METHODS OF ANALYSIS Introduction

When developing a method for stress-testing studies, it is useful to think about what would constitute an ideal method. An ideal method would enable the accurate quantifi cation of the parent compound as well as all of its degradation products. While this sounds simple in theory, it is nearly impossible to develop an ideal method early in the development cycle of a new drug when few if any of the potential degradation products are known. After all, one of the reasons for conducting stress-testing studies is to discover the potential degradation products. The ideal chromatographic method will resolve all degradation products from the parent as well as from each other, all degradation products will be detected, and the relative response of the degradation products with respect to the parent will be known. Since it is diffi cult to develop a chromatographic method that satisfi es all of these requirements, the focus should be on devel- oping a primary screening method that has the highest likelihood of resolving and detecting a diverse set of degradation products. In many cases, it may be impossible to come up with an ideal method; however, the use of two “orthogonal” analysis methods often will give adequate results in these cases.

Specifi c versus Generic Methods

Chromatographic methods used for analysis of stress-testing samples can be developed for each specifi c drug substance or product being stressed or they can be “generic” methods, which are used for analysis of samples from many drug substances. The major advantage to using a generic method is the reduced amount of method development time for each compound. This is something that can be particularly important for laboratories that conduct stress testing on many compounds such as early-phase development labs. Some of the disadvantages include a less-specifi c method and the greater likelihood that major degradation products may be missed. If one is developing a generic HPLC method for stress testing, gradient elution will almost certainly be required since different drug substances likely will have different polarities. In our experience, we have found that just a few reversed-phase methods work for greater than 80% of the compounds for which we have conducted stress-testing studies. These generic method conditions are outlined in Table 10 . Refer to the reversed-phase HPLC discussion later in this chapter for more information on selection of chromatographic parameters.

Figure 5 HPLC related substances chromatograms (UV detection, 205 nm) obtained on solutions of LY334370

hydrochloride in water exposed to intense fl uorescent light for 14 days ( bottom ) or simulated sunlight produced by a xenon arc lamp ( top ).

AU 0.00 0.10 0.20 0.30 0.40 0.50 Minutes 5.00 10.00 15.00 20.00 25.00 30.00 35.00 E 9.5% 4.4% J 5.1% 0.5% I 2.4% 0.7% K 0.2% 0.8% F 0.5% D 0.6%

Validation of Methods

According to the ICH guideline on validation of analytical methods ( 5 ), the objective of valida- tion of an analytical procedure is to demonstrate that it is suitable for its intended purpose. The reader should keep in mind that stress-testing methods are screening methods to be used to help understand the degradation chemistry of a drug and therefore do not need to (nor, in gen- eral, can they) be validated to the extent of fi nal control methods. In addition, stress-testing methods are usually only used in a limited number of laboratories without a formal method transfer. The overall validation should be signifi cantly abbreviated when compared to the vali- dation of fi nal control methods, since stress-testing methods are investigational methods. The concepts in the ICH guideline on validation of analytical methods are a good starting point for validation of stress testing methods. The ICH guideline gives parameters to be considered when validating methods. These parameters include accuracy, precision, specifi city, detection limit, quantitation limit, linearity, and range. All of these parameters should be addressed to some extent when validating stress-testing methods; however, the overall validation should be kept to a minimum since stress-testing methods are investigational methods.

Accuracy normally should not be a problem with stress testing methods as long as the method is linear and samples are completely dissolved prior to analysis. The specifi city of methods cannot be fully validated since one normally does not know all of the possible degra- dation products during initial stress testing. Specifi city can be addressed by using any known impurities and the degradation products produced in the method development samples. Preci- sion (repeatability) of the assay of the main component can be evaluated by preparing a limited number of assay samples (e.g., 5 to 10) and using simple statistics to estimate the standard deviation. Estimation of intermediate precision and reproducibility should normally not be necessary for stress testing methods. Detection and quantitation limits for degradation prod- ucts can be determined by using the parent compound and assuming that the responses of degradation products will be similar. Although there is no requirement to reach any specifi c detection limit a reasonable goal is 0.1%, since the goal of stress testing is to detect the major degradation products in samples ∼10–20% degraded. The linearity of the method should be validated over ranges for both assay and impurity determination. A typical assay range might be from 50% to 110 % of nominal sample concentration while a typical range for impurity deter- mination might cover a range from the quantitation limit to a few percent. If one wishes to quantitate impurities versus the parent peak then linearity (range) should be demonstrated from the quantitation limit to at least 100% of nominal sample concentration.

