Symeres Weert
Vliesvenweg 1, 6002 NM Weert, The Netherlands Telephone +31 495 549072, Telefax +31 495 549074
XRF Spectrometry and ICP-MS for the Analysis of Elemental Impurities in Active Pharmaceutical Ingredients
Literature study
Examiner Prof. dr. Govert Somsen
Second examiner dr. Rob Haselberg Contact person Symeres dr. Remy Litjens
Subject XRF Spectrometry and ICP-MS for the Analysis of Elemental Impurities in Active Pharmaceutical Ingredients
Date 02 March 2021
2 Contents 1 Abbreviations ...3 2 Abstract ...4 3 Background ...5 4 Elemental impurities ...6 4.1 ICH classes...6 4.2 Specifications...8 5 X-Ray Fluorescence ...9
5.1 Energy Dispersive and Wavelength Dispersive X-Ray Fluorescence ... 10
5.2 XRF Measurement ... 14
5.3 XRF used in the Pharmaceutical Industry ... 16
6 Inductively Coupled Plasma – Mass Spectrometry ... 19
6.1 Ionization Potential ... 20
6.2 ICP-MS Plasma ... 21
6.3 Measurement of Impurities by ICP-MS ... 22
6.4 ICP-MS used in the Pharmaceutical Industry ... 23
7 Other Techniques for Detection of Elemental Impurities ... 25
7.1 Inductively Coupled Plasma – Optical Emission Spectroscopy ... 25
7.2 Sulfide Precipitation ... 26
7.3 Atomic Absorption Spectroscopy ... 27
8 Comparison XRF versus ICP-MS ... 29
9 Conclusion ... 30
10 Bibliography ... 31
11 Appendix 1: Relative isotopic abundance table [73] ... 37
3 1 Abbreviations
AAS Atomic Absorption Spectroscopy API Active Pharmaceutical Ingredient CeO Cerium Oxide
DDP Digital Pulse Processor EDXRF Energy Dispersive XRF EMA European Medicines Agency HCl Hydrochloric acid
HF Hydrogen Fluoride HNO3 Nitric acid
ICH International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use
ICP-AES Inductively Coupled Plasma – Atomic Emission Spectroscopy ICP-MS Inductively Coupled Plasma – Mass Spectrometry
ICP-OES Inductively Coupled Plasma – Optical Emission Spectrometry JP Japanese Pharmacopoeia
LOD Lower Limit of Detection MES Multi Element Standard
MIC Microwave – Induced combustion PDE Permitted Daily Exposure
Ph. Eur. European Pharmacopoeia RCC Residual Carbon Content S/N Signal to Noise Ratio TDS Total Dissolved Solids
USP United States Pharmacopoeia WDXRF Wavelength Dispersive XRF XRF X-Ray Fluorescence
4 2 Abstract
In the production of active pharmaceutical ingredients (APIs), elemental impurities can be introduced into the matrix, the API. The levels of elemental impurities in drug product (and therefore also in drug substance) should be within acceptable limits. These limits are based on the toxicity and the permitted daily exposure of the element, as described in regulatory guidelines. Inductively coupled plasma – mass spectrometry (ICP-MS) and X-Ray fluorescence (XRF) are two different types of techniques which both can be used to detect elemental impurities. The former is commonly used in the pharmaceutical field, while the latter is not. For ICP-MS, often a sample destruction is required to minimize matrix effects and to solubilize the sample. This makes the sample preparation time consuming. XRF on the other hand, does not require a time-consuming sample preparation. There are two main types of XRF spectroscopy, energy dispersive (ED) and wavelength dispersive (WD) XRF. An EDXRF instrument directly measures the different energies while a WDXRF instrument physically separates the fluorescence signals using a rotating crystal or monochromator. Due to this additional technology, WDXRF is more expensive and it takes longer to measure a spectrum, but the limit of detection and resolution are better as compared to EDXRF.
Besides XRF and ICP-MS there are several other techniques to determine the content of elemental impurities. Inductively coupled plasma – optical emission spectroscopy (ICP-OES), sulfide precipitation or atomic absorption spectroscopy (AAS) can also be used to analyze elemental impurities. The latter is not commonly used for elemental impurities in APIs since only one target element can be measured at the time, while ICP-OES is commonly used since it can detect multiple elements at the time. Before ICP-MS and ICP-OES became the main techniques, sulfide precipitation was used. However, this technique was not able to differentiate between different elemental impurities.
Generally, the use of ICP-MS is preferred over the use of XRF for the detection of elemental impurities in the pharmaceutical field due to its low detection limits, good precision and accuracy. XRF however, is gaining more interest since it is, compared to ICP-MS; cheap, fast and easy to use. The main issue with XRF was its detection limits and resolving power, but those are improving. Nevertheless, compared to ICP-MS, XRF is still not commonly used in the analysis of APIs. Several articles have been published about the use of XRF in the pharmaceutical field, but none of those studies showed a method which could be used as a generic method to analyze multiple elemental impurities in different types of APIs. This is mainly because the matrix, the API, can influence the results in an XRF analysis and therefore for each API a suitable calibration curve must be prepared.
5 3 Background
In the production of active pharmaceutical ingredients (APIs), elemental impurities can be introduced into the API through for example metal catalysts, raw materials or from the production environment (e.g. equipment). As these elemental impurities do not provide any therapeutic benefit to the patient and may even pose toxic side effects, their levels in drug product (and therefore drug substance) should be controlled within acceptable limits. Based on the toxicity of the element, the route of administration and daily dose, acceptable limits are determined.
Palladium for example is often used as a metal catalyst in coupling reactions. After the reaction the catalyst is removed. However, small quantities of catalyst may remain in the reactor. It is known that exposure to toxic levels of palladium can cause acute toxicity or an allergic reaction with respiratory symptoms. Nickel on the other hand can lead to chronic bronchitis and cancer of the lung and nasal sinus. The possible toxic effects of the elemental impurities are described in Table 1. [1], [2]
All medicines to be marketed must follow the pharmaceutical guidelines which apply in the country were the product is (or will be) on the market. According to the United States Pharmacopoeia (USP), Inductively Coupled Plasma - Mass Spectrometry (ICP-MS) and Inductively Coupled Plasma – Optical Emission Spectrometry (ICP-OES) could be used to measure elemental impurities in APIs accurately at ppm level (or even lower). Other instruments can be used if the validation requirements are met. The European Pharmacopoeia (Ph. Eur.) does not describe a specific method. However, it mentions that several techniques may be used, including ICP-MS and XRF, considering that the method has been verified by a system suitability test or validation.
Based on the applicable guidelines, quantification of the elemental impurities requires analytical testing using ICP-MS, ICP-OES, or an alternative procedure which must be validated and meet the acceptance criteria, on the API (final release sample). ICP-MS is a rather time-consuming analysis. Prior to analysis, the material needs to be digested to reduce matrix influences. Since XRF is more readily accessible, the technique can be used for research and development (R&D) support and in process controls (IPC). The elemental impurities of interest, in the context of elemental impurity determination, are the analytes. The analytes are present in the API, which in turn means that the sample matrix is the API.
The main focus of this study is on ICP-MS and XRF spectrometry for the analysis of elemental impurities in active pharmaceutical ingredients to illustrate two completely different techniques to measure elemental impurities in APIs at ppm level. Both techniques are described in detail, including the advantages and disadvantages of each technique in the pharmaceutical field. The main analytical challenge is the accurate detection of elemental impurities at ppm level in active pharmaceutical ingredients.
