degradation problems: Case studies

SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES K A R E N M. A L S A N T E , T O D D D. HATAJIK, LINDA L. LOHR, D I N O S SANTAFIANOS, A N D T H O M A S R. SHARP Pfizer, IncGroton, CT 06340

INTRODUCTION AND BACKGROUND A. The Drug Development Time Line B. The Process C. The ICH Guidelines D. Evaluation of Known Standards E. Cases with No Standard Match F. LC/MS Analysis G. Preparative Isolation Versus Small-Scale Synthesis H. NMR Characterization . CASE STUDIES A. Degradation Case Studies B. Process-Related Impurities Case Studies 1. SUMMARY AND CONCLUSIONS APPENDIX—LESSONS LEARNED REFERENCES

INTRODUCTION AND BACKGROUND The objective of this chapter is to provide guidance for isolating and identifying process-related impurities and degradation products from pharmaceutical drug candidates using actual case studies. The identification of process-related impurities and degradation products can provide an understanding of impurity formation and define degradation mechanisms. If the identification process is performed at an early stage of drug development, there is adequate time to address the drug substance process and drug product formulation to prevent or control these impurities and degradants from forming long before the filing stage. Impurity and degradant structure elucidation is a collaborative effort involving the analytical chemist, process chemist and/or formulator, as w^ell as the degradation, mass spectrometry, and NMR experts. The process described in this chapter uses a designed approach for the impurity and/or degradant identification, which focuses on

361

362

K.M. ALSANTE et al.

efficiency so that the success of data collection is maximized and project time lines are met. There are a number of activities other than collecting experimental data, even though the experiments are central to the process. Some of these key activities include collecting project background information prior to pursuing experimental work, asking the right questions, and meeting v^ith project analysts and structure elucidation experts. The activities associated with the overall process are captured in the process flowchart provided in Figures 1-3.^

Unknown Degradant Impurity Problem Identified

Contact Technology Groups Assess Time Lines for Completion

Discuss w/ Project Team

Confirm Structure

HPLC/UV LC/MS

Yes

„-''1*Ri/Deg.Stl -".^RRT Match*

No

.^^-I^Dato^ • — , ^ MW?

JNO Develop LC-MS Method

No

^ MS compatibtftv. methocP ^ ^

jConfirm Structure! (MS and RRT)

Discuss possible structures with project team. Determine if information is suitable or if isolation is required.

Impurityi Evaluate Process Stage

Synthesize (Must be most efficient route)

isolate Degradant/lmpurity for NMR Studies

FIGURE I Impurity/degradant isolation and Identification process flowchart (RRT =relative retention time; PRI = process-related impurity).

363

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES

Obtain Bulk

Assess Separation (NP TLC {Methods & RP HPLC IVIethods)

Determine Optimum Resolution of Deg/ Impurity: RP or NP? Normal Phase

Reverse Phase

Develop Scaleable TLC Method

Method Development

<1%

Isolate and Perform LCMS

>1% Perform Flash Chromatography

Determine Max. Analytical Load

Scale to Prep. HPLC

Isolate Approx. 25 mg for NMR FIGURE 2

I m p u r i t y / d e g r a d a n t isolation and identification process f l o w c h a r t .

A. The Drug Development Time Line

One of the most important aspects of the project that determines approach is where the pharmaceutical drug candidate is in the drug development time line. Typically, the focus will be on collecting LC/MS data only through the phase 1 clinical stage because of limited analytical resources compared to the number of compounds in early development. At the phase 2 clinical stage and beyond, more time is invested in isolation, synthesis, and structural identity using NMR characterization of forced degradation products and impurities of concern. At this stage, the pharmaceutical drug candidate has a better chance of making it to market. If an unknown degradant or impurity is at a critical threshold and drug substance and/or drug product release for clinical supplies could be impacted, the most efficient technique for identification will be used. For these cases, isolation, synthesis, LC-NMR, and/or traditional NMR are used no matter what stage of development: phase 1 all the way through registration. B. The Process

As already emphasized, the process of identification of impurities and/or degradants begins early in drug development. Early brainstorming sessions should involve the analytical chemist, process chemist, formulator, and degradation chemist, as well as the mass spectrometry and NMR experts. It is imperative to involve all that are familiar with the project of interest. The

364

K.M.ALSANTEetal. Determine purity, solubility & quantity Discuss time lines

Routine characterization 1-D proton & carbon acquiredj for structural confirmation

No

Yes

Identify LC method Optimize method for LC-NMR Perform appropriate LC-NMR expts

Select appropriate probe: submicro^ micro, or traditional Perform appropriate expts

Yes

No

Run expts on standards as needed for comparison

Interpret NMR data Identify possible structures consistent with data Discuss results with project team to confirm structure and to determine plausible synthetic pathway FIGURE 3

N M R spectroscopy identification process f l o w c h a r t .

group meets to assess the time lines for completion and to gather all pertinent information. This initial planning and discussion effort can save significant time in the experimental stage. A few questions that need to be answered at this early stage are • • • •

Is this an impurity or degradant problem? Is it a drug substance, drug product, or excipient-related problem? At what level is the impurity/degradant present? Is it a process-related impurity (PRI); and if so, at what step of the process is it formed? • Is it a degradant; and if so, under what degradation condition is it formed? • Are enriched samples with the unknown impurity/degradant available? By gathering all relevant information, the most efficient method of isolation and identification can be selected. Many more unexpected questions

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES

36S

and insights can arise at the early meetings that can improve the efficiency and quahty of the identification process. C The ICH Guidelines The first step of the process is to determine at what level the unknown is present. According to the ICH Guidelines on Impurities in New Drug Substances:^ The studies conducted to characterize the structure of actual impurities present in the new drug substance at a level greater than 0.1% (depending on the daily dose, calculated using the response factor of the drug substance) should be described. Note that all specified impurities at a level greater than the identification threshold in batches manufactured by the proposed commercial process should be identified. Degradation products observed in stability studies at recommended storage conditions should be similarly identified. When the identification of an impurity is not feasible, a summary of the laboratory studies demonstrating the unsuccessful effort should be included in the application. According to the ICH Guidelines on Impurities in New Drug Products:^ Degradation products observed in stability studies conducted at recommended storage conditions should be identified when present at a level greater than the identification thresholds (1% for a maximum daily dose of 2g). Identification of impurities below the 0 . 1 % level is generally not necessary unless the potential impurities are expected to be unusually potent or toxic.*^ Therefore, it is imperative to determine the level of the unknown impurity and/or degradant early in the process. If the unknown is below the 0.1% threshold, then a discussion will need to take place among the project team members in order to determine if isolation and identification are necessary. However, if the unknown is at or above the 0.1% limit, then effort should be put forth for identification. D. Evaluation of Known Standards Once a decision has been made to identify an unknown, the next logical step is to evaluate all known process-related impurities, precursors, intermediates, and degradation products. By observing the relative retention times (HPLC) of all known process-related impurities, precursors, and intermediates (if available), one can quickly determine whether or not the impurity of interest is truly unknown. If the relative retention time of the unknown impurity matches that of a standard, the unknown can be identified using HPLC with ultraviolet (UV) photodiode array detection as well as mass spectrometry (MS) detection. The identity can be considered confirmed by correlating the retention time, UV spectra, and mass spectra of the unknown impurity with that of the standard. The time and energy saved by analyzing data that may be already available can be considerable.

366

K.M. ALSANTE et al.

E. Cases with No Standard Match

Identifying an unknown by using a standard, as described in the above paragraph, is a quick and easy process. However, what happens when the relative retention time of an unknown does not match that of a standard? The next step is to obtain molecular mass and fragmentation data via HPLC/ MS. It is essential to determine the molecular mass of the unknown. Not only does the molecular mass help in the identification of the unknown, but it also enables one to track the correct peak by HPLC if isolation becomes necessary. In order to run LC/MS, a mass spectrometry compatible HPLC method must be available. The mobile phase should contain volatile buffers that are HPLC/MS-compatible. If such a method is not available, then one must be developed, which adds time to the identification time frame.

F. LC/MS Analysis

When a new unknown impurity is observed, as a first experiment, the sample is analyzed by LC/MS. If the structure of the unknown impurity cannot be conclusively elucidated by LC/MS data, LC-NMR can be employed to analyze the sample. If the sample is not suitable for LC-NMR analysis, the impurity needs to be purified for NMR characterization. If the mass spectrometry data evaluation yields sufficient structural information, this eliminates the need to isolate the impurity in question. If standards of the proposed structures are available, they can be correlated with the unknown as previously described. If standards are not available, which is usually the case, the proposed structures can be discussed with the project team. The project team can then decide if the information is suitable for their needs or if isolation is required. It is essential for a thorough analysis of the pros and cons of each ionization technique. For example, knowing that a degradant might be unstable enables the selection of a delicate ionization technique that will not destroy the degradant prior to mass spectral detection.

G. Preparative Isolation Versus Small-Scale Synthesis

Preparative HPLC is typically the technique of choice for impurity purification. It is often necessary to enrich the impurity before preparative HPLC purification. Various techniques such as soHd-phase extraction can be used to enrich the low-level impurity. To ultimately confirm the structure of a new impurity, it may be necessary to synthesize the compound and compare its spectroscopic characteristics to those observed in the original sample. A very effective means of getting useful structural information is to conduct a degradation study on the purified impurity. An alternative to isolation is small-scale synthesis. If possible, structures have been proposed from the mass spectrometry data, one can study the synthetic chemistry and determine at which step of the process the impurity

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES

367

and/or degradant is most likely to be formed. By knowing the synthetic chemistry, the feasibility of the proposed structures can be evaluated. Proposed structures can then be synthesized if a reasonable synthesis is available. It is easier to synthesize and identify the unknown if the chemistry works quickly (i.e., one-step/straightforward chemistry). If small-scale synthesis is chosen, the synthesis must be the most efficient route. At this stage of the process, it is frequently necessary to isolate and characterize the unknown. One of the most important factors to consider when approaching an isolation experiment is the sample origin. It is vital to determine whether the unknown is an impurity and/or degradant and to locate a sample that contains an enriched quantity of the unknown. Isolating low-level impurities can prove to be very cumbersome and timeconsuming. Therefore, the ultimate goal is to find a sample that contains an enriched quantity of the unknown. Two great resources of enriched samples are retained mother liquor samples and purposeful degradation/ stability samples. If the unknown is a drug substance degradant, then the degradation reaction can be scaled up to generate a large quantity of the unknown. If it is a drug product degradant, then effort should be put forth to form the degradant in the drug substance so that extraction from the excipients is not required. Whenever enriched samples are not available, the unknown must be isolated from the bulk drug substance or drug product. A number of methods can be used for isolating impurities and/or degradants. Three of the most utilized techniques discussed earlier in Chapters 9 and 10 are thin-layer chromatography (TLC), flash chromatography (column chromatography), and preparative high-performance liquid chromatography (HPLC). The actual technique used depends on the nature of the impurity and/or degradant, including the amount present in the original material from which it must be isolated. A good starting point is to assess the separation that is currently being used by analytical chemists. Does the current methodology provide optimum resolution of the impurity/degradant from the main band and other impurities, and if so, is that method by TLC or HPLC? This is a key factor in determining which technique to utilize.

