The Applicability of Universal Detection in the Pharmaceutical Industry

MSc Chemistry

Analytical Sciences

Literature Thesis

The Applicability of Universal Detection in the

Pharmaceutical Industry

by

André Koelewijn

August 2014

Supervisor:

dr. W. Kok

i

THE APPLICABILITY OF UNIVERSAL DETECTION IN THE PHARMACEUTICAL INDUSTRY

A Literature Study to the Graduate School of the University of Amsterdam

in partial fulfillment

of the requirements for the degree of

MASTER OF SCIENCE

August 2014

Chemistry, Analytical Sciences

by

ii

© Copyright by André K. Koelewijn 2014

All Rights Reserved

THE APPLICABILITY OF UNIVERSAL DETECTION IN THE PHARMACEUTICAL INDUSTRY

Author: André K. Koelewijn, BASc. Abbott Healthcare Products B.V. Supervisor: dr. Wim Kok University of Amsterdam

iii

Abstract

In the absence of suitable reference materials for impurity quantitation, laboratories have developed techniques using mass detectors such as the evaporative light scattering detector, the charged aerosol detector (CAD) and the chemical luminescence detector (CLND), to normalize the UV response of each impurity of interest by their molar ratios and thus generate relative response factors without requiring isolated and purified compound-specific standards.

Quantitative NMR has been gaining acceptance as a tool for relative response factor determination and can be considered a “universal” detector. While effective, q-NMR is limited in applicability and sensitivity. Alternatively, ELSD, CAD and CLND are very promising detectors with respect to “universal” detection. These detectors are affordable and when able to determine a relative response factor with satisfactory accuracy, it is a very cost efficient alternative to synthesizing and characterizing reference standards.

In this literature study, the background of the different detection techniques is presented in relation to universal response. Focus has been put on pharmaceutical analysis.

iv

Table of Contents

1.1 DETERMINATION OF RELATIVE RESPONSE FACTORS ... 1

2 UV-VISIBLE DETECTORS ... 3

2.1 RELATIVE RESPONSE FACTORS ... 5

3 EVAPORATIVE LIGHT SCATTERING DETECTOR (ELSD) ... 6

3.1 UNIVERSAL RESPONSE ... 7

3.2 OPERATING PRINCIPLES ... 7

3.2.1 Nebulizing and evaporation ... 7

3.2.2 Light Scattering ... 9

3.3 CHARACTERISTICS OF ELSD RESPONSE ... 10

4 CORONA CHARGED AEROSOL DETECTOR (CAD) ... 12

4.1 OPERATING PRINCIPLE ... 13

4.2 THE RESPONSE OF CAD SYSTEMS... 14

4.3 THE EFFECT OF ADDITIVES TO THE MOBILE PHASE ... 15

4.4 REDUCING COLUMN BLEED IN MS,CAD AND ELSD DETECTION ... 16

4.5 MOBILE PHASE COMPENSATION IN ELSD AND CAD ... 17

5 CHEMILUMINESCENT NITROGEN DETECTOR (CLND) ... 18

5.1 OPERATING PRINCIPLES ... 19

5.2 THE RESPONSE OF THE CLND DETECTOR ... 20

5.3 EQUIMOLARITY OF CLND RESPONSE ... 22

6 SELECTED APPLICATIONS OF CAD ... 23

6.1 UNIVERSAL QUANTITATION OF PDE-5 INHIBITORS USING LC-CAD ... 24

6.2 ANALYSIS OF DRUGS WITHOUT A NATURAL UV CHROMOPHORE ... 26

7 SELECTED APPLICATIONS OF CLND ... 27

7.1 ANALYSIS OF STREET DRUGS WITHOUT THE USE OF PRIMARY REFERENCE STANDARD ... 27

v

7.3 QUANTIFICATION OF DRUGS IN PLASMA AND WHOLE BLOOD BY HPLC-CLND ... 28

8 APPLICATIONS FOR DETERMINING RELATIVE UV RESPONSE FACTORS WITHOUT THE USE OF A REFERENCE STANDARD ... 28

8.1 DETERMINATION OF RELATIVE RESPONSE FACTORS WITH UV AND CAD ... 28

8.2 DETERMINATION OF RELATIVE RESPONSE FACTORS WITH UV AND CLND ... 30

9 CONCLUSION ... 31

10 BIBLIOGRAPHY ... 33

List of Tables and Figures TABLE 1: PARAMETERS FOR LC-CAD METHOD. LINEAR AND POWER FIT EQUATIONS, LOD AND LOQ FOR SILDANAFIL AND TADALAFIL.DATA TAKEN FROM REFERENCE 101 ... 26

TABLE 2:RELATIVE RESPONSE FACTORS OF PACLITAXEL-RELATED IMPURITIES.DATA TAKEN FROM REFERENCE 60 ... 30

TABLE 3:RRF AT 230 NM AS DETERMINED FROM KNOWN COMPOUND WEIGHTS OR FROM CLND PEAK AREAS.DATA TAKEN FROM REFERENCE 1 ... 31

FIGURE 1:SCHEMATIC OF A UV DETECTOR (LEFT) AND A PDA DETECTOR (RIGHT).REPRODUCED FROM REFERENCES 15,16 ... 4

FIGURE 2:SCHEMATIC REPRESENTATION OF A LIGHT GUIDED DETECTOR FLOW CELL.REPRODUCED FROM REFERENCE 17 ... 5

FIGURE 3:SCHEMATIC DEPICTION OF THE MAIN STEPS OF THE ELSD OPERATION.DESIGN (A)ELSD TYPE A AND (B)ELSD TYPE B. REPRODUCED FROM REFERENCE 23 ... 8

FIGURE 4:APPARENT (AREA%) LEVELS OF AN IMPURITY IN A MAJOR COMPONENT THAT WOULD BE OBTAINED BY HPLC USING ELSD.THE ASSUMPTION WAS MADE THAT BOTH MAJOR AND MINOR COMPONENTS HAVE IDENTICAL VALUES OF RESPONSE FACTOR A AND COEFFICIENT OF REGRESSION B.THE EFFECT OF THE VALUE B IS SHOWN.REPRODUCED FROM REFERENCE 44 .. 12

FIGURE 5:SCHEMATIC DEPICTION OF A CAD DETECTOR.REPRODUCED FROM REFERENCE 53 ... 14

FIGURE 6:CHROMATOGRAMS WITH SUPERIMPOSED ORGANIC MODIFIER COMPOSITION OBTAINED FOR THE SEPARATION OF SIX SULFONAMIDES.(A)UV SIGNAL IN NORMAL GRADIENT RUN;(B)CAD SIGNAL WITHOUT MOBILE-PHASE COMPENSATION IN A GRADIENT RUN; (C) CAD SIGNAL WITH MOBILE-PHASE GRADIENT COMPENSATION. PEAK IDENTIFICATION: 1, SULFAGUANIDINE; 2, SULFAMERAZIN; 3, SULFAMETHAZINE; 4, SULFAMETHIZOLE; 5, SULFAMETHOXAZOLE; 6, SULFADIMETHOXIN.REPRODUCED FROM REFERENCE 51. ... 18

FIGURE 7:SCHEMATIC OVERVIEW OF CLND INSTRUMENTATION.REPRODUCED FROM REFERENCE 1 ... 20

FIGURE 8:PROPOSED EXCEPTIONS TO CLND EQUIMOLAR RESPONSE RULE.REPRODUCED FROM REFERENCE 92 ... 23

FIGURE 9:CHROMATOGRAMS OBTAINED FOR THE MIXTURE SILDANEFIL AND TADALAFIL 10 µGML-1 EACH BY DIFFERENT TYPES OF DETECTOR: A)UV, B)MS POS/NEG, C)CAD.REPRODUCED FROM REFERENCE 101 ... 25

vi

FIGURE 10:CAD AND UV RESPONSE FACTORS OF PACLITAXEL-RELATED IMPURITIES.REPRODUCED FROM REFERENCE 60 ... 29

FIGURE 11: MOLECULAR STRUCTURES A) ACETAMINOPHEN, B) 2-QUINOXALINOL, C) 2-HYDROXYQUINOLINE, D) DANSYL -PHENYLALANINE.REPRODUCED FROM REFERENCE 1 ... 31

vii

Glossary

BP British Pharmacopoeia

CAD Corona Charged Aerosol Detector CLND Chemiluminescent Nitrogen Detector

CLNSD Condensation Nucleation Light Scattering Detector ELSD Evaporative Light Scattering Detector

