IL FARMACO, 49 (6), 431-435 (1994)
`
`This material may be protected by Copyright low (Title 17 U.S. Code)
`
`AUTOXIDATION OF DRUGS: PREDICTION OF DEGRADATION IMPURITIES
`FROM RESULTS OF REACTION WITH RADICAL CHAIN INITIATORS (*)
`
`GIOVANNI BOCCARDI
`Sanofi Recherche, Centro Ricerche Sanofi-Midy S.p.A,
`via Piranesi 38, 20137 Milan, Italy.
`
`Summary— In the study of the degradation of drug substances by molecular oxygen, their specific reaction
`mechanisms must be taken into account. The rate-determining step is usually the reaction of the substrate with
`a radical chain initiator, which is often an unknown impurity. The reactivity and selectivity of autoxidation
`can be controlled better by using a radical chain initiator, such as AIBN, than by changing the temperature
`or the oxygen pressure. In this paper the products profiles of four pharmaceutical substances in a simple oxidation
`test with AIBN are compared with the results of long term natural stability tests or with already established
`stabilities.
`
`Stress testing is the basis of all studies of the stability
`of a new drug substance. The first aim of this kind of
`investigation is to discover the chemical and physical
`factors that can affect the stability of the molecule
`adversely, in order to design stable formulations. The
`second aim is to obtain samples of the drug
`contaminated with all possible and significant
`degradation impurities, in order to validate the
`analytical methods for the long term stability studies
`and to isolate the main impurities. While standard
`experimental conditions for the study of accelerated
`and long term stability are defined in all the regulatory
`guidelines, the protocol for the reactivity study must
`be fit to the particular chemistry of the molecule being
`examined. For investigation of the hydrolytic pathway
`of degradation, the general protocol is to study the
`effect of acids or bases at elevated temperatures on the
`stability of aqueous solutions of the drug substance,
`because it is well known that hydrolysis reactions are
`catalyzed by acids and bases. Oxidation is a more
`complex reaction, and the pharmaceutical literature
`describes stress testing with various oxidizing agents,
`such as hydrogen peroxide, heavy metal ions, acids,
`bases, high oxygen pressure, high temperature and, in
`some instances, strong oxidants such as potassium
`permanganate and chromic anhydride. Very often this
`literature emphasizes the poor predictiveness of this
`kind of stress testing. One reason for this poor
`predictiveness is that the operating mechanisms of the
`oxidation with the above reagents are completely
`different from the radical chain mechanism of
`autoxidation. Long term, room temperature
`degradation of an organic chemical is better simulated
`by using a radical chain initiator to accelerate the rate-
`controlling step of autoxidation. Use of this approach
`in the reactivity study has been described in the recent
`
`(*) Presented at the V Convegno su recenti Sviluppi ed Applicazioni
`nell'Analisi Farmaceutica, Alghero, October 13-16, 1993.
`
`pharmaceutical literature". In this paper the
`experimental conditions for use of some radical chain
`initiators and the predictivity of this kind of reactivity
`test for four examples will be discussed.
`In the electronic structure of molecular oxygen",
`the highest occupied molecular orbitals are two degener
`7r* orbitals in which there must be two electrons. The
`ground state, according to the Hund rule, is the state
`in which each of these two orbitals is occupied by one
`electron, and the spins are parallel: this is the triplet
`ground state (3Eg) of the atmospheric molecular
`oxygen. Triplet dioxygen can be excited, both
`chemically and photochemically, to the first excited
`state with spin multiplicity 0, the singlet state 'Ag, 22
`kcal higher than the ground states. The triplet ground
`state is the state of dioxygen involved in autoxidation.
`The reactivity of triplet dioxygen toward organic
`molecules can be summarized as follows.
`Electron-rich molecules such as pyrroles6, «,(3-
`unsaturated enamines9, carbanions', 9,10-
`cyclopentane-4a,4b-dihydrophenanthrenes, strained
`cycloalkenes9 and, under more drastic conditions,
`tertiary amines, sulfoxides, alkenes and alkynes" can
`react with oxygen in an electron-transfer reaction:
`(eq. 1)
`R- + (3Eg) 02—*R• + 02- • (cid:9)
`In addition, triplet oxygen reacts very fast with organic
`radicals:
`(eq. 2)
`R. + (3Eg) 02—,R00. (cid:9)
`and this reaction is very important in propagation of
`radical chains.