H. NMR Characterization Nuclear magnetic resonance (NMR) spectroscopy is used as a complementary technique to LC/MS. It is a powerful analytical tool for structural elucidations, particularly for small molecules. It provides key structural information and intramolecular connectivity not readily obtainable by any other analytical methods. While mass spectrometry typically uses fragmentation patterns and total mass to suggest a general family of plausible structures, NMR spectroscopy refines the molecular connectivity to elucidate a single, specific structure and, if desired, stereochemistry. Recent technology advancements in the field of magnetic resonance have made significant strides in improving sensitivity limits of this technique. This becomes particularly critical in the structural elucidation of pharmaceutical drug impurities and

368

K.M.ALSANTEetal.

degradants, which are often extremely mass- and sometimes solubilityhmited. The nondestructive nature of NMR makes it an especially valuable tool for impurity and degradant characterization, since these are typically valuable samples that may need to be revisited in the future. The NMR strategy previously outlined serves as a guideline for the approaches taken in the following case studies.

II. CASE STUDIES

Actual case studies using these isolation and syntheses as well as mass spectral and NMR approaches are outlined in the remainder of this chapter. This collaborative multidisciplinary strategy is applied to the structure elucidation of all the impurity and degradant case studies presented. The successful elucidation of these structures was essential to predict potential toxicity, set threshold limits, and identify ways to prevent or greatly reduce their formation. The case studies have been organized into degradation and process-related impurities examples.

A. Degradation Case Studies I. Case Study A.I: N-Oxide Versus Sulfoxide Differentiation Using Preparative HPLC, LC/MS, and NMR Characterization Drug Substance

Oxidative Degradant

G .Ar TV-Oxide or Sulfoxide?

Xlf

©

Ar

g)

In this first case study, HPLC analysis of multiple drug substance samples (bulk release, stability, and forced degradation) indicated the presence of a very polar impurity/degradant, eluting just after the solvent front. A twomonth time frame was allotted for identification. In collecting background information on the project, it was noted that the degradant steadily increased in stability samples as well as in oxidative forced degradation experiments.^ LC/MS data indicated a molecular weight of M+16. LC/MS could not confirm the site of oxidation (N-oxide versus sulfoxide); therefore, NMR work was required. Since forced degradation generated the degradant, a scaled-up oxidative challenge was performed on the drug substance to prepare an enriched quantity of IM+16] degradant for isolation and identification. Reaction conditions for the scaleup were selected based on previous forced degradation oxidative data. Oxidative degradation for ~10 days, with an anhydrous methanol solvent system, yielded the most

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES

369

abundant quantity of the degradant (20%). Reversed-phase preparative HPLC isolation was problematic since the M+16 degradant w^as extremely polar and eluted in the void volume. The LC/MS data confirmed that the isolated degradant had a molecular mass of M+16. Removal of the solvents by evaporation unfortunately resulted in a high level of residual ammonium acetate in the NMR sample that greatly compromised the detection of the target degradant signals. Even more important, the 87:1 saltidegradant content dramatically changed the magnetic susceptibility of the sample, making the NMR probe untuneable. Running the NMR experiments using an untuned probe would greatly compromise the achievable sensitivity. Therefore, the sample was further purified. The sample was placed in a lyopholizer to try to remove the ammonium acetate, but the attempt was not successful. An alternate method of isolation was developed to obtain a cleaner sample. Normal-phase chromatography was attempted because of problems experienced with reversed-phase chromatography. A drug substance oxidation solution was analyzed by TLC using an ethyl acetate/methanol/triethylamine solvent system. This mobile phase was found to be suitable for a large-scale purification by column chromatography on siUca gel. The solution was purified by medium pressure (^40 psi) column chromatography on silica gel using ethyl acetate/methanol/triethylamine as the eluent. Fractions containing the desired material were concentrated by evaporation and dried under high vacuum, to give a clear colorless oil. The oil was dissolved in a small amount of methanol/ ethyl acetate and concentrated by evaporation. This step was repeated twice. The oil was dissolved in methanol/ethyl acetate cooled to 0°C (affording traces of a white solid) and treated with sufficient hexanes (added slowly) to effect slow precipitation of a white solid. The solid was collected by filtration and dried under high vacuum overnight to give a white sofid (142 mg). A second crop was obtained and dried overnight under high vacuum to give (52 mg). Samples were both demonstrated to have the same Rf by TLC analysis. The LC/MS data confirmed that the isolated degradant had a molecular mass of M H - 1 6 , a critical step to confirm no further degradation on isolation/ concentration prior to detailed NMR analysis and interpretation. The integrity of the sample was confirmed by correlating the retention time and mass spectra with that of the unknown acid M + 1 6 degradation sample. A complete NMR characterization^ was performed on this purified sample, as well as on the parent free base and related sulfone degradant for comparison. Methanol was chosen as the solvent, since solubility in methanol was demonstrated in the LC method development. The parent was also soluble in methanol, which allowed for a straightforward comparison of chemical shifts. Enough sample was isolated (30 mg) to acquire a carbon spectrum, which was particularly useful to detect the quaternary carbon resonances in the proposed oxidation region. Compared to the parent carbon spectrum, significant downfield chemical shifts were observed for both quaternary carbons neighboring the sulfur entity. In addition, upfield chemical shifts were observed for the aromatic methines immediately next to these quaternaries. The NMR data therefore confirmed oxidation had occurred at the sulfur site, and hence the sample was identified as the sulfoxide of the parent structure.

370

K.M.ALSANTEetal.

2. Case Study A.2: N-Oxide Versus Sulfoxide Differentiation Using Synthesis/Preparative HPLC Drug Substance

Oxidative Degradant

^4

A^^^R2

In case study A.2, a degradant was observed at a greater than 0 . 1 % level in drug product stability studies using ICH storage conditions. A oneweek time frame was allotted for the identification process to meet the project time line. Oxidation of the compound was suspected by LC/MS by the presence of a M+16 degradant, suggesting the addition of oxygen; however, as in case study A.l, LC/MS analysis could not differentiate between an N-oxide or sulfoxide. Further NMR analysis was required to characterize the degradant. Since time was an important factor, the most efficient route to obtaining a pure sample for NMR was determined to be a synthetic reaction. When considering synthesis as an option in the identification of impurities, having a high degree of certainty in the structure of the impurity is necessary. A one- or two-step reaction can afford grams of material sufficient for mass spectrometry and NMR analyses, correlation by HPLC, and for use as a standard for HPLC method development. Most important, the synthesis has to be short and efficient. Multistep syntheses can become extensive research projects when complicated reactions do not work as expected. With this example, a single-step mild oxidation was planned in order to selectively synthesize the sulfoxide or the N-oxide. HPLC analysis of the crude reaction mixture would confirm if the correct degradant was being formed. An aggressive oxidant would probably afford both degradants nonselectively; hence, a mixture of the drug substance was treated with a less aggressive excess of hydrogen peroxide-urea complex. Analysis of the reaction mixture indicated that the desired degradant was forming. The reaction afforded 94 mg of crude degradant. The purity was lower than desirable for NMR analysis, so further purification was undertaken to facilitate structure elucidation by NMR. The existing analytical methodology contained perchloric acid that was not suitable for preparative HPLC isolation and LC/MS analysis because of safety concerns with concentration of perchloric acid and high probability of damage to the mass spectrometer over time. A mass spectrometry compatible method using a 0 . 1 % acetic acid buffer was selected as a starting point. Minor method development produced a suitable method to separate the reaction components: the drug substance, the sulfoxide, and the N-oxide.

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES

371

By determining the maximum analytical load (load that maintains adequate resolution of the unknown component), the chromatographic conditions were scaled up and a semipreparative chromatographic method was developed for the isolation of the degradant. The degradant of interest was then isolated by preparative HPLC. Concentration of the isolate by evaporation and drying under high vacuum afforded 153 mg of degradant. The theoretical yield was 42 mg; therefore, the remaining 111 mg was attributed to ammonium acetate and solvent impurities from the mobile phase, a complication to be aware of in preparative HPLC. Analytical HPLC analysis of the concentrated sample showed a single peak with the retention time of the proposed N-oxide. It is essential to confirm the retention time with the actual analytical method for purity and potency used by the project team before further NMR and mass spectrometry structure elucidation begins. Despite the presence of extremely large amounts of ammonium acetate, structure elucidation of the N-oxide by NMR spectroscopy was successful with no further cleanup. This is the exception rather than the norm in the isolation process. Of the 150 mg of sample yielded from the isolation process, less than l l m g of isolate was attributed to the degradant. The remaining mass was due to residual ammonium acetate. As with Case A.l, this extremely high salt content significantly shifted the magnetic susceptibility of the NMR probe, thus making probe tuning quite difficult. In addition, the presence of a sizeable background component put high demands on the dynamic range of the NMR spectrometer's receiver, making detection of the relatively small amount of degradant challenging. Methanol was chosen as the preferred solvent because it is commonly used for chromatography and has the ability to remove the solvent easily if needed. Unfortunately, the sample demonstrated poor solubility in methanol, thus reducing the sample concentration and hence the detected signal size. Methanol also prohibits the detection of labile protons, although this was not forecasted as an issue for this project. The proposed N-oxide structure is a charged species, so spectral simulations would not be typically reUable, since usually only neutral species are included in simulation databases."^ Despite these challenges, a full set of NMR experiments was performed. The data were interpreted and the N-oxide structure was successfully elucidated within 24 hours. Most noteworthy NMR observations included major chemical shift changes near the nitrogen site plus a broadening of the N-methyl resonance. This NMR evidence facilitated a subsequent structural elucidation of another degradant of the same parent species, which was determined to be a chloromethyl adduct binding at the same nitrogen site. This clearly demonstrates the potential reactivity of the N-methyl site. In terms of impact on the project, the N-oxide structural elucidation allowed for an appropriate specification of the degradant, and cHnical time lines were not impacted. It is advised in all preparative HPLC isolations that an analytical or preparative scale reinjection is performed to clean up the analyte of interest from the salt. This can include washing the analyte by reversed-phase HPLC (preparative or analytical scale depending on isolate amount) with aqueous phase and ramping up the organic phase to elute the desalted analyte of interest.