HPLC High Pressure Liquid Chromatography KF Karl-Fisher

LLSD Laser Light Scattering Detector MALS Multi Angle Light Scattering MS Mass Spectrometry

NCE New Chemical Entity

NMR Nuclear Magnetic Resonance Ph.Eur European Pharmacopoeia PDA Photo Diode Array RI Refractive Index

RRF Relative Response Factor RSD Relative Standard Deviation TGA Thermogravimetric Analysis USP United States Pharmacopoeia UV Ultra Violet

1

1 Introduction

There is an urgent need for detection technologies that enable accurate and precise quantification of solutions containing small organic molecules in a manner that is rapid, cheap, non-labor-intensive, readily automated, and without a requirement for specific analysis standards.

Modern pharmaceutical research programs produce and screen hundreds of thousands of new chemical entities (NCEs) per year. Before they can be used either as synthetic intermediates or for initial biological testing, NCEs must be shown to have a minimal standard of purity to ensure they are fit for the business purpose. At the time of initial synthesis (often sub-milligram amounts), the identity of many impurities will be unknown and so it will not be possible to economically establish their response factors for typical HPLC detectors used in high-throughput analysis mode. In these circumstances, purity assessments are often made on the basis of relative “Area Percentage” using detectors such as UV, ELSD, and MS, in the knowledge that there is potential for response factors of unknown impurities to differ significantly from those of each other and of the target compound. Accordingly, purity assessments are inevitably compromised to some degree.

1.1 Determination of Relative response factors

In pharmaceutical analysis, the determination of drug purity generally includes quantitation of all detected peaks based on relative peak areas obtained at a specific wavelength. In order for such quantitation to accurately reflect weight percentages of impurities, the relative responses (absorbance per unit weight) at the given wavelength must be similar. If, as is often the case, the response factors differ significantly between impurities and the parent compound, then correction factors need to be applied to the impurity peak areas in order for them to be correlated to weight percentages. Thus, the determination of relative response factors (RRF) of impurities is integral for assessing purity of a given sample by HPLC-UV.

Traditionally, the process for establishing response factors involves the use of ‘standards’, or isolated samples of individual impurities, from which accurate concentrations can be prepared. The purity of each such sample must, therefore, be known. This is a relatively straightforward procedure if such impurities can be readily synthesized and recrystallized. The purity of these synthetically-prepared samples is generally estimated using a combination of HPLC (with UV, light-scattering, or other appropriate detection), nuclear magnetic resonance (NMR), and some method

2

to determine volatile impurities (e.g. TGA, Karl Fischer). The process often requires assumptions about the identity and levels of contaminants that may be present, and involves a significant amount of time and effort (i.e. expense). The effort required is typically even greater for impurities for which synthetic samples are not readily available (e.g. low-level process impurities and degradation products). Such impurities need to be isolated and purified using standard techniques such as preparative TLC or HPLC. The risk of having adventitious contaminants in the isolated impurities is greater than for synthetic samples because of the large amounts of solvents used, the possibility of non-chromophoric contaminants (e.g. solvents or column bleed), the presence of counter-ions (e.g. trifluoroacetic acid (TFA), acetate, etc.), and the impracticality of using crystallization (because of the low amounts isolated) to enhance the purity. Moreover, in many cases the impurities are unstable and thus very difficult to purify. In order to determine sample purity, amounts of 50 mg or more are typically needed. Isolation of these amounts of impurities can be very time-consuming and costly. Thus, a simpler method for determining response factors that avoids the need to isolate and characterize impurity samples would save considerable time and effort 1_.

A potential alternative for determining UV response factors would use two HPLC detectors: a standard UV absorbance detector, and a detector that has a response proportional to weight. For example, if a second detector could provide accurate information on the relative amounts of the impurities and parent compound in a sample mixture, then this information, combined with the UV peak areas, would supply the desired RRF information without the need for a purified impurity sample. Unfortunately, most HPLC detectors do not provide accurate quantitative information unless a standard of the given analyte is available for comparison. Some ‘universal’ HPLC detectors will give a response for most compounds, but that response is not uniformly related to weight. For example, evaporative light scattering (ELSD) allows detection of most non-volatile substances, but the detector response can be highly variable because it depends on the quantity and nature of the particles produced upon the desolvation process occurring in the detector. Only for compounds of very similar structures can one expect similar responses per unit mass by ELSD, and even then, the variability is 10–20%. For compounds of widely varying structures, charges, or vapor pressure; or for varying mobile phase compositions (e.g. gradient HPLC), the ELSD response can vary markedly 2-7_{. Similarly, mass spectral (MS) detectors are universal, but the response per unit weight depends} greatly on the ionization type (e.g. electrospray, atmospheric-pressure chemical ionization, etc.) and on the ionization efficiency of the analyte under the given conditions. Refractive index is

3

another universal detector, but it too suffers from variability in response depending on the mobile phase composition, temperature, and dissolved gases; furthermore, it is relatively insensitive 4, 8_.

Also the use of LC-NMR, which provides additional useful information for identification and characterization of the degradation products 9-12_{, is employed to a lesser extent due to the higher} analyte concentration required for routine use.

These days a great number of articles have been published dealing with these challenges. Three types of detectors, ELSD, CAD and CLND have been focused on in this literature study in particular because they are the choice of detector in the most promising publications with respect to ‘universal’ detection. The next chapters will discuss these detectors in depth to explain the challenges related to each detection technique and the correspondence to relative UV response factors. As for pharmaceutical analysis, the UV detector remains the preferred type of detector for multiple reasons but especially because it is well-known, economic and very robust. The latter is very important when analytical methods need to be transferred.

2 UV-Visible Detectors

The UV-vis absorbance detector is the most common HPLC detector in use today because many compounds of interest absorb in the UV (or visible) region (from 190 to 600 nm). Sample concentration, output as absorbance, is determined by the fraction of light transmitted through the detector cell by Beer's law:

𝐴 = log(𝐼0⁄ ) = 𝜀𝑏𝑐 𝐼

Where 𝐴 is absorbace, 𝐼0 is the incident light intensity, 𝐼 is the intensity of the transmitted light,

𝜀 is the molar extinction coefficient of the sample, 𝑏 is the path length of the cell in cm and 𝑐 is the molar sample concentration.

UV absorbance occurs as a result of the transition of electrons from 𝜋 → 𝜋∗_{, 𝑛 → 𝜋}∗ _{or 𝑛 → 𝜎}∗

molecular orbitals; most aromatic compounds absorb strongly at or below 260 nm, compounds with one or more double bonds (for example, carbonyls and olefins) at ~215 nm, and aliphatic compounds at ~205 nm 13, 14_{. Mobile phase solvent and buffer selection is also important for} optimum UV sensitivity and linearity; UV cutoffs (the wavelength at which the solvent absorbs) become particularly significant at low wavelengths. There are three different types of UV detectors:

4

fixed-wavelength detectors that rely upon distinct wavelengths, and variable-wavelength and photodiode array detectors that rely upon one or more wavelengths generated from a broad spectrum lamp. Fixed-wavelength detectors, the backbone of early HPLC systems, are cheap and simple, but are in limited use today. The most common fixed wavelength detectors use the 254-nm output from a low-pressure mercury lamp, the reason many variable-wavelength and photodiode-array applications today still use this wavelength out of sheer habit. Variable wavelength detectors can be tuned to operate at the absorbance maximum of an analyte or at a wavelength that provides more selectivity. They can also be programmed to change wavelengths during a chromatographic run to compensate for response of different analytes. In a variable-wavelength detector, light from a broad-spectrum (for UV deuterium is common, tungsten for visible) lamp is directed through a slit to a diffraction grating that spreads the light out into its constituent wavelengths. The grating is then rotated to direct a single wavelength of light through a slit, through the detector cell, to a photodiode (see Figure 1). Photodiode-array detectors have an optical path similar to variable-wavelength detectors except the light passes through the flow cell before hitting the grating, allowing it to spread the spectrum across an array of photodiodes. Z-path or tapered detector cells designs are commonly used in most UV detectors for HPLC. When using a 100 mm x 4.6 mm column (3-µm particle size), detector cell volumes of 8-10 µL are required (for example, 1-mm diameter, 10-mm path length). An 8-µL flow-cell volume can lead to unacceptable extra column band broadening using smaller column diameters and particle sizes; however, reducing the path length to decrease extra column volume will decrease the signal.