The following questions were answered: Is the technique able to detect at the limits described in ICH guideline? Are there signals which interfere with elements of interest? Is there a matrix effect and can the matrix effect be removed or reduced? Is the technique suitable to be used in a pharmaceutical production and/or R&D environment?
Furthermore, other techniques like ICP-OES, sulfide precipitation and atomic absorption spectroscopy (AAS), which can also be used to analyze elemental impurities are briefly discussed. [3], [4]
6 4 Elemental impurities
Elemental impurities can be introduced during the production of APIs through environmental contaminants, by using catalysts in synthesis which could not be removed entirely at the end of the process(step), by residues in starting materials or they may be present as impurities (e.g. through interactions with processing equipment or container closure systems). Traces of inorganic impurities can reduce drug stability and shelf life of some pharmaceutical products by catalyzing the degradation of the API. [5]
Furthermore, as these metallic contaminations (as metals or as metalloids) do not provide any therapeutic benefit to the patient and may be toxic at certain levels, their levels in drug product and therefore in drug substance, intermediates and raw materials, should be controlled within acceptable limits and be documented. These limits are based on the toxicity of the element, the route of administration (e.g. oral and injectable) and permitted daily exposure (PDE). For each API a risk assessment must be conducted to verify which elemental impurities have to be monitored throughout the production process. [6]
Before 2008, the elemental impurities limits were based on a maximal concentration in the drug substance. This approach did not take the drug dosage into account and therefore, people who received a higher dose (or multiple doses) of a drug were allowed to be exposed to a higher amount of elemental impurities than people who received a lower (or single) dose. Thus, to incorporate the dose, the European Medicines Agency (EMA) introduced the concept of PDE in 2008. Based on toxicology data, the PDE of each elemental impurity was determined. Using the maximal administered dose, the maximal concentration of the elemental impurities can be calculated in the drug product (e.g. API and excipients). The elemental impurities chapters in the Ph. Eur., USP and Japanese Pharmacopoeia (JP) were harmonized in 2009 by initiation of the International Counsel for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH). This new approach generally led to lower maximum levels in APIs which requires the use of analytical methods that are able to measure those lower levels qualitatively and quantitatively. [7], [8]
Raw materials are generally the main source of introducing elemental impurities and these elemental impurities are often hard to remove during the production process. Therefore, elemental impurities determination is often part of raw material testing as well. However, the acceptable limits of the concentration of elemental impurities are only set for the drug product since that contains the content of elemental impurities where the patient is exposed to. Acceptable limits of the concentrations in the drug product vary between 0.15 and 100 ppm, depending on the element. The elemental impurities are classified in different classes (class 1, 2A, 2B and 3; see paragraph 4.1). [3], [4]
PDEs of iridium, osmium, rhodium and ruthenium are determined based on the similarities with palladium since there is insufficient published data which describes the toxicity effects of these elements. The PDEs of all other elemental impurities are based on the route of administration, relevant (animal) studies, safety data, and the likely oxidation state of the element in the drug product.
4.1 ICH Classes
The ICH guideline divides the elemental impurities in four different classes. This is based on the natural abundance, probability of being used in the manufacturing process, probability of being co-isolated and the toxicity of the element (see Table 1). The toxicity of elemental impurities can lead to various toxic effects on the human body, which include but are not limited to various types of cancer, damage to the cardiovascular system and damage to the brain. [3], [4], [9]
7
Table 1, Overview of the ICH Class, Permitted Daily Exposure Limits and Toxic Effects for Elemental Impurities in Drug Products [2] – [10]
Element ICH Class Oral PDE (µg / day) Parenteral PDE (µg / day) Inhalational PDE (µg / day)
At toxic levels it can cause
Cd 1 5 2 2 Damage to lungs,
kidneys and bones
Pb 1 5 5 5 Damage to brain,
kidneys, liver and bones
As 1 15 15 2 Cancer and skin lesions
Hg 1 30 3 1 Damage to nervous,
digestive and immune systems, lungs, kidneys,
skin and eyes
Co 2A 3.000 300 30 Lung and cardiac
problems
V 2A 100 10 1 Irritation of the eyes,
nose and throat
Ni 2A 200 20 5 Chronic lung disease
Ti 2B 8 8 8 Yellow nail syndrome
and hypertension
Au 2B 100 100 1 Kidney damage and skin
rash
Pd 2B 100 10 1 Acute toxicity
Ir 2B 100 10 1 Damage to eyes
Os 2B 100 10 1 Damage to lungs, thorax
and respiratory system
Rh 2B 100 10 1 Damage to respiratory
system
Ru 2B 100 10 1 Cancer
Se 2B 150 80 130 Brittle hair, deformed
nails, lose feeling and control in arms and legs
Ag 2B 150 10 7 Discoloration of skin
Pt 2B 100 10 1 Liver damage
Li 3 550 250 25 Tremor, increased
reflexes, dizziness, kidney damage and
altered level of consciousness Sb 3 1.200 90 20 Irritation to respiratory system, pneumoconiosis and gastrointestinal symptoms Ba 3 1.400 700 300 Gastroenteritis, hypopotassemia, hypertension, cardiac arrhythmias and skeletal
muscle paralysis
Mo 3 3.000 1.500 10 Damage lung, kidneys
8
Cu 3 3.000 300 30 Nausea, vomiting,
stomach cramps, diarrhea and kidney and
liver damage
Sn 3 6.000 600 60 Nausea, vomiting,
abdominal cramps and diarrhea
Cr 3 11.000 1.100 3 Cancer, respiratory
disease and skin irritation
Arsenic, cadmium, mercury and lead are the four elements which have no or limited use in the manufacturing process of pharmaceuticals and are considered as human toxicants. These four elements belong to Class 1 and their presence in drug substances is most likely introduced by (mined) raw materials. Due to the toxic nature of these elements they must always be included in the risk assessment of the product and thus always require determination of their levels in APIs. Lead for example, is known to be harmful for the human brain, especially for children. Furthermore, it increases the blood pressure and it can develop tumors in the kidney(s). [9] Class 2 is divided in two subclasses, Class 2A and 2B. Both classes are route-dependent toxicants. In subclass 2A are the elements copper, nickel and vanadium which have a high probability of occurrence in drug substances. While subclass 2B elements (e.g. silver, gold, palladium and platinum) have a reduced probability of occurrence in the drug substance due to a lower chance of co-isolation with other materials and a low natural abundance. These elements only must be measured if it is used during the process or when it is present in starting materials. However, the guidelines do recommend measuring the three Class 2A elements even though they are not used in the process.
Class 3 elements have a low toxicity and therefore these elements do not have to be monitored. Barium, lithium and tin are examples of Class 3 elements.
Elements of the fourth group are amongst others: aluminum, calcium, iron, potassium, magnesium, manganese, sodium, wolfram and zinc. These elements are not categorized since the DPE for these elements have not been established globally. These elements are, due to differences in regional regulations and/or their low toxicities, covered in guidelines of regional regulations and not by the generic ICH guideline. [3]
4.2 Specifications
Until 2018, USP chapter <231> was effective. That chapter described the heavy metal test by sulfide precipitation (see paragraph 7.2). Increasing knowledge about the toxicity of various elemental impurities and the limited qualitative and quantitative capabilities of the sulfide precipitation led to replacement of chapter <231> by chapters <232> and <233>. In USP chapter <232> Elemental Impurities – Limits, the required specifications are described. In USP chapter <233> Elemental Impurities – Procedures, the analytical procedures for the evaluation of elemental impurities that are suitable for detecting the limits are described. In these two chapters it is stated that the procedure used must be able to detect the target element(s) unequivocally in the sample and the relative standard deviation of the drift (standard measured before and after the samples) must be ≤ 20% for each target element. Furthermore, the recovery for spiking experiments, which indicates if there is a matrix effect, at various concentrations (50 – 150%), must be within 70 – 150%. These two USP chapters are harmonized with Ph. Eur. chapter 2.4.20 and JP chapter 2.66.
