Direct evidence of 2-cyano-2-propoxy radical activity during AIBN-based oxidative stress testing in acetonitrile-water solvent systems

Article abstract of DOI:10.1002/jps.23500

Oxidative susceptibility testing was performed on a drug substance containing a methoxy-naphthalene moiety. 2,2′-azobisisobutyronitrile (AIBN) was employed to initiate peroxy radical oxidation to mimic autoxidation processes. In acetonitrile (ACN)-water solvents, three major degradation products are formed. However, addition of small amounts of methanol to the solvent system completely eliminated the observed degradation products. To understand this effect, the structures of the three degradants have been elucidated using nuclear magnetic resonance, liquid chromatography-tandem mass spectrometry, and accurate mass Fourier transform ion cyclotron resonance mass spectrometry. One degradant structure definitively proves the degradation resulted from alkoxy radicals (2-cyano-2 propoxy radical) arising from the disproportionation of the tertiary AIBN-derived peroxy radicals, rather than from the intended action of the AIBN peroxy radicals themselves. The reaction occurs over a wide range of AIBN and drug substance concentrations. This protective effect of several percent methanol by volume is rationalized by known methanol H atom donation rates to similar tert-butoxy and cumyloxy radicals (ca. 10 M^-1 s^-1) and the high methanol concentration relative to the dilute substrate being investigated. This work confirms recent proposals for addition of at least about 10% methanol to the standard ACN-water AIBN stress testing diluent to insure that only the desired peroxy radical activity is present during the oxidative stress test.

Full text of DOI:10.1002/jps.23500

Direct Evidence of 2-Cyano-2-Propoxy Radical Activity During
AIBN-Based Oxidative Stress Testing in Acetonitrile–Water
Solvent Systems
MICHAEL A. WATKINS, STEVEN PITZENBERGER, PAUL A. HARMON
Merck & Company, Merck Research Laboratories, West Point, Pennsylvania 19486
Received 12 September 2012; revised 11 January 2013; accepted 12 February 2013
Published online 11 March 2013 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.23500
ABSTRACT: Oxidative susceptibility testing was performed on a drug substance contain-
ing a methoxy-naphthalene moiety. 2,2ꢀ-azobisisobutyronitrile (AIBN) was employed to initiate
peroxy radical oxidation to mimic autoxidation processes. In acetonitrile (ACN)–water sol-
vents, three major degradation products are formed. However, addition of small amounts of
methanol to the solvent system completely eliminated the observed degradation products. To
understand this effect, the structures of the three degradants have been elucidated using nu-
clear magnetic resonance, liquid chromatography–tandem mass spectrometry, and accurate
mass Fourier transform ion cyclotron resonance mass spectrometry. One degradant structure
definitively proves the degradation resulted from alkoxy radicals (2-cyano-2 propoxy radical)
arising from the disproportionation of the tertiary AIBN-derived peroxy radicals, rather than
from the intended action of the AIBN peroxy radicals themselves. The reaction occurs over
a wide range of AIBN and drug substance concentrations. This “protective effect” of several
percent methanol by volume is rationalized by known methanol H atom donation rates to
similar tert-butoxy and cumyloxy radicals (ca. 10 M−1 s−1) and the high methanol concentra-
tion relative to the dilute substrate being investigated. This work confirms recent proposals
for addition of at least about 10% methanol to the standard ACN–water AIBN stress testing
diluent to insure that only the desired peroxy radical activity is present during the oxida-
tive stress test. © 2013 Wiley Periodicals, Inc. and the American Pharmacists Association
J Pharm Sci 102:1554–1568, 2013
Keywords: analysis; antioxidants; chemical stability; forced conditions; free radicals;
oxidation
stress test results has remained challenging. The pri-
mary reason for this is that the causative oxidants in
the solid dosage form may not be well understood, in
combination with a less than complete understanding
of the actual “oxidants” generated in oxidative stress
testing systems. Thus improving the predictive power
of oxidative stress testing methodologies continues to
INTRODUCTION
Pharmaceutical stress testing has become an integral
part of the drug development process, and numerous
literature reports1–5 and several book contributions
have been dedicated to this subject.6,7 A major aspect
of such stress testing is carrying out simple solution-
based studies, which can reveal potential hydrolytic
and oxidative liabilities of the active pharmaceutical
ingredient (API). Predicting the oxidative behavior
of formulated API’s in solid dosage forms under long
term stability conditions from the solution oxidative
be a subject of interest8–16
.
The work herein is rooted in this latter topic,
and in particular in understanding the actual ox-
idants present during early stage, solution phase
oxidative screening experiments designed to expose
drug molecules to peroxy radicals. Peroxy radical-
based oxidation (chain oxidation or sometimes re-
ferred to as free radical autoxidation) is a common
oxidative pathway as demonstrated by the ubiqui-
tous use of phenolic chain breaking antioxidants
such as 2,6-di-tert-butyl-4-methoxyphenol and 2,6-
di-tert-butyl-4-methylphenol, which quench peroxy
Correspondence to: Michael A. Watkins (Telephone:
+3176550643; Fax: +3176516510; E-mail: watkins michael a@
lilly.com)
Michael A. Watkins’ present address is Eli Lilly & Company,
Lilly Research Laboratories, Indianapolis, Indiana 46285
Journal of Pharmaceutical Sciences, Vol. 102, 1554–1568 (2013)
© 2013 Wiley Periodicals, Inc. and the American Pharmacists Association
1554
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 5, MAY 2013