5

Early PDA detectors were not as sensitive as single-wavelength UV detectors, but the gap has been closing in recent years thanks to a new detector cell type, as illustrated in Figure 2. This detector cell consists of a light-guided flow cell equivalent to an optical fiber. Light is transferred efficiently down the flow cell in an internal reflectance mode that still maintains a 10-mm flow-cell path length with a volume of only 500 nL, extending the path length while maintaining low dispersion.

Figure 2: Schematic representation of a light guided detector flow cell. Reproduced from reference 17

In the UV detector, the incident light beam is bounced off a movable grating and imaged onto a slit which allows only a limited range of wavelengths to pass through the sample. In the PDA, the incident light beam (including all wavelengths) is passed through the sample and then bounced off a fixed grating and imaged onto a linear array of photo detectors.

2.1 Relative response factors

Relative Response Factor (RRF) is an analytical parameter used in chromatographic procedures to control impurities / degradants in drug substance and drug product. RRF is used to correct the difference in detector response of impurities with the analyte peak. RRF is established by slope method with linear range of solutions. Different Pharmacopoeias refer the term RRF differently.

As per European Pharmacopoeia (Ph.Eur) The Relative detector response factor, commonly referred as Response Factor, expresses the sensitivity of a detector for a given substance relative to a standard substance. The correction factor is reciprocal of the response factor. Ph.Eur refers RRF as Correction factor or Response factor. As per United States Pharmacopoeia (USP) The Relative

6

response factor, is the ratio of the responses of equal amounts of the Impurities and the drug substance. USP refers RRF as Correction factor or Response factor or Relative response factor. As per British Pharmacopoeia (BP) The Response Factor is a relative term, being the response of equal weights of one substance relative to that of another in the conditions described in the test. BP refers RRF as Response factor.

Establishment of RRF is required to avoid the stability issues with standards, to reduce the cost on preparation of Impurity Standards, to reduce Maintenance of Impurity Standards, due to the lack of donation of Impurity Standards, difficulty in synthesis and isolation of Impurity Standards, for convenience and time saving. Relative Response factor (RRF) is used in different stages: Clinical phase 1 to phase 4 studies, in drug purity tests, Mass balance tests, in limit tests and in stability indicating methods etc.

3 Evaporative Light Scattering Detector (ELSD)

ELSD was originally introduced in 1966 by Ford and Kennard 18 _{at Union Carbide’s Australian} laboratories, and the theory of its operation was developed by Charlesworth 19_{, Stolyhwo et al.} 20, 21 and Mourey and Oppenheimer 22_{. The detector became commercially available since the early} 1980s 23_{. It has been used extensively for qualitative and quantitative analysis of several classes of} compounds, especially poorly chromophoric materials 4, 21, 24-28_{. Quantitative work has employed} well-characterized standards, either of the specific analyte, or of closely related compounds, and the latter practice has perforce been followed when using ELSD to quantify combinatorial libraries. These are normally comprised of chromophoric compounds, and one attraction of ELSD is that it generally provides a more uniform response than does UV. Thus, when standards that were sufficiently representative of the whole library were employed, quantification errors of as low as 10-20% were found for limited sample sets 6, 29_.

ELSD serves as a complement to other detection methods. Although not the most sensitive, it is sensitive enough for most applications, commonly offering limits of detection in the hundreds of picograms on-column. Used in this way, evaporative light-scattering detectors frequently reveal more components in a sample than can be seen with light-absorbing detectors. Also, ELSD offers gradient compatibility and relatively simple operation. At other times, ELSD is the detection method of choice in terms of sensitivity to weakly chromophoric or non chromophoric compounds such as

7

carbohydrates and phospholipids. A very powerful detector combination is an in-series connection of a photodiode-array UV-vis detector for multi-wavelength determination of chromophoric constituents and an evaporative light-scattering detector for determining both chromophoric and non chromophoric compounds with relative abundance profiling 30_.

3.1 Universal response

Because they respond to the quantity of light-scattering particles, evaporative light-scattering detectors provide a more uniform response to structurally similar analytes than light-absorbing detectors. For many analyte classes, such as lipids, users can create a universal calibration set from a single analyte to quantify all analytes of the same class 31-34_.

Evaporative light-scattering detectors 'also are nearly universal detectors, with the exception that they will miss truly volatile sample constituents such as low molecular weight alcohols. ELSD will respond to all light-scattering particle aggregates that remain after the mobile phase has evaporated. In this regard, ELSD offers a more accurate record of the relative abundance of compounds in a sample than RI detection. Because it is not influenced by the bulk properties of the solvent, ELSD is fully gradient compatible 22_.

3.2 Operating principles

The operation principle of ELSD 19, 22 _{mainly consists of three successive processes, depicted in} Figure 3. First is nebulization of the chromatographic effluent, second is evaporation of the mobile phase, and third is detection of the non-volatile residual particles, by means of the measurement of the scattered light.

3.2.1 Nebulizing and evaporation

In the first step of ELSD detection mechanism, the effluent from a chromatographic column enters a Venturi-type nebulizer, where it is transformed into an aerosol. These nebulizers create a high flow of carrier gas (air or inert gas, such as nitrogen, carbon dioxide, argon or helium) over the liquid surface producing a high amount of droplets with remarkably uniform size. Distribution and mean values of droplets diameter are considered to be very critical parameters, which strongly influence the analytical characteristics (i.e., detectability, sensitivity and repeatability) of the ELSD methods 20, 35, 36_{. The size distribution of the aerosol droplets can be successfully described by the} Mugele–Evans upper-limit log-normal size distribution 22, 37 _{and it is mainly determined by nebulizer}

8

geometry as well as liquid and gas flow rates. Narrow distribution is a requirement for good repeatability and ELSD sensitivity. Furthermore, it has been indicated that mean diameter of aerosol droplets strongly influences ELSD response and in fact an increase of the mean diameter of aerosol droplets results in ELSD response enhancement.

Figure 3: Schematic depiction of the main steps of the ELSD operation. Design (a) ELSD type A and (b) ELSD type B. Reproduced from reference 23

ELSDs are classified in two types according to their structure after the nebulization unit 36, 38_{. In} ELSD of type A (non-aerosol splitting, Figure 3a), the entire aerosol immediately enters the heated evaporation tube (drift tube), where the evaporation process begins. In ELSD of type B (aerosol splitting, Figure 3b), the aerosol, before the evaporation step, enters a glass chamber or a focusing cone (nebulization chamber), in which the droplets of high size are condensed on the walls of the chamber and diverted to waste. The proportion of the wasted aerosol depends on mobile phase volatility and varies from >90% (aqueous mobile phases) to <10% (highly volatile organic solvents).