9 5 X-Ray Fluorescence
X-Ray fluorescence is a technique which can be applied to detect elementals impurities in active pharmaceutical ingredients. X-Rays are on the high energy side of the electromagnetic radiation spectrum (see Figure 1). Elements from sodium to uranium can generally be detected using XRF in tablets, liquids and powders. For APIs the latter is most frequently measured since APIs are often provided in powder form to be able to mix it (with for example binding agent, fillers, lubricant, stabilizer, flavors, colors, etc.) to form the drug product. The detection limit is ranging from parts per million to percentages depending on the instrument, the matrix and the element of interest.
Figure 1, Electromagnetic Radiation Spectrum from Radio Waves to Gamma Rays
Generally, the XRF source consists of an in-situ X-Ray generator. The X-Ray beam is aimed at the sample, with or without a filter to modify the beam. The X-Ray beam has a certain penetration depth, which is depending on the elemental composition, density of the sample, thickness of the sample and the power of the X-Ray beam (the more powerful the beam, the deeper the penetration depth). The atoms which are hit by the beam, with more energy than the binding energy of the involved electrons, are excited from the ground state to a higher energy state. The electron from the inner shell which is removed makes the atom unstable. As a result, an electron from the outer shell transfers to the inner shell to fill the vacancy and a specific amount of energy, a secondary Ray, is emitted (see Figure 2). The secondary X-Ray, which is a fluorescence signal, is emitted and will result in a peak. The energy of the fluorescence signal is characteristic for a specific element. The amount of counts of a specific energy is then plotted against the energy (in KeV). The peak area correlates with the concentration of an element and therefore XRF can also be used as a (semi) quantitative method. [11]
10
Figure 2, Effect of an X-Ray Beam on an Atom. An Electron is Ejected by the X-Ray Beam and the Vacancy is filled by an Electron from an Outer Shell
The vacancy of the electron which is ejected from the inner shell will be filled with an electron from an outer shell. This electron can be originated from different shells resulting in different types of transition energies (see Figure 3). They are labeled depending on the shell where the electron is transferred to (e.g. K, L, or M). If the electron moves one shell closer to the nucleus it is an α-transition and if the electron came from two or three shells further, it is respectively a β or γ transition. Since each element has multiple transition energies, some can overlap with those of other elements. Often the software of the XRF can correct for the overlapping energies since each transition energy is characteristic for an element and the ratio of the different types of transitions are known (see Appendix 2).
Figure 3, Electronic Transitions of filling a Vacancy from an Outer Shell after an incident X-Ray hit the Atom [12]
5.1 Energy Dispersive and Wavelength Dispersive X-Ray Fluorescence
There are two general types of XRF spectroscopy: Energy Dispersive XRF and Wavelength Dispersive XRF (WDXRF). An Energy Dispersive XRF (EDXRF) instrument directly measures the different energies and plots the relative number of X-Rays (counts, number of X-Rays with a specific energy level that is detected) against the energy. The individual X-Rays, which are formed after exciting the sample with a focused electron beam, are picked up by a detector and converted into proportional electrical voltages.
Each time an X-Ray hits the detector, the specific energy which is detected is plotted. Since the X-Rays are not physically separated in EDXRF, unlike with WDXRF, sum peaks can be observed. Sum peaks are the result of two X-Rays hitting the detector at the exact same time.
11 Because they hit the detector at the same time the pulse processor is not able to register them as individual events and the detector will sum their energies (see Figure 4). In Table 2 it is shown that the sum of X-Rays of some elements overlap with the X-Ray energy of another element. The probability of sum peaks decreases with a lower current and shorter processing time.
Figure 4, Sum Peak of two Fe Kα Photons [13]
Table 2, Sum Peaks which will appear in the Spectrum as a different Element
Sum Peaks Element
Mg + Ca V
Si + Ca Cr
Mg + K Ti
Al + Si K
Ca + Ca Ni
In Figure 5 it is shown that the X-Ray source irradiates the sample which then absorbs the light and subsequently emits energy at a longer wavelength (fluorescence). This fluorescence signal is detected and processed by the digital pulse processor (DPP) and the counts (number of X-Rays at a specific energy level) are displayed on the computer resulting in an XRF spectrum. To improve the signal to noise ratio (S/N) in the measured voltages the noise is reduced by averaging the signal over a period of time, known as the processing time. Since the noise is random, it is averaged out and the signal is increased. Thus, the longer the processing time, the more the noise is reduced, and the element signal is increased. However, a longer processing time also increases the need for dead time. The dead time is related to the input and output rate of the detector. The input rate is the rate at which the X-Ray pulses are detected and the output rate (acquisition rate) is the rate at which process pulses are send to the computer for analyzing the data. If the input rate is greater than the output rate, the pulses will pile-up since the detector is not able to process each signal before the next pulse is detected. The dead time is a period of time where the pulses are not measured. After an X-Ray hits the detector, the dead time is started. During the dead time the signal is converted to a voltage. After that, the dead time is stopped allowing a new X-Ray pulse to hit the detector. In this way a pulse pile-up in the detector is minimized, but this also results in fewer pulses which can be measured and thus leads to a lower intensity. If the dead time is set too long this can lead to inaccurate data.
Depending on the detector energy, the resolution is ranging from 150 eV to 300 eV. This is a measure of how closely lines can be resolved in a spectrum, a lower resolution number result in sharper peaks. The detection limits are usually in the range of 1000 to 10000 ppm. Sodium is often the lightest element which can be measured using EDXRF. This is due to fact that light elements have low energy levels (< 1 keV). So it is harder to eject an electron without the electron being absorbed before it reaches the detector, or the ejected electron does not have enough energy to reach the detector at all.
12
Figure 5, Schematic overview of an Energy Dispersive X-Ray Fluorescence System
Figure 6, Schematic overview of a Wavelength Dispersive X-Ray Fluorescence System
Wavelength Dispersive XRF (WDXRF) on the other hand physically separates the X-Rays according to their wavelengths using a rotating crystal or monochromator. In Figure 6 the setup with a crystal is shown. The fluorescent signals originating from the sample are scattered in different layers of atoms in the crystal causing some beams to travel over a longer optical path. These beams have a specific angle and are all in phase which means that constructive interference takes place. By moving the crystal, the angle is changed which allows measuring a wide range of different wavelengths. The detector converts the X-Ray signals to proportional voltages.