DIRECT EVIDENCE OF 2-CYANO-2-PROPOXY RADICAL ACTIVITY
1555
Figure 1. Thermal degradation of AIBN to form peroxy radicals in solution.
radicals.10,17–19 A survey of stress testing industry
practices nearly 10 years ago found a growing trend
in the use of azonitrile “radical initiators” such as
AIBN (2,2ꢀ-azobisisobutyronitrile) to create peroxy
radical mediated oxidation in the oxidative screening
experiment.20 This trend has likely continued since
that time. These experiments rely on thermal de-
composition of an azonitrile compound to generate a
peroxy radical oxidant (2-cyano-2-peroxy radicals) as
shown in Figure 1 for AIBN. Generation of significant
degradant products from this type of AIBN oxidative
screening experiment is then interpreted as signal-
ing the drug molecule’s susceptibility toward autoxi-
dation. In this regard, the absence of degradant peaks
in this oxidative test is a highly desired outcome, as it
proves no chemical bonds are reactive with the highly
selective peroxy radical.16,17,21 This, in turn, portends
no peroxy radical driven oxidation will be observed in
formulated dosage forms over longer term drug prod-
uct ICH stability studies.
The desire to appropriately signal “nonreactivity”
in the subject selective oxidative screening test is the
central motivation for the current work. The poten-
tial problem comes from the way the azonitrile-based
experiment has evolved in the early phase, phar-
maceutical oxidative screening context. Dilute API
concentrations are typically used (ca. 0.1–1 mg/mL)
along with 20–100 mol % (or higher) of the azonitrile
compound. Typical solvents used historically were
acetonitrile (ACN)–water mixtures. Many of the com-
pounds being screened will not be reactive with per-
oxy radicals. Under these conditions, the dispropor-
tionation of the 2-cyano-2-peroxy radicals formed in
Figure 1 cannot be neglected. The issue is that the
2-cyano-2-peroxy radicals in Figure 1 are tertiary,
and as such disproportionation will result in forma-
tion of a new, much stronger oxidant as shown in
Figure 2.21 The 2-cyano-2 propoxy radical thus formed
is an alkoxy radical. Similar alkoxy radicals such as
tert-butoxy and cumyloxy radicals have reaction rates
with organic substrates which are about 104–105
times faster than the analogous peroxy radicals.22
These rates reflect the ∼20 kcal/mol higher bond dis-
sociation energies (BDEs) of the RO H bonds rel-
ative the ROO H bond as shown in Table 1. The
presence of this alkoxy radical in solution could de-
stroy the intended peroxy radical selectivity of this
oxidative screening test giving “false positive” re-
sults. This problem is not unique to AIBN but is
general for all of the azonitrile compounds com-
monly used for oxidative susceptibly studies in that
they all generate tertiary peroxy radicals similar to
that shown for AIBN in Figure 2. Werber et al.23
recently analyzed the degradation products of 2,2ꢀ-
azobis(2-amidinopropane) dihydrochloride in aque-
ous solution in the absence of any added substrate,
and concluded that the alkoxy radical (analogous to
that shown in Figure 2 for AIBN) was the promi-
nent radical species in solution. Nelson et al.15 have
recently reported AIBN and ACVA [4,4ꢀ-azobis(4-
cyanovaleric acid)] oxidative screening of cumene
(used as a model drug substance). Through several
lines of indirect evidence, they attributed oxida-
tive degradation products of cumene to the activ-
ity of 2-cyano-2-propoxy radical—rather than the
expected peroxy radicals. Similar to our observa-
tions with developmental drug compounds,24,25 Nel-
son and coworkers showed a 5–10-fold reduction in
the cumene oxidation products attributed to alkoxy
radical activity upon the addition of only a few volume
percent methanol to the ACN solvent-based AIBN
experiment.
Figure 2. Disproportionation of tertiary 2-cyano-2-peroxy radical.
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 5, MAY 2013

1556
WATKINS, PITZENBERGER, AND HARMON
Table 1. Calculated (∗) and Experimental Bond Dissociation
from Sigma–Aldrich (St. Louis, Missouri) and was
used as received. Nuclear magnetic resonance (NMR)
solvents CD3CN and d6-DMSO were obtained in
99.96%D isotopic purity from Cambridge Isotope
Labs. All other reagents were of analytical grade qual-
ity or higher.
Energies (BDEs)
Compound
Bond
BDE (kcal/mol)
2-Cyano-2-hydroperoxide NC(CH3)2COO
H
H
83a
106a
105b
∼103c
96d
2-Cyano-2-propanol
tert-Butanol
Cumyl alcohol
Methanol
NC(CH3)2CO
(CH3)3CO
C6H6(CH3)2CO
H
H
HO(CH2)
NC(CH2)
H
H
Acetonitrile
94b
METHODS
aCalculated as described in Experimental section.
High-Performance Liquid Chromatography and
Oxidation Conditions
bRef. 28.
cRef. 29.
dRef. 30.
2,2ꢀ-azobisisobutyronitrile and ACVA oxidative
stresses were typically performed in 50:50 ACN–
water solvents with 0%–20% methanol by vol-
ume added unless otherwise indicated. Azonitrile
compound was present at 5 mM (0.8 mg/mL) and
compound 1 present at 0.1 mM (0.05 mg/mL) unless
otherwise specified. The stress solutions were held at
40◦C for 3 days to provide adequate time for oxidation
to occur. An Agilent 1100 HPLC equipped with a
photodiode array detector and a thermally controlled
column compartment and autosampler was used
for all chromatographic analysis. Chromatographic
analysis was performed using an isocratic mobile
phase [29/71 (v/v) 0.1% formic acid–ACN] using
an Agilent Zorbax 4.6×100 mm, 1.8 :m SB-C18
column held at 60◦C with a mobile phase flow rate of
1.5 mL/min and a run time of 4 min.
Fourier transform ion cyclotron resonance mass
spectrometry (FT-ICR–MS) and NMR analyses of
each degradation product were conducted using iso-
lated and purified samples of each degradation prod-
uct prepared by preparatory scale high-performance
liquid chromatography (HPLC). Preparatory HLPC
was performed with an Agilent 1100 preparatory
HPLC system. Degradant 1 was isolated using
a Phenomenex Luna 21.2 × 250 mm, 5 :m C8
column held at ambient temperature with an isocratic
The current work for the first time unambiguously
proves the presence and activity of the 2-cyano-2-
propoxy radical in ACN–water solvent systems dur-
ing pharmaceutically relevant oxidative screening ex-
periments. AIBN and ACVA oxidative stressing of
compound 1 (Fig. 3) in ACN–water mixtures show
about 20% degradation of compound 1, which is
eliminated by addition of only a few volume per-
cent methanol. Full structural elucidation of the
degradation products provides direct evidence of 2-
cyano-2-propoxy radical activity. The dependence of
degradant yields on AIBN and drug concentration (in
the absence of methanol) is examined and discussed.
The protective effect of methanol is rationalized, and
the potential impact of methanol addition to the wide
range of mole percent AIBN levels commonly used
in oxidative screening experiments is discussed and
recommendations made.
EXPERIMENTAL
Materials and Reagents
The drug substance compound 1 was synthesized in
98% purity by Merck and Company, Inc. (Rahway,
New Jersey). AIBN was obtained in 98% purity
Figure 3. Compound 1 - (S)-3-(4-(1-(3-(3,5-dichlorophenyl)-5-(6-methoxynaphthalen-2-yl)-1H-
pyrazol-1-yl)ethyl)benzamido)propanoic acid; an API containing a methoxy-naphthalene moiety.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 5, MAY 2013
DOI 10.1002/jps