`However the vast majority of organic molecules are
`in the singlet state, and their reaction with triplet
`dioxygen:
`RH + (3Eg) 02—,ROOH (cid:9)
`(eq. 3)
`is spin forbidden. For this reason, a great many organic
`molecules, in spite of the large negative value of the
`Gibbs free energy of oxidation, are kinetically inert
`
`NOVARTIS EXHIBIT 2050
`Noven v. Novartis and LTS Lohmann
`IPR2014-00550
`Page 1 of 5
`
`

`
`432 (cid:9)
`
`Giovanni Boccardi
`
`SCHEME I
`
`toward triplet oxygen. In the latter case, that of a singlet
`organic molecule, inert toward triplet dioxygen, the well
`established mechanism of autoxidation at "normal"
`room temperature is depicted in equations 4-7:
`In- + RH —)InH + R. (cid:9)
`Initiation (cid:9)
`(eq. 4)
`R. + 02 —)R00. (eq. 5)
`ROO• + RH —)ROOH + R• Propagation (eq. 6)
`2 ROO. —)Inert products (cid:9)
`Termination (cid:9)
`(eq. 7)
`
`The rate-determining step of the radical chain is the
`initiation reaction (eq. 4), whereas the propagation step
`(eq. 5) is very fast. This is why the oxidation rate does
`not depend on the oxygen partial pressure". In this
`case it is absolutely useless to increase oxygen partial
`pressure to accelerate the oxidation: only the initiator
`In•, often a trace contaminant or impurity, controls
`the overall oxidation rate. To obtain reproducible
`oxidation, the concentration of the initiator must be
`controlled/2, for instance by adding known amounts
`of chemically defined radical chain initiators.
`We could try to accelerate autoxidations just by
`increasing the temperature, exploiting the Arrhenius
`law. But in the complex mechanism depicted above,
`each step has its own activation energy, and the
`Arrhenius law can break down because the rate-
`determining step, and thus the mechanism, changes as
`temperature increases. Thermally labile compounds,
`such as hydroperoxides, do not only decompose at high
`temperature, but become efficient catalysts of the
`oxidation, giving kinetic profiles and chemical
`selectivity very different from those at room
`temperature".
`With combined high temperature and high oxygen
`pressure, the reaction path and the reaction products
`may change dramatically, because new mechanisms
`become effective. Amines and thioethers at ambient
`oxygen pressure and temperature undergo oxidation
`only in the a position to the heteroatom, often leading
`to a complex mixture of products. This a-oxidation fits
`with the mechanism of eqs. 4-7, corresponding to a
`reaction catalyzed by a foreign initiator. Under high
`oxygen pressure (70 bar) and at elevated temperature
`(100 °C), a specific oxidation of the heteroatom was
`obtained in high yields". The mechanism of this
`oxidation involves the formation of an oxygen complex
`with the organic molecule, hence the need for high
`oxygen pressure, and a radical chain oxidation initiated
`by an electron transfer reaction.
`
`RESULTS AND DISCUSSION
`
`TETRAZEPAM
`Tetrazepam 1 (Scheme 1) is a benzodiazepine used
`in therapy as myorelaxant. 1 is quite stable in solid
`pharmaceutical forms, but is oxygen-sensitive in
`solution. The main pattern of degradation of 1 (3) is
`the oxidation of the 3' carbon atom to give the 3'-
`
`hydroperoxide 2 and the 3'-keto-derivative 3. Minor
`degradation impurities are the epoxide 5 and the
`product 4. We carried out an oxidation test with AIBN
`as a first approach to the profile of autoxidation
`impurities. An acetonitrile solution of the substance
`and of the radical chain initiator AIBN (2,2'-azobis[2-
`methylpropanenitrile]) was stored 48 h in the dark at
`40 °C. The profile of degradation was very similar to
`that of a tablet sample in an accelerated stability study,
`i.e., after storage for 6 months at 55 °C and 8507o
`relative humidity. Other stress tests (oxidation with
`Cu+ + and Fe+ + + , thermal stress of the bulk
`substance) did not produce the important degradation
`impurity 2. Reaction with hydrogen peroxide, widely
`used in preformulation studies, yielded only the minor
`impurity 5. Our conclusion was that, since AIBN
`oxidation occurs by the same radical chain mechanism
`as natural autoxidation, it is the best test to use to
`predict the oxidative behaviour'.