Due to random fluctuations in the voltage signal, so-called shot noise, the voltage signal has a relatively large noise signal compared to the element signal. By averaging the incoming signal, the noise is reduced by the pulse processor. The longer the processing time, the more noise can be reduced resulting in an improved spectral resolution as described earlier. The obtained counts are then plotted against the energy level to get a figure which shows the characteristic peaks. These peak areas correspond to the concentration of the elements. [14]
WDXRF has a high spectral resolution which is between 5 eV to 20 eV, depending on the system set up and used crystal. By using a WDXRF set up, the spectrum is built up point by point and therefore it takes longer to measure the spectrum. This also requires a higher source power compared to EDXRF and the source tube must be cooled since most of the power is dissipated as heat. Cooling the detector will minimize the electric noise. To reduce the measuring time multiple crystals and detectors can be used to measure multiple wavelengths simultaneously. However, generally only one detector is used due to extra costs of multiple crystals and detectors. [15]
Compared to EDXRF, WDXRF is relatively expensive due to the system setup with the moving crystal(s). Furthermore, the acquisition of the spectrum is limited by the number of simultaneous detectors. But the advantages are that the peaks are narrower (0.2 – 0.3 keV, full width at half maximum) and the resolution with WDXRF is significantly higher as compared to EDXRF (see Figure 7). Due to this higher resolution, WDXRF provides better sensitivity and therefore lower detection limits are achieved, also towards low atomic number elements which are known to have a lower resolution since a smaller molecule has less internal energy and is therefore less detectable. Using WDXRF, beryllium to uranium can be measured versus sodium to uranium for EDXRF. [16]
13
Figure 7, Comparison of the Resolution of an EDXRF spectrum with a WDXRF Spectrum [17]
To increase the resolution of EDXRF, the X-Ray beam can be placed at an angle below the critical total reflection angle (generally set between 0.05 – 0.5°) instead of 45° in an EDXRF instrumental set-up. This angle can be calculated using equation 1. Due to this small angle, most of the excitation beam photons are reflected towards the detector (see Figure 8). The background signal, however, is lower compared to the conventional EDXRF because of less scattering. This XRF set-up is called Total Reflection XRF (TXRF) and is gaining increasing interest. Using TXRF requires only a thin layer of sample and this set-up is suitable for trace level analysis, including the analysis of elemental impurities in APIs. Due to this thin layer, matrix effects are negligible. A comparison of the spectra of an Multi element standard (MES) using TXRF and EDXRF is shown in Figure 9. [18]
𝛼𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙≈ 1.65
𝐸 × √
𝑍
𝐴 × 𝜌 (1)
Where:
αcritical = Critical angle
E = Photons energy in keV
Z = Atomic Number of the reflector A = Atomic Mass in g/mol
ρ = Density in g/cm3
Figure 8, Instrumental set-up of a Total Reflection XRF [19]
14
Figure 9, Comparison of an TXRF and EDXRF spectrum of a 1000 ppm multi element standard (MES) [20] 5.2 XRF Measurement
There is little to no sample preparation necessary for an XRF measurement of an API powder. The sample only requires homogenization and the sample needs to have a smooth surface, since XRF only measures the surface of the sample. Typically, a few grams are needed for an XRF measurement. If needed the sample is grinded to obtain a homogeneous sample of fine powder. The sample is placed in a sample cup and compressed to obtain a smooth surface. XRF is a non-destructive technique which means that the sample could be used for different purpose (e.g. analyses) after the XRF analysis. [22]
An external calibration curve must be measured in order to quantify elements. Typically, cellulose is used as matrix for the calibration curve. The calibration curves must be prepared after installation of the equipment or if (an essential part of the) equipment is replaced. It must be considered that the calibration curve is dependent on the sample matrix. Thus, it would be best if the calibration curve is measured in each matrix (API). However, often it is not possible to receive the matrix without the element of interest. Furthermore, it is time consuming to prepare a calibration curve for each new API for a company which only produces a limited number of batches per API, while XRF its main selling point is its fast analysis time.
The sample preparation for a TXRF experiment is slightly different compared to EDXRF and WDXRF. First, the solid sample is grinded into small particles to obtain a homogeneous sample. Second, the sample is diluted to a suspension and an internal standard is added to be able to quantify the elements. The use of an internal standard eliminates the need for an external calibration curve which can be a time-consuming process since a suitable matrix has to be used to prepare the calibration curve in. After that a few microliters of the sample suspension is pipetted on a disc (e.g. quartz glass). Then, the sample is allowed to dry under heat and/or vacuum. Subsequently, the sample is measured by the TXRF. [21]
Commonly used X-Ray sources are rhodium, silver and palladium. The energies of the used X-Ray source will always be present in the spectrum unless a filter is used to remove those lines. A filter absorbs the fluorescence signal of the source. Since the signal of the source has a relative high intensity, the removal of the source element will not only remove the signal from the source but also the signal originating from the sample. Therefore, it is not recommended to
15 measure the palladium content of a sample while making use of a palladium source. However, if there is only one system available that makes use of a palladium source it is possible to correct for the background signal by using a blank measurement. Nevertheless, this will lead to a less accurate result compared to a palladium measurement using a different type of X-Ray source. [23]
There are different types of filters commercially available depending on the element which must be removed. It must be noted that the material of the filter will be observed in the spectrum. For example, if the filter material consists of titanium, a titanium signal will appear in the spectrum. [24], [25]
The lower limit of detection (LOD) is in the ppm range and generally lighter elements have a higher LOD compared to heavier atoms since lighter elements have lower energy levels (< 1 keV). If the API material is too dense, the X-Ray beam is not powerful enough to excite the entire sample. This is caused by the fact that the energy of the electron source is consumed in the outer part of the sample, so the penetration depth is less and consequently less sample is excited and detected. The linearity of the XRF is dependent on the element and the sample matrix. Generally, the XRF is linear up to 100% of the element of interest. However, this is not necessary since the elemental impurities are only allowed to be present at ppm level and no pure samples must be measured for elemental impurities detection.
An example of an EDXRF spectrum of an API is shown in Figure 10. This API has sulfur in the structure and zinc bromide is used in the last synthesis step to remove the tert-butyloxycarbonyl (Boc) group from the structure. This protecting Boc group is often used in the organic chemistry to protect an amine during the production process. The bromine and sulfur peaks are identified in the spectrum with respectively pink and light blue vertical lines. The blue curve is the background signal. It is shown in the spectrum that the sulfur signal is overlapping with other peaks. Sulfur shows two peaks in this picture. The largest peak is the Kα peak while the smaller peak is Kβ. The Kα of bromine is also the largest peak while the Kβ and Lα are smaller. Based on the peak areas compared to the calibration curve, the concentration of bromine and sulfur can be determined. [26]
16
Figure 10, Example of an EDXRF Spectrum with the Bromine and Sulfur Peaks identified 5.3 XRF used in the Pharmaceutical Industry
Compared to ICP-MS, XRF is barely used for the quantification of elemental impurities in APIs. Recently however, the use of XRF in the pharmaceutical field has increased. This is mainly due to its fast sample preparation compared to other techniques like ICP-MS and its improvements regarding detection limits. Margui et al. reported about the use of WD-XRF to detect palladium in active pharmaceutical ingredients. No chemical sample preparation was required, and the calibration standards were made of cellulose spiked with palladium. This resulted in a reliable calibration curve and the acceptance criteria described by the ICH guidelines were met. However, it should be noted that only one API was described in this article for the detection of only one element (palladium).