DIRECT EVIDENCE OF 2-CYANO-2-PROPOXY RADICAL ACTIVITY
1557
mobile phase consisting 60/40 ACN–0.1% formic acid
pumped at a flow rate of 5 mL/min with a 5 mL in-
jection volume. Degradants 2 and 3 were isolated us-
ing an Agilent Zorbax 9.4 × 250 mm, 5 :m SB-CN
column held at ambient temperature with an isocratic
mobile phase consisting 50:50 ACN–0.1% formic acid
pumped at a flow rate of 5 mL/min with a 5 mL
injection volume. Solvent was removed from the
preparatory fractions by rotary evaporation followed
by lyophilization.
were determined using density functional theory
calculations (BLYP/AUG-cc-pVDZ level of theory)
performed using a Linux High-Performance Com-
puting cluster and the Gaussian 09 computational
package.26 An isodesmic cycle with the experimen-
tally determined BDE of tert-butanol was used as a
reference.
RESULTS
Solvent Dependence of AIBN-Induced Degradation
Products
Liquid Chromatography–Mass Spectrometry
Three major degradation products are generated from
the AIBN oxidative stress of Compound 1 in ACN–
water solvent systems as shown in the upper chro-
matogram in Figure 4 (5 mM AIBN, 0.05 mg/mL
compound 1). Degradant 1 is the most abundant
and elutes just before compound 1; the total com-
pound 1 degraded is near 20% over the 3-day exper-
iment. The lower chromatogram in Figure 4 is from
an identical experiment except using a 50:50 (vol %)
methanol–water solvent system. No degradants are
observed. Figure 5 demonstrates just how little
methanol needs to be present to significantly quench
the formation of degradants 1–3. Adding 1% methanol
has a large effect, whereas 10% methanol has essen-
tially eliminated all degradant formation. A different
Azonitrile initiator, 4,4ꢀ-azobis (4-cyanovaleric acid)
(ACVA), was also used in 50:50 ACN–water solvent
to determine if the observed oxidation was specific to
AIBN. Even larger yields of degradant 2 and 3 were
observed, degradant 1 was not observed and there
were several new lower level peaks throughout the
chromatogram. All of the ACVA-induced degradants
disappeared upon addition of methanol at levels
similar to that shown for the AIBN case shown in
Figure 5.
A similarly equipped Agilent 1100 HPLC system was
interfaced with a Finnigan LCQ Deca-XP quadrupole
ion trap for nominal mass and MSn analysis of
degradation products. A Bruker Daltonics 7T-Actively
Shielded BioApex II FT-ICR/MS equipped with elec-
trospray ionization (ESI) for accurate mass analysis
of degradation products and MS–MS fragment ions.
Ions of interest were radio-frequency isolated in the
Infinity ICR cell before infrared multiphoton disso-
ciation for structural analysis. External/internal cal-
ibration was accomplished using polypropylene gly-
col oligomers. The Finnigan LCQ mass spectrometry
method for LC–MS–MS analysis employed positive
ion ESI with data dependent scanning (most abun-
dant ion was selected) for MS–MS analysis with a
collision-induced dissociation energy maintained at
20% (relative collision energy) while scanning the in-
strument from 100 to 1000 m/z. The ion source condi-
tions are as follows: ESI emitter voltage held at 5 kV,
capillary voltage of 5 V heated to 200◦C with a sheath
gas flow rate of 70 (arbitrary units) and an auxiliary
gas flow rate of 5 (arbitrary units).
Nuclear Magnetic Resonance
Nuclear magnetic resonance analyses were conducted
on VarianUNITY INOVA 600 MHz spectrometers, one
equipped with an indirect detection, “carbon en-
hanced,” cryogenically cooled, HCN triple resonance
probe and the other with a standard, indirect detec-
tion, HCN triple resonance probe. Compound 1 and
degradant 1 were analyzed by NMR as CD3CN solu-
tions. Because of solubility reasons, degradants 2 and
3 were analyzed as d6-DMSO solutions. The NMR so-
lutions were each spiked with tetramethylsilane to
serve as a chemical shift reference. The samples were
maintained at 25◦C for all NMR experiments. Stan-
dard Varian pulse sequences (1D, HMQC, gHMBC,
and ROESY) were used to acquire the NMR data.
Degradation Time-Course and Drug–AIBN
Concentration Study
The analysis of compound 1 degradation over time
was performed by preparing a 5 mM AIBN/0.05 mg/
mL compound 1 in 50:50 ACN–water stress solution,
adding this solution to multiple HPLC vials stored on
an HPLC autosampler heated to 40◦C and analyzing
one HPLC vial every 5 h for nearly 100 h. The degra-
dation time-course data in Figure 6 show no interde-
pendence of reaction rate for any of the degradation
products. In a separate experiment, each degradation
product (degradants 1, 2, and 3) was isolated in abun-
dance by preparative HPLC and was subsequently
subjected to a 3 day, 5 mM AIBN stress. None of the
degradation products observed during the AIBN ox-
idative stressing of compound 1 were formed from
stressing isolated degradation products.
Theoretical Computations
Bond dissociation energy values for the RO H bonds
of 2-cyano-2-hydroperoxide and 2-cyano-2-propanol
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 5, MAY 2013

1558
WATKINS, PITZENBERGER, AND HARMON
Figure 4. AIBN stress (5 mM) of 0.05 mg/mL compound 1 in ACN–water and MeOH–water
diluents.
Figure 7 shows the effect of ranging the AIBN con-
centration from 1 to 5 mM while holding drug con-
stant at 0.05 mg/mL (solvent 50:50 ACN–water). The
upper portion of Figure 7 shows the percent compound
1 degraded, ranging from near 20% with 5 mM AIBN
to near 4% at approximately 1 mM AIBN. The lower
portion of Figure 7 shows the individual degradant
area percent (relative to the 0.05 mg/mL compound
1 standard peak area) formed during the same ex-
periments. Overall, the responses in Figure 7 appear
linear. In Figure 8, the compound 1 concentration
was ranged from 0.05 to 1.0 mg/mL (Upper, using
50:50 ACN–water solvent system) and from 0.10 to
10 mg/mL using 100% ACN solvent (lower portion of
Fig. 8). After stressing, the samples were diluted to
0.05 mg/mL compound 1 and assayed as previously.
The y-axis shows the summed area percent (relative
to 0.05 mg/mL compound 1 standard peak area) for all
three degradants. The decreasing summed degradant
area percent data in Figure 8 clearly show that there
is a limit to the absolute amount (mg/mL) of com-
pound 1 degradants that can be formed.
Figure 5. AIBN stress (5 mM) of 0.05 mg/mL compound 1 in ACN–water–MeOH diluents.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 5, MAY 2013
DOI 10.1002/jps