372

K.M.ALSANTEetal.

3. Case Study A.3: Degradant Isolation Using Preparative HPLC, LC/MS, and N M R Characterization

Drug Substance

Thermal/Humidity Degradant

Case study A.3 involves the same drug substance detailed in case study A.2. The drug substance was stored at ICH storage conditions of 40°C/75% relative humidity for six days. Samples analyzed by HPLC indicated the presence of an unknown degradant at the 0.6% level. The aim of this project was to isolate and identify this degradation product. This required a team effort from the following groups: Degradation, NMR, Mass Spectrometry, and Discovery. The molecular weight of the degradant was confirmed by LC/ MS (M-262); however, very little structural information was obtained from LC/MS/MS fragmentation. As a result, isolation and further NMR analysis were required to complete the identification project. In order to obtain a pure sample for traditional NMR, the degradant was isolated by preparative HPLC. The existing analytical method had the advantage of being mass spectrometry-compatible from the start with a mobile phase of 0 . 1 % acetic acid pH adjusted with ammonium hydroxide solution and acetonitrile as the organic modifier. Additional method development was not required, which can be time-consuming and adds to the isolation and characterization time-frame. On the preparative HPLC scale, degraded material at a concentration of 50mg/ml gave a split peak. This was determined to be the result of the high concentration of the material loaded on the preparative HPLC column, another occurrence to keep in mind in the scale-up process. Sufficient preparative runs were conducted to afford 6mg of the degradant. LC/MS analysis on the pure isolated sample afforded greater sensitivity and confirmed the molecular weight of the degradant (M-262). NMR structure elucidation indicated that a cyclized compound had formed. In particular, COSY-type proton-proton correlations gave the general ring structure shown. Key observations included a change from the parent proton spectrum of a methylene multiplet to a broad singlet, in addition to the maintenance of an electron-rich methine singlet. Parallel to this characterization effort, the Discovery project team had resource capacity and synthesized the proposed molecule. The NMR data for the isolated degradant and synthesized compound were in excellent agreement and hence supported the structure for the isolated degradant.

373

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES

Proposed Mechanism of Formation

V R3v J

6^' , /

R,/^

H

R.-^e

Y •H

r°

-i-

^ io

v 1

Me

(a) Formation of a radical cation at the sulfur atom, (b) allylic radical can exist in two canonical forms, (c) the second form can undergo a [1,5] hydrogen abstraction to transfer the radical to the side chain, (d) at some point loss of a proton can regenerate the vinylic sulfide, (e) the radical is well positioned to cyclize onto the vinylic double bond and the resultant radical can be quenched by radical abstraction from solvent, (f) the sulfur atom can oxidize to the sulfoxide, (g) the sulfoxide can form a double bond by synelimination resulting in loss of the sulfur containing side chain. 4. Case Study A.4: Scaled-up Oxidative Degradation and Isolation Using Solid-Phase Extraction and Preparative HPLC

Drug Substance

Drug Product Degradant

374

K.M. ALSANTE et al.

A degradant at a critical level ranging from 0.6% to 2.8% was observed in drug product stability samples stored at 30°C/60%RH for six weeks. The drug product vehicle was an acidified polymeric matrix making this case study A.4 particularly challenging. Identification was required for the following reasons: (1) this degradant increased at 5°C, the storage condition for the drug product, (2) after six weeks at 5°C, the degradant was 0 . 1 % , and (3) after 12 weeks at 5°C, the degradant was 0.18%. In order to set a higher specification for this degradant so that a longer expiration date may be obtained for the drug product, isolation and identification were needed. LC/MS indicated that the molecular weight of the unknown degradant was M H - 1 4 , a possible ketone functionality. Since the oxidation could reasonably occur at different benzylic positions that are present in the parent molecule, further isolation and NMR analysis were required to identify the exact site of oxidation. This degradant was observed only during stability and forced degradation studies of the drug substance in the presence of the polymeric matrix. All stress testing studies on the drug substance "as is" without the polymer showed that the drug substance was stable. Therefore, isolation work could not be conducted on the drug substance alone, without the complex polymeric matrix. It is always advantageous when a drug product degradant can be repHcated in drug-substance-only stress samples, eliminating the need to further isolate the degradant from the excipients. In this case, it required cleaning up the degradant from the complex polymeric matrix. To generate adequate amounts of the degradant of interest and decrease isolation time, a pressurized forced degradation oxidation reaction was performed on the polymer formulation in the presence of a free radical initiator. Approximately 14% of the degradant was generated by intentionally degrading the formulation. To isolate the degradant from the polymer formulation, Waters Oasis MCX ion-exchange solid-phase extraction (SPE) cartridges were utilized to separate the degradant from the complex polymers. Oasis MCX cartridges contain a mixed-mode polymeric sorbent with reversed-phase and cation exchange functionalities that are highly selective for basic compounds. The oxidation solution was placed on 100 Waters 6cc Oasis MCX cartridges prewashed with 0 . 1 % hydrochloric acid. The loaded cartridges were washed sequentially with 0 . 1 % hydrochloric acid, water, and finally methanol to remove the polymer. The remaining organic material was then eluted from the cartridges by washing with a solution of 5% ammonium hydroxide in methanol. Combining the ammonium hydroxide in methanol fractions afforded a clear colorless solution, which was concentrated by evaporation to give a white semisoUd (1.64 g). This SPE cleanup successfully extracted the drug substance-related compounds from the polymer. The desired degradant was then isolated by preparative chromatography. The isolated degradant was analyzed by NMR and mass spectrometry for structural identity. Based on LC/MS data collected, the isolated degradant was confirmed to have an addition of 14 Da to the molecular weight of the drug substance, indicative of a ketone moiety. NMR analysis confirmed that the M + 1 4 degradant contained a carbonyl group at the benzyfic site shown.

14

375

SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES,^

Comparing the proton and carbon spectra of the parent and degradant, it was apparent that the proton resonances at the benzyhc site had disappeared, and the carbon resonance was no longer observed in the aliphatic region. Additionally, the proton and carbon resonances of the adjacent carbon atom were deshielded, with respect to the parent, in the degradant. The ketone oxidation product was confirmed by both NMR and mass spectrometry, and the degradation mechanism outlined in Scheme 1 was proposed. Since the degradant was formed at higher levels with pressured oxygen/free-radical forced degradation conditions, a free-radical autoxidation mechanism involving Russell termination was proposed.^ Proposed Mechanism of Formation

1$

HO^O

5. Case Study A.5: Isolation of a Degradant with Same Molecular Weight as the Drug Substance Requiring N M R Characterization for Structure Elucidation

Drug Substance

Degradant

V (CH2)n The current manufacturing process for the synthesis substance in case study A.5 generated an impurity at the Identification was required for a drug safety assessment on within a 1-month time line. LC/MS analysis suggested that the

of the drug 1-2% level. this impurity impurity had

376

K.M. ALSANTE et al.

a different HPLC retention time but had an identical molecular weight. It was decided to isolate and thoroughly identify this impurity, using NMR in order to elucidate the structure. Isolation and identification of the degradant was complicated by the structural similarity of the impurity with the parent compound, having the same molecular weight. This resulted in very similar HPLC retention times for the impurity and the main band. An enriched process-retained mother liquor sample containing approximately 4 - 5 % impurity was obtained, and a new preparative-scale HPLC method was developed. The impurity was isolated by preparative-scale HPLC. According to LC/MS analysis, an early step in the bulk manufacturing process for the synthesis of the drug substance generated a "precursor impurity" with an identical molecular weight to the desired product. This precursor impurity was tracked through the synthesis starting at a step in which a cyclobutyl precursor was coupled to a dichloro substituted intermediate. It was assumed that the impurity was the result of the cyclobutyl precursor displacing the undesired chloro group. The completion of the synthesis led to the presence of impurity in the final drug substance. This assumption suggested that the impurity was a structural isomer of the desired product. The 1-month time line allowed method development for isolation. Normal-phase silica gel TLC was investigated briefly. However, a suitable separation was not achieved. The analytical HPLC method was not suitable for preparative-scale HPLC chromatography. The method utilized a Symmetry Shield RP-8 packing material that was not readily available as a preparative-scale column. A sample containing 5% of the desired impurity was obtained and used for HPLC method development and impurity isolation. Due to the similarity of the retention times, this problem was not trivial. The first preparative HPLC method using 0.05% trifluoroacetic acid/ acetonitrile/methanol appeared to work well on an analytical scale; however, when the method was scaled up on a Symmetry C-8 column, adequate resolution was not achieved. This method was attempted using a Waters radial compression technology as well as a stainless steel preparative column containing Symmetry C-8. The method was also adjusted to 0 . 1 % trifluoroacetic acid; however, no separation was achieved. A new method was developed on an analytical scale using 0 . 1 % formic acid in water/ tetrahydrofuran/methanol. This method worked well on a preparative scale using a radial compression Symmetry C-8 column. A solution containing the impurity dissolved in 50% tetrahydrofuran/50% water was purified by preparative HPLC. The combined fractions were concentrated by evaporation to give a white solid (47.5 mg). This sample was analyzed by HPLC and found to match the retention time of the impurity. The sample was dissolved in tetrahydrofuran and purified by preparative-scale HPLC using a longer elution time in order to effect a final analytical cleanup. This sample was analyzed by HPLC and found to match the retention time of the impurity with a purity of 97.4%. A sample of the impurity was analyzed by LC/MS and NMR. Based on ^H and ^^C NMR spectroscopy, the unexpected ring enlargement structure was assigned. To facilitate spectral comparisons to the parent, DMSO was

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES^.