In 2017, an article was published about the analysis of Class 1 and Class 2A elements, except for mercury since it evaporates easily, using an EDXRF. A spectrum of 100 µg/g lead standard in cellulose is shown in Figure 11. Here it is shown, as expected, that the α- and β-transition happen more frequently than the γ-transition. The γ-transition peak is too small to be used for quantification, therefore, the α- and β-transition peaks are used. A different recent article from September 2020, did describe a quantitative EDXRF method which was able to detect multiple
17 elements. Using this method, Class 1 (Pb, Cd, Hg and As) and Class 2A (Co, V, Ni) elements were analyzed according to the ICH guidelines. Nevertheless, this method was also developed for only one API (angiotensin receptor blocker). [27], [28], [29]
Figure 11, XRF spectrum of 100 µg/g lead in cellulose matrix
Another study describes the use of XRF during the development process. Here ICP-OES is used for the final release sample, but during the development of the production process, XRF was used since it is a rapid technique which is, relative to ICP-OES and ICP-MS, cheaper and does not require highly trained personnel. However, it is described in literature that a significant matrix effect is observed in the XRF analysis. This matrix effect must be considered when selecting an appropriate compound for the calibration standards. As a result, for each new compound a new calibration curve must be made using a suitable matrix and this is a time-consuming process. [30], [31]
Furukawa and Davis described an EDXRF method for the analysis of Class I, Class 2A elements and Ir, Pt, Rh, Ru and Pd of Class 2B. These Class 2B elements were chosen since they are commonly used in the production process of APIs. Compared to other XRF studies regarding analysis of elemental impurities, this was an extensive study (see Table 3). This method was able to detect the mentioned elements if 1 gram per day is the maximum daily intake of the drug product. [32]
Table 3, Overview of various studies using XRF for the analysis of elemental impurities
XRF type Element(s) Number
of APIs
Reference
WDXRF Pd 1 [27]
EDXRF Pb, Cd, Hg, As, Co, V, Ni 1 [28]
EDXRF Cd, Pb, As, V, Co, Ni 1 [29]
WDXRF Fe, Zn, Cr, Ni 2 [30]
EDXRF Pb, Cd, Hg, As, Co, V, Ni, Ir, Pt, Rh, Ru, Pd
5 [32]
TXRF Rh, Pd, Ir, Pt 2 [33]
Generally, EDXRF is preferred over WDXRF for the analysis of elemental impurities because multiple elements can be measured at once instead of sequentially. Furthermore, EDXRF is more suitable for the measurement of volatile elements like mercury owing to the lower X-Ray
18 intensity which is required. In another article, Margui et al. described a TXRF method to analyze Rh, Pd, Ir and Pt in APIs according to the ICH guidelines. It was found that spectral interferences disturbed the quantification of palladium using the Pd-L shell line. However, when the Pd-K shell line was used, palladium could be quantified. The three other elements of interest did not show overlapping lines using the W – anode tube. However, for other elements the W – anode might not be suitable due to overlapping lines, but another source (e.g. Mo) is preferred. Some instruments can handle multiple sources which is beneficial if multiple elements must be measured which require a different source. However, this setup is more expensive. [33], [34], [35]
19 6 Inductively Coupled Plasma – Mass Spectrometry
In an ICP-MS instrument, the liquid sample is pumped into the sample introduction system which consists of a nebulizer, spray chamber and an ICP torch (see Figure 12). The sample is mixed with argon gas in the nebulizer and is thereby forming aerosols. While the larger droplets are removed by the cooled spray chamber, the smaller droplets are swept into the central channel of the plasma. As they travel through the high temperature plasma, the aerosol droplets are dried, decomposed, atomized and then ionized resulting in positively charged ions. [36]
Figure 12, Schematic overview of an Inductively Coupled Plasma – Mass Spectrometer
The ions are extracted from the plasma into the sample cone using high vacuum. The ions are focused by electrostatic lenses which remove the neutral atoms by bending the ions towards the reaction cell while the direction of the neutral atoms remains unchanged as is shown in Figure 13.
Figure 13, Ions from the ICP Beam are bended by the Electrostatic Lenses towards the Reaction Cell
In the reaction cell, the ions collide with a collision gas. Only ions with sufficient amount of energy will enter the mass filter. Both helium and nitrogen can be used as collision gases since they are inert gasses. However, more scattering is observed if nitrogen is used due to the larger size compared to helium and therefore helium is preferentially chosen as collision gas. Furthermore, helium is a light gas and therefore has a higher diffusion rate. This results in the removal of polyatomic interferences. Due to the fact that polyatomic compounds are larger than the analyte, they will collide more often with helium which results in a lower amount of kinetic energy. A potential barrier at the collision cell exit only allows the higher energy analyte ions to pass through to the mass analyzer (see Figure 14). [7]
20
Figure 14, Schematic effect of using Helium gas in the Collision Cell on the Analyte and Polyatomic Compound
The analyte ions leaving the collision cell enter the mass filter and are separated based on the mass to charge ratio. Only the target ion is stable and passes along the axis of the mass filter. Rapidly, all the different masses are scanned from low to high throughout the run sequentially and send to the electron multiplier detector. The electronic signals are processed by the detector creating a mass spectrum where the concentration is proportional to the intensity of the mass.
6.1 Ionization Potential
Each atom has a different ionization potential as is depicted in Figure 15. Some elements (e.g. calcium, strontium, barium and cerium) may have a small population of doubly charged species due to a relatively low second ionization potential (< 12.5 eV). Since the mass filter separates the ions based on their mass to charge ratio, any doubly charged ion will appear at half the nominal mass (i.e. 88Sr2+ interferes with 44Ca+ at mass 44). Most atoms have several natural isotopes so these can also be detected at different masses. The natural abundances of all atoms can be found in Appendix 1. The higher the ionization potential of an element, the harder it is to ionize it and therefore the sensitivity is of that element is lower. [37]
21
6.2 ICP-MS Plasma
The plasma has a high density of free electrons. To obtain an efficient ionization, the droplets which enter the plasma should be small, so they have a lower density and are therefore less likely to overload the plasma. Furthermore, the temperature of the plasma also influences the ionization. In Figure 16, it is shown that a higher temperature is able to ionize elements with a higher ionization potential. The higher the temperature, the better the atoms will ionize. The plasma temperature is directly related to the formation of Cerium oxide (CeO) and therefore the ionizability of the sample. This is expressed in the CeO/Ce ratio. The lower the CeO/Ce ratio, the hotter the plasma. The formation of Cerium oxide is determined during the tuning of the system. The tuning solution contains cerium and it is used to test the quality of the plasma.
Figure 16, Degree of Ionization at different Plasma Temperatures
Operating at a low CeO/Ce level (< 3%) also ensures good dissociation of sample matrix which means less deposition of matrix on the interface. This in turns means better stability during long runs of high matrix samples. Since Ce – O is a strong bond, the CeO/Ce ratio indicates the efficiency of the plasma to decompose the Ce – O bond and other oxide interferences.
A higher CeO/Ce ratio indicates that the plasma is overloaded by the matrix and has insufficient residual energy to fully ionize the analytes. Signal loss due to suppression is greater for poorly ionized elements. To illustrate the effect of the CeO/Ce ratio the signal suppression of 0.3% NaCl at two different CeO/Ce ratios is shown in Figure 17. Here it is shown that the suppression at a higher CeO/Ce is more severe than at a lower ratio. Suppression of the signal leads to a lower limit of detection and less accurate results. [38]
22
Figure 17, Suppression of Signal in 0.3% NaCl at 1% CeO/Ce ratio and 1.7% CeO ratio
Besides the spectral interferences discussed earlier also physical interferences (e.g. spray formation and ionization) may affect ICP-MS performance. These are caused by the matrix and can cause a drift in the results, loss of sensitivity and accuracy. [39]
6.3 Measurement of Impurities by ICP-MS
To measure elemental impurities in APIs, the USP recommends digesting the sample by microwave destruction. Hereby, the matrix (the API) is decomposed/destroyed using concentrated acids (e.g. nitric acid and hydrochloric acid) to suppress matrix effects. Furthermore, the addition of acids is used to solubilize all elements. Some compounds (e.g. talc) require the addition of hydrogen fluoride (HF) to solubilize the sample. However, extra safety precautions might be needed due to the highly corrosive and toxic properties of HF which are more severe than with nitric acid (HNO3) and hydrochloric acid (HCl). To minimize loss of volatile elemental impurities (e.g. and Hg), closed vessels are used when digesting the sample. It is known that mercury, due to its volatility, can be absorbed in the introduction system of the ICP. Often HCl is used to stabilize mercury in the solution. Therefore, most destruction sample preparation uses a combination of nitric acid and HCl. Two kinds of destruction tubes can be used: quartz and an inert polymer (e.g. PFA or TFM). The latter should be used compounds which contain fluoride since HF is formed during the sample digestion and it can etch the quartz material. Generally, 100 – 250 mg sample is used for an ICP-MS analysis. [7], [40] – [42] If the residual carbon content (RCC) is too high, matrix effects can occur. Therefore, it is recommended to test if the digestion was complete by measuring the amount of nitrogen and carbon the first time an API is measured. To obtain a complete digestion the carbon counts should be low indicating that the sample is completely digested, and the matrix effect is minimized. Furthermore, since nitric acid is consumed during the digestion, the nitrogen counts should be high indicating that there was enough acid available to decompose the material. [43] The elements are identified based on their masses and the elements are quantified by using an internal standard. The internal standard is added equally to all samples (e.g. blanks, standards and samples). If signal suppression occurs, it will impact both the internal standard signal and analyte signal. However, matrix effects are not linear over the mass range so not all ions are equally suppressed. Therefore, an internal standard with a mass near the analyte of interest should be used to correct for signal suppression and enhancement interferences. If multiple elements are measured, the internal standards need to be spread across the mass range of the analytes being determined. This typically includes low, mid and high mass ranges.