9

Each type appears its own benefits, while the appropriate choice depends on the nature of the analyte and the composition of the mobile phase. ELSD of type B requires lower evaporation temperature than type A and thus it is more sensitive for volatile, semi-volatile or thermo-sensitive analytes. On the other hand, for non-volatile analytes, ELSD of type A appears to be more sensitive, since the entire quantity of analyte reaches the optical cell. Considering the composition and flow rate of the mobile phase, ELSD of type A is incompatible with gradient elution and requires low flow rates and highly volatile mobile phases (non-aqueous or low water portion), while ELSD of type B appears of a wider compatibility 39_{. In the stage of evaporation of Mobile Phase, which is performed} in a heated drift tube, the size of the aerosol droplets is reduced. Ideally, the purpose of this stage is to completely vaporize the mobile phase, without any analyte loss (due to evaporation or thermal decomposition). The completeness of the mobile phase evaporation and the extent of loss of analyte are mainly determined by the evaporation temperature, which should be selected in accordance to the mobile phase and analyte volatility, to the mobile phase flow rate and to the ELSD (A or B). Inappropriate selection of the evaporation temperature results, in case of low temperature, in an excessive noise or baseline with spiked sharp peaks, or in case of high temperature, in reduced sensitivity. Apart from the analyte loss, high evaporation temperature causes rigorous solvent evaporation, which destroys uniformity of particle size. For low-melting analytes, higher evaporation temperatures favors the formation of liquid (melted) rather than solid particles 40_{. Both effects result in decrease of ELSD sensitivity. The evaporation temperature is} usually set between 30◦C and 100◦C. Decrease of the required evaporation temperature can be obtained with nebulizing gas of high thermal conductivity (helium was found to require at least 30◦C lower evaporation temperature than carbon dioxide), which in cases of volatile and thermo-sensitive compounds results in enhancement of detector sensitivity 41_{. On the other hand, for} stable and non-volatile compounds, the ELSD response factor has been found to be independent of the nature of the nebulizing gas 42_.

Efficiency of chromatographic separations does not appear to be affected by the length of the evaporation tube 36_{, providing that an appropriate level of vacuum is applied at the end of the flow} path in order to establish a stable flow of liquid and solid aerosol.

3.2.2 Light Scattering

The aerosol, after the evaporation process, ideally composed by solid particles of analyte, enters the optical cell and passes through a light beam. The scattered light is measured by a

10

photomultiplier or a photo diode, providing the output signal. Light scattering processes are classified in two types: elastic scattering, in which the scattered radiation is of the same frequency as the incident radiation, and inelastic scattering, in which the scattered radiation is of a different frequency. In ELSDs, inelastic scattering is considered to be negligible and it is not further examined. Elastic scattering is classified in three types: Rayleigh, Debye and Mie. Refraction-reflection mechanism, which has its origin in the induced secondary emission of particles in the path of the incident beam, has also been reported as a potential mechanism of scattering in the ELSD optical cell 19, 20, 36_{. The type of interaction between the light and the particles depends on the} size, shape and surface properties of the particles and the wavelength (λ) of the incident light. Reyleigh-Debye scattering occurs with the smallest particles (D/λ < 0.1), Mie scattering becomes the predominant mechanism for 0.1 < D/λ < 1, while the refraction-reflection mechanism occurs in case that the particle size is greater than the wavelength (D/λ > 1). In the case of λ = 0.35μm (the wavelength with the maximum emission of tungsten lamp), Mie and refraction-reflection processes prevail, since analyte particles with radius smaller than 100 nm is usually not detectable. Furthermore, refraction is relatively more important compared to reflection in the refraction-reflection domain. Actually, in most cases, more than one scattering mechanisms occur in the ELSD optical cell, due to: (a) variations of the aerosol droplet diameter D, which is dependent on the nebulization and evaporation processes, (b) the polychromatism of the light source and (c) dependence of the mean droplet diameter on the sample concentration. Since scattering and not absorbing phenomenon is intended to occur when the light interacts with the analyte particles, a tungsten filament or halogen lamp that produces a continuous spectrum of wavelengths, rather than a monochromatic laser emitting diode, is favored as a light source. In some instruments a secondary gas, independent of the nebulizing gas, is used to concentrate the particles in the center of the detection cell and to prevent the deposition on the cell inner surfaces. The power of scattered light is controlled by the particle diameter, the light wavelength and the angle of scattered light. It has been observed that the ELSDs sensitivity is higher, but the dynamic range narrower, for low detection angle, with wide angular acceptance and the use of vertically polarized or non-polarized light 35, 43_.

3.3 Characteristics of ELSD response

The response of the ELSD detector has a logarithmic nature. Although this nature of the main working regime of the ELSD is well known, the extent to which the suppressive effect at low levels

11

of analyte can cause quantitative errors is poorly appreciated. The area response A of the ELSD is known to be related to the mass M of the analyte by the following relationship. In most applications, a non-linear response has been observed for the ELSD. The area of the chromatographic peak (A) appears good correlation with the analyte mass (M) according to the exponential relationship:

𝐴 = 𝑎𝑀𝑏

where a is the response factor and b is the coefficient of regression of the response curve and are dependent on the ELSD instrumentation and on nebulization and evaporation processes (flow rates of the nebulization gas and mobile phase, composition of the mobile phase, evaporation temperature, etc.) 22_{. Coefficient b generally varies between 0.9 and 2} 2, 36_{. In case that b is close to} unity (b = 1) (mainly observed for high water content in mobile phase) a linear calibration curve can be constructed, otherwise it is necessary to use double logarithmic coordinates in order to obtain a linear calibration curve.

Considering the case of two compounds denoted 1 and 2. The equation may be restated as log 𝐴1= log 𝑎1+ 𝑏1log 𝑀1

log 𝐴2= log 𝑎2+ 𝑏2log 𝑀2

log (𝐴2⁄𝐴1)= log 𝑎2− log 𝑎1+ 𝑏2log 𝑀2− 𝑏1log 𝑀1

In an idealized situation, the two compounds may exhibit the same response factors a and coefficient b. The equation may be rewritten as

log (𝐴2⁄𝐴1)= 𝑏 log (𝑀2⁄𝑀1)

Suppose the object of the experiment is to measure the absolute level of compound 2 in compound 1 as a percentage of the total sample, so that 𝑀2+ 𝑀1= 100, then the areas returned

by the ELSD will vary as a function of exponent b:

𝐴2⁄𝐴1= [𝑀2⁄(100 − 𝑀2)]𝑏

Setting 𝐴2+ 𝐴1= 100, the area of compound 2 is 𝐴2, when

12

This relationship is illustrated in Figure 4, which shows that underestimations of purity will increase with increasing exponent b, the value of which is expected to be 1.33 in an idealized situation of pure Mie scattering 3_{. Note that for detectors such as UV, CLND and MS where one} normally works in a linear response domain, b would be unity (b=1).

Figure 4: Apparent (Area%) levels of an impurity in a major component that would be obtained by HPLC using ELSD. The assumption was made that both major and minor components have identical values of response factor a and coefficient of regression b. The effect of the value b is shown. Reproduced from reference 44

Apart from the exponential relationship, second and third order polynomial regressions have also been utilized in order to achieve a good correlation between the peak area and the analyte mass 45_.

4 Corona Charged Aerosol Detector (CAD)

Corona charged aerosol detection (CAD) is a unique technology gaining in popularity in which the HPLC column eluent is first nebulized with a nitrogen (or air) carrier gas to form droplets that are then dried to remove mobile phase, producing analyte particles 46, 47 _{The primary stream of} analyte particles is met by a secondary stream of nitrogen (or air) that is positively charged as a result of having passed a high-voltage, platinum corona wire. The charge transfers diffusional to the opposing stream of analyte particles, and is further transferred to a collector where it is measured

13

by a highly sensitive electrometer, generating a signal in direct proportion to the quantity of analyte present 48_{. Because the entire process involves particles and direct measurement of charge,} CAD is highly sensitive, provides a consistent response, and has a broad dynamic range, which offers advantages when analyzing compounds lacking UV chromophores, as illustrated in Figure 5. Often compared to other universal-type HPLC detection methods, like RI and ELSD, CAD has been shown to be much easier to use, and unlike RI, can accommodate gradients. In addition, CAD response is not dependent upon the chemical characteristics of the compounds of interest, but on the initial mass concentration of analyte in the droplets formed upon nebulization, providing a much more uniform response as opposed to, for example, UV, where responses can vary dramatically according to the wavelength used and the extinction coefficient 49_{. CAD has ELSD-like} quantification features 50, 51_{. Corona-charged aerosol detector (CAD) has recently demonstrated its} usefulness as an “universal” detector due in part to its sensitivity and equimolar response to non-volatile analytes 52_.