377

used as the NMR solvent for both the parent and the impurity. This has the additional advantage over protic solvents of permitting detection of labile protons. Unfortunately, DMSO is difficult to remove from a sample if an alternative solvent has to be used. The NMR experiments v^ere acquired on only 1.7 mg of sample. Since this predated our purchase of microprobes optimized for such smaller sample sizes, a standard 5-mm probe w^as used, thus reducing potential sensitivity gains from concentrating the sample. The presence of several consecutive quaternary carbons and heteroatoms made long-range correlations to protons difficult to observe. Therefore, a carbon spectrum w^as desired, but was challenging to acquire because of sample mass limitations. Both hetero- and homonuclear correlation experiments were used to elucidate the final structure. After isolation and identification of the impurity, the structure was found to be different from the originally proposed structure. This example demonstrates the need to obtain the scientific evidence rather than rely on suggestions and proposals. Fortunately, in this case study the time line allowed ample time for method development, and this is not always the case. It is impossible to predict the complications that may arise in the isolation or characterization. Proactive involvement prior to crisis mode is desired. 6. Case Study A.6: Scaled-Up Oxidative Degradation, Preparative HPLC, and Characterization by LC/MS and N M R

Drug Substance

Degradant

Oxidative forced degradation of the drug substance generated an unknown degradant that was observed in ICH stability samples of both the drug substance and drug product (tablets). This pharmaceutical drug candidate was at a later stage of development; therefore, it was decided to proceed through the identification process in a proactive manner under no strict deadlines. From forced degradation studies, it was determined that the unknown degradant was present at approximately 29% in the oxidative challenge degradation sample. Without having a proposed structure, preparing an enriched sample and isolation by preparative HPLC was concluded to be the most efficient route. A large-scale oxidation degradation using a free-radical initiator and molecular oxygen was performed on the drug substance in order to produce a large amount of the unknown oxidative degradant for preparative HPLC isolation. By determining the maximum analytical load of the unknown oxidative degradant, the chromatographic conditions were scaled up and a semipreparative chromatographic method was developed. Thirty-two injections of the scaled-up oxidative challenge degradation sample were performed.

378

K.M. ALSANTE et al.

The collected fractions were combined and evaporated to dryness yielding a yellow solid. The isolated material was submitted for LC/MS and NMR analyses. LC/MS of the primary degradant indicated a molecular mass increase of 16 Da. Initial thoughts suggested an N-oxide, but the fragmentation induced by coUisional activation in the tandem mass spectrometer was inconsistent with this proposal, and two other structures were proposed. The initial NMR results indicated that the sample was not pure and contained a large amount of residual ammonium trifluoroacetate from the mobile phase, which greatly effected the electronic environment and dominated the experimental spectra. This made both probe tuning and impurity detection extremely challenging. Therefore, the sample was further purified. HPLC reinjection was performed to remove any residual TFA (trifluoroacetic acid) salt that might have been present. The sample was dissolved in a minimum amount of methanol and reinjected onto the semipreparative column. In order to avoid forming a TFA salt, TFA was not used in the mobile phase during the cleanup procedure. The cleanup procedure yielded 170mg of the unknown oxidative degradant, which was analyzed by NMR. Of the three proposed structures, the NMR data supported only the N-oxide. The standard set of NMR experiments was acquired. Based on comparison of corresponding proton and carbon chemical shifts, the local environment of the degradant was more electron-rich than the parent at the proposed oxidation site, as is expected for an N-oxide. This observation is consistent with oxygen addition. The remaining proton and carbon resonances were essentially the same for the parent compound and the degradant, thus demonstrating no other structural differences. The N-oxide structure is also consistent with the parent MS fragmentation, but inconsistent with the daughter fragmentation. This is a case where interpretation of the fragmentation led to false conclusions about the position of the oxygen atom. 7. Case Study A.7: Scaled-Up Acid Degradation and Isolation Using Preparative HPLC: Characterization by LC/MS, N M R , and IR Spectroscopy

Drug Substance

Acid Degradant

Acid-forced degradation of the drug substance generated an unknown degradant that was not observed using ICH storage conditions (40°C/75% RH), so identification was not time-critical. This pharmaceutical drug candidate was at a later stage of development; therefore, it was decided to proceed through the identification process in a proactive manner under no

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES

379

Strict deadlines. Based on LC/MS data, the molecular mass of the unknown acid degradant was determined to be M+30. Degradant structures matching the molecular weight were proposed, but further structural information based on NMR analysis was required. From forced degradation studies, it was determined that the unknown degradant was present at approximately 6% in the acid challenge degradation sample. Without having a proposed structure, preparing an enriched sample and isolation by preparative HPLC was concluded to be the most efficient route. A large-scale acid degradation using 1.0 N HCl at 70° C was conducted using the drug substance in order to produce a large amount of the drug substance unknown acid degradant for preparative HPLC isolation. By determining the maximum analytical load of the unknown acid degradant, the chromatographic conditions were scaled up and a semipreparative chromatographic method was developed for the isolation of the unknown acid degradant. Ninety injections of the scaled-up acid-challenged degradation sample were performed. The collected fractions were combined and evaporated to dryness, yielding a yellow solid. HPLC reinjection was performed to remove any residual TFA (trifluoroacetic acid) salt that might have been present. It was dissolved in a minimum amount of methanol and reinjected onto the semipreparative column. In order to avoid forming a TFA salt, TFA was not used in the mobile phase during the cleanup procedure. The cleanup procedure yielded 50 mg of the unknown acid degradant. The isolated sample was analyzed by LC/MS, and the LC/MS data confirmed a M+30 degradant after isolation and concentration confirming the pure isolated degradant sample was the correct species. The identity was confirmed by correlating the retention time and mass spectra with that of the unknown acid degradant peak in the ten-day acid degradation sample. Based on NMR analysis, the structure shown was proposed for the drug substance-related unknown acid degradant. In order to provide further evidence for the presence of the carbonyl in the proposed structure, infrared spectroscopy was performed. The IR spectrum showed a C = 0 stretch at 1670 cm~^ (consistent with an aryl ketone as in the proposed structure of the acid degradant). The IR and NMR spectral data indicated no significant TFA in the acid degradant isolate. The MS, NMR, and IR data were all consistent with the acid degradant structure proposed. 8. Case Study A.8: Scaled-Up Oxidative Degradation and Preparative Isolation: N M R Characterization Critical to Differentiate Between Two Possible Structures

Degradant

380

K.M. ALSANTE et al.

An unknown degradant formed in the drug product during stability studies using ICH conditions (40°C/75% RH) and was projected to be present at approximately 0 . 1 % at the end of shelf-life date. Hence, the identification team proactively tackled the identification effort. LC/MS of the unknown degradant indicated a molecular mass decrease of 44 Da. Several M-44 structures were proposed. In evaluating the degradation data, the degradant was also observed in forced degradation oxidative challenge. An oxidative degradation study was performed on the drug substance in order to generate the unknown drug product degradant. The unknown degradant was formed by free-radical oxidation of the drug substance only, a significant advantage. This eliminated the need to isolate the low-level degradant from the drug substance as well as the excipients. HPLC analysis of the pressurized oxygen/ radical initiator-forced degradation study indicated that 5% of this degradant was produced after 20 days. The degradant was identified by analytical HPLC as being the drug product degradant. A large-scale oxidative degradation was performed on drug substance in order to produce a large amount of unknown degradant for isolation. A suitable semipreparative HPLC method was developed for isolation of the unknown degradant from the drug substance. Fractions containing the desired degradant were combined and concentrated by evaporation. The isolated material was purified in a final analytical cleanup to yield the pure desired degradant for NMR analysis. A complete NMR characterization was performed to elucidate the structure. For comparison, the same set of experiments was collected for the parent drug substance. Initial NMR characterization data were consistent with a proposed structure, but reevaluation was performed when a new reaction pathway was proposed. A second structure was proposed that was also consistent with the NMR data. Additional NMR experiments were performed to differentiate between the two structures. These included nOe difference experiments to probe spatial relationships within the molecule. Possible Degradant Structures O JrllN

o

NH

R^

J

K^

9. Case Study A.9: Scaled-Up Light Degradation and LC/MS and N M R Characterization

Drug Substance

Degradant

o

o OH

r'^'^N

^-^^"-^

o ^OH

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES

381

In a related case study (A.9), an unknown degradant formed in the drug product during stability studies using ICH conditions of 40°C/75% RH and was projected to be present at approximately 0.1% at the end of shelf-life date using ICH conditions. LC/MS of the unknown degradant indicated a molecular mass increase of 14 Da, indicative of oxidation to a ketone moiety. Since several ketone products were possible, isolation of the major degradant was required to confirm the exact structure. A light degradation study was performed on a mixture of drug substance N-oxide and polymer excipient (1:9) in an attempt to generate an enriched quantity of the unknown degradant that had been observed in drug substance and drug product stability samples. HPLC analysis of the UV-challenged forced degradation sample indicated that 4 % of the unknown degradant was produced. Therefore, a large-scale UV degradation was performed so that the degradant could be isolated using semipreparative chromatography. A suitable semipreparative HPLC method was developed for isolation of the unknown degradant from the drug substance N-oxide/polymer excipient mixture. Fractions containing the desired degradant were combined and concentrated by evaporation. The isolated material was purified in a final analytical cleanup to yield the desired impurity. The structure of the degradant was determined by LC-MS and NMR spectroscopy. Two possible structures were proposed based on the molecular weight. NMR characterization was performed to elucidate the structure. For comparison, the same set of experiments was collected for the parent drug substance. Unfortunately, because of solubility differences, the parent was dissolved in dimethyl sulfoxide, while the degradant was dissolved in methanol. This somewhat complicated comparison of the corresponding proton resonances. Sensitivity was an issue with this sample because of mass limitations; only 0.6 mg were available for the NMR experiments. Preliminary interpretations of the NMR data revealed an inconsistency between the NMR and MS results. As always, a proton spectrum was acquired at the start and end of the NMR experiments to ensure sample integrity during the full acquisition period. A synthetically plausible structure was proposed based on the NMR observations. This structure differed from that originally expected by the project team. The mass spectral data were then revisited. The recorded total molecular mass of the impurity was exactly 28 Da higher than that of the proposed structure. This suggested that a sodium adduct had formed during the LC/MS experiment, a relatively common anomaly. Repeating the LC/MS work with this in mind demonstrated consistency between the NMR and MS data, thereby confirming the proposed structure. 10. Case Study A.IO: Adduct Degradant Synthesized and Characterized by LC/MS and N M R

Drug Substance

Succinic Acid Adduct Degradant

H

N^ Ri

• R2

o / Rr

\ "R2

382

K.M. ALSANTE et al.