23 Furthermore, it is important that the internal standards are fully ionized and are not present in the sample itself (e.g. 6Li, 45Sc, 89Y and 209Bi).
The matrix, in this case the API, can influence the signal suppression. Each matrix can have a different effect and therefore it is of vital importance to test if the method (digestion method and selection of internal standards) is suitable for each matrix. This is tested by conducting spike recovery experiments on each sample matrix (API). To minimize difference in signal suppression between the samples, standards and blanks, it is important that the content has the same concentration of acids. [44]
ICP-MS peaks have a Gaussian shape and are typically about 0.8 mass units wide which results in baseline separated peaks. When a full mass spectrum is shown the peaks are generally depicted as vertical lines (see Figure 18).
Figure 18, Example of a Mass Spectrum obtained with ICP-MS of a Standard Solution containing 50 µg/L Multi-Element Standard, 50 µg/L Internal Standard Rhenium, 0.5 µg/L Mercury Internal Standard and 500 µg/L Lutetium Internal Standard [45]
6.4 ICP-MS used in the Pharmaceutical Industry
ICP-MS is described in the literature as a robust and reliable method. Often a generic method is used for multiple elemental impurities and this method can be used to a wide variety of APIs (see Table 4). Class 1 and Class 2A elements are most often requested and therefore these elements are generally analyzed in a generic method which can be used for multiple APIs. [46], [47]
Table 4, Overview of various studies using ICPMS for the analysis of elemental impurities
Element(s) Number
of APIs
Microwave destruction?
Reference
Class 1, Ir, Os, Pd, Pt, Rh, Ru, Cr, Mo, Ni, V, Cu, Mn, Fe, Zn
2 Yes [46]
All USP elements 1 Yes [48]
Class 1 3 Yes [49]
Class 1, Se, In, Sn, Sb, Bi, Ag, Pd, Pt, Mo, Ru
53 No [50]
Cd, Cr, Cu, Ir, Mo, Ni, Os, Pb, Pd, Pt, Rh, Ru, V
6 Yes [51]
An example of a generic method is described by Chahrour et al. Here the development and validation of an ICP-MS method for all the elemental impurities described by the USP measured
24 in one analysis, is meeting the requirements described in the ICH guideline. In the development of this method, extra attention was given to the analysis of osmium. It is known that osmium easily oxidizes in various forms during the sample digestion in the presence of nitric acid. Due to its volatility osmium is lost, resulting in low recoveries and underestimation of the true osmium content in the sample. However, hydrochloride is often used in combination with nitric acid to digest the sample and chloride is capable of stabilizing osmium. By the addition of hydrochloride, OsCl62- is formed which is soluble and less volatile. To improve the recovery of osmium further, ascorbic acid and thiourea were added owing to their reducing properties. Osmium is a Class 2B element and therefore, the osmium content only must be tested if osmium is used in the production process. Consequently, the use of a stabilizing mixture for the analysis of osmium is generally not used in a generic method. Hence, this mixture is only used if the detection of osmium is required. [48]
A study from 2014 showed that, for the determination of As, Cd, Hg and Pb (Class 1 elements), a digestion in closed vessels is preferred over a digestion method using open vessels (e.g. dry ashing method). When a dry ashing method is used, the sample is placed in an oven at circa 500°C. This method is efficient in decomposing the API, however, volatile elements like mercury are lost since an open vessel is used. To circumvent the loss of volatile elements the microwave – induced combustion (MIC) method or acid digestion in a closed vessel can be used (see Figure 19). With MIC, the sample is placed in a closed vessel and the digestion is initiated by microwave radiation. Due to the use of a closed vessel, the evaporation of the volatile elements is avoided. [49]
Figure 19, Schematic view of a closed vessel for microwave destruction [52]
Muller et al. described an article about the importance of testing the RCC in the digested material. In some cases, the digestion method is not able to decompose the entire sample resulting in inaccurate results. Furthermore, polyatomic ions containing carbon (e.g. 40Ar13C+) can occur. These polyatomic ions can be removed by using a collision or reactor cell, as described in Figure 14. The effectivity of the microwave digestion can be measured by the RCC. If the residual carbon content is too high, the temperature/pressure in the microwave vessels can be increased or the amount of sample can be decreased. Another option to improve the destruction of the API is too use more concentrated acids or stronger acids (e.g. HF). [51]
25 7 Other Techniques for Detection of Elemental Impurities
Besides XRF and ICP-MS, there are other techniques available to measure elemental impurities. Since February 2009, ICP-MS and Inductively Coupled Plasma – Optical Emission Spectrometry (ICP-OES) are mentioned as a suitable technique for elemental analysis according to the European pharmacopoeia. Before that, the heavy metals test was based on sulfide precipitation using thioacetamide. Furthermore, the USP states that if both ICP-OES and ICP-MS are not available for a specific application, another atomic spectroscopic technique, like Atomic Absorption Spectroscopy (AAS), could be used if the method meets the requirements regarding precision and the detection limit, amongst others.
7.1 Inductively Coupled Plasma – Optical Emission Spectroscopy
ICP-OES also referred to as Inductively Coupled Plasma – Atomic Emission Spectroscopy (ICP-AES) is also used to analyze elemental impurities. ICP-OES is less influenced by matrix effect compared to ICP-MS since it has a higher tolerance for samples with high total dissolved solids (TDS). Therefore, the samples do not have to be diluted as much as for an ICP-MS analysis. Compared to ICP-MS, ICP-OES is a faster technique; circa 2500 samples per 24 hours can be measured compared to 1000 samples for ICP-MS. However, this calculation does not include sample preparation which is the same for both techniques. Generally, the sample preparation of APIs requires microwave destruction which is time consuming and thus mainly determines the total analysis time.
The principle of ionizing elements is similar to ICP-MS, but instead of detecting the elements based on their masses it is detected based on their emission energy. For a schematic overview of the basic components of an ICP-OES, see Figure 20. The plasma, which is 6000 to 7000 Kelvin, excites the elements after fragmentation to a higher energy state. When the atom falls back to the ground state a specific amount of energy is released. This energy results in an emission spectrum which is characteristic for each element. Based on the intensity of each wavelength the concentration can be calculated for all elements. The excitation and emission process happen within nanoseconds while a mass spectrum takes about microseconds. Therefore, an ICP-MS measurement takes longer than an ICP-OES measurement. [53]
Figure 20, Schematic overview of the Basic Components of an Inductively Coupled Plasma – Optical Emission Spectrometer
ICP-MS is able to reduce molecular interferences in the mass detector while this is not possible in ICP-OES and therefore spectral interferences (e.g. polyatomic interferences like 40Ar16O+) are more severe for ICP-OES.