DIRECT EVIDENCE OF 2-CYANO-2-PROPOXY RADICAL ACTIVITY
1559
analysis of the AIBN stressed solution using a
quadrupole ion trap mass spectrometer. The LC–MS
data revealed the mass (relative to the mass of Com-
pound 1) of degradant 1 to be +83 Dalton (Da), the
odd number indicating the net gain of an odd num-
ber of nitrogen atoms. Nominal masses indicated
that degradant 2 and 3 were compound 1 plus 30
and 48 Da, respectively, indicating the possibility of
adding two oxygen atoms (followed by elimination of
two hydrogen atoms) and the addition of three oxygen
atoms, respectively.
Degradant 1 was further examined with tandem
mass spectrometry. Nominal mass losses of 27 Da
followed 41 Da provide evidence that degradant 1
contains an added nitrogen. The unusual 27 Da frag-
ment is indicative of the loss of HCN from a nitrile-
containing species. A definitive molecular formula
was obtained for degradant 1 through FT-ICR–MS
accurate mass measurement. The molecular formula
of the API was found to have increased by C4H5NO.
Figure 6. AIBN degradation time course for compound 1.
Each degradant possess a linear abundance–time relation-
ship.
Nuclear Magnetic Resonance
Degradant 1 NMR. The proton NMR spectrum of
Degradant 1 was compared with that of the API.
Significant differences were found only in the res-
onances of the naphthyl system. In particular, the
1-position proton of the naphthyl moiety was missing
and a new singlet, integrating for 6 h, was observed
at 1.79 ppm. The naphthyl moiety carbon and pro-
ton resonances were assigned via analysis of the two-
and three-bond C H couplings (gHMBC experiment).
The chemical shift of the 1-position carbon was found
to be 137.7 ppm, consistent with heteroatom sub-
stitution (likely oxygen). No C–H or H–H couplings
were observed between the naphthyl moiety and the
group attached at the 1-position. Fortunately, NOEs
(ROESY) were observed between the 1.79-ppm pro-
tons and the 8-position naphthyl proton resonance at
8.13 ppm and between the 1.79-ppm protons and the
methoxymethyl group resonance at 3.99 ppm. Several
lines of evidence point to the 1.79-ppm resonance be-
ing due to a geminal pair of methyl groups. First, the
1.79-ppm resonance integrates for six protons. Sec-
ond, an autocorrelation in the gHMBC experiment
(1.79 ppm H to 28.3 ppm C) is consistent with three-
bond C–H coupling from one methyl to the other
methyl. The methyls have degenerate proton and car-
bon chemical shifts. A carbon chemical shift of 28.3
ppm is consistent with a methyl group that has sev-
eral beta substituents. The only other long-range C–H
correlations involving the geminal methyl groups are
between the methyl protons and the carbon reso-
nances at 75.4 and 122.4 ppm. Because a correlation
was not observed to a second downfield carbon, it is
likely that the 122.4-ppm carbon is not olefinic and
the methyl groups are therefore attached to the 75.4-
ppm carbon. The intensity of the gHMBC correlation
Figure 7. Upper, % compound 1 degraded at 0–5 mM
AIBN (compound 1 at 0.05 mg/mL). Lower, area% (rela-
tive to 100% compound 1 peak area) of Degs. 1 (circles),
Deg. 2 (squares), or Deg. 3 (triangles) in same experiment.
Solvent = 50:50, ACN–water.
Structural Elucidation of the Degradants
LC–MS
The nominal molecular weight of the degradation
products was obtained by positive ion ESI/LC–MS
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 5, MAY 2013

1560
WATKINS, PITZENBERGER, AND HARMON
Figure 8. Upper, 5 mM AIBN, compound 1 ranges from 0.05 to 1.0 mg/mL (50:50, ACN–water
solvent). Summed area% (relative to 100% 0.05 mg/mL compound 1) decreases from near 14%
to 5%. Lower, similar plot except ranging 10× higher in compound 1 (10 mg/mL) but in 100%
ACN solvent. Summed degradant area decreases to 0.6% at 10 mg/mL compound 1.
from the methyl protons to the 122.4-ppm carbon
strongly suggests that it is due to a carbon three
bonds away from the methyl protons. Giving addi-
tional consideration to chemical shift, the 122.4-ppm
resonance is likely due to a nitrile group. The sub-
stantial downfield shift of the 75.4-ppm carbon sug-
gests that its fourth attachment is to an oxygen atom.
Thus, the moiety at the 1-position of the naphthyl sys-
tem is the 2-hydroxy-isobutylnitrile group shown in
Figure 9, nicely in accord with the added molecular
formula identified by accurate mass measurement.
system were found shifted substantially downfield,
one to 178.9 ppm and the other to 183.8 ppm. Using
the ring attached to the pyrazole as a starting point,
the proton at 7.94 ppm showed a three-bond gHMBC
correlation to the 183.8-ppm carbon. The proton on
the opposite side of the ring at 8.11 ppm showed a
similar correlation to a carbon at 178.9 ppm. Both of
these downfield carbons exhibited long-range correla-
tions to the olefinic proton at 6.41 ppm. The correla-
tion from 6.41 to 183.8 ppm was found to be rather
weak, as one would expect for a two-bond correlation.
The two downfield carbons must be carbonyls based
on their chemical shift. The only structure that fits
all the NMR data and the nominal mass increase of
30 Da over compound 1 is the para-quinone structure
shown in Figure 10.
Degradant 2 NMR. Degradant 2 was dissolved in d6-
DMSO and its proton NMR spectrum compared well
except for the resonances in the naphthalene region.
A 1,2,4-trisubstitued aryl group was found attached to
the pyrazole by observation of NOEs from 7.35 ppm to
protons at 7.94 and 7.87 ppm. The proton–proton cou-
plings easily established the 1,2,4 relationship of the
substituents. The methoxyether moiety of the parent
compound was found to be intact, but from the ROESY
data, adjacent only to a isolated aromatic proton at
6.41 ppm (singlet). HMQC and gHMBC experiments
were next used to assign the carbon resonances of
the degradant. The carbon framework of the naph-
thalene was found preserved in the degradant struc-
ture, although two carbons of the naphthalene ring
Degradant 3 NMR. d6-DMSO was used as the sol-
vent for NMR analysis of this degradant as it was not
fully soluble in CD3CN. Proton NMR resonances of
the degradant 3 were compared with those of com-
pound 1. Some chemical shift differences were ex-
pected because of the solvent change. Significant dif-
ferences were found in the resonances of the naphthyl
system even after accounting for the change of sol-
vent. The most significant change in the proton spec-
trum was the appearance of a vicinal set of protons
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 5, MAY 2013
DOI 10.1002/jps

DIRECT EVIDENCE OF 2-CYANO-2-PROPOXY RADICAL ACTIVITY
1561
Figure 9. NMR chemical shift assignments for compound 1 and degradant 1, (S)-3-(4-(1-(3-(3,
5-dichlorophenyl)-5-(6-methoxynaphthalen-2-yl)-1H-pyrazol-1-yl)ethyl)benzamido)propanoic
acid and (S) - 3 - (4 - (1 - (5 - (5 - ((2 - cyanopropan - 2 - yl)oxy) - 6 - methoxynaphthalen - 2 - yl) - 3 - (3,
5-dichlorophenyl)-1H-pyrazol-1-yl)ethyl)benzamido)propanoic acid, respectively.
at 8.54 and 6.22 ppm that share a 16.2-Hz coupling.
These resonances are suggestive of a transoid ",$-
unsaturated carbonyl structure. The 0.24-ppm upfield
shift of the methoxynaphthalene protons from 3.93 to
3.69 ppm also signified a major structural change.
HMQC and gHMBC experiments were next used to
assign most of the degradant carbon resonances. Not
all resonances were observed because of the limited
amount of sample. The methoxy group was found
to be part of a methyl ester based on a three-bond
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 5, MAY 2013