In case study A. 10, the drug substance generated a degradant at greater than 0.6% under ICH conditions of 40°C/75% RH for 10 days. Mass spectral analysis indicated the degradant had a molecular weight of M+lOO, leading to the suggestion that the succinic acid had condensed with the drug substance to form an amide (succinic acid adduct). In deciding the most efficient identification route, only one adduct degradant structure could be proposed with a high degree of confidence. In evaluation of isolation versus synthesis, a synthetic one-step route with well-precedented chemistry was available; hence, the synthetic route was selected. A mixture of the drug substance free base from the succinate salt in chloroform was treated with succinic anhydride and heated at reflux. The solution was cooled to room temperature, concentrated by evaporation, and purified by column chromatography on silica gel to give the product (0.95 g, 69% yield) as an off-white foam. The synthesized material was shown by HPLC to elute at the same relative retention time as the impurity and co-eluted when spiked with a sample of degraded drug substance containing the degradant. When synthesizing, the impurity/degradant spiking experiments are a key piece of confirmation data. The molecular weight and LC/MS fragmentation data were consistent with a succinamide derivative. A complete NMR characterization was performed to elucidate the structure. Chloroform was chosen as the preferred solvent since it allows for detection of labile protons, and it is easily removed if necessary. However, NMR experiments on the parent species were run, using dimethyl sulfoxide as the solvent. This therefore made comparison of proton resonances between parent and adduct degradant less straightforward. The presence of two new carbonyl entities strongly supported the proposed succinamide structure. Long-range proton-carbon correlations observed between the methylenes and the carbonyl neighboring the nitrogen of interest confirmed the covalent bonding of the succinate moiety. The NMR data obtained have been informative for comparison to subsequent adduct degradants and impurities of this drug substance requiring identification. One such example is the analogous tartrate adduct. Archiving of data and capturing lessons learned have saved valuable time in minimizing duplication of effort. II. Case Study A.I I: Adduct Degradant Synthesized and Characterized by LC/MS and N M R

Drug Substance

Tartaric Acid Adduct Degradant OH

H N Ri

^ \

In case study A. 11, a degradant was observed in drug product-forced degradation (thermal/humidity) studies of the drug substance salt. This compound was at a later stage of development and proactive efforts were

14 SOIYING IMPURIT^DEGRADATION PROBLEMS: CASE STUDIES.^,,.^,

383

devoted to identification. LC/MS evidence indicated that the degradant had a molecular weight of M+132 Da, vs^hich corresponded to the condensation of the tartaric acid with the drug substance to form a tartrate adduct. Since there was a reasonably high degree of confidence in the proposed structure, synthesis was determined to be the most efficient route. Ideally, a one-step condensation of tartaric acid and amine would be the shortest route. However, since tartaric acid contains two carboxylic acid groups, it would be difficult to stop the amine from condensing with both groups to form an unwanted bis adduct in which one tartaric acid condenses with two drug molecules. It was decided to investigate the reaction of tartaric acid anhydride. An acid anhydride can react with an amine to give the desired adduct. The second acid group is hidden in the anhydride as a leaving group; therefore, it did not interfere with the reaction. HQ

i)H

Ha

OH

\ cm NR OH ^^^

HNR2

However, it is possible for a hydroxyl group to undergo the same reaction to form an ester group instead of an amide. For this reason it is necessary to remove or protect any hydroxyl group in the reaction mixture. HQ

HO. ^R

OH

HQ.

OH

\ R

The hydroxyl groups of tartaric acid anhydride as shown above had to be protected. Diacetyl tartaric anhydride was available from Aldrich. It was determined that this anydride would react to form the desired diacetyl adduct, and the removal of an acetyl group tends to be a relatively trivial task in organic chemistry. The degradant was prepared in two steps by the condensation of the drug substance and (-h)-diacetyl-L-tartaric anhydride followed by hydrolytic removal of the acetate groups. The first step was conducted in chloroform at 50°C and was catalyzed by the addition of dimethylaminopyridine. The reaction was complete in approximately 2 days. The solvent was then removed by evaporation, and the crude organic material was purified by column chromatography on silica gel using 1:1, methanolrethyl acetate as the mobile phase to afford 634 mg of the diacetate. A solution of the diacetate adduct in methanol was treated with potassium carbonate. The resultant hazy solution was stirred at room temperature for 4 hours. HPLC analysis of the reaction mixture indicated that the reaction was complete. The solution was concentrated by evaporation to remove methanol, treated with water, and neutraHzed to approximately pH 7 by the addition of formic acid. The resultant solution was concentrated by

384

K.M. ALSANTE et al.

evaporation, and the condensate purified by preparative scale HPLC. The main peak was collected and the solvent was removed from the combined fractions by evaporation to give a glassy solid. The soUd was dissolved in methanol, filtered through a short plug of cotton wool, and concentrated by evaporation to give an oil. The oil was treated with hexane and manipulated to give a white sohd that was dried under high vacuum overnight. Once a proposed degradant/impurity has been synthesized, it is important to take extra care to ensure that the correct compound has been made. It is vitally important that the sample is analyzed by HPLC and UV and shown to co-elute with the degradant to ensure that the compound is identical to the degradant. The sample was analyzed by LC/MS and found to have a molecular weight of M+132, corresponding to the molecular weight of the proposed tartaric acid adduct. HPLC and ^H and ^^C NMR spectroscopy were also used to fully characterize the compound. The standard set of NMR experiments were collected using deuterated methanol as the chosen solvent. This provided consistency with the corresponding parent spectra, yet prohibited the detection of exchangeable protons such as N - H and O-H groups. The spectra were complicated by the presence of a sample impurity. In addition, rotamers were observed, as evidenced by pairs of peaks apparent for several resonances. This work supported the hypothesis that the tartrate salt generated an adduct in stability studies. Unexpectedly, based on interpretation of the NMR results, the complexation of tartrate disrupted the molecular symmetry observed for the parent. Assignment of ^H and ^"^C NMR resonances demonstrated a clear match with the corresponding impurity data, thus confirming the proposed tartrate adduct formation. 12. Case Study A.I2: Adduct Degradant Synthesized and Characterized by LC/MS: Description of N M R Complications

Drug Substance

o

Succinic Acid Adduct Degradant

o

O OH

OH

OH

This case study (A. 12) describes the synthesis and identification of an adduct degradant that had been observed in degradation (thermal/humidity) studies of the amorphous drug. For this project, there was plenty of notice provided and resource allocated since the compound was at a later stage in development. LC/MS data showed the adduct degradant had a molecular weight of M+204, suggesting addition of succinic acid (the drug substance salt) to the parent. However, LC/MS data was unable to assign which of the three hydroxy 1 groups had reacted with succinic acid groups. Since this was not time-critical, synthesis was evaluated as the optimum route rather than isolation at a low level, which would have been extremely difficult and time-consuming.

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES

385

It was decided that the most efficient method of identification was to couple succinic anhydride with the drug substance and isolate the peak of interest by preparative-scale HPLC. A solution of the drug substance was treated with dimethylaminopyridine and succinic anhydride. The resultant solution was stirred at room temperature for 48 hours. The reaction mixture was partitioned between ethyl acetate and water, and the aqueous layer was then treated with I N HCl. The two layers were shaken well, and the aqueous layer was removed. The organic layer was then washed with water, saturated sodium chloride solution, dried with magnesium sulfate, and concentrated by evaporation to afford a clear colorless oil (1.81 g). A suitable preparative HPLC method using a volatile mobile phase of 0.1% formic acid in water/ methanol was developed, and the crude reaction mixture was purified by preparative-scale HPLC. The solution was concentrated by evaporation and the water was removed from this solution by freeze-drying to afford a white lyophilate (40 mg). Analysis of the isolated compound by LC/MS determined that the succinic acid had added to a hydroxyl group on the pyroUidine ring. The two hydroxyl groups on the 5-membered pyrollidine ring are so similar that they were considered the same reactivity. The isolate was analyzed by mass spectrometry and found to have a molecular weight corresponding to the succinic acid adduct. MS/MS analysis of this sample indicated that a fragment with a molecular weight of 204 Da was being formed. This LC/MS evidence suggested that the succinic acid had added to a hydroxyl group on the pyrollidine ring; however, it was not possible to determine which of the two hydroxyl groups reacted with succinic acid. NMR spectral analysis did not yield a definitive answer because of complications caused by limited sample mass, contaminants present in the sample, and most significantly, a rotamer effect arising from two separate amides. This gave rise to very complex NMR spectra that were not easily interpreted. In this study, the structure of the impurity was based on the LC/MS/MS data only. B. Process-Related Impurities Case Studies I. Case Study B.I: Identification Using a Combination of Enriched Forced Degradation Samples and Normal-Phase HPLC

Drug Substance

Impurity A HO^^^O

Impurity C H

Ar

- " ^ ^ ^

1 ^

NH2 OH

NH2 OH

^"v^O Me*

^^

Impurity B Aryl Compound

NH2 OH

OH

386

K.M. ALSANTE et al.

A drug substance sample contained three unknown impurities in case study B.l. Impurity A was present at 0.3% and Impurities B and C were both present at 0 . 1 % . The time Une for completion of the identification process was 1 month. At the initial stage of the identification project, it is critical to look at the degradation data and process-related impurities to determine if unknowns are actually known standards. From forced degradation studies, Impurities A and B were identified in the acid degradation experiment (3 hours in 1 N HCl) at the 13% and 5% level, respectively, based on HPLC relative retention time comparison. Acid degradation provided a key enriched sample. At the 5% and 13% level, this made the isolation task an order of magnitude less complex. The identities of Impurities A and B were confirmed by correlating retention times and mass spectra of the unknown impurities in the acid degradation sample with the retention times and mass spectra of known standards. LC/MS of the acid degradant gave molecular weights that were found to be identical to the standards. The standards were run on LC/MS and spiked into the sample. Impurity A was confirmed as the carboxylic acid analog of the amide drug substance, and Impurity B was confirmed as the aryl portion of the drug substance (as shown above). This background work saved valuable intensive isolation efforts. Impurity C did not have a standard match. Additionally, it could not be enriched by degradation or found in retained synthesis mother liquor samples, so isolation at the 0.1% level was required. In order to identify Impurity C, a normal-phase TLC method was modified for isolation since it gave adequate separation of all three impurities. The analytical TLC method utilized a number of solvents. One of the major components of the mobile phase was acetic acid. Since acetic acid is an aggressive acidic solvent that can degrade compounds, it was deemed necessary to develop a method in the absence of acetic acid. Acetic acid is also difficult to remove in the isolation process. If evaporation is used to remove solvents from the isolate, less polar solvents can evaporate preferentially, leaving the isolated compound in concentrated acetic acid solution. Bearing this in mind, it was decided to develop a milder mobile phase. The stability of the impurity should always be a concern in developing an isolation/identification strategy. The isolate should be checked by HPLC and LC/MS at many different stages of the process. Previous TLC analysis of the drug substance afforded two method solvent systems that were able to separate the desired impurities. The old conditions involved a methyl ethyl ketone (or methyl butyl ketone)/acetic acid/water solvent system. Hence, after experimentation, it was found that the solvent system of methyl ethyl ketone/water/acetic acid (80/19/1 by volume) afforded separation of the main band from the desired impurities and reduced the amount of acetic acid present. Using this new solvent system, the impurity band of interest was separated from the main band and the silica gel containing the impurity was removed from the plate by scraping with a spatula. Extraction of the compound from the silica gel was accomplished using acetonitrile. The isolated material was analyzed by HPLC, and the Impurity C isolate was identified.

