0.74
0.02
0.13
generic A
(lactose)
0.43
0.03
generic Z
(lactose)
0.30
Figure 6sGeneric N, a fluoxetine HCl/lactose capsule at 85 °C.
To provide the link between the drug-excipient interac-
tion under study and the cited chemistry (which was done
mainly with monosaccharides and simple amines or amino
acids), we have examined the volatile products of the solid-
state reaction of lactose monohydrate and fluoxetine HCl
by GC/MS as reported in the experimental section. The
data from six such experiments is summarized in Table 6.
These include lactose-fluoxetine HCl with and without
added magnesium stearate (5% relative to lactose), two
generic capsule formulations, and two positive controls with
dimethylamine HCl and piperidine HCl. The top part of
the table contains products derived from the disaccharide
while the lower portion contains products derived from the
amine. Major products are indicated by +++ and minor
ones by a single +; the quantities are thus relative within
an experiment and between experiments but were not
accurately determined.
Three “negative" controls, lactose alone, fluoxetine HCl
alone, and a Prozac capsule (starch-based formulation)
were also performed. Essentially no decomposition was
observed; none of the compounds in Table 6 were produced
in these control experiments.
All of the products derived from the sugar have been
observed by previous workers in the Maillard-Amadori
arena, thus clearly linking the present study to their
Figure 7sGeneric O, a fluoxetine HCl/lactose tablet at 85 °C.
tion of the stress stability study with a capsule (generic
N) and a tablet (generic O). The results are shown in
Figures 6 and 7, respectively. Both products displayed
Maillard browning and produced 2, 3, and many other
impurities. The results from the tablet (Figure 7) are
comparable with decomposition profiles of the simple
mixtures of lactose, drug, and magnesium stearate. How-
ever, the capsule was significantly less stable, showing
formation of nearly 4% of 2 after 122 h at 85 °C, and a
total impurity level of 13%. This implies the presence of
additional catalysts in this capsule, the identities of which
are unknown.
The third type of stability study on lactose formulations
of fluoxetine HCl utilized the more standard accelerated
conditions of 40 °C and 75% relative humidity used to
support regulatory submissions. In this study the gradient
HPLC assay was used to assess the stability of Prozac and
two generic lactose-containing capsules, designated generic
A and generic Z. The results from 9 months are given in
Journal of Pharmaceutical Sciences / 37
Vol. 87, No. 1, January 1998

Table 6sVolatile Products from Pyrolysis (100 °C, 20 h) of Lactose and Secondary Amine Hydrochlorides
GC/MS
retention
(min)
fluox HCl
lactose
Me2NH HCl
piperidine HCl
fluox HCl
+
generic
A
generic
Z
+
lactose
Mg St
+
lactose
compound
+
lactose
+
Mg stearate
+
+ Mg St
vinyl acetatea
acetic acida
3.1
3.8
5.47
6.70
8.04
8.43
9.12
9.36
10.21
11.79
14.19
14.79
10.12
13.45
29.34
7.09
+
++
+
+
+++
+
+
+++
+
+
+++
+
+++
+
2,3-pentanedionea
4-methyl-3-pentenal
3-furaldehydea
2-furaldehydea
2-furfuryl alcohola
3-furfuryl alcohola
2-acetylfurana
isomaltol
+
+
++
+
+
+
+
+
+
++
++
+
+
+
+
+
+
+
+
+
++
++
+
+++
+
+
+
maltola
+
+++
+
+++
+
++
2,3-dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one (5)a
styrenea
+
+++
+
4-(trifluoromethyl)cresola
N-formylfluoxetine (3)a
N,N-dimethylformamidea
N,N-dimethylacetamidea
N-formylpiperidinea
+++
+
++
+
9.15
14.53
15.74
++
+
N-acetylpiperidine
a Matched both the mass spectrum from a library and the retention time of authentic compound.