1562
WATKINS, PITZENBERGER, AND HARMON
Figure 10. NMR chemical shift assignments for degradants 2 and 3, (S)-3-(4-(1-(3-(3,5-
dichlorophenyl)-5-(6-methoxy-5,8-dioxo-5,8-dihydronaphthalen-2-yl)-1H-pyrazol-1-yl)ethyl)
benzamido)propanoic acid and (S,E)-4-(1-(1-(4-((2-carboxyethyl)carbamoyl)phenyl)ethyl)-3-(3,
5-dichlorophenyl)-1H-pyrazol-5-yl)-2-(3-methoxy-3-oxoprop-1-en-1-yl)benzoic acid, respectively.
coupling from the methyl protons to a carbonyl car-
bon at 166.7 ppm. Also found to be coupled to the
166.7-ppm carbonyl are the 8.54 and 6.22-ppm res-
onances mentioned above. This observation estab-
lished substructure A (Fig. 11). A three-bond coupling
from the 6.22-ppm proton resonance to 131.5 ppm as-
signed the carbon ( to the carbonyl. An additional
long-range correlation from the 8.54-ppm proton to
the protonated carbon at 125.9 ppm further defined
the attachment point. The 135.2-ppm carbon shown in
substructure A was not observed via correlation from
7.53 ppm, indicating that the resonance for the car-
bon at that position may be sufficiently broad to not
show up in the low signal-to-noise gHMBC data set.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 5, MAY 2013
DOI 10.1002/jps

DIRECT EVIDENCE OF 2-CYANO-2-PROPOXY RADICAL ACTIVITY
1563
C10 position of compound 1. Figure 12 depicts our view
that degradant 1, 2, and 3 all in fact derive from the
same initial addition reaction to C10. A detailed mech-
anistic rationalization of each degradant is not central
to the main thrust of the current work. Figure 12 pro-
vides a general rationalization, and highlights that
after 2-cyano-2-propoxy addition at C10, oxygenation
at either C9 (giving peroxy radical intermediate A) or
C7 (peroxy radical intermediate B) is possible. Inter-
mediate A could lead to degradant 1 if elimination (of
water or hydrogen peroxide) across the C9C10 bond
occurs, whereas C9C10 bond scission leads ultimately
to degradant 3. Intermediate B leads to degradant 2
and requires additional oxidation at C10 and C7.
Note ACVA similarly produced degradants 2 and
3, but the ACVA analogue of degradant 1 (contain-
ing an additional CH2COOH group compared with
degradant 1) did not appear to be stable enough to ac-
cumulate under the experimental conditions utilized.
Figure 11. Degradant 3 substructure A.
Extremely weak correlations (possibly noise peaks)
were observed coming from the 8.54-ppm and the
7.25-ppm resonances to the tentatively assigned
135.2-ppm carbon (see Fig. 10). Additional gHMBC
and HMQC correlations were used to establish that
the left side of substructure A is a phenyl moiety that
remains after fragmentation of the parent naphtha-
lene ring structure. Evidence for a carboxyl group be-
ing attached to the tentatively assigned 135.2-ppm
carbon is a strong three-bond correlation from the
proton resonating at 7.56 ppm to a carbonyl carbon at
169.6 ppm. A carboxyl group is also the only moiety
at this position that gives a structure consistent with
the observed nominal mass increase of 48 Da over
compound 1.
Mechanism of Methanol Quenching
Figure 5 and the degradant 1 structure show explic-
itly that low levels of methanol serve to eliminate
the ability of the 2-cyano-2-propoxy radical to dif-
fuse long enough to encounter compound 1. The ef-
fect of only 1% methanol in Figure 5 is remarkable. A
chemical reaction with methanol is implicated rather
than a decrease in yield of 2-cyano-2-propoxy radi-
cal (Fig. 1 followed by Fig. 2). A minor solvent com-
position change in this system is expected to have
a weak impact on 2-cyano-2-propoxy radical yield.27
The only plausible methanol chemical reactivity to ac-
count for this effect would be a hydrogen atom dona-
tion from methanol to the 2-cyano-2-propoxy radical
to yield acetone cyanohydrin as shown in Figure 13.
Note that methanol radical would be formed, which
would be expected to oxygenate rapidly in oxygen-
saturated solution, giving a methanol peroxy radical
(see Fig. 13). In this way, the 2-cyano-2-propoxy radi-
cal activity would be “converted” to methanol peroxy
radical activity. In fact, Nelson et al.15 showed that
low levels of methanol oxidation products formalde-
hyde and formic acid where found in their cumene
experiments, consistent with Figure 13.
The protective effect of a few percent methanol by
volume in Figure 13 is readily rationalized by con-
centration and reactivity arguments. The H atom
abstraction rate of 2-cyano-2 propoxy radical from
methanol can be estimated from analogous data for
known alkoxyl radicals of similar structure and RO–H
BDE values. The tert-butoxy and cumyloxyl radi-
cals provide convenient comparisons. Table 1 shows
the relevant RO–H BDE values are within a few
kcal/mole.26–28 Table 2 shows the H atom abstrac-
tion rates of tert-butoxy and cumyloxy radicals from
methanol are very similar31,32 and a similar rate for
the 2-cyano-2-propoxy radical is expected. Table 2 also
DISCUSSION
Degradant 1: Definitive Proof of 2-Cyano-2-Propoxy
Radical Formation and Activity
The incorporation of the 2-cyano-2-propoxy radical
into the methoxy-napthalene ring moiety of com-
pound 1 (Fig. 9, degradant 1) is the first definitive
proof of the formation and activity of this AIBN
derived alkoxy radical under pharmaceutically rel-
evant oxidative susceptibility screening conditions (if
methanol is not used as a cosolvent). Figures 5 and
6 show that degradant 1 (and the related degradants
2 and 3, below) persist at 1–5 mM AIBN levels and
throughout the entire 0.05–10 mg/mL compound 1
range studied. Recently, the potential presence of
the 2-cyano-2-propoxy radical was implicated15 when
studying the AIBN-induced oxidation of cumene (as
a simple model for a drug molecule) under similar
conditions as described here. However, in that case,
all oxidation products observed derive from an ini-
tial H atom abstraction of the single benzylic C H
bond of cumene. Thus, the oxidation product struc-
tures formed do not unambiguously distinguish be-
tween an initial H atom abstraction by an alkoxy rad-
ical or by a peroxy radical. In the present case, it is
clear that degradant 1 results from an initial addi-
tion reaction of the 2-cyano-2-propoxy radical to the
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 5, MAY 2013

1564
WATKINS, PITZENBERGER, AND HARMON
Figure 12. General mechanistic rationalization of degradants 1–3. All three degradants are
formed by addition of the 2-cyano-2-propoxy radical at C10. Oxygenation at C9 leads to degradant
1 and degradant 3, whereas oxygenation at C7 ultimately leads to degradant 2.
Figure 13. 2-Cyano-2-propoxy reaction with methanol to form acetone cyanohydrin and
methanol radical. Note methanol, ACN, and dilute drug molecules all are potential substrates.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 5, MAY 2013
DOI 10.1002/jps