`The choice of temperature and solvent is very
`important in oxidative reactions. We never carry out
`the AIBN test on drug substances at a temperature over
`40 °C, to prevent homolytic decomposition of species
`which are stable at ambient temperature, i.e., the
`hydroperoxide 2. Table I shows the solvent effects on
`the AIBN test. Acetonitrile is our preferred solvent
`because of its inertness to oxidants and its neutrality.
`Indeed, degradation of 1 is maximal in this solvent.
`Furthermore, replications of the test in acetonitrile gave
`
`NOVARTIS EXHIBIT 2050
`Noven v. Novartis and LTS Lohmann
`IPR2014-00550
`Page 2 of 5
`
`

`
`Atttoxidation of drugs (cid:9)
`
`433
`
`TABLE 1 - SOLVENT EFFECTS ON PRODUCT DISTRIBUTION FOR TETRAZEPAM
`(35 uM) AFTER DEGRADATION CATALYZED BY RADICAL CHAIN INITIATORS.
`(PERCENT COMPOSITION OF THE REACTION MIXTURE AFTER 48 h at 40°C)
`
`Solvent
`
`Catalyst
`(a)
`
`Composition (07o)
`1
`2 3
`4 (cid:9)
`5
`
`AIBN 65.8 8.4 11.2 1.9 (cid:9) 2.8
`Acetonitrile (b)
`Acetonitrile+water (20'7o) (b)
`AIBN 68.1 9.6 11.2 1.9 (cid:9) 2.8
`AIBN 75.2 8.3 6.6 2.4 (cid:9) 1.8
`Methanol
`AIBN 84.0 6.3 1.0 5.2 trace
`Ethanol
`AIBN 89.6 4.3 0.3 4.8 trace
`2-Propanol
`Acetonitrile+aq. buffer (C) 1:1 ACVA 89.4 3.3 5.0 1.3 (cid:9) 1.0
`
`(°) 60 mM. ACVA.---4,4'-azobis[4-cyanovaleric acid].
`(b) From ref. 3.
`(") Phosphate buffer 0.022 M pH 7.0
`
`very reproducible results. The presence of water up to
`20 % (V/V) has no important effect on the rate of
`degradation nor the distribution of the degradation
`product. Alcohols slow the reaction: in methanol the
`reaction is slightly slower than in acetonitrile, and in
`ethanol and isopropanol the oxidation is still slower.
`Alcohols probably inhibit oxidation by competing with
`the test substance for the initiator radicals. Indeed, we
`found acetone, the product of the solvent oxidation
`(0.2%) in an isopropanol solution of AIBN stored 48
`h at 40 °C. Table I shows another important feature
`of the solvent effect in oxidation. The ratio between
`keto-impurity 3 and hydroperoxide impurity 2 decreases
`in parallel with the total degradation of 1. Under neutral
`conditions and at moderate temperature, the
`hydroperoxide 2 is stable in solution. This means that
`the origin of the keto-impurity 3 is a reaction other than
`from decomposition of the hydroperoxide 2, probably
`the termination reaction between two hydroperoxide
`radicals. According to this hypothesis, the most
`inhibitory solvent, isopropanol, is also the most
`powerful hydrogen-radical donor, which is able to
`efficiently quench the hydroperoxide radicals,
`decreasing formation of the keto- impurity.
`AIBN is not soluble in water, and this may prevent
`using this reagent with very polar organic substances.