variations such as inclusion of the nitrogen within rings
or the presence of functionality which would greatly
diminish the nucleophilicity of the alkaloid drug. The
present findings suggest a relatively simple experimental
design to probe this question, namely, the use of the
formamide derivative as a chemical marker for the Mail-
lard reaction. In all systems studied here (and most in the
literature in which their presence would have been de-
tected) N-formyl compounds have been observed as a major
product of the Maillard process. Therefore, simple reaction
of an alkaloid drug with acetic formic anhydride should
provide samples of N-formyl compound(s) which may or
may not be isolated and purified, but which are usually
stable and usually separable from the drug on chromato-
graphic systems such as GC, HPLC, or electrophoresis.
These derivatives will often be readily available and
detectable and their formation in stressed samples of
potential formulation mixtures will be a strong indication
of the Maillard reaction. This chemical marker will be
more accessible than Amadori rearrangement products
since the latter are often unstable and are difficult to
prepare, isolate, and characterize. The more traditional
method of screening simply for color and odor is still
possible; however, these sensory clues are not present in
early stages of the Maillard process when impurity levels
may exceed the level of 0.1%.18 An alternative screening
program for the presence of Maillard-based drug-excipient
interactions using GC/MS detection of volatile carbohydrate-
derived compounds such as given in Table 6 could also be
useful.
Scheme 3sDecomposition of ARP, 2.
findings. Of particular note is the widespread formation
of 2-furaldehyde and other furans and 2,3-dihydro-3,5-
dihydroxy-6-methyl-4H-pyran-4-one, 5; see Scheme 3. The
identity of this compound was proven by comparison with
authentic material, prepared according to a published
method.40 Kim and Baltes recently reported that many
cyclic and acyclic Maillard products, including several listed
in Table 6, can be derived from thermolysis of this pyra-
none.41 The possible presence of 1,2-dicarbonyl compounds
(such as 2,3-pentanedione) in thermally stressed generic
fluoxetine HCl products is of concem since some of them,
including glyoxal and maltol, have been reported to be
weak bacterial mutagens.42
It is also noteworthy that many of the same products
were produced from the reactions with the HCl salts of
dimethylamine and piperidine as with fluoxetine HCl. For
example, the formamides of all three amines were found
as well as furaldehyde and maltol. This proves that
fluoxetine is not unique or unusual in this regard; nearly
all secondary amines whether cyclic or acyclic, should
undergo similar reactions with lactose, in agreement with
the bulk of the scientific literature.
The results of the studies on solution-phase experiments,
solid-phase excipient interaction screening, stability studies
on fluoxetine HCl and lactose and formulated products, and
the analysis of products clearly prove that the Maillard
reaction occurs in this system. Although the clinical
significance of this drug-excipient interaction is unknown,
the relevance of these findings to formulation scientists are
more straightforward. Namely, the Maillard reaction of
secondary amines and lactose should be considered when
selecting formulation ingredients and when examining the
stability of such products. However, in many cases, the
formulation scientist will not know whether a specific
nitrogen-containing drug will be compatible with reducing
carbohydrates or not, usually due to significant structural
Acknowledgments
Technical advice from Fred M. Perry, Richard A. Berglund,
Richard Frank, Douglas P. Kjell, Douglas E. Dorman, Bernard A.
Olsen, and Kerry J . Hartauer is gratefully acknowledged. Ad-
ditional assistance such as data manipulation and analytical
support was provided by Willis V. Bell III, Sharon V. Snorek, Gary
L. Thomas, Melinda R. Wilhelm, Patrick J . J ansen, and David G.
Robbins. Elizabeth S. LaBell provided the sample of 3.
References and Notes
1. (a) Maillard, L.-C. C. R. Acad. Sci. Ser. 2, 1912, 154, 66. (b)
Maillard, L.-C. C. R. Acad. Sci. Ser. 2 1912, 155, 1554-6.
38 / Journal of Pharmaceutical Sciences
Vol. 87, No. 1, January 1998

2. (a) Ellis, G. P. The Maillard Reaction. Adv. Carbobydr. Chem.