14 SOLVING IMPURIT^DEGRADATION PROBLEMS: CASESTUDIES,^^^

387

A major concern was the presence of silica gel in the sample. It was found that the presence of silica contaminants in crude unfiltered solutions was complicating the LC/MS identification/confirmation of the presence of Impurity C. It was decided to attempt to remove the silica gel by filtering the extraction solution through a pad of celite. In a subsequent experiment, it was found that extraction of the impurities from the silica gel using methanol/dichloromethane, followed by removal of the silica gel by filtration through a short pad of celite afforded samples with very little organic material present. It was presumed that the organic impurity adhered to the celite and did not elute from the solid phase. Normally an organic chemist can remove solid contaminants such as silica gel by filtration through celite without any problems; however, the extreme difference in the scale involved in this case study complicated the procedure. Synthetic organic chemists normally deal with gram-to-milligram quantities of material, and sometimes techniques do not transfer from the milligram scale to the submilligram scale. Hence, the experiment was repeated and celite was eliminated in the removal of the silica gel from the sample. The extraction solution was filtered through a pin/head-sized plug of prewashed cotton wool wedged into a pipette. This example demonstrates the need to perform isolation and extraction experiments at a small scale before committing precious isolate samples and the need to check for the presence of the desired impurity/degradant at multiple stages along the isolation process. After completing the TLC method development, a concentrated solution of the drug substance was applied to 10 TLC plates (20 x 20 mm). The plates were eluted using 80% methyl ethyl ketone/19% water/1% acetic acid (using top phase only). The band of silica gel containing the impurity of interest was removed from the TLC plates by identifying the band under UV light and carefully scraping the band. The silica gel was extracted by stirring with methanol/dichloromethane followed by crude filtration through glass filter, then filtration through a short prewashed plug of cotton wool. The resultant clear, colorless solution was concentrated by evaporation to give a sample of the impurity for mass spectral analysis. Because of the weak intensity of the bands (more polar than the main band by TLC) according to UV light, a highly concentrated solution was applied in order to observe the impurity bands. LC/MS analysis of the sample was aided by the greater abundance and purity of the sample, and more sophisticated techniques were utilized to elucidate the structure. Impurity C showed an abundant doubly charged ion with a molecular weight consistent with addition of an extra nucleus. Based on this LC/MS evidence. Impurity C was proposed to originate by incomplete protection of a synthetic precursor. Further confirmation of structure by traditional NMR was not possible because of limited amount of sample. This project demonstrated the need to bring in new microsample NMR technologies. These technologies have been purchased and utilized successfully to solve subsequent degradation issues as demonstrated in some of the following case studies.

388

K.M. ALSANTE et al.

2. Case Study B.2: Isolation Using Normal-Phase Preparative HPLC and Characterization by LC/MS and N M R Characterization Within I Week

Drug Substance

Process-Related Impurity

For case study B.2, an impure starting material gave rise to an impurity at the end of the drug substance synthesis. An unknown impurity at the 0.3% level was discovered after process colleagues prepared the freebase form of the drug from the salt form. The time line for completion of the identification was 1 week and posed quite a challenge. There was no time available for lengthy HPLC method development and isolation, as well as challenging spectral structure elucidation experiments and data interpretation. As illustrated in this series of case studies, these factors are very difficult to predict at the beginning of the project. Because of the difference in polarity, normal-phase chromatography was identified as the technique of choice under the strict time line. A sample of the drug substance was analyzed by a normal-phase TLC technique already developed, which detected the impurity. A suitable solvent system for column chromatographic separation was developed based on the analytical TLC method (10% ethyl acetate/90% hexane). The drug substance sample was eluted on two large ( 2 0 x 2 0 cm) TLC plates using 10% ethyl acetate/90% hexanes as the mobile phase. The silica gel containing the impurity was removed from the TLC plates, and the compound of interest was extracted by washing with 1% triethylamine/99% acetonitrile. The material was analyzed by HPLC and found to be identical to the impurity of interest. The stability of the impurity on silica gel was investigated, and no problems were identified — another critical step after a lengthy isolation process. It is essential that the degradant does not further degrade in the concentration step. Up-front stability investigation is a worthwhile investment and can save valuable time reisolating an impurity or degradant, especially in cases in which the project is sample-limited. The impurity was then isolated at a larger scale by column (170 x 4 5 mm) chromatography on silica gel using 10% ethyl acetate/90% hexane as the mobile phase. Fractions containing mostly the required impurity were combined and concentrated by evaporation to give a clear, colorless oil (~0.9mg). By LC/MS, the isolated impurity showed an abundant doubly charged ion. The molecular weight suggested that one of the starting materials was contaminated by a dimer, which propagated through the synthesis. Less than 1 mg of degradant was isolated for NMR characterization. Hexanes present in the sample further complicated the analysis. Therefore, traditional 5-mm tube NMR would have been excessively time-consuming.

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES

389

A 3-mm NMR "microprobe" was used to obtain high-quality data in a fraction of the time. In addition, the enhanced sensitivity for mass-limited samples afforded by microprobe NMR technology enabled the acquisition of a carbon spectrum, v^hich would not have been readily achievable using a traditional 5-mm probe. This was advantageous for this project because of the large number of nonprotonated carbons present. The symmetry of the molecular structure led to pairs of magnetically equivalent carbon sites giving rise to single peaks in the carbon spectrum. Proton integration showed that the unusual ratio of a bridging methylene to the remaining groups was due to a single methylene group that joins the two identical ends of the dimer. An alternative proposed structure included an additional oxygen neighboring the bridging methylene, which was disproved by the observed proton and carbon chemical shifts in this molecular region. Simulations of the proton and carbon spectra (using ACDLabs^^ software) further supported the reported dimer structure. Based on NMR and MS analyses on the isolated impurity, an unexpected dimer structure was confirmed within the one-week time frame required. Ideally, one should allocate a larger amount of time to characterize such mass-limited samples.

3. Case Study B.3: Isolation Using Preparative HPLC: LC/MS and N M R Characterization Required to Differentiate Between Four Proposed Structures

Drug Substance

Process-Related Impurity

OH

OH CF3

jL-^^f^\^CF2H

* Y A bulk lot sample of drug substance contained an unknown impurity present at 0.96% in case study B.3. At the request of the project team, this impurity was isolated for structural identification within a 1-month time frame. Previous mass spectral data indicated that the impurity had a molecular weight of M-18. Two possible structures were proposed based on the mass spectral data: a dehydration product and a process-related -CF2H product. NMR data were required to differentiate between these proposed structures. Given the deadline, the most efficient technique for sample isolation was reversed-phase preparative HPLC isolation, scaling up the existing analytical reversed-phase method. With two or more possible structures, synthesis is typically too time- and resource-intensive. This would be twice the effort in that the CF2H and dehydration products would both need to be synthesized.

390

K.M.ALSANTEetal.

Potential M-18 Dehydration Impurities

Alternative Proposed M-18 impurity OH

CF3

Since the existing analytical method contained phosphate buffer, method conditions needed to be modified for preparative isolation requiring a volatile mobile phase (0.1% formic acid in methanol). The bulk drug substance lot was purified by preparative HPLC using the modified preparative chromatographic conditions. The component of interest eluted from 5-6 minutes and was collected over multiple runs. The fractions containing the impurity were combined and concentrated by evaporation to give a crude oil. The oil was further purified in a final analytical cleanup to afford 1 mg of sample for NMR analysis. Despite the mass limitations and other challenges faced, a full set of NMR experiments were acquired on the sample, using a 3-mm microprobe. Methanol was chosen as the NMR solvent-based on solubility determinations from the chromatography analysis. Unfortunately, this prohibited the detection of exchangeable protons, including N - H and O-H groups. It also yielded relatively poor proton spectral resolution. In addition, this differed from acetone used for the parent, hence complicating chemical shift comparisons of the proton spectra. Racemization at the chiral center for both the parent and the impurity led to pairs of resonances in the carbon spectra. Peaks arising from a sample contaminant were also apparent in both the proton and carbon spectra of the impurity. Signal-to-noise ratios for carbon resonances near the fluorinated group were low because of ^^C-^^F splittings. Carbon quartets observed in the parent spectrum for the fluorinated quaternary and two nearby methylenes appeared as triplets in the impurity carbon spectrum. This clearly demonstrated that the trifluoro group had become a difluoro group in the impurity. ^H and ^^C NMR and mass spectral data were consistent with the -CF2H product, and so the dehydration product was ruled out. This approach yielded the identification required within the time frame allotted. 4. Case Study B.4: Identification Using a Combination of Preparative HPLC and Synthesis Process-Related Impurities