`Oxidation of tetrazepam with 4,4'azobis [4-cyanovaleric
`acid] at pH 7.0 in a 1:1 mixture of acetonitrile and water
`(Table I) gave qualitatively similar results. There was,
`however, a small difference in reactivity, seen as the
`formation of the minor impurity 3'-hydroxytetrazepam,
`an impurity never observed in the standard AIBN test.
`
`PHENYLBUTAZONE
`To investigate the predictivity of the products of the
`oxidation with radical chain initiators, we carried out
`the test on drug substances with known degradation
`profiles and of different chemical classes, comparing
`the results with data in the literature.
`Phenylbutazone 6 (Fig. 1) is an oxygen-sensitive
`substance and oxydation yields 4-hydroxy-
`phenylbutazone 7'4. The AIBN test on 6 yielded 7 as
`
`the main product (6%) after 48 h at 40 °C and an
`unknown peak with an HPLC relative retention time
`of 0.78 (Fig. 2). After treatment with methionine (used
`
`HO
`
`N-N
`
`10
`
`'N-CH3
`HBr, H2O
`
`PHENYLBUTAZONE (6)
`
`DEXTROMETHORPHAN
`HYDROBROMIDE (8)
`
`TRIFLUOPERAZINE DIHYDROCHLORIDE (10)
`
`Fig. 1 - Structures of phenylbutazone, dextromethorphan
`hydrobromide and trifluoperazine dihydrochloride.
`
`as peroxide scavenger), the unknown peak disappeared,
`with an increase in the peak corresponding to 7. The
`unknown peak probably corresponds to 4-hydroperoxy-
`phenylbutazone, which is the presumed but never
`isolated intermediate of 6 autoxidation. This example
`shows not only the predictivity of the results of reaction
`with the radical chain initiator, but also the possibility,
`because of its mild conditions, to accumulate labile
`intermediates.
`
`DEXTROMETHORPHAN HYDROBROMIDE
`Dextromethorphan hydrobromide 8 (Fig. 1) is a very
`stable drug substance. By photochemical oxidation, 8
`yields 10-ketodextromethorphan 9'5,16. The same
`impurity was found in trace amounts during
`preformulation of an antitussive syrup containing 8'7.
`In the AIBN test, after 72 h at 40 °C, 8 underwent a
`slight degradation to give the impurity 9 (1 o7o). In the
`case of dextromethorphan, the low reactivity in the
`AIBN test reflects the good stability of the substance.
`
`TRIFLUOPERAZINE HYDROCHLORIDE
`Trifluoperazine dihydrochloride 10 (Fig. 1) is known
`to give the sulfoxide 11 by thermal and photochemical
`degradation18. Phenothiazines, and in particular 10,
`are also known to directly react with oxygen by an
`electron-transfer mechanism. The AIBN test yielded
`
`NOVARTIS EXHIBIT 2050
`Noven v. Novartis and LTS Lohmann
`IPR2014-00550
`Page 3 of 5
`
`

`
`434
`
`Giovanni Boccardi
`
`LC A 235,4 450,80
`
`of A1019C.D
`
`PHENYLBUTRZONE
`
`1
`
`12
`
`O
`
`I
`O
`O
`
`8 (cid:9)
`G
`Ti me ( m n . )
`
`10 (cid:9)
`
`20-
`
`E
`
`10-
`
`0-
`
`Fig. 2 - The HPLC chromatograms of phenylbutazone 6 in acetonitrile after 48-h degradation at 40°C in the presence of AIBN.
`
`11 (67 %) as the only degradation product after 48 h
`at 40 °C. This result deserves some discussion. We
`expected 10 not to react in the AIBN test, or to give
`a different product, because AIBN acts as an hydrogen
`radical acceptor and the formation of the sulfoxide 11
`must follow a different mechanism. Probably a
`peroxide, namely 2-hydroperoxy-2-methyl-
`propanenitrile and hydrogen peroxide, are products or
`intermediates of AIBN decompostion" and these
`oxidants can be the active species, via an ionic
`mechanism, in our AIBN test on 10. Indeed, 10 reacts
`fast with hydrogen peroxide to give 11. The ability of
`the AIBN test to predict the formation of the "good"
`degradation impurity is probably is due to a secondary
`reaction and not to a radical hydrogen extraction.