DIRECT EVIDENCE OF 2-CYANO-2-PROPOXY RADICAL ACTIVITY
1565
Table 2. Tert-butoxy and Cumyloxy Radical (Parenthesized) H
Atom Abstraction Rates from Methanol and Other More Reactive
Organic Substrates
ward solvent in ACN–water systems give it a long
enough diffusional lifetime to encounter and react
with mM levels of dilute, more reactive drug sub-
stances.
Compound KRH (M−1 s−1
)
Compound
2-Propanol
KRH (M−1 s−1
18 × 105
)
Methanol
Toluene
Cumene
2.9 × 105
(3.0 × 105)
2.3 × 105
Linear and Nonlinear Behaviors in Figures 5 and 6
Cyclohexene
57 × 105
(101 × 105)
83 × 105
Although not central to the primary focus of this work,
the data in Figures 5 and 6 have practical and mecha-
nistic interest to practitioners of the oxidative screen-
ing experiment and will be discussed very briefly. The
linear behavior in Figure 7 might not at first seem
expected, given the species reacting is the 2-cyano-
2-propoxy radical which derives from disproportion-
ation (Fig. 2) and might increase with the square of
the initial AIBN concentration used. However, appli-
cation of steady state equilibrium considerations re-
veals that the steady state concentration of the peroxy
radicals in Figure 2 will only increase by the square
root of the factor increase in initial AIBN concentra-
tion being used. Thus, the linear behavior in Figure 7
is expected.
Figure 8 highlights that the absolute amount
(mol/L) of total degradation products being observed
reaches a maximum value as the drug concentration
increases at fixed AIBN levels. The total moles of
degradants being formed at the 10 mg/mL compound
1 data point in Figure 8 can be estimated. The total
moles of the 2-cyano-2-propoxy radical formed can be
estimated from the AIBN Arrhenius parameters in
our experiments as described in detail by Boccardi36
if an efficiency of disproportionation in Figure 2 is
assumed. Using an efficiency value of approximately
50% in Figure 2 (allowing for peroxide formation for
example) then the moles of 2-cyano-2-propoxy radical
formed is only a factor of about 2 greater than the
moles of compound 1 degradants formed. These esti-
mations show that the yield of the 2-cyano-2-propoxy
radical itself is clearly the “limiting reagent” in
Figure 8.
8.7 × 105
Tetrahydrofuran
(22.7 × 105)
(104 × 105)
All tert-butoxy values are from Paul et al.31 Cumyloxy data are from
Baingnee et al. 29 with the exception of the methanol value which derives
from Banks et al. 32
compiles similar data for substrates with activated
C H bonds such as those found in tetrahydrofuran
and the allylic C H bond in cyclohexene.27,31 These
more reactive substrates should be viewed as surro-
gates for dissolved drug molecules in the oxidative
test. These substrates show 20–30-fold higher reac-
tion rates (compared with methanol) with tert-butoxy
and cumyloxyl radicals (Table 2) and by analogy with
the 2-cyano-2-propoxy radical. However, given that
methanol even at 2% by volume corresponds to a con-
centration of approximately 500 mM, the protective
effect of 2%–10% added methanol on dilute submil-
limolar substrates with a wide range of C H bond
reactivities is readily rationalized.
It is interesting to note that ACN solvent on a vol-
ume basis does not appear able to be as effective an
H-atom donor as methanol. This would not appear
generally consistent with the relevant BDE values
shown in Table 1. However, this relatively “inert”
property of ACN has been consistently observed
and reported for tert-butoxy and cumyloxy radical
reactions.29–35 In this context ACN is considered more
comparable to benzene. A rate constant for tert-butoxy
radical reaction with ACN of about 8.3 × 103 M−1 s−1
can be estimated from Paul et al.31 and the B-
scission rate for tert-butoxy radical in ACN reported
by Tsentalovich.33 This is approximately 35 times
slower than the rates shown for methanol in Table 2.
The B-scission product distribution of cumyloxy rad-
ical was recently used15 to similarly estimate an ap-
proximately 35-fold slower H atom abstraction rate
from ACN compared with methanol. This work sug-
gests that 2-cyano-2-propoxy radical also exhibits a
similarly slow rate of reaction with ACN. The strik-
ingly different reaction kinetic profile for H atom
abstraction from methanol and ACN by 2-cyano-2-
propoxy radical, even though they have nearly identi-
cal H atom bond energies, points to factors that influ-
ence the reaction transition state. Transition state
polarity is one such factor that can influence bar-
rier height and thus the reaction kinetics of these
systems.15 This relative decrease in H atom abstrac-
tion reaction rate for 2-cyano-2-propoxy radical to-
Implications for Azonitrile-Based Oxidative
Susceptibility Screening Experiments: Addition of 10%
Methanol to the Solvent System and Higher AIBN–Drug
Molar Ratios
This work confirms the addition of ≥10% by volume
methanol to all azonitrile compound-based oxidative
susceptibility screening test solutions as recom-
mended by Nelson et al.15 to quench any 2-cyano-
2-propoxy radical activity that may be present. In
Figure 5, the true reactivity of compound 1 with per-
oxy radical—the original point of the AIBN or ACVA
oxidative screening test–-is revealed only after the
addition of methanol. Without methanol added, and
without full structural elucidation of all degradants,
the 20% of compound 1 degraded in Figures 2 and 3
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 5, MAY 2013