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES

391

For case study B.4, a bulk lot sample of drug substance contained an unknown Impurity A present at 0.82% and an unknown Impurity B present at 1.12%. Identification of both impurities was requested within a 1-month time frame. It was confirmed that the impurities did not match any known standards. LC/MS data indicated that Impurity A had a molecular weight of M+60, suggesting propanol addition, and Impurity B had a molecular weight of M+32, suggesting methanol addition. The next step was to look into the synthesis to find possible propanol and methanol addition reactions. In this investigative step, it was found that the reagent 2-propanol was utilized in step 11 of the process synthetic route. For Impurity A, a process-enriched sample was not available; therefore, reversed-phase preparative HPLC was used to clean up the analyte. The drug substance bulk lot was purified by preparative HPLC, using a chromatographic method modified for isolation with a volatile mobile phase containing a formic acid buffer and methanol organic modifier. Impurity A was collected over multiple runs. The fractions were combined and concentrated by evaporation to give a crude oil that was further purified in a final analytical cleanup to provide a clean sample for NMR. HPLC reinjection was performed to remove residual peaks that were present and could complicate the NMR analysis. The isolate was dissolved in a minimum amount of solvent, and reinjected onto the semipreparative column. The resultant formic acid salt sample was converted to the freebase by an analytical reversed-phase HPLC cleanup using water followed by a water/methanol gradient as the mobile phase. This process afforded 1.9 mg of Impurity A. The molecular mass of the isolated Impurity A was confirmed to be 60 Da higher than the drug substance. This mass spectral confirmation on the isolate is critical prior to NMR analysis to assure that the correct peak has been isolated and that there are no changes that are due to contamination or degradant of the sample. Typically, this should be performed on the fraction collected from first preparative injections to confirm peak identity. NMR characterization was performed on approximately 1 mg of sample, which was particularly challenging because of contaminants present in the sample. Methanol was used as the solvent based on chromatographic solubihty studies. Exchangeable protons were thus not observable, although this was not deemed to be of consequence in this project. Comparison to the corresponding parent proton spectrum was somewhat difficult because of the solvent change from dimethyl sulfoxide. It was observed that most of the structure of the impurity was identical to the parent and that the acetylene of the parent had changed. The observation of a dimethyl methine resonance exhibiting a long-range proton-carbon correlation to a methine was key to the structural elucidation. The observed proton-proton coupling constant of 7 Hz for the protons on either side of a double bond pointed to a cis configuration, consistent with vinylic coupling constant values reported in the literature.^ For Impurity B, process colleagues synthesized methanol adducts of the drug substance, using a literature cesium hydroxide-catalyzed method.^ NMR analysis of the synthesized M + 3 2 indicated that the impurity was the methyl (versus isopropyl) vinyl ether analog of the drug substance. The

392

K.M. ALSANTE et al.

synthesis yielded a cis-trans mixture of enol ether isomers according to NMR and LC/MS. One of these isomers corresponded to the M H - 3 2 impurity. Based on NMR and mass spectrometry, Impurity B was identified as the methyl vinyl ether analog of the drug substance. 5. Case Study B.5: Preparative HPLC Isolation from a Retained Process Sample and Characterization Using LC/MS and N M R Characterization

Drug Substance

Process-Related Impurity

A bulk lot sample of the drug substance contained an unidentified process-related impurity present at 0.2% in case study B.5. Based on mass spectral data, the impurity had a molecular weight of M-h92, suggesting addition of phenol. Mass spectral fragmentation data suggested that phenol added in the left-hand portion of the molecule and the right-hand portion of the molecule was intact. At this low level, it was critical to look for an enriched sample. Review of forced degradation data had not produced this impurity. Fortunately, a retained process sample containing the desired impurity at the 2.6% level was available. This sample, enriched in the desired impurity, was used for isolation by preparative-scale HPLC. A suitable preparative HPLC method using 0 . 1 % formic acid in water and methanol afforded lOmg of the desired impurity. The NMR analysis confirmed the phenol addition proposed from mass spectral evidence. NMR spectral interpretation was complicated because of the absence of an expected long-range proton-carbon correlation peak pertaining to the complicated right-hand portion of the molecule. This suggested a potential inconsistency between the NMR and MS observations. Since a lack of an observation is not the preferred approach for a structural proof, the NMR experiments were repeated in an alternative solvent. The sample previously run in deuterated acetonitrile was dried and then redissolved in deuterated methanol. The new long-range heteronuclear correlation spectrum showed the anticipated proton-carbon coupling, hence supporting the proposed impurity structure. The original version of the NMR experiment in question was optimized for a standard long-range coupling constant of 6 Hz, as observed in the parent drug substance. However, the long-range coupling for the impurity structure was observed to be weaker than this, perhaps because of changes in molecular geometry. As noted, mass spectral analysis of the impurity suggested that the righthand portion was intact, and more in-depth NMR experiments proved consistent with this conclusion as well. The source of this impurity was subsequently investigated. Mass spectral analysis of the process intermediates indicated that the M-H92 impurity was present at the start of the synthesis,

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES

393

before the right-hand portion of the molecule is added to the molecule. This case study is a prime example that no technique can stand alone. N M R and mass spectrometry are critical techniques in structure elucidation, and any apparent discrepancies must be further probed. 6. Case Study B.6: MultidisciplinaryTeam Approach Using LC/MS, Process Enrichment, Scaled-Up Forced Degradation, Preparative HPLC, Synthesis, and N M R

Drug Substance

Process-Related Impurity O

HO

' \

Ar Ar

N' H

\r

A multidisciplinary team approach was used to conclusively elucidate the unknown impurity in case study B.6. A bulk sample of drug substance contained an unknown impurity at the 0.15% level. The identification required a number of techniques already discussed in this chapter (all incorporated into this case study): LC/MS, enrichment of a process sample, scale-up of forced degradation, preparative isolation, NMR characterization, and small-scale synthesis. The instability of the impurity under acidic chromatographic conditions presented a tremendous challenge in purifying and identifying the impurity. However, this acidic instability problem turned into an advantage, conducting a degradation study of the impurity that provided very rich information about its structure. Extensive LC/MS analysis suggested that the molecular weight was M + 3 7 1 . Several structures were proposed with a low degree of confidence; therefore, isolation and spectral characterization were required to conclude the exact structure. Isolation and identification of the impurity was complicated by the acid instability of the drug substance and impurity. Standard workup of the preparative HPLC fractions decomposed the impurity before spectral data could be recorded. A new high pH isolation strategy was developed. The drug substance main band was removed from the bulk drug substance containing the impurity by solid-phase extraction technology, and the impurity was isolated by preparative-scale HPLC. Attempts to directly isolate this impurity using preparative HPLC proved to be unsuccessful. SoHd-phase extraction technology was employed to remove the main band and more polar impurities resulting in a concentrated solution of the desired impurity. The NMR analysis of this compound was extremely challenging both experimentally and in data interpretation. Only 3 mg of sample were available, which was low given the relatively high molecular weight. Sample concentration was further limited by very poor solubility. The final solvent chosen was deuterium oxide, doped with a small amount of sodium hydroxide to adjust the pH to a more basic environment. The use of a protic solvent eliminated the ability to detect labile protons, including what proved to be a critical N - H group involved in a newly formed molecular bridge. The parent

394

K.M. ALSANTE et al.

species was run in dimethyl sulfoxide, so comparison of proton spectra between parent and impurity was not straightforward. Spectral resolution was not ideal because it was later shown to be a pseudo-dimer structure yielding very similar proton and carbon chemical shifts for the two halves of the molecule. Additionally, two-dimensional heteronuclear correlation experiments were not as useful as they may otherwise have been because of several consecutive quaternary carbons and heteroatoms. Sample integrity during NMR acquisitions was verified by acquiring a proton spectrum both initially and at the conclusion of the full set of experiments. No change in the spectrum confirmed that the sample remained intact. However, preliminary analysis of the NMR data supported a structure vastly different from the mass spectral results, conflicting significantly with the determined molecular mass. Since sample integrity during NMR acquisition was verified, the mass spectral experiments were rerun to evaluate data reproducibility. It was subsequently shown that the sample integrity was compromised by acidic chromatographic conditions used for the LC/MS experiments. This explained the apparent inconsistency in results and provided a clue of the molecular structure. In addition, running LC/MS on the NMR sample verified the presence of an exchangeable proton in the middle of the molecular structure, since the molecular weight of deuterium increased the fragment size in this region by one mass unit. Combined with observed NMR heteronuclear correlations and the MS "nitrogen count rule," it was determined that this was an N - H group. Combined interpretation of the NMR and mass spectral observations yielded the bulk of the molecular structure as a pseudo-dimer. However, the carbon NMR spectrum appeared to be "missing" an anticipated quaternary carbon resonance, based on the measured total mass. Removing the small degree of line broadening commonly used for processing carbon spectra revealed a distinct shoulder on one of the observed carbon resonances. This was the "missing" carbon and demonstrated the necessity of fully appreciating one's detection and resolution limitations. Analysis of the process synthetic pathway used to prepare this molecule afforded insight into the potential mechanism of formation of this processrelated impurity. It was observed that an early step in the synthesis of the molecule was forming a small amount of side product. It was subsequently observed that the side product was still present at later stages in the manufacture of the drug. This side product in turn reacted with one of the starting materials. The impurity that was formed at this stage was not the isolated impurity. This emerging impurity was modified by the subsequent steps of the synthesis to form the impurity that was finally isolated and identified. Early step in the synthesis: generation of the side product The reagents used in this step were designed to convert a functional group on the aryl ring and were not intended to react with the ester group. Starting Material

Desired Product

Side Product

O

0

0

A

- - ^ ^

Arx

^OEt

^

A

^

Ary

'

^

A

^OEt

Ary

^

^

NH

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES

395

Later step in the synthesis; the side product reacts with the precursor Side Product

Ary-^

^NH2

Precursor

HO

Coupled Precursor Impurity

Avy^

\

Ary

^^ H

<

\ry

This multidisciplinary team approach toward impurity identification was successfully applied to the identification of this impurity. Each and every discipline played a very important role. Without SPE enrichment to remove the drug substance from the impurities, the impurity could not be purified by preparative HPLC. The novel acidic degradation study of the impurity provided very valuable information of the structure of the impurity. Mass spectrometry and nuclear magnetic resonance spectroscopy were the ultimate tools in this structure elucidation. Furthermore, the formation mechanisms were concluded by a careful examination of the process. 7. Case Study B.7: Preparative HPLC on an Enriched Drug Substance Lot and Characterization by LC/MS and N M R