`
`CONCLUSIONS
`
`Oxidation by atmospheric dioxygen is a complex
`reaction, and its actual mechanism must be taken into
`account when carrying out stress tests to obtain the
`degradation impurities that might arise during long term
`storage of a drug and its pharmaceutical forms. Using
`a radical chain initiator such as AIBN or its soluble
`analogues is often a valid way to accelerate
`autoxidations.
`
`EXPERIMENTAL PART
`
`MATERIALS AND REAGENTS
`
`Tetrazepam (Sanofi Chimie), phenylbutazone (CFM, Milan,
`
`Italy), dextromethorphan hydrobromide (Roche) and trifluoperazine
`dihydrochloride (ICM, Rozzano, Italy) were pharmaceutical grade
`compounds. 7", 9"and 11" were prepared according to the
`references cited. AIBN (2,2'-azobis[2-methylpropanenitrile]) was
`obtained from Merck and ACVA (4,4'-azobis[4-cyanovaleric acid])
`from Aldrich. All other chemicals were reagent or HPLC grade.
`
`INSTRUMENTS
`The HPLC instrument consisted of a Perkin-Elmer series 3 pump,
`a Rheodyne 7161 injector with a 20-pl loop, a Hewlett Packard 1040
`A diode array UV detector and a Hewlett Packard 79994A data
`station. The identity of each impurity in the chromatograms of the
`test solutions was confirmed by comparison of the retention time
`and of the UV spectrum at the peak maximum with the
`corresponding data obtained with an authentic sample of the
`impurity.
`
`AIBN TEST
`The substance under examination was dissolved in acetonitrile
`or other solvent to give a solution about 10' M. A quantity of
`AIBN 1:1 w/w or mol/mol, as indicated in the individual
`experiments, was dissolved in the same solution. The solution was
`stored in a 25-m1 pyrex vial fit with a screw cap, and the vial was
`stored in the dark at 40 ± 0.5 °C. After the appropriate time, each
`solution was diluted with HPLC mobile phase to a suitable
`concentration and injected directly.
`
`TETRAZEPAM AIBN TEST
`Tetrazepam (35 mM) and AIBN (60 mM) were dissolved in
`acetonitrile. HPLC analysis: column RoSil C18 HL (4.6 mm x 25
`cm) 5 pm (R.S.L.); analytical wavelength 254 nm; mobile phase a
`mixture of 50 vol. of a 0.01 M aqueous potassium dihydrogen
`phosphate (pH 4.65) and 50 vol. acetonitrile; flow rate 1.1 ml.
`
`PHENYLBUTAZONE AIBN TEST
`Phenylbutazone (32 mM) and AIBN (60 mM) were dissolved in
`acetonitrile. HPLC analysis: column ABondapack C18 (Waters) 30
`
`NOVARTIS EXHIBIT 2050
`Noven v. Novartis and LTS Lohmann
`IPR2014-00550
`Page 4 of 5
`
`(cid:9)
`

`
`Autoxidation of drugs (cid:9)
`
`435
`
`cm x 3.9 mm; analytical wavelength 235 nm; mobile phase a mixture
`of 50 vol. of 0.1M phosphate buffer (pH 3.7) and 50 vol. acetonitrile;
`flow rate 1.2 ml/min.
`
`DEXTRONIETHORPHAN HYDROBROMIDE AIBN TEST
`Dextromethorphan hydrobromide (14 mM) and AIBN (30 mM)
`were dissolved in the desired solvent (see table I). HPLC analysis:
`column /iBondapack C18 (Waters) 30 cm x 3.9 mm; analytical
`wavelength 280 nm. Mobile phase: dioctyl sulfosuccinate (2.9 g )
`in 680 ml methanol and 290 ml water, with phosphoric acid (1 ml)
`added and pH adjusted to 3.8 with diluted ammonia; flow rate 1.3
`ml/min.
`
`TRIFLUOPERAZINE DIHYDROCHLORIDE AIBN TEST
`Trifluoperazine dihydrochloride (21 mM) and AIBN (21 mM)
`were dissolved in acetonitrile. HPLC analysis: column Bondapack
`C18 (Waters) 30 cm x 3.9 mm; analytical wavelength 278 nm; mobile
`phase a mixture of 20 vol. of 0.006 M aqueous sodium
`hexanesulfonate (p11 3.8) and 80 vol. methanol.