1959, 14, 63-134. (b) Hodge, J . E. Dehydrated Foods,
Chemistry of Browning Reactions in Model Systems. J . Agric.
Food Chem. 1953, 1, 928-43. (c) Yaylayan, V. A.; Huyghues-
Despointes, A. Crit. Rev. Food Sci. Nutr. 1994, 34, 321-369.
3. (a) Ledl, F. Chemical Pathways of the Maillard Reaction. In
The Maillard Reaction in Food Processing, Human Nutrition
and Physiology; Finot, P. A., Aeschbacher, H. U., Hurrell,
R. F., Liardon, R., Eds.; Birkhauser Verlag: Basel, 1990; pp
19-42. (b) Ledl, F.; Schleicher, E. New Aspects of the
Maillard Reaction in Foods and in the Human Body. Angew.
Chem., Int. Ed. Engl. 1990, 29, 565-94.
22. Aue, W. P.; Bartholdi, E.; Ernst, R. R. Two-Dimensional
Spectroscopy. Application to Nuclear Magnetic Resonance.
J . Chem. Phys. 1976, 64, 2229-46.
23. Bax, A.; Subramanian, S. Sensitivity-Enhanced Two-Dimen-
sional Heternuclear Shift Correlation NMR Spectroscopy. J .
Magn. Reson. 1986, 67, 565-9.
24. Bax, A.; Summers, M. F. Proton and Carbon-13 Assignments
from Sensitivity-Enhanced Detection of Heteronuclear Mul-
tiple-Bond Connectivity by 2D Multiple Quantum NMR. J .
Am. Chem. Soc. 1986, 108, 2093-4.
25. Tjan, S. B.; van den Ouweland, G. A. M. PMR Investigation
into the Structure of some N-Substituted 1-Amino-1-deoxy-
D-fructoses (Amadori Rearrangement Products). Tetrahedron
1974, 30, 2891-7.
26. (a) Bailey, K., Butterfield, A. G. Structure of Condensation
Products between Isonicotinylhydrazine and Reducing Sug-
ars by 13C NMR Spectroscopy. Can. J . Chem. 1981, 59, 641-
6. (b) Platzer, N.; Davoust, D.; Lhermitter, M.; Bauvy, C.;
Meyer, D. M.; Derappe, C. Structural Analysis of Five
Lactose-Containing Oligosaccharides by Improved High-
Resolution, Two-Dimensional Proton NMR Spectroscopy.
Carbobydr. Res. 1989, 191, 191-207.
4. Paulsen, H.; Pflughaupt, K.-W. Glycosylamines. In The
Carbobydrates, Chemistry and Biochemistry; Horton, D.,
Pigman, W., Eds.; Academic Press: New York, 1980; Vol 1b,
pp 881-927.
5. Hodge, J . E. The Amadori Rearrangement. Adv. Carbobydr.
Chem. 1955, 10, 169-205.
6. Ledl, F.; Fritsch, G.; Hiebl, J .; Pachmayer, O.; Severin, T.
Degradation of Maillard Products. In Amino-Carbonyl Reac-
tions in Food and Biological Systems; Fujimaki, M., Namiki,
M., Kato, H., Eds.; Elsevier: Amsterdam, 1986; pp 173-82.
7. Pischetsrieder, M.; Severin, T. The Maillard Reaction of
Disaccharides. In Maillard Reactions in Chemistry, Food,
and Health; Labuza, T. P., Reineccius, G. A., Monnier, V.,
O’Brien, J ., Baynes, J ., Eds.; Spec. Pub. 151; Royal Soc.
Chem.: London, 1994; pp 37-42.
27. Hayase, F.; Kato, H. Low-Molecular Maillard Reaction
Products and Their Formation Mechanisms. In Amino-
Carbonyl Reactions in Food and Biological Systems; Fuji-
maki, M., Namiki, M., Kato, H., Eds.; Elsevier: Amsterdam,
1986; pp 39-48.