Drug Substance

Process-Related Impurity

A bulk lot sample of drug substance contained an unknown processrelated impurity at 0.4% in case study B.7. To meet project time lines, it was necessary to identify the impurity. Mass spectral data indicated that the impurity had a molecular weight of M + 1 2 (compared to the parent). LC/MS/ MS data indicated that the change was located in the sugar portion of the molecule. Isolation and NMR analysis was required to provide more structural information on the impurity. This again was a process-related impurity not observed in forced degradation studies; therefore, effort was placed on finding an enriched mother liquor sample. An enriched bulk lot containing 16% of the impurity of interest was identified and used to reduce the time required for isolation by a factor of 40. Isolation was required for further NMR analysis. The enriched bulk lot containing 16% impurity was used for isolation by preparative HPLC using 45 500-|aL injections. A suitable reversed-phase preparative HPLC method using a volatile mobile phase (0.1% formic acid in water and acetonitrile) was developed based on the analytical purity and potency assay. The fractions containing the impurity were combined and concentrated by evaporation. A final analytical cleanup was performed to remove salts and

396

K.M.ALSANTEetal.

solvent impurities for NMR analysis. The resultant formic acid salt sample was converted to the freebase by the approach used in case study B.4. Preparative scale reversed-phase HPLC using water followed by a water/ acetonitrile gradient as the mobile phase afforded 12 mg of the impurity. It was not possible to determine the nature of the counterion; therefore, X~ has been used to denote a counterion of unknown composition. The process-related impurity structure was confirmed by LC/MS and NMR spectroscopy. The mass spectral data pointed to an unusual M+12 total mass, implying an implausible "raw" carbon addition. Fragmentation patterns and comparison with the parent data showed the sugar ring to be the altered region of the molecular structure. NMR experiments were run on a mass limited sample, which was challenging because of the relatively high molecular weight. Deuterated methanol was chosen for high solubility and ease of removal. This also facilitated comparison of proton spectra between impurity and parent, since the parent was also run in methanol. Unfortunately, the proton and carbon spectra were quite complex, since virtually all of the resonances were clustered in the aliphatic spectral region. The two methyl groups were observed to be magnetically equivalent in the parent structure, yet were unequivalent in the impurity, demonstrating a change in the electronic environment at the attached nitrogen site. A new methylene resonance was observed, which accounted for the expected carbon addition. However, the total mass was not readily explained. Long-range heteronuclear correlations provided the necessary insight. Correlations were observed between the two methyl groups and the methine neighboring the nitrogen. The dimethyl entity also showed a correlation to the new methylene, establishing proximity between the two species. Finally, the new methylene showed a long-range correlation to the oxygenated methine, thereby confirming the unexpected closed-ring structure. Chemical shift changes compared to the parent spectra provided further support for this novel structure. 8. Case Study B.8: Preparative HPLC Isolation from Retained Process Samples and Characterization by LC/MS and N M R

Drug Substance R

I ^xxN^^"

"O

Process-Related Impurity R

I x-\s.vvNH

' "O

A bulk lot sample of drug substance in case study B.8 contained an unknown impurity at the 0.3% level. This unknown process-related impurity was less polar than the parent in the chromatographic assay. Based on mass spectral data, the impurity had a molecular weight of M + 4 , suggesting over-reduction of the benzene ring. Attempts to obtain mass spectral

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES

397

fragmentation were unsuccessful; therefore, isolation was required for structural identification. Considering the low level of impurity, it was critical to look for an enriched sample. A retained process mother liquor containing approximately 13% of the unknown impurity was prepared, and a suitable semipreparative HPLC method was developed for isolation of the unknown impurity from the mother liquor sample. Fractions containing the desired impurity were combined and concentrated by evaporation. The isolated material was purified in a final analytical cleanup to yield a 162mg of the desired impurity for NMR analysis. A complete NMR characterization was performed to elucidate the structure. Assignment of the ^H and ^^C spectra was able to determine the placement of the double bond in the partially saturated ring. 9. Case Study B.9: Conversion to Freebase and Preparative HPLC Isolation, Followed by LC/MS and N M R Characterization

Drug Substance

R^-^N^

Process-Related Impurities

R^-^N-^R^

R^-^N-^R,

A bulk lot sample in case study B.9 contained an unknown impurity present at 0.2%, and identification was required. An enriched sample was not available. In addition, the drug substance was a succinate salt. In deciding the isolation approach and in order to avoid contamination of the isolated impurity with succinic acid, the impurity was isolated from the freebase. To prepare the freebase, the succinic acid salt was dissolved in dichloromethane and treated with a 1 N solution of sodium hydroxide in water. The aqueous layer was removed, the organic layer was then washed with brine and dried with sodium sulfate; and the solvent was removed by evaporation to give a pale orange solid. The solid was dried in vacuo at room temperature overnight to give the freebase as a pale orange solid (6.15 g, 96%). HPLC analysis against a standard indicated that the sample was 99.4% freebase. A semipreparative HPLC method was developed for isolation of the unknown impurity from the freebase. Fractions containing the desired impurity were combined and concentrated by evaporation. The isolated material was purified in a final analytical cleanup to yield 25 mg of the desired impurity. LC/MS of the isolated impurity indicated the presence of two peaks, corresponding to a dimer-like degradant and the parent drug. A full set of NMR experiments was performed to elucidate the structure. Despite having an adequate amount of sample, sensitivity was limited by low solubility. Deuterium oxide as the solvent eliminated the abiUty to detect exchangeable protons, although this was not deemed to be a significant limitation for this structural elucidation. The NMR analysis confirmed that there were two primary components present in a ratio of approximately 2.5 to 1,

398

K.M. ALSANTE et al.

respectively. The spectra were consistent with the parent and the structures given above. III. SUMMARY AND CONCLUSIONS The process of characterization of impurities described in this chapter uses a designed approach for the isolation of unknown impurities and degradants in pharmaceutical drug substances. This approach focuses on efficiency, so that the success of data collection is maximized. The isolation of pure material is crucial when trying to identify the structure of an unknown impurity/degradant. Once the unknown has been isolated, it can be submitted for structure elucidation using mass spectrometry and NMR spectroscopy. Identification of degradants and impurities aids in the understanding of degradation mechanisms and impurity formation. Identification conducted at an early stage allows for improvement in the drug substance/drug product development process to prevent these degradants/impurities long before the filing stage. The multidiscipUnary team-based approach outlined in this chapter is most efficient for solving impurity/degradant problems. One can often be misled by a single technique, and multiple characterization techniques are essential. The following appendix summarizes key lessons learned in the process of isolating and characterizing degradant and impurities. APPENDIX—LESSONS LEARNED Isolation/Synthesis • UtiUze all available sources of information such as TLC, LC/MS, NMR archive data, project lab background, as well as process, formulation, and discovery chemistry. • Ask questions and conduct meetings when needed; even the projects that look easy can be hard. To collect all background data can be of significant value with tight time lines. • Find out what methodologies are in place. If a method without a volatile mobile phase is not available, development time is required which adds time to the identification. • The mobile phase used for isolation should be selected carefully to avoid salt complexation. • It is advised to perform isolation and extraction experiments at a small scale before committing precious isolate samples. • Shortcuts should not be taken in isolation efforts for NMR analysis. An analytical HPLC cleanup/desalting is required for NMR analysis. This improves the sample purity and quality of NMR data. It is especially difficult to characterize dimers and oxidations of heteroatoms if a sample is not of high purity. • Clean glassware is essential in the isolation process when isolating |ig/mg-sample amounts. Glassware impurities can contaminate precious isolation samples and complicate NMR analysis.

14 SOLVING IMPURITY/DEGRADATION PROBLEMS: CASE STUDIES

399

• Up-front stability investigation is a worthwhile investment and can save valuable time reisolating an impurity/degradant, especially in cases in which the project is sample-limited. • Attention should be given to HPLC and collection tubing selection as well: Tygon tubing that can leach phthalates should not be used in HPLC and fraction-collection equipment. • In a last step prior to NMR analysis, the isolated/synthesized sample should be washed with deuterated solvents several times to purify the sample for NMR. • Recovery is never what you expect it to b e . . . assume 50%. Also, the project completion time line is never what you expect it to be. Project teams need to be aware of this and notify the isolation/characterization experts as soon as possible to ensure there is adequate time to complete the project. • Use chemistry knowledge to assess the stability of a compound prior to isolation, and determine if special collection conditions may be necessary. • For preparative TLC and column chromatography, it can be critical to remove silica gel and other interferences in a filtration step. Mass Spectrometry

• For mass spectrometry-compatible mobile phases, volatile acid salts are acceptable (i.e., TFA, formic acid, acetic acid). The ammonium counterion works best. • TEA and other organic amines as mobile-phase additives tend not to work well and suppress ionization. • Less than I m g of sample can be analyzed, and with LC/MS, timeintensive isolation is generally not required. NMR Spectroscopy

• Deuterated solvents are required for NMR analysis. • For drug substance analysis, the freebase is preferable to avoid excessive salt levels, which complicate data interpretation. • Greater than 5-mg sample is preferred, but less than 1 mg is possible in some cases. A microprobe can be used for the less than 1-mg scale sample cases. • Nonprotic organic solvents minimize undesirable proton exchange (i.e., CDCI3, d6-DMSO). • Molecular weight and proposed structures based on LC/MS analysis expedite the NMR structure elucidation.

REFERENCES 1. Alsante, K. M., Friedmann, R. C , Hatajik, T. D., Lohr, L. L., Sharp, T. S., Snyder, K. D., Szczesny, E. J. Degradation and Impurity Analysis for Pharmaceutical Drug Candidates. In Handbook of Modern Pharmaceutical Analysis (Ahuja, S. and Scypinski, S., eds.), Academic Press, 2001.

400

K.M. ALSANTE et al. 2. International Conference on Harmonisation, "Draft Revised Guidance on Impurities in New Drug Substances," Federal Register 65(140):45085-45090, 2000. 3. International Conference on Harmonisation, "Draft Revised Guidance on Impurities in Ntw Drug Products," Federal Register 65(139):44791-44797, 2000. 4. ACDLabs™ Software. 5. Russell, G. A. Deuterium-isotope Effects in the Autoxidation of Aralkyl Hydrocarbons. Mechanisms of the Interaction of Peroxy Radicals. / . Am. Chem. Soc. 79:3871, 1957. 6. Bellucci, G. and Chiappe, C. Crown Ether Catalyzed Stereoselective Synthesis of Vinyl Ethers in a Solid Liquid Two-Phase System. Synlett. 880, 1996. 7. Knochel, P. Cesium Hydroxide Catalyzed Addition of Alcohols and Amine Derivatives to Alkynes and Styrene. Tetrahedron Lett. 40:6139, 1999.