4.1 Operating principle

The Corona CAD detector measures charge that is imparted to analyte particles, with the charge being in direct proportion to the amount of the analyte in the sample. This charge is measured accurate and consistent, regardless of the analyte. The result is that the Corona CAD detector can quantify any nonvolatile analyte. This includes those without chromophores or those that do not ionize, thus providing a consistent response that is independent of chemical structure. With charged aerosol detection, many semi-volatile analytes can be even measured.

14

Figure 5: Schematic depiction of a CAD detector. Reproduced from reference 53

Charged aerosol detection begins by converting the eluent into droplets which are subsequently dried, forming particles. The particle size increases with the amount of analyte. A stream of positively charged gas collides with the analyte particles. The charge is then transferred to the particles; the larger the particles, the greater the charge. High mobility species (free ions) are removed by an ion trap 54 _{while the remaining charged particles are then transferred to a collector} where the charge is measured by a highly sensitive electrometer. This generates a signal in direct proportion to the quantity of analyte present 46_.

Operation is simple, requiring only setting of a few controllable parameters; among them the gas input pressure, the temperature or the temperature range and signal output range. However, this could actually limit the room for optimization of detection.

4.2 The response of CAD systems

Like ELSD, CAD system is a mass-dependent detector and the generated response does not depend on the spectral or physicochemical properties of the analyte as in a specific UV detector, which is a concentration-dependent detector. Theoretically this means that CAD and ELSD systems as bulk property detectors generate a similar response for identical amounts of different analytes. For example, it was observed that only a slight variation of the response for equal amount of compounds analyzed, over a test set of 17 chemically different compounds under isocratic elution

15

conditions. 52 _{However this variation was about 7% relative standard deviation (RSD) between all} responses of all 17 chemically different compounds, which indicates that CAD response depends upon analyte volatility 52_{. The response of the CAD} 52, 55 _{is similar as the response of the ELSD} 56 _and have been described by the equation

𝐴 = 𝑎𝑀𝑏

Where A is the output signal from the detector, M is the mass injected and a and b are constants that are dependent on analyte and chromatographic conditions (a represents the response intensity and b represents the response shape). The CAD has been found to give a lower response for particles greater than 10 nm in diameter and this explains the non-linear response at higher analyte concentration 55_{. If linear calibration curves are desired, the equation can be} converted to a linear relationship by taking the logarithm of both sides. This allows accurate quantification when using a two- or three-point calibration curve with CAD. The study of recovery performed on 10 components by Vervoort et al. 57_{, showed that recovery for a high concentration} sample was always in the 98–102% interval when using a two- or three-point calibration curve. The same was true for low concentrations, except for one compound (isoconasole). With the latter, the area response of the data point with the lowest concentration used for calibration (0.005 mg/ml) deviated markedly from the values based on the other data points, for no obvious reason 57_. However, when the concentration level is very low or when the concentration range is small, the calibration curve is close to linear 58_{. In HPLC determination of enantiomer ratios} 59_{, in contrast to} the ELSD response, the CAD signal is nearly linear in the range of interest for many routine analytical studies (5–250 μg/ml). This study also proved the higher sensitivity of CAD relative to ELSD 59_{, as the signal generated by CAD is much less influenced by the aerosol droplet size or its size} distribution, which is influenced by the analyte concentration of the droplets. The sensitivity of CAD was poorer than with CNLSD in some experiments; however it was improved and was comparable to the CNLSD when CAD was coupled with reversed-phase HPLC as shown in the study of Dixon and Peterson 55_{. Furthermore, over a narrow concentration range, good linearity of the CAD response} was observed when determining the relative response factors of paclitaxel-related impurities 60_.

4.3 The effect of additives to the mobile phase

The response of CAD system is also sensitive to contaminants or additives to the mobile phase. With varied concentrations of ammonium acetate (5, 10 and 20 mM) in the mobile phase

16

(water/acetonitrile, 60/40), CAD system performed significantly better than ELSD system at low buffer concentrations but, at higher buffer concentrations the S/N ratio for CAD system dropped markedly. Using volatile acids such as formic or acetic acid did not pose any problem for CAD 57_.

Moreau 61 _{designed two experiments to measure the effect of various common HPLC solvents} on the CAD baseline at a constant flow rate at 0.5 ml/min, without a column and without injecting any samples. In the first experiment, methanol produced the highest CAD background of the four solvents hexane, isopropanol, methanol and water. In the second, acetonitrile produced the highest CAD background of hexane, isopropanol, acetonitrile and water. However, more studies are needed to fully investigate the effects of these solvents on the noise and performance of CAD system 61_.

4.4 Reducing column bleed in MS, CAD and ELSD detection

If the column is implicated, late-eluted sample components might cause high background signals. Flushing the column with a strong solvent can solve this problem. Sometimes the column might bleed light-scattering material from its own degradation. For example, traditional silica based columns are stable only within the pH range of 2–7.5. A silica-based column under acid pH conditions (pH < 2) can potentially lose bonded phase because of hydrolysis, whereas basic conditions (pH > 7.5) can result in silica dissolution. Either situation can cause poor peak shapes for basic analytes 62-65_{. Modern Type-B silica–based materials are less prone to such instability. Column} degradation could result in mobile-phase contaminants that show up with ELSD but that would not be apparent with UV or RI detection 66_.

Teutenberg et al. 50 _{used CAD system for detecting column bleed, as an indicator of induced} degradation of the stationary phase. Five HPLC columns (from Phenomenex, ZirChrom, Thermo, Polymer Laboratories, and ZirChrom-Sachtleben) were heated to 200°C using a homemade heating system. The results were also evaluated by ultraviolet diode array detection at different wavelengths. CAD system was better for detecting HPLC column bleed than the UV detector because, in charged aerosol detection, peak area is not dependent on the extinction coefficient of the analytes. For high temperature HPLC, the prototype carbon clad titanium dioxide column shows the least bleed independent of the detection technique 50_{. For normal temperature HPLC, novel} polymer based columns show the least bleed.

17

4.5 Mobile phase compensation in ELSD and CAD

As with many aerosol processes, the response of CAD system is influenced by the diameter of the generated particles 55_.

The dependence of the ELSD and CAD response on the mobile phase composition is a significant drawback, especially in applications in which the universal and uniform response of the detector is crucial (e.g. in impurity profiling in pharmaceutical analysis). The response of CAD system is influenced by the diameter of the generated particles 55_{. Since the droplet diameter is} related to several factors, including density and viscosity of the mobile phase, it is also dependent on the mobile-phase composition. In gradient elution chromatography the response factor will vary significantly with the mobile-phase composition, which is the main drawback of this detector 51, 67_. Higher organic content in the mobile phase leads to greater transport efficiency of the nebulizer, which results in a larger number of particles reaching the detector chamber and a higher signal 68_.

Mathews et al. 69 _{corrected for this phenomenon in ELSD by developing a generic calibration} routine. A 3-D calibration surface was constructed including the varying response with the mobile-phase composition, as opposed to the 2-D calibration normally used. The main drawbacks of this approach are the large number of analyses that must be performed for calibration and the fact that the results are method-specific. An empirical and more universal approach is demonstrated to solve this problem 51_.

Gorecki et al. 51 _{proposed an elegant approach based on mobile-phase compensation to solve} this problem. The principle is, for a gradient elution, to provide the detector at all times with a constant composition of the mobile phase. A separate pump compensates the organic content in the mobile phase by delivering exactly an inversed gradient prior to the detection to ensure a constant mobile-phase composition at the detector inlet. This resulted in constant response, independent of the mobile-phase composition in the column 51_.