26 Various articles are written about the development of a suitable ICP-OES method for the quantification of elemental impurities in APIs according to the ICH guidelines. Støving et al. were able to develop a method for the detection of all Class I elements and various others, like the commonly used catalysts palladium and nickel. Here it was described that it was necessary to add additional rinsing steps to prevent memory effect of mercury (Class 1 element) which can be absorbed by the introduction system of the ICP. The memory effect, which was observed however, was only present at a concentration that was significantly lower than the limit described by the authorities and therefore this was not considered as an issue. The memory effect of osmium (Class 2A element) on the other hand, did result in issues. This is caused by the fact that osmium easily forms OsO4 which can be absorbed by the ICP resulting in a memory effect or it will vaporize due to its volatility. Since this memory effect arises in the ICP, it is not only a challenge for the ICP-OES, but also for the ICP-MS. To prevent the oxidation of osmium from happening, the use of HNO3 as destruction acid must be avoided. Furthermore, less OsO4 is observed if the plasma temperature, and thus efficacy, is higher as discussed in paragraph 6.2. Klose et al. described a method to stabilize osmium after the digestion by addition of a combination of ascorbic acid, thiourea and EDTA. [54], [55]
7.2 Sulfide Precipitation
Before ICP-MS and ICP-OES were commercially available, heavy metals were measured for years by sulfide precipitation, the so-called heavy metal test. Before circa 1950, hydrogen sulfide was used for the hydrolysis, but due to safety reasons it was replaced by thioacetamide (C2H5NS). Thioacetamide is stable, both as solid state and as solution, but it hydrolyzes quickly at a temperature above 80°C and it is used to precipitate metallic impurities. After the hydrolysis reaction, the metal impurities colors the solution by forming metallic sulfides. Metals that are colored by sulfide ions are lead, mercury, bismuth, arsenic, antimony, tin, cadmium, silver, copper, molybdenum and selenium. The color of the sample is visually compared to a lead standard (10 ppm) and based on that it is concluded if the sample contains less or more metals than the lead standard. [56], [57], [58]
This technique assumes that all samples react in the same manner as the lead standard and therefore this technique assumes to have no matrix effect. However, it is known that the colloids from different metals can behave differently. Besides that, this method cannot differentiate between metallic impurities if there are multiple metal impurities present in the solution. Another issue with this technique is that colored metals like copper can color the solution leading to a false positive result. Furthermore, Rosenthal and Taylor described already in 1957 that the reaction was also pH dependent which makes it harder to compare it to a standard. In 1995, Blake reported that frequently 50% of the metals, or even more, were lost during the sample preparation and that especially the analysis of the toxic element mercury was prone to have extremely low recoveries (towards 0% recovery). Therefore, from 2018 onwards, the authorities do not consider this method as a suitable method for the qualitative and quantitative analysis of elemental impurities in APIs anymore. [6], [59], [60]
Compared to XRF a similar sample amount is needed to perform the sulfide precipitation, do note that sulfide precipitation is a destructive analysis. The reaction time, which is generally 2 min, is important since it must be sufficient to precipitate the sample, but not too long for the sample to become unstable. Furthermore, it is important that only one analyst compares the solutions at a predefined reaction time since visual interpretation can differ from analyst to analyst while the reaction continues. Due to all these restrictions it is not possible to detect metal impurities at ppm level. Therefore, since ICP-MS and ICP-OES have been commercially available, the guidelines recommend using ICP-MS or ICP-OES instead of sulfide precipitation. [61]
27
7.3 Atomic Absorption Spectroscopy
Flame Atomic absorption spectroscopy (flame AAS) is a technique where the sample is ionized using a flame. Using a constant gas flow in a nebulizer the sample solution is converted into aerosols and transferred into the flame. Depending on the analyte the light source is chosen. The radiation source sends light to the source and into the monochromator. In the monochromator the light is separated by mirrors and gratings (see Figure 21). Another option to ionize the sample is using a graphite tube instead of a flame. This graphite tube has a heating program designed to remove the solvent and atomize the sample. Since the entire sample is atomized the sensitivity is increased compared to flame AAS. However, due to the heating program the measurement time takes longer.
Figure 21, Schematic set up of a Flame Atomic Absorption Spectrometer
Each element absorbs light at a specific wavelength and based on this, the analyte is isolated by the monochromator and measured by the detector. Depending on the concentration of analyte in the flame the light is absorbed. The higher the concentration of atoms in the sample, the more light will be absorbed, and a decrease of light is detected. The concentration of the analyte will be determined based on a linearity curve. Lamberts Beer’s law is used to calculate the concentration (Equation 2). [62]
𝐿𝑎𝑚𝑏𝑒𝑟𝑡 𝐵𝑒𝑒𝑟′𝑠 𝐿𝑎𝑤 𝐴 = 𝜀 ∗ 𝑐 ∗ 𝑙 = log(I0/𝐼) (2) Where: 𝐴 = Absorbance 𝜀 = Absorption coefficient (M-1 * cm-1) 𝑐 = Concentration (M) 𝑙 = Path length (cm)
I0 = Initial light intensity before absorption 𝐼 = Light intensity after absorption
AAS is a suitable technique for measuring complex matrixes and the analysis costs are only half the costs of an ICP-MS analysis. However, the detection limit is often not enough to meet the acceptance limits described by authorities for final drug product testing, but this technique could be used to characterize high concentrations in raw materials. Sample preparation is often quite similar for AAS and ICP-MS. For both analyses, the sample must be in solution and acidification is commonly used for APIs. The main disadvantage of AAS compared to both XRF and ICP-MS, is that the technique is not able to measure all Class 1 and 2A elements at the same time. [6], [47], [63]
Zhang and Wang described a suitable flame AAS method for the analysis of copper in Wenglitong capsules. These capsules are used for the treatment of prostatitis. This study,
28 which was published in 2008, met the acceptance criteria for copper. However, this method was only suitable for the detection of copper and it was tested for only one API. Gomez et al. did describe a flame AAS method which was used for multiple elements. Here the samples were digested using strong acids and for each element different settings were used, illustrating the inability of the AAS to detect all elements simultaneously. [64], [65]
In 2018, an article was published about the use of a high-resolution continuum source graphite furnace atomic absorption spectrometer for the measurement of elemental impurities. This method was developed as a screening method to analyze if an elemental impurity is present in the API or not. After the screening the present elemental impurities were quantified. This method was developed for twelve elements including Class 1 and 2 (A and B) elements. Unfortunately, mercury and arsenic of Class 1 were not tested in this study even though all Class 1 elements must be tested in an API according to the guidelines. Furthermore, cobalt is also not included in this study while it is strongly recommended by the authorities to test all three Class 2A elements. For this study a xenon arc lamp was used which can measure multiple elements using the same instrument settings. The method was suitable for measuring the twelve selected elements according to the ICH guideline. [66], [67]
29 8 Comparison XRF versus ICP-MS
Only a few articles describe the use of XRF for the determination of elemental impurities. Usage of XRF is not common because the achieved detection limits are often not good enough. Margui
et al. did report about the successful use of (WD)XRF for the determination of tin, iron,
chromium and nickel in APIs according to the ICH guidelines. However, they also described that the results are significantly influenced by the API matrix (e.g. neutral organic compound or salt). [68]
In Table 5 a comparison of the XRF and ICP-MS techniques is shown based on this report. ICP-MS analysis is relatively expensive due to the use of helium and argon gas. In addition, the instrument itself is more expensive, and in contrast to XRF analyses, chemicals are needed (e.g. HNO3, HCl, tuning solution and internal standards). XRF requires low maintenance since the system only has a few moving parts and the XRF does not require a highly trained analyst. On the other hand, maintenance for ICP-MS is more labor intensive and specialized personnel is needed for the analysis.