8. Van Boekel, M. A. J . S.; Berg, H. E. Kinetics of the Early
Maillard Reaction During Heating of Milk. In Maillard
Reactions in Chemistry, Food, and Health; Labuza, T. P.,
Reineccius, G. A., Monnier, V., O’Brien, J ., Baynes, J ., Eds.;
Spec. Pub. 151; Royal Soc. Chem.: London, 1994; pp 170-5.
9. Hodge, J . E.; Rist, C. E. The Amadori Rearrangement under
New Conditions and its Significance for Nonenzymatic
Browning Reactions. J . Am. Chem. Soc. 1953, 75, 316-322.
See also refs 2-4.
10. Colaco, C. A. L. S.; Collett, M.; Roser, B. J . Pharmaceutical
Formulation Instability and the Maillard Reaction. Chem.
Oggi 1996, 14, 32-7.
11. Kumar, V.; Banker, G. S. Maillard Reaction and Drug
Stablity. In Maillard Reactions in Chemistry, Food, and
Health; Labuza, T. P., Reineccius, G. A., Monnier, V.,
O’Brien, J ., Baynes, J ., Eds.; Spec. Pub. 151; Royal Soc.
Chem.: London, 1994; pp 20-27.
12. Wong, D. T.; Bymaster, F. P.; Engleman, E. A. Prozac
(Fluoxetine, Lilly 110140), The First Selective Serotonin
Uptake Inhibitor and an Antidepressant Drug: Twenty
Years Since its First Publication. Life Sci. 1995, 57, 411-
41.
13. Goodhart, F. W. Lactose, In Handbook of Pharmaceutical
Excipients, 2nd ed.; Wade, A., Weller, P. J ., Eds.; American
Pharmaceutical Association: Washington DC, 1994; pp 252-
7.
28. Mills, F. D.; Baker, B. G.; Hodge, J . E. Thermal Degradation
of 1-Deoxy-1-piperidino-D-Fructose. Carbobydr. Res. 1970, 15,
205-13.
29. Hayase, F.; Kim, S. B.; Kato, H. Maillard Reaction Products
Formed from D-Glucose and Glycine and the Formation
Mechanisms of Amides as Major Components. Agric. Biol.
Chem. 1986, 49, 2337-41.
30. Castello, R. A.; Mattock, A. M. Discoloration of Tablets
Containing Amines and Lactose. J . Pharm. Sci. 1962, 51,
106-8.
31. Hammouda, V.; Salakawy, S. A. Lactose-induced Discolora-
tion of Amino Drugs in Solid Dosage Form. Pharmazie 1971,
26, 181.
32. (a) van Dooren, A. A. Design for Drug-Excipient Interaction
Studies. Drug Dev. Ind. Pharm. 1983, 9, 43-55. (b) Chrza-
nowski, F. A.; Ulissi, L. A.; Fegely, B. J .; Newman, A. C.
Preformulation Excipient Compatibility Testing. Application
of a Differential Scanning Calorimetric Method versus a Wet
Granulation Simulating, Isothermal Stress Methodol. Drug
Dev. Ind. Pharm. 1986, 12, 783-800.
33. Hartauer, K. J .; Guillory, J . K. A Comparison of Diffuse
Reflectance FT-IR Spectroscopy and DSC in the Character-
ization of a Drug-Excipient Interaction. Drug Dev. Ind.
Pharm. 1991, 17, 617-30.
34. Blau, G. E.; Kuenker, K. E.; Agin, G. L; Steiner, E. C.
Chemical Reactor Design and Analysis with Simusolv,
presented at the 7th International Symposium on Large
Chemical Plants, October 6-7, 1988, Brugge, Belgium.
Simusolv is a trademark of The Dow Chemical Company.
35. Cohn, G. Reactions in the Solid State. Chem. Rev. (Wash-
ington, D.C.) 1948, 42, 527-73.