The combination of this approach with the CAD provides a powerful tool for fast quantitative or semi quantitative analysis of analytes, for which, for example, no pure standards are available. In this way, calibration and analysis can be done with a limited number of analyses and a universal response is approached. Figure 6 illustrates the effect of mobile phase compensation.

18

Figure 6: Chromatograms with superimposed organic modifier composition obtained for the separation of six sulfonamides. (A) UV signal in normal gradient run; (B) CAD signal without mobile-phase compensation in a gradient run; (C) CAD signal with mobile-phase gradient compensation. Peak identification: 1, sulfaguanidine; 2, sulfamerazin; 3, sulfamethazine; 4, sulfamethizole; 5, sulfamethoxazole; 6, sulfadimethoxin. Reproduced from reference 51_.

5 Chemiluminescent Nitrogen Detector (CLND)

Since it was first used with HPLC in 1992 by Fujinari and Courthaudon 70_{, Chemiluminescent} nitrogen detection (CLND) has found rapid application in the areas of combinatorial chemistry 71-77 and metabolic studies 78_{. Taylor et al.} 72 _{found the absolute error for a set of chemically and} structurally diverse compounds averaged approximately 10%. Shah et al. 74 _{used flow injection} analysis (FIA) in conjunction with CLND and MS to analyze up to 1000 compounds per day.

Unlike ELSD, CLND can provide molar information, although to convert CLND data into an amount of analyte, it is necessary to know the molecular weight and number of nitrogen atoms in the molecule. These data are normally available from the registered structure and may be confirmed with concurrent MS techniques. In case of an unspecified impurity rising above the reporting threshold and obviously the structure of the compound is not known, accurate mass

19

determination using high-resolution mass spectrometry (HRMS) can give the mass and the number of nitrogen atoms 79-82_{, however, information about the location of the nitrogen atoms in the} molecule must be known (see also paragraph 5.3). HRMS in combination with CLND is a powerful tool, especially in the pharmaceutical industry, where these compounds often contain nitrogen atoms.

Diminished CLND response can occur for compounds where adjacent nitrogen atoms may give diatomic nitrogen during combustion 83_{. CLND is incompatible with nitrogenous chromatographic} solvents and additives, and baseline chromatographic separation is necessary for accurate quantification.

Mutton et al. 44 _{indicated that molar nitrogen response in CLND is sufficiently stable and robust} to consider it in preference to ELSD as a high-throughput quantitative tool without the need to weigh any materials other than a single primary reference standard such as caffeine or indole. It was reported that, over a linear range of 2 orders of magnitude, the CLND detector exhibited an equimolar response with ±10% average error for compounds studied 72_.

5.1 Operating principles

The chemiluminescent nitrogen detector is an element specific detector in which the column effluent is nebulized with oxygen and a carrier gas of argon or helium and pyrolyzed 84-86_{. Nitrogen} containing compounds react with oxygen at 1050°C to yield CO2, H2O, NO and other combustion products. The products are then routed through a membrane dryer to remove H2O to a detector, where NO is reacted with ozone too form excited NO2*. As the NO2* decays to the ground state, photons are emitted and detected by a photomultiplier tube 87_{. The resultant signal is directly} proportional to the amount of nitrogen in the original analyte. An advantage of LC-CLND compared to other techniques is that calibration with a single nitrogen-containing secondary standard is sufficient, and the standard need not be chemically similar to the analytes. Equimolar response follows from the principle of LC-CLND 88_{. Care must be taken in choosing nitrogen-free mobile} phase components and additives (for example, no acetonitrile!).

20

Figure 7: Schematic overview of CLND instrumentation. Reproduced from reference 1

Single calibrant quantification by LC-CLND is straightforward in case of relatively simple materials, requiring no extraction, such as combinatorial chemistry library products 89_, nitrogen-containing anions in seawater 90 _{or seized street drugs} 82 _{and blood samples} 88_.

5.2 The response of the CLND detector

The chemiluminescent response is proportional to the number of moles of nitric oxide, and correspondingly to the number of moles of nitrogen originally present in the analyte. For virtually any nitrogen-containing compound (with the exception of N2 and compounds containing N=N bonds), the signal is independent of structure. Amines, amides, nitrates, nitrogen-containing heterocycles, etc. all produce a signal that is directly related to the number of moles of nitrogen present. Provided the molecular formula of the analyte is known, one can thus determine its relative weight in the sample from the amount of nitrogen in the HPLC peak. Quantitation, then, requires only a single nitrogen-containing standard, which need not be structurally related to the analyte. Thus, the CLND provides a unique and demonstrated capability for determining sample concentrations without the need for standards of each analyte 71, 72_{. Moreover, for determination of} relative amounts, no standard whatsoever is necessary. All that are needed are the relative CLND peak areas and the molecular formulas of the analytes. Once the relative amounts are found, it is a simple matter to use UV peak areas (e.g. from a UV detector in series with the CLND) to determine the RRF. Thus, UV response factors (per unit weight) for impurities, relative to parent compounds, can be determined by means of the following equation

21

𝑅𝑅𝐹 = [𝑈𝑉 𝑝𝑒𝑎𝑘 𝑎𝑟𝑒𝑎 𝐶𝐿𝑁𝐷 𝑝𝑒𝑎𝑘 𝑎𝑟𝑒𝑎]⁄ 𝑖 [𝑈𝑉 𝑝𝑒𝑎𝑘 𝑎𝑟𝑒𝑎 𝐶𝐿𝑁𝐷 𝑝𝑒𝑎𝑘 𝑎𝑟𝑒𝑎]⁄ _𝑝×

𝑀𝑊 # 𝑛𝑖𝑡𝑟𝑜𝑔𝑒𝑛⁄ _𝑖 𝑀𝑊 # 𝑛𝑖𝑡𝑟𝑜𝑔𝑒𝑛⁄ 𝑝

Where I is impurity; p denotes parent compound; MW is molecular weight; # nitrogen is the number of nitrogen molecules in the molecular formula.

For unknown impurities, accurate mass LC–MS can be used to determine the molecular formula 79-81_{. For molar (rather than weight) RRF values, one needs only the relative number of nitrogens} per molecule and not the molecular formula or weight. By use of the CLND, then, relative UV response factors can be determined without fraction collection or purification, without standards, and even without preparations of known concentrations of analytes. As a result, sample preparation is greatly simplified, and the stability of the impurity is not an issue. The CLND is limited, of course, to mobile phases that do not contain nitrogen. Acetonitrile and amine modifiers, commonly used in HPLC, are, therefore, precluded. Also, the CLND is not readily amenable to non-volatile buffers in the mobile phase.

However, it is still possible to determine relative UV response factors for samples run under these non-CLND-compatible HPLC conditions. One option is to analyze a sample with the CLND first (using a compatible mobile phase) to determine the relative amounts of the analytes present. Then the same sample is analyzed with UV detection using non-CLND-compatible HPLC conditions, and the RRF calculated using Eq. (1). Such an approach assumes that the sample is unchanged (i.e. no degradation or change in analyte concentration) between the two analyses. Often, especially when multiple laboratories and conditions are involved, it may be preferable to use a second approach. In this case, a CLND compatible mobile phase is used to separate the compounds of interest and determine RRF values under the given conditions (RRF1). Separately (e.g. at a laboratory without access to the CLND) and as needed, a sample is assayed by HPLC-UV using both the original CLND-compatible mobile phase and the non-CLND-compatible mobile phase of interest. The relative UV peak areas are then used to correct the RRF value for the change in conditions. Thus, RRF1 obtained originally with the CLND-compatible conditions can be used to determine RRF2 for any different set of conditions by multiplying by the ratio of the relative UV areas obtained under each.