Chiral Molecules

Chiral analysis is required for drugs being developed as single enantiomers possessing a single chiral center capable of undergoing racemization. Molecules that have multiple chiral centers may not necessarily require a chiral method since racemization of one of the chiral centers will result in the formation of a diastereomer. Diastereomers can typically be resolved on achiral chromatography systems, while chiral impurities from a molecule with a single chiral center

Table 10 Generic Reversed-Phase HPLC Conditions

Column C₈ or C₁₈

Weak solvent 0.1% trifl uoroacetic acid in water

Phosphate buffer, pH 2–3 for acids or pH 6–7 for bases

Ammonium acetate, pH 4.5 to 5.5 (works well when using MS detection) Strong solvent Acetonitrile or methanol

Column C₈ or C₁₈

Detection UV-PDA

Elution Linear gradient adjusted such that the parent elutes near the middle of the chromatographic run

will require a dedicated chiral method for analysis ( 6 ). Aside from chiral HPLC methods, chiral capillary electrophoresis (CE) is a commonly used technique ( 7 ).

Chromatographic Methods

There are numerous analytical techniques that can be used to analyze stress test samples. Some of the more common ones include reversed-phase HPLC, normal-phase HPLC, thin layer chro- matography (TLC), CE, and gas chromatography (GC). The following paragraphs contain brief discussions on the use of these techniques for analysis of stress test samples.

Reversed-Phase HPLC

Since a large majority of pharmaceutical products are amenable to reversed-phase HPLC, this is usually the method of choice. The development of reversed-phase HPLC [or ultra high pres- sure liquid chromatography (UHPLC)] methods is a broad subject with many research articles and books devoted to it and it is not practical to try to cover this topic in depth in this chapter. There are, however, some major points to consider when developing an HPLC/UHPLC method for analyzing stress test samples ( 8 ).

HPLC versus UHPLC

UHPLC is becoming the standard in many modern analytical laboratories. UHPLC systems allow for rapid analyses (typically 5 minutes or less) on smaller columns with smaller (sub 2 µm) particle sizes. These systems can handle the higher pressures associated with smaller particles while also delivering higher numbers of theoretical plates ( 9 ). Although UHPLC requires different hardware, the development of UHPLC methods is really no different than the development of HPLC methods and the same factors must be considered.

Isocratic versus Gradient Elution

Since one does not know what degradation products will form, gradient elution should be used. This signifi cantly increases the chances that degradation products, which are much more polar than the parent compound will be pulled away from the solvent front and those which are much less polar than the parent will elute from the column. Often the use of a multistep gradient is particularly benefi cial. The fi rst segment of the multistep gradient starts at very low organic concentration and ramps rapidly to the second segment in which the parent is eluted either under essentially isocratic conditions or with a very shallow gradient in order to maxi- mize resolution of the degradation products from the parent compound. The third segment of the gradient begins after the parent elutes and is a rapid ramp to high organic concentration to elute any less polar degradation products. The following example illustrates the benefi ts of using gradient elution for stress testing. Figure 6 illustrates a chromatogram obtained under isocratic conditions on a sample of drug substance in 50/50 acetonitrile/pH 7 phosphate buffer stressed at 70°C for 14 days. The assay result of 95.1% indicates that signifi cant degradation has occurred. The related substances result of only 0.7%, however, suggests a potential mass bal- ance problem. Reanalysis of the same sample using gradient elution enabled the detection of a number of degradation products not visible with the isocratic method ( Fig. 7 ).

Column Selection

For the majority of pharmaceutical products, C ₈ or C ₁₈ , bonded phases will give adequate separations and one of these two stationary phases should be tested before moving to other stationary phases. Sometimes improved selectivity of the separation can be achieved using other nonpolar stationary phases such as phenyl. Phenyl stationary phases are used since these phases can provide both hydrophobic interactions as well as pi–pi stacking interactions with double bonds and aromatic groups. Occasionally, polar analytes cannot be adequately retained on C ₁₈ or C ₈ phases. In these cases the use of mixed-mode, polar-embedded stationary phases can be very useful. Polar embedded phases contain a polar group that is embedded

within the nonpolar ligand chain. Common polar groups used in these phases include ethers, acrylates, carbamates, amides, and urea. The polar-embedded phases also can be useful for obtaining separation of compounds that cannot be separated using conventional C ₈ or C ₁₈ stationary phases.