14. Wolfrom, M. L.; Cavalieri, L. F.; Cavalieri, D. K. Chemical
Interactions of Amino Compounds and Sugars. II. Methyl-
ation Experiments J . Am. Chem. Soc. 1947, 69, 2411-3.
15. Duvall, R. N.; Koshy, K. T.; Pyles, J . W. Comparison of
Reactivity of Amphetamine, Methamphetamine, and Di-
methylamphetamine with Lactose and Related Compounds,
J . Pharm. Sci. 1965, 54, 607-611.
36. Carstensen, J . T. Stability of Solids and Solid Dosage Forms.
J . Pharm. Sci. 1974, 63, 1-14.
16. Simon, H.; Kraus, A. Fortschr. Chem. Forsch. 1970, 14, 430-
70.
37. Mills, F. D.; Hodge, J . E. Amadori Compounds: Vacuum
Thermolysis of 1-Deoxy-1-L-Prolino-D-Fructose. Carbohydr.
Res. 1976, 51, 9-21.
17. Simon, H.; Palm, D. Chem. Ber. 1965, 98, 433-45.
18. (a) Labuza, T. P. Interpreting the Complexity of the Kinetics
of the Maillard Reaction. In Maillard Reactions in Chemistry,
Food, and Health; Labuza, T. P., Reineccius, G. A., Monnier,
V., O’Brien, J ., Baynes, J ., Eds.; Spec. Pub. 151; Royal Soc.
Chem.: London, 1994; pp 176-181. (b) Labuza, T. P.; Baisier,
W. M. The Kinetics of Nonenzymatic Browning. In Physical
Chemistry of Foods; Schwartzberg, H. G., Hartel, R. W., Eds.;
Marcel Dekker: New York, 1992; pp 595-649.
38. Birch, E. J .; Lelievre, J .; Richards, E. L. Influence of the
amine moiety on the thermal degradation of N-substituted
1-amino-1-deoxyfructoses. Carbobydr. Res. 1984, 128, 351-
5.
39. Shigematsu, H.; Shibata, S.; Kurata, T.; Kato, H.; Fujimaki,
M. Thermal Degradation Products of Several Amadori Com-
pounds. Agric. Biol. Chem. 1977, 41, 2377-85.
40. Van Den Ouweland, G. A. M.; Peer, H. G. Synthesis of 3,5-
Dihydroxy-2-methyl-5,6-dihydropyran-4-one from Aldohex-
oses and Secondary Amine Salts. Rec. Trav. Chim. Pays-Bas
1970, 89, 750-4.
19. Pischetsrieder, M.; Severin, T. Advanced Maillard Products
of Disaccharides: Analysis and Relation to Reaction Condi-
tions. In Chemical Markers for Processed and Stored Foods;
Lee, T-C, Kim, H-J ., Eds.; ACS Symposium Series 631; Am.
Chem. Soc.: Washington, DC, 1996; pp 14-23. See also refs
1-10.
20. Wirth, D. D.; Olsen, B. A.; Hallenbeck, D. K.; Lake, M. E.;
Gregg, S. M.; Perry, F. M. Screening Methods for Impurities
in Multi-sourced Fluoxetine Hydrochloride Drug Substances
and Formulations. Chromatographia 1997, 46, 511-23.
21. US Pharmacopeia 1997, 23, Seventh Supplement, 3890-2.
41. Kim, M-O.; Baltes, W. On the Role of 2,3-Dihydro-3,5-
dihydroxy-6-methyl-4(H)-pyran-4-one in the Maillard Reac-
tion. J . Agric. Food Chem. 1996, 44, 282-9.
42. Bjeldanes, L. F.; Chew, H. Mutagenicity of 1,2-Dicarbonyl
Compounds: Maltol, Kojic Acid, Diacetyl and Related Sub-
stances. Mutation Res. 1979, 67, 367-71.
J S9702067
Journal of Pharmaceutical Sciences / 39
Vol. 87, No. 1, January 1998