𝑅𝑅𝐹2= 𝑅𝑅𝐹1×_{(𝑈𝑉 𝑝𝑒𝑎𝑘 𝑎𝑟𝑒𝑎)}(𝑈𝑉 𝑝𝑒𝑎𝑘 𝑎𝑟𝑒𝑎)𝑖,2⁄(𝑈𝑉 𝑝𝑒𝑎𝑘 𝑎𝑟𝑒𝑎)𝑖,1 𝑝,2⁄(𝑈𝑉 𝑝𝑒𝑎𝑘 𝑎𝑟𝑒𝑎)𝑝,1

22

where i is impurity; p denotes parent compound; and 1 and 2 represent CLND-compatible and non-compatible HPLC conditions, respectively. The approach described with Eq. (2) permits running a separate sample at any time at a site removed from the CLND, albeit that sample must be run with UV detection under both sets of HPLC conditions. In pharmaceutical development, where HPLC conditions may be modified across several sites and over a period of years, such flexibility is quite valuable. The RRF values for key impurities can thus be determined at a core laboratory early in development using a given set of CLND-compatible conditions. As development progresses and HPLC conditions change, the RRF values can be continuously corrected and updated simply by use of Eq. (2), which requires only an HPLC-UV system and access to the original data. Finally, the RRF value is independent of whether the given impurity is present at trace levels or at significant amounts. Thus, in pharmaceutical development, forced degradation may often be used to generate relatively large amounts of key impurities, such that both UV and CLND peak areas may be precisely measured. The accurate RRF values thus obtained may then be applied widely to other situations in which the impurities are present at much lower levels. Previous reports have demonstrated the unique capability of the CLND to accurately measure concentrations of nitrogen-containing compounds without the need for standards of the same or similar compounds 71, 72_{. The application} of this attribute to the simple and accurate determination of relative UV response factors for a variety of compound structures, by gradient as well as isocratic HPLC have been demonstrated. Also the capability in resolving mass balance issues, which occur frequently during forced degradation studies, using photo degradation of nifedipine as an example 1_.

5.3 Equimolarity of CLND response

The major advantage of the CLND detector at this moment is the equimolarity for the nitrogen response. However, N=N containing compounds have a tendency to yield N2 on combustion instead of nitrogen oxide. This has an effect on the CLND signal because N2 does not react with ozone during the chemiluminescent reaction and a less then equimolar response is observed. For those types of compounds the area will be not reliable and the concentration will be underestimated.

The obvious limitation of CLND is that it can only be applied to compounds containing nitrogen. However, since an estimate of 90% of the developmental and marketed drugs contain nitrogen, CLND can be widely used in pharmaceutical analysis 91, 92_.

23

Although previous reports have reported equimolar response with a 10% error 72_{, it was also} found that the sole exception to equimolar response of the CLND arises from chemical structures containing adjacent nitrogen atoms (Figure 8), such as azo or azide groups. It was speculated that this deviation was predictable and can be corrected. A proposed rule of thumb is that the response should be 0 when adjacent nitrogen atoms are connected by a double bond and 0.5 when adjacent nitrogen atoms are connected by a single bond (Figure 8).

Figure 8: Proposed exceptions to CLND equimolar response rule. Reproduced from reference 92

In compounds with adjacent nitrogen atoms connected by a single bond, the CLND response is highly structure dependent 83, 92_{. Substitutions on the nitrogen atoms or nearby in the molecule can} increase the CLND response to approach a value higher than the predicted value 0.5. Without substitution on nitrogen or in the molecule, much lower values than predicted are obtained. The substitution effect can be understood in term of how surrounding chemical bonding energy affect the N2 formation inside CLND. If more chemical bonds or higher bonding energy surround the nitrogen atoms, they tend to be pulled toward the mother molecule rather than form N2 and leave the molecule 92_.

Based on these studies, a structurally similar calibration compound should be used for this class of compounds in the quantitative analysis using CLND. Nevertheless, careful interpretation of the results is needed when dealing with N=N or N-N containing compounds 83_.

6 Selected applications of CAD

CAD has been applied for the analysis of nonvolatile and semi volatile neutral, acidic, basic, and zwitterionic compounds, both polar and nonpolar. These include lipids, proteins, steroids, polymers, carbohydrates, peptides and other compounds with weak chromophores used in the pharmaceutical, chemical, food, and consumer products industries and in life science research 93_. Some applications are described in the following paragraphs.

24

6.1 Universal quantitation of PDE-5 inhibitors using LC-CAD

A significant part of all reported cases of illicit supplements in Europe concerns herbal aphrodisiacs adulterated with phosphodiesterase 5 (PDE-5) inhibitors 94_{. Because they are} composed of natural ingredients, herbal supplements are perceived as a safe alternative to synthetic active substances. In order to avoid detection of such forgery in screening tests, patent infringement or coming under Pharmaceutical Law, analogues of PDE-5 inhibitors or their mixtures are frequently seen as adulterants.

For quantitative measurements, multiple reaction monitoring (MRM) scan mode is generally applied 95-98_{. Usually, the content of active compounds is measured by} 1_{H NMR using TSP-d4} (trimethylsilyl-2,2’,3,3’-tetradeuteropropionic acid) as internal reference 99_{. In other published} studies, the content of analogue is estimated using UV response of the approved PDE-5 inhibitor’s standard 100_{. However, this approach is justifiable only when the analogue and the reference} standard possess an identical chromophore. It cannot be applied to the quantification of all analogues, as the recorded response could differ even for very similar compounds depending on the chromophoric group. The present study focuses on the introduction of an alternative analytical method for the quantitative analysis of PDE-5 inhibitors or their analogues detected in counterfeit herbal supplements for erectile dysfunction. The application of CAD for PDE-5 inhibitors determination has not been reported, however, this detector appears to be suitable for such quantification, when the NMR technique cannot be used.

Chromatograms of the mixture of sildenafil and tadalafil at a concentration of 10 µg ml−1 _of each standard were recorded by three different detectors: UV, MS and CAD using identical chromatographic conditions. Relative response factors for tadalafil (RRF) were calculated (compared to sildenafil response). Obtained results indicate that the responses for each standard differ largely depending on the wavelength (Figure 9a) and it is not possible to select the optimal wave length simultaneously for standards, analogues and novel compounds of an unknown structure. RRFs have to be determined, which is difficult in case of lack of standards. A chromatogram recorded by the MS detector shows that the ability of ionization in the set conditions is different for these two standards (Figure 9b). Peak areas calculated from extracted ion chromatogram (EIC) were significantly different. The ionization of analogues may also be different. Hence, the MS measurements cannot be used for universal quantification. Peak areas measured

25

with CAD detector were almost similar, although peak shapes of standards were significantly different (Figure 9c).

Figure 9: Chromatograms obtained for the mixture sildanefil and tadalafil 10 µgml-1 each by different types of detector: a) UV, b) MS pos/neg, c) CAD. Reproduced from reference 101

Indirect determination was performed for sildenafil, tadalafil, vardenafil and 7-methyl vardenafil analogue standards. The error was ±3% in both ranges. Moreover, the linearity test shows that the slope and the intercept values for the calibration functions for sildenafil and tadalafil are nearly uniform (Table 1). RRFs for sildenafil (calculated as a ratio of slopes of sildenafil and tadalafil calibration curves) for a linearity range, basic range and auxiliary range are 1.02, 1.02 and 1.00 respectively and tadalafil RRFs are 0.98, 0.98 and 1.00 respectively. The results are satisfactory for the purpose of which the method was developed, and indicate the utility of the CAD for universal calibration 101_.