Mobile Phase Selection

There are a number of factors to consider when selecting a mobile phase. For compounds with ionizable functional groups, it is important that the pH of the mobile phase is con- trolled. The type of mobile phase additive used depends upon the type of detector used.

Figure 6 Isocratic HPLC chromatogram obtained on a sample of drug substance stressed in pH 7 phosphate

buffer/ACN. AU 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 Minutes 5.00 10.00 15.00 20.00 25.00 30.00 35.00 40.00 Assay = 95.1% Related substances = 0.7% A B C

Figure 7 Gradient HPLC chromatogram obtained on a sample of drug substance stressed in pH 7 phosphate

buffer/ACN. AU 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 Minutes 5.00 10.00 15.00 20.00 25.00 30.00 35.00 40.00 Assay = 95.1% Related substances = 4.3% Gradient artifact A B D C E F G

Detectors such as MS and CAD require the use of volatile mobile phase additives. Some of the more common volatile mobile phase additives include formic acid, trifl uoroacetic acid, acetic acid, ammonium acetate salts, ammonium bicarbonate, and ammonium hydroxide. These mobile phase additives also work with UV or PDA detectors as well although some of them tend to obscure the lower wavelength range due to absorbance in the 200 to 220 nm range. When there is a desire to obtain chromatograms at lower wavelengths, phosphate buffers are a good choice due to their very low UV cutoffs. In addition, phosphate has three p K values and possesses good buffering capacity at low, neutral, and basic pH values. The organic component of the mobile phase is typically either acetonitrile or methanol. Acetoni- trile offers some minor advantages over methanol like a lower UV cutoff and lower viscosity, but methanol typically works nearly as well.

Detection

There are many types of HPLC detectors available today, with the most popular ones includ- ing UV and UV-photodiode array (PDA), fl uorescence, refractive index, evaporative light scattering (ELSD), charged aerosol (CAD), and the mass spectrometer. Of these, the most com- monly used detector for pharmaceutical analytical methods is the UV detector since a major- ity of pharmaceutical compounds have some type of chromophore. Multiple detectors in series can also be utilized in order to obtain more information per chromatographic run. For example, a PDA detector can be combined with a mass detector to give both UV and mass spectral information on impurities ( 10 ). The use of a mass detector (LC-MS) contributes to the structure-elucidation process for degradation products in that structures may be proposed based on mass. For compounds that do not have any UV absorbance, the ELSD, mass spectro- metric, and CAD detectors are generally the most useful and can be incorporated individually or in series.

The CAD is a detector that is described as a nearly universal detector for nonvolatile and semivolatile compounds, with a response that is refl ective of the total mass passing through the detector and has been shown to be useful for the determination of response factors ( 11 ). Fluorescence detection is not desirable because there is no guarantee that the degradation prod- ucts of a fl uorescent molecule will fl uoresce. Most refractive index detectors are best suited for isocratic elution and are therefore not particularly useful for stress-testing methods that utilize gradient elution. Since the UV detector is the most widely used, the remainder of this discussion will focus on the aspects of UV detection. The use of UV-transparent buffers and organic modi- fi ers (e.g., phosphate buffers and acetonitrile or methanol) for the HPLC mobile phase is desir- able since it enables chromatograms to be acquired at relatively low wavelengths near 200 nm. Typically, monitoring at these low wavelengths increases the likelihood that all of the degrada- tion products will be detected since most compounds possessing a chromophore will absorb at these low wavelengths. The use of a PDA-UV detector signifi cantly increases the amount of information obtained from the chromatographic run. The use of the PDA detector allows for the extraction of chromatograms at multiple wavelengths from a single chromatographic run as well as providing the UV spectra of individual peaks. These UV spectra can be used to correlate peaks arising from different stress conditions or from different chromatographic systems. The use of a PDA detector also enables the determination of the UV homogeneity of individual peaks, thereby giving an indication of their purity.