`
`REFERENCES
`
`(1) G. BOCCARDI, G. PALMISANO and G. RIVA, 4° Conregno
`Nazionale di Chinzica Farmaceutica, Milan, 1988, poster.
`(2) A.R. OYLER, R.E. NALDI, K.L. FACCHINE, D.J. BURINSKY,
`M.H. COZINE, R. DUNPHY, J.D. ALVES-SANTANA, M.L.
`COTTER, Tetrahedron, 47, 6549 (1991).
`(3) G. BOCCARDI, C. DELEUZE, M. GACHON, G. PALMISANO and
`J.P. VERGNAUD, J. Pharr: Sci., 81, 183 (1992).
`(4) G.B. SMITH, L. DIMICHELE, L.F. COLWELL, G.C. DEZENY,
`
`A.W. DOUGLAS, R.A. REAMER, T.R. VERIIOEVEN, Tetrahedron,
`49, 4447 (1993).
`(5) H. KNOPF, E. MUELLER, A. WEICKMANN, in: Houben-Weyl
`Methoden der Organischen Chemie, vol. 4, part la, G.Thieme,
`Stuttgart, 1981, p.69.
`(6) B.D. BEAVER, J. V. COONEY, J.M. Jr WATKINS J.M.,
`Heterocycles, 23, 2847 (1985).
`(7) S.K. MALHOTRA, J.J. HOSTYNEK and A.F. LUNDIN, J. An:.
`Chem. Soc., 90, 6565 (1968).
`(8) A. BROMBERG, K.A. MUSZKAT, J. An:. Chem. Soc., 91, 2860
`(1969).
`P.D. BARTLETT, R. BANAVALI, J.Org. Chem., 56, 6043 (1991).
`P. CORREA, G. HARDY, D.P. RILEY, J.Org. Chem., 53, 1695
`(1988).
`R.A. SiiEwoN, J.K. Komi', Metal-catalyzed Oxidations of
`Organic Compounds, Academic Press, New York, 1981, p. 18.
`G.W. BURTON and K.U. INGOLD, J. Am. Chem. Soc., 103, 6472
`(1981).
`K.A. CONN'ORS, G.L. AMIDON and V.J. STELLA, Chemical
`Stability of Pharmaceuticals, 2nd ed., Wiley, New York, 1986,
`p. 92.
`D.V.C. AWANG, A. VINCENT, F. MATSUI, J. Pharr: Sci., 62,
`1673 (1973).
`Q. HAEFIGER, A. BROSSI, L.H. CHOPARD-DIT-JEAN, Q. WALTER,
`Q. SCHNIDER, Hely. Chin: Acta, 39, 2953 (1956).
`G. BOCCARDI, P. IVIEZZANZANICA, U. Guzzl, G. LESMA, G.
`PALMISANO, Chem. Pharnz Bull., 37, 308 (1989).
`G. BOCCARDI, unpublished results.
`Analytical Profile of Drug substances, Florey K., Ed., Academic
`Press, San Diego, CA, vol. 9, 1980, p. 543.
`A. GOOSEN, C.W. MCCLELAND, D.H. MORGAN, J.S. O'CONNEL,
`A. RAMPLIN, J. Chem. Soc. Perkin Trans. 1, 1993, 401.
`
`NOTE ADDED IN PROOF - We became aware of an error in the paragraph "Phenylbutazone". The synthesis
`and spectroscopic characterisation of 4-hydroperoxy-phenylbutazone has been previously reported by Von P.
`MENZ, M. SCHULZ and R. KUGE, ATZ/ZeiM. - Forsch./Drug Res. 37(11), 1229 (1987). This compound was also
`detected as a major degradation product of phenylbutazone in a sample of 4-years old tablets.
`
`NOVARTIS EXHIBIT 2050
`Noven v. Novartis and LTS Lohmann
`IPR2014-00550
`Page 5 of 5