1566
WATKINS, PITZENBERGER, AND HARMON
would be interpreted as a molecule with signifi-
cant peroxy radical reactivity that might require
careful excipient/formulation choices. However, with
methanol added, Figure 5 shows there is no reactiv-
ity of compound 1 with the 2-cyano-2-peroxy radicals
generated from Figure 1. The methanol results bet-
ter predict the subsequent actual long-term stability
data, which showed that conventional solid dosage
formulations of compound 1 show no detectable oxida-
tive degradation over long term accelerated stability
conditions. Use of methanol will insure the appropri-
ate peroxy radical activity of the subject test, facili-
tate better correlations with real stability data, and
will allow for better delineation between oxidatively
“reactive” and “nonreactive” molecules.
proximately 750 mol % AIBN would correlate to 50%
drug lost for a reactive compound (chain length = 1).
Similarly, a 24 h experiment with three fold higher
AIBN levels (∼2300 mol % AIBN) would also cor-
relate to 50% drug loss for a reactive molecule as
defined above. In our laboratories, we routinely use
≥50% methanol as solvent, drug at 0.1–1 mM, and
AIBN (40◦C) at 1000–5000 mol % over a 1–3-day pe-
riod. Many compounds are found to degrade a few
percent or less under these conditions, and seldom do
we encounter a 50% drug loss over the 3-day test pe-
riod. In our view, these results reflect the well known
selectivity of the peroxy radical reaction.
CONCLUSIONS
This last delineation is worthy of some considera-
tion, and has some bearing on the use of methanol
as described here. What is a reasonable definition of
a “reactive” molecule in this context? Boccardi36 sug-
gested that if a molecule reacts with every AIBN (or
ACVA) derived peroxy radical produced by Figure 1
during the experiment, that would be classed as “re-
active,” and further propagation of the drug-peroxy
radical thus formed (i.e., chain lengths greater than
1) would indicate a “very reactive” drug substance.
However, such a “reactive” molecule would be hard to
observe under the some of the oxidative screening test
conditions being utilized in the industry. Currently,
in the early oxidative screening paradigm, there is a
large range of mol % azonitrile compound used, rang-
ing from approximately 10 to 5000 mol %.8,11,13,15
Drug concentrations are typically in the 0.1–1 mg/mL
range, whereas AIBN or ACVA are used from approx-
imately 0.2–5 mM. Many practitioners use 20 mol %
or less, and 5–20 mol % values have been recently
recommended.37 However, using even 20 mol % AIBN
in a 40◦C experiment for 3 days would lead to degra-
dation of only 1.5% of the drug substance present
for a compound thus classified as “reactive.”36 Very
reactive compounds would degrade more than 1.5%,
and “nonreactive” compounds would not degrade any
drug. This is a relatively poor “dynamic range” for
oxidative classification in our view. This type of prac-
tical consideration is what lead Boccardi12 to recom-
mend approximately 100 mol % azonitrile levels. The
only “danger” from using higher and higher azonitrile
compound levels, in our view, is possible creation of
the strong alkoxy radical oxidants from the dispro-
portionation in Figure 2. However, as demonstrated
here, that danger is completely obviated with the use
of ≥10% methanol. Compound 1 is clearly very reac-
tive with the 2-cyano-2-propoxy radical. Yet Figure 5
shows complete elimination of the 2-cyano-2-propoxy
derived degradants, even at the 5000 mol % AIBN
(or ACVA) used. Addition of methanol thus allows for
larger, more convenient dynamic ranges. For exam-
ple, for a 3-day AIBN experiment at 40◦C, use of ap-
Oxidative susceptibility screening of compound 1 us-
ing AIBN or ACVA in an ACN–water solvent has
been shown to generate alkoxy radical activity. Struc-
ture elucidation of degradant 1 (Fig. 9) proves, un-
equivocally, the presence and activity of the 2-cyano-
2-propoxy radical as the oxidizing species, which is
formed from the disproportionation of the 2-cyano-
2-peroxy radical (Fig. 2). The absolute amount of
degradants (1–3) formed is limited by the amount of
the 2-cyano-2-peroxy radical, which can form given
Figure 2. The generation of the 2-cyano-2-propoxy
radicals is a general problem for the subject oxida-
tive test, as this alkoxy radical is more reactive, and
less selective, than the desired peroxy radical. This
situation has the potential to yield misleading in-
formation by giving the appearance that the drug
substance is susceptible to chain oxidation processes
when in fact that may not be the case. Small amounts
of methanol were found to quench the undesired 2-
cyano-2-propoxy radical activity in the present study
(Fig. 5), similar to that reported previously.15 The pro-
tective methanol effects can be rationalized by large
relative concentrations and competitive H atom dona-
tion rates relative to dilute drug substance substrates.
It is strongly recommended that all azonitrile-based
oxidative stress testing experiments maintain at least
10% methanol (by volume) in the reaction cosolvent.
The addition of methanol should lead to a reexam-
ination of commonly used azonitrile mol % levels in
the early oxidative screening paradigm. Higher azoni-
trile mol % values of 100%–1000% relative to the drug
substance being screened are readily warranted.
ACKNOWLEDGMENTS
The authors would like to thank Mr. Vincent Van
Nostrand for performing the FT-ICR–MS experi-
ments to aid in structure elucidation of degradation
products. The authors also wish to thank Dr. Steven
Baertschi of Eli Lilly and Company for his useful
discussion and encouragement.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 5, MAY 2013
DOI 10.1002/jps