Normal-Phase HPLC

Normal-phase HPLC is a good complementary technique to reversed-phase HPLC in that it often gives different selectivity. It is also more effective in separating geometric isomers than reversed-phase HPLC. The main problem with normal-phase HPLC is that aqueous samples are not normally compatible with the technique. Since many of the stress-testing samples contain water, normal-phase HPLC is rarely used as the primary analytical technique for stress test samples. Nonetheless, normal phase can be a useful complementary technique to

reversed-phase HPLC. A detailed discussion on the development of normal-phase HPLC methods is beyond the scope of this chapter.

Thin-Layer Chromatography (TLC)

TLC is one of the oldest chromatographic methods and is widely used in the pharmaceutical industry. TLC and high-performance TLC (HPTLC) are complementary to reversed-phase HPLC. TLC and HPTLC are typically carried out under normal-phase conditions and therefore can be very useful for separating impurities that cannot be easily separated under reversed phase conditions. One signifi cant advantage of TLC is that detection is carried out on the entire plate following development. This ensures that all of the impurities can be detected whether or not they migrate from the origin, as long as they are separated from the parent and the correct visualization technique is used. Another signifi cant advantage of TLC is the possibility of run- ning multiple samples in parallel rather than running sequentially as is done in HPLC. Some of the disadvantages of TLC include generally decreased sensitivity, resolving power, and the ability to provide accurate quantifi cation as compared to HPLC. For additional details on TLC and HPTLC, see one of the many literature references ( 12 ).

Capillary Electrophoresis (CE)

CE is another complementary technique to reversed-phase HPLC. Since there are a signifi cant number of resources available describing CE, a detailed discussion of CE will not be pre- sented. Although a number of detectors are available for CE, the most useful detector for analysis of stress-testing samples is the UV detector. Commercial CE instruments are also available with a PDA detector, which helps when correlating peaks between CE and HPLC. A number of articles have been published describing the use of CE for detection of pharmaceuti- cal impurities ( 13 ). A signifi cant amount of work has been done to develop “generic” CE methods that can be used for a wide variety of compounds. For basic solutes, Altria has sug- gested the use of a phosphate buffer at pH 2.5 ( 14 ) and for acidic solutes a borate buffer at a pH of 9.3 ( 15 ). Hilhorst has suggested a MEKC strategy for impurity profi ling which involves the use of an SDS system and a CTAB system ( 7a ). Analysis of samples on these two systems guarantees, in principle, that all compounds will pass the detector in at least one of the two systems. Altria has also developed a generic MEKC method utilizing lithium dodecyl sulfate and beta cyclodextrin ( 16 ). One of the major benefi ts of CE is that its separation mechanism is different from that of HPLC and it will often give different selectivity. This is illustrated in the analysis of a partially degraded drug sample ( Fig. 8 ). The drug contains two carboxylate groups and is therefore negatively charged under the CE analysis conditions. The top trace is a reversed-phase HPLC chromatogram and the bottom trace is a CE electropherogram. The analysis conditions are given on the fi gures. The peaks detected using the two techniques were correlated by comparing UV spectra obtained using PDA detectors. Clearly, the two techniques give signifi cantly different selectivities.

Gas Chromatography (GC)

GC is a good choice for analysis of volatile drug substance stress test samples. It is also a com- plementary technique to HPLC when volatile degradation products are suspected. For exam- ple, see the LY297802 example described in chapter 6, where volatile and nonchromophoric degradation products were missed by the HPLC-UV detection scheme but readily detected by extraction and analysis using GC with a fl ame ionization detector (FID). The FID is usually the GC detector of choice for analysis of stress test samples since it is a nearly universal detector for carbon-containing compounds and has the required sensitivity. The mass spectrometer is another widely used GC detector and can give structural information on the peaks as they elute from the column. One major difference between HPLC and GC is the typical requirement of an internal standard for quantitative analysis using GC.

CONCLUSIONS

Stress testing is an important part of the drug development process as it provides knowledge about the degradation chemistry of drug compounds. This knowledge is used primarily to develop stability-indicating analytical methods but is also useful for other purposes such as formulation development, package development, and the design of offi cial stability studies. Very little formal guidance is available for the design and execution of stress-testing studies