26

Table 1: Parameters for LC-CAD method. Linear and power fit equations, LOD and LOQ for Sildanafil and Tadalafil. Data taken from reference 101

Compound name Range (µgml-1_{) Linear Power LOD (µgml}-1_{) LOQ (µgml}-1₎

𝑦 = 𝑎𝑥 + 𝑏 r2 _{𝑦 = 𝑎𝑥}𝑏 _r2

Sildanafil 0.8-60.0 𝑦 = 0.3332𝑥 + 0.2247 0.9992 𝑦 = 0.3936𝑥0.9615 _{0.9995 0.25 0.80}

Tadalafil 1.0-60.0 𝑦 = 0.3258𝑥 + 0.2679 0.9993 𝑦 = 0.3892𝑥0.9586 _{0.9995 0.35 1.00}

6.2 Analysis of drugs without a natural UV chromophore

CAD has been applied for the analysis of polyketide 102 _{and bisphosphonate compounds} 103 _that do not possess a natural chromophore. Mass spectrometry is a powerful tool for quantifying low-titer and poorly identified polyketide compounds, as it has a low limit of detection and is able to provide sample molecular weight information. Pistorino and Pfeifer 102 _{compared the analytical} capabilities of MS, ELSD and CAD analyzers for quantifying a model polyketide target compound (6-deoxyerythronolide B. The limit of detection (LOD), within-run precision, dynamic range and linearity obtained with all three detection systems were compared. The results showed that CAD is a cost-efficient alternative to MS for research and commercial use challenged by low natural product titers, since the limit of detection is lower than that of MS while maintaining a comparable dynamic range. In addition, the aerosol detectors were more precise and provided greater accuracy over the measurement range. However, ELSD exhibited a well-defined calibration curve over its analytical range and consistently the best precision of the three analyzers, but it is restricted by its limit of detection. ELSD could be primarily applicable for titers ≥ 1 mg/L 102_.

Traditional analysis of bisphosphonates is very time-consuming and various methods are needed to analyze for assay and degradation products. These involve indirect UV detection 104-106 _or pre-column derivatization to obtain better sensitivity and selectivity 107-113_{. Fang and co-workers} 103 developed a rapid, direct and stability-indicating method for analysis of etidronate, a model bisphosphonate compound without a UV chromophore. A mixed-mode column was used to separate etidronate (2) from its impurities and no time-consuming derivatization was needed with CAD 112, 114, 115_{. The method was successfully validated for specificity, linearity, accuracy, precision,} sensitivity and stability, and it may be used to analyze dissolution samples as well as assay/degradation products of etidronate for both release and stability testing purposes. Similar methodologies may be applied for pharmaceutical analysis of other bisphosphonates with significantly improved analytical efficacy and accuracy.

27

7 Selected Applications of CLND

7.1 Analysis of street drugs without the use of primary reference

standard

A novel approach was used to analyze street drugs in seized material without primary reference standards. Identification was performed by liquid chromatography/ time-of-flight mass spectrometry (LC/TOFMS), essentially based on accurate mass determination using a target library of 735 exact monoisotopic masses. Quantification was carried out by LC-CLND with a single secondary standard (caffeine), utilizing the detector’s equimolar response to nitrogen. Sample preparation comprised dilution, first with methanol and further with the LC mobile phase. Altogether 21 seized drug samples were analyzed blind by the present method, and results were compared to accredited reference methods utilizing identification by gas chromatography/mass spectrometry and quantification by gas chromatography or liquid chromatography. The 31 drug findings by LC/TOFMS comprised 19 different drugs-of-abuse, byproducts, and adulterants, including amphetamine and tryptamine designer drugs, with one unresolved pair of compounds having an identical mass. By the reference methods, 27 findings could be confirmed, and among the four unconfirmed findings, only 1 apparent false positive was found. In the quantitative analysis of 11 amphetamine, heroin, and cocaine findings, mean relative difference between the results of LC/CLND and the reference methods was 11% (range 4.2-21%), without any observable bias. Mean relative standard deviation for three parallel LC/CLND results was 6%. Results suggest that the present combination of LC/TOFMS and LC/CLND offers a simple solution for the analysis of scheduled and designer drugs in seized material, independent of the availability of primary reference standards 82_.

7.2 Analysis of biogenic amines

Analysis of biogenic amines is critical to pharmaceutical and food industry due to their biological importance. For many years, the determination of biogenic amines has relied on HPLC coupling with pre-, on-, or post-column derivatization procedures to enable UV or fluorescent detections.

Biogenic amines were separated on a Phenyl-Hexyl column by an ion-pair liquid chromatography method using perfluorocarboxylic acids as ion-pair reagents and detected by

28

CLND. This direct separation and detection HPLC method eliminated the time consuming and cumbersome derivatization procedures. Compared with HPLC–UV (post-column derivatization with ninhydrin) and HPLC-CAD methods, this HPLC–CLND technique provided narrower peaks, better baselines, and improved separations and detections. Excellent linearity was acquired by CLND for each of the 14 biogenic amines ranging from less than 1 ng to about 1000 ng (on-column weights). The relative response factors determined by this LC–CLND method were proportional to the numbers of nitrogen atoms in each compound, which has been the characteristic of the equimolar determinations by CLND. In addition, a number of samples including beer, dairy beverage, herb tea, and vinegar were analyzed by the LC–CLND method with satisfactory precision and accuracy 116_.

7.3 Quantification of drugs in plasma and whole blood by HPLC-CLND

LC–CLND was employed as the analytical technique, based on the detector’s equimolar response to nitrogen and using caffeine as single secondary standard. Liquid–liquid extraction recoveries for 33 basic lipophilic drugs were first established by LC–CLND in blood specimens spiked with the respective reference substances. The mean recovery by butyl chloride–isopropyl alcohol extraction for plasma and whole blood was 90 ± 18 and 84 ± 20%, respectively. The validity of the generic extraction recovery-corrected single-calibrant LC–CLND was then verified with proficiency test samples, including 20 different analyses. The mean accuracy was 24 and 17% for the plasma and the whole blood samples, respectively, and the maximum error was 31% for both specimens. All 20 analyses results by LC–CLND fell within the confidence range of the reference concentrations. LC–CLND proved to be an easy-to-use and robust technique, allowing analysis of 1000 injections of biological extracts without a need for major maintenance operations 88_.

8 Applications for determining relative UV response factors

without the use of a reference standard

8.1 Determination of relative response factors with UV and CAD

The relative response factors (RRFs) of paclitaxel-related impurities were determined by high performance liquid chromatography (HPLC) equipped with an Ultraviolet (UV) detector and charged aerosol detector (CAD) in tandem. The peak response using CAD was independent of analyte structure in an isocratic analysis for this application. After a sample containing known and unknown

29

impurities was analyzed with HPLC–UV–CAD, an empirical approach was developed to calculate the RRFs for all impurities. The RRFs of known impurities were also determined by linear calibration curves. For known impurities, the RRFs values determined with two approaches are comparable. The new approach is effective yet simpler to determine the RRFs for unknown impurities or degradation products since the need for obtaining authentic pure materials was eliminated. Linear calibration curves for all nine compounds were constructed using the peak areas and analyte concentrations in the range of 0.1–2.0 µg/mL by linear regression analysis.

In pharmaceutical studies, the UV RRFs for an unidentified compound is often assumed to be one for the purposes of quantitation. This approximation could lead to potential mass balance problems. Due to the fact that the CAD generates nearly constant response under isocratic conditions for compounds of similar concentration, the relative magnitude of the CAD response correlates to the relative mass of analytes. Based on the UV peak areas and the relative response, RRFs for impurities could be determined by the following equation:

𝑅𝑅𝐹 =_{(𝐶𝐴𝐷 𝑝𝑒𝑎𝑘 𝑎𝑟𝑒𝑎)}(𝑈𝑉 𝑝𝑒𝑎𝑘 𝑎𝑟𝑒𝑎)𝑖⁄(𝑈𝑉 𝑝𝑒𝑎𝑘 𝑎𝑟𝑒𝑎)𝑠

𝑖⁄(𝐶𝐴𝐷 𝑝𝑒𝑎𝑘 𝑎𝑟𝑒𝑎)𝑠

where I and s stand for impurity and standard, respectively.

The linearity data of all the analytes, summarized in Table 2 show the structure independent response of CAD and is also represented in Figure 10.