DIRECT EVIDENCE OF 2-CYANO-2-PROPOXY RADICAL ACTIVITY
1567
18. Foti MC. 2007. Antioxidant properties of phenols. J Pharm
Pharmacol 59:1673–1685.
REFERENCES
19. Hovorka SW, Schoneich C. 2001. Oxidative degradation
of pharmaceuticals: Theory, mechanisms and inhibition. J
Pharm Sci 90:253–269.
1. Baertschi SW. 2006. Analytical methodologies for discover-
ing and profiling degradation-related impurities. Trends Anal
Chem 25:8:758–767.
20. Alsante KM, Baertschi SW. 2003. A stress testing benchmark-
ing study. Pharm Tech Feb 60–72.
2. Raillard SP, Bercu J, Baertschi SW, Riley CM. 2010. Pre-
diction of drug degradation pathways leading to structural
alerts for potential genotoxic impurities. Org Proc Res Dev
14:1015–1020.
3. Alsante KM, Ando A, Brown R, Ensing J, Hatajik TD, Kong W,
Tsuda Y. 2007. The role of degradant profiling in active phar-
maceutical ingredients and drug products. Adv Drug Deliv Rev
59:29–37.
4. Reynolds DW, Facchine KL, Mullaney JF, Alsante KM,
Hatajik TD, Motto MG. 2002. Available guidance and best
practices for conducting forced degradation studies. Pharma
Technol 26(2):48–54.
5. Singh S, Bakshi M. Guidance on conduct of stress tests to de-
termine inherent stability of drugs. Pharma Technol. Online,
April 2000 page 1-14.
6. Baertschi SW, Ed. 2005. Drugs and the pharmaceutical
sciences. In Pharmaceutical stress testing: Predicting drug
degradation, Vol. 153. Boca Raton, Florida: Taylor & Francis.
7. Baertschi SW, Alsante KM, Reed RA, Eds. 2011. Drugs and
the pharmaceutical sciences. Pharmaceutical stress testing:
Predicting drug degradation, vol. 210. 2nd ed. New York: In-
forma Healthcare.
8. Boccardi G, Deleuze C, Gachon M, Palmisano G, Vergnaud
JP. 1992. Autoxidation of tetrazepam in tablets: Prediction of
degradation impurities from the oxidative behavior in solu-
tion. J Pharm Sci 81:183–185.
9. Boccardi G. 1994. Autoxidation of drugs: Prediction of degra-
dation impurities from results of reaction with radical chain
initiators. Il Farmaco 49:431–435.
10. Waterman KC, Adami RC, Alsante KM, Hong J, Landis
MS, Lombardo F, Roberts CJ. 2002. Stabilization of pharma-
ceuticals to oxidative degradation. Pharm Dev Technol 7:1–
32.
21. Ingold KU. 1969. Peroxy radicals. Acc Chem Res 2:1–9.
22. Hendry DG, Mill T, Piszkiewicz L, Howard JA, Eigenmann
HK. 1974. A critical review of H atom transfer in the liq-
uid phase: Chlorine atom, alkyl, trichloromethyl, alkoxy, and
alkylperoxy radicals. J Phys Chem Ref Data 3:937–978.
23. Werber J, Wang YJ, Milligan M, Li X, Ji JA. 2011. Analy-
sis of 2,2’-azobis (2-amidinopropane) dihydrochloride degra-
dation and hydrolysis in aqueous solutions. J Pharm Sci
100:3307–3315.
24. Watkins MA, Templeton AC, Harmon PA. 2006. Structure
and mechanism elucidation of non-characteristic forced degra-
dation reaction products from an azonitrile radical initiator.
Oral Presentation at the Institute for International Research
3rd Annual Conference on Forced Degradation Studies, Short
Hills, New Jersey, February 28.
25. Harmon PA, unpublished data.
26. Gaussian 09, Revision B.01,Frisch MJ., Trucks GW, Schlegel
HB, Scuseria GE, Robb MA, Cheeseman JR, Scalmani G,
Barone V, Mennucci B, Petersson GA, Nakatsuji H, Caricato
M, Li X, Hratchian HP, Izmaylov AF, Bloino J, Zheng G, Son-
nenberg JL, Hada M, Ehara M, Toyota K, Fukuda R, Hasegawa
J, Ishida M, Nakajima T, Honda Y, Kitao O, Nakai H, Vreven
T, Jr Montgomery JA, Peralta JE, Ogliaro F, Bearpark M,
Heyd JJ, BrBKs E, Kudin KN, Staroverov VN, Kobayashi R,
Normand J, Raghavachari K, Rendell A, Burant JC, Iyengar
SS, Tomasi J, Cossi M, Rega N, Millam NJ, Klene M, Knox
JE, Cross JB, Bakken V, Adamo C, Jaramillo J, Gomperts
R, Stratmann RE, Yazyev O, Austin AJ, Cammi R, Pomelli
C, Ochterski JW, Martin RL, Morokuma K, Zakrzewski VG,
Voth GA, Salvador P, Dannenberg JJ, Dapprich S, Daniels AD,
¨
Farkas O, Foresman JB, Ortiz JV, Cioslowski J, Fox DJ. 2009.
Wallingford, Connecticut: Gaussian, Inc.
11. Lombardo F, Campos G. 2004. How do we study oxidative
chemical stability in discovery? Some ideas, trials, and out-
comes. In Pharmaceutical profiling in drug discovery for lead
selection; Kerns EH, Lipinski CA, Thakker DR, Wang B,
Eds. Virginia: AAPS Press, pp 183–194.
12. Boccardi G. 2005. Oxidative susceptibility testing. In Drugs
and the pharmaceutical sciences: Pharmaceutical stress test-
ing: Predicting drug degradation; Baertschi SW, Ed, Vol. 153.
Boca Raton, Florida: Taylor & Francis, pp 205–234.
13. Nelson ED, Harmon PA, Szymanik RC, Teresk MG, Li L,
Seburg RA, Reed RA. 2006. Evaluation of solution oxygena-
tion requirements for azonitrile based oxidative forced degra-
dation studies of pharmaceutical compounds. J Pharm Sci
95:1527–1539.
27. Hammond GS, Sen JN, Bozer CE. 1955. The efficiency of rad-
ical production from azobis-isobutyronitrile. J Am Chem Soc
77:3244–3248.
28. McMillen DF, Golden DM. 1982. Hydrocarbon bond dissocia-
tion energies. Ann Rev Phys Chem 33:493.
29. Baignee A, Howard JA, Scaiano JC, Stewart LC. 1983.
Absolute rate constants for reactions of cumyloxy in solution.
J Am Chem Soc 105:6120–6123.
30. Blanksby SJ, Ellison GB. 2003. Bond dissociation energies of
organic molecules. Acc Chem Res 36:255–263.
31. Paul H, Small RD, Scaiano JC. 1978. Hydrogen abstraction
by tert-butoxy radicals. A laser photolysis and electron spin
resonance study. J Am Chem Soc 100:4520–4526.
32. Banks JT, Scaiano JC. 1993. The laser drop method: A new
approach to induce multiple photon chemistry with pulsed
lasers. Examples involving reactions of diphenylmethyl and
cumyloxyl radicals. J Am Chem Soc 115:6409–6413.
33. Tsentalovich YP, Kulik LV, Gritsan NP, Yurkovskaya AV.
1998. Solvent effect on the rate of $-scission of the tert-butoxyl
radical. J Phys Chem A 102:7975–7980.
34. Avilla DV, Brown CE, Ingold KU, Lusztyk J. 1993. Solvent
effects on the competitive $-scission and hydrogen atom ab-
straction reactions of the cumyloxyl radical. Resolution to a
long standing problem. J Am Chem Soc 115:466–470.
35. Karki SB, Treemaneekarn V, Kaufman MJ. 2000. Oxidation
of HMG-CoA reductase inhibitors by tert-butoxyl and 1,1 di-
phenyl-2-picrylhydrazyl radicals: Model reactions for predict-
ing oxidatively sensitive compounds during preformulation.
J Pharm Sci 89:1518–1524.
14. Harmon PA, Kosuda K, Nelson ED, Mowery M, Reed RA.
2006. A novel peroxy radical based oxidative stressing sys-
tem for ranking the oxidizability of drug substances. J Pharm
Sci 95:2014–2028.
15. Nelson ED, Thompson GM, Yao Y, Flangan HM, Harmon
PA. 2009. Solvent effects on the AIBN forced degradation
of cumene: Implications for forced degradation practices. J
Pharm Sci 98:959–969.
16. Harmon PA, Boccardi G. 2011. Oxidative susceptibility test-
ing. In Drugs and the pharmaceutical sciences: Pharmaceu-
tical stress testing: Predicting drug degradation; Baertschi
SW, Alsante KM, Reed RA, Eds. Vol. 210, 2nd ed. New York:
Informa Healthcare, pp 168–191.
17. Johnson DM, Gu LC. 1988. Autoxidation and antioxidants.
In Encyclopedia of pharmaceutical technology; Swarbuck J,
Boylan JC, Eds. Vol. 1. New York: Wiley, pp 415–450.
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 5, MAY 2013

1568
WATKINS, PITZENBERGER, AND HARMON
36. Boccardi G. 2005. Oxidative susceptibility testing. In
Drugs and the pharmaceutical sciences: Pharmaceutical
stress testing: Predicting drug degradation; Baertschi SW,
37. Baertschi SW, Jansen PJ, Alsante KM. 2011. Stress testing:
A predictive tool. In Drugs and the pharmaceutical sciences:
Pharmaceutical stress testing: Predicting drug degradation;
Baertschi SW, Alsante KM, Reed RA, Eds. Vol. 210, 2nd ed.
New York: Informa Healthcare, pp 25.
Ed. Vol. 153. Boca Raton, Florida: Taylor
pp 216.
& Francis,
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 102, NO. 5, MAY 2013
DOI 10.1002/jps