Both techniques can measure a wide range of elements. ICP-MS analysis has a lower limit of detection compared to XRF analysis due to the elimination of matrix effect. For an XRF experiment it is important that the sample is homogeneous and that the surface is smooth since the X-Ray beam only penetrates the outer part of the sample.
Table 5, Comparison XRF and ICP-MS Analysis [69], [70]
XRF ICP-MS
Sample preparation No or little sample preparation Destruction using microwave
Destructive No Yes
Sample amount 1 – 2 g 100 – 250 mg
Matrix effect Yes Little (after destruction)
Analysis time 5 min (excl. preparation of calibration curve)
5 min (excl. two hours sample preparation)
Price Relatively low Relatively high
Safety No safety concerns Use of strong acids
Maintenance Low High
Limit of detection 100 ppm – 10.000 ppm 0.1 ppm – 1 ppm Application in
pharmaceutical field
Research and development Release testing & research and development
Both XRF and ICP-MS techniques experience matrix effects. Digestion of the sample using microwave and strong acids reduce the matrix effect in case of ICP-MS significantly. Furthermore, interferences of doubly charged ions at half nominal mass can be formed in the plasma, but these interferences are easy to recognize. In Figure 10 and Figure 18 it is shown that the XRF spectrum shows broader peaks compared to the ICP-MS spectrum and has therefore a larger chance of overlapping peaks which lead to less accurate results. [71], [72] Multiple lines are observed per element in an XRF spectrum leading to a higher chance of overlapping peaks, but on the other hand, this gives the ability to select a different line for a specific element if less overlapping occurs at that line. For example, Sauer et al. selected the V-Kβ line, since the V-Kα line overlapped with the Ti-Kβ line because a high level of titanium was present in the sample. A disadvantage of this approach is that the intensity of the V-Kβ line is lower than that of the V-Kα line resulting in a higher limit of detection. ICP-MS has a better resolving power and therefore less interferences influences the analysis. Polyatomic interferences, however, can overlap with another element, but this can be avoided using a collision cell. [34]
30 9 Conclusion
XRF and ICP-MS techniques both have their advantages and disadvantages when compared to each other. When measuring elemental impurities in APIs, ICP-MS is preferred since the detection limits which are currently requested by regulatory agencies in the drug product are met using ICP-MS but are often not met when XRF is used. And these demands by regulatory agencies are ever growing. It should be noted that, in order to measure volatile elements by ICP-MS, a sample destruction in a closed vessel is required. Furthermore, a collision or reaction cell is needed to reduce polyatomic interferences. Another advantage of ICP-MS over XRF is the fact that the acid digestion as sample preparation of ICP-MS reduces the matrix effect and therefore a generic method can be developed which is able to measure multiple elemental impurities in various kinds of APIs, while there is matrix effect with an XRF experiment. Due to this matrix effect with an XRF analysis, it is important to verify if the external calibration curve is suitable for each matrix, the API.
XRF could be used for R&D purposes, for example for a batch comparison. However, only a few articles are written about the XRF analysis of a final release sample, especially methods which can determine multiple elements in various types of APIs or even in only one API are hardly reported. Nevertheless, XRF is gaining more attention due to its relatively easy, fast and cheap instrumentation and improvement of the resolving power and detection limits.
31 10 Bibliography
[1] "Palladium: Exposure, Uses and Human Health Effects," in Encyclopedia of
Environmental Health, Michigan, Elsevier, 2011, pp. 307 - 314.
[2] Agency for Toxic Substances and Disease Registry, "Public health statement Nickel," Department of Health and Human Services, Public Health Service, 2005.
[3] International Conference on Harmonisation, Guideline for elemental impurities Q3D, ICH, 2013.
[4] United States Pharmacopoeia, "<232> Elemental Impurities - Limits," USP. [5] K. A, Fliszar, D. Walker and A. Leonardo, "Profiling of Metal Ions Leached from
Pharmaceutical Packaging Materials," Journal of pharmaceutical science and
technology, vol. 60, no. 6, pp. 337 - 342, 2006.
[6] V. Balaram, "Recent advances in the determination of elemental impurities in pharmaceuticals - Status, challenges and moving frontiers," Trends in Analytical
Chemistry, vol. 80, pp. 83 - 95, 2016.
[7] G. Li, D. Schoneker, K. L. Ulman, J. J. Sturm, L. M. Thackery and J. F. Kauffman, "Elemental Impurities in Pharmaceutical Excipients," Pharmaceutics, Drug Delivery and
Pharmaceutical Technology, vol. 104, pp. 4197 - 4206, 2015.
[8] P. Pohl, A. Bielawska-Pohl, A. Dzimitrowicz, P. Jamroz and M. Welna, "Impact and practicability of recently introduced requirements on elemental impurities," Trends in
Analytical Chemistry, vol. 101, pp. 43 - 55, 2018.
[9] M. Maithani, R. Raturi, S. V. Gupta and P. Bansal, "Elemental impurities in pharmaceutical products adding fuel to the fire," Regulatory Toxicology and
Pharmacology, vol. 108, pp. 1 - 11, 2019.
[10] I. Netto and V. H. Phutane, "Reversible Lithium Neurotoxicity: Review of the Literature,"
The Primary Care Companion for CNS Disorsers, vol. 14, pp. 1 - 14, 2012.
[11] Bruker, "Handheld XRF: How it works," Bruker, [Online]. Available: https://www.bruker.com. [Accessed 16 May 2019].
[12] Matsusada, "Technical knowledge," Matsusada, [Online]. Available:
https://www.matsusada.com/news/2019/02/column-x-ray-fluorescence.html. [Accessed 05 August 2019].
[13] Thermo Scientific, "Quant'X EDXRF training manual," Thermo Scientific.
[14] B. Beckhoff, B. Kanngiesser, N. Langhoff, R. Wedell and H. Wolff, Practical X-ray Fluorescence Analysis, Berlin: Springer-Verlag, 2006.
[15] R. Jenkins, X-Ray Fluorescence Spectrometry, Vols. 51-53, Wiley-Interscience, 1999. [16] R. Jenkins, R. W. Gould and D. Gedcke, "Quantitative X-Ray Spectrometry," Journal of
Applied Crystallography, vol. 15, p. 360, 1982.
[17] Bruker, "S6 JAGUAR – Full WDXRF on a Bench," Bruker, [Online]. Available: https://www.bruker.com/products/x-ray-diffraction-and-elemental-analysis/x-ray-fluorescence/s6-jaguar.html. [Accessed 05 January 2020].
[18] A. R. Chowdhury, N. Maheshwari, J. Soni, M. Kapil, T. Mehta and A. Mukharya, "Quantitative X-ray fluorescence analysis: Trace level detection of toxic elemental impurities in drug product by ED-XRF spectrometer," Journal of Pharmaceutical and
Biomedical Analysis, vol. 189, pp. 1 - 16, 2020.
[19] Rigaku, "Nanohunter II," 8 June 2017. [Online]. Available:
