UNIVERSITY GENT FACULTY PHARMACEUTICAL SCIENCES Department of Pharmaceutical Analysis Drug Quality and Registration (DruQuaR)

Transcription

1 UNIVERSITY GENT FACULTY PHARMACEUTICAL SCIENCES Department of Pharmaceutical Analysis Drug Quality and Registration (DruQuaR) Academical year DEVELOPMENT OF A NEW HIGH-THROUGHPUT METHOD: ION MOBILITY SPECTROMETRY VERSUS HIGH PERFORMANCE LIQUID CHROMATOGRAPHY Grégoire Bouchez Master Industrial Pharmacy Promotor Prof. Dr. B. De Spiegeleer Members of the jury Prof. J. Hoogmartens Prof. R. Kemel Prof. S. Apers Prof. A. Van Schepdael Prof. E. Adams Dr. A Van Eeckhaut Prof. Y. Michotte

2 COPYRIGHT The author and the promotors give the authorization to consult and to copy parts of this thesis for personal use only. Any other use is limited by the laws of copyright, especially concerning the obligation to refer to the source whenever results from this thesis are cited. June 1, 2010 Promotor Author Prof. Dr. Bart De Spiegeleer Grégoire Bouchez

3 Acknowledgments I would primarily like to thank Prof. Dr. Bart De Spiegeleer, my promoter, as well as PhD student Bram Baert for guiding me through my experimental work and through the writing of this thesis. Thank you for challenging me through experimental problems and building the pressure during the last days. Another word of thanks to the other PhD students at DruQuaR and the laboratory staff for the daily help and atmosphere in the laboratory. I d also like to thank my fellow master students, for the leisure time during this year. With special regard to fellow student Apr. Céline Thierens, who also worked on a thesis at DruQuaR, and shared the same ups and downs during daily work. I m grateful to my family and my environment in Ronse, for the many support during my studies. Last but not least, I d like to thank my girlfriend Valérie for believing in me and recomforting me when I needed it.

4 List of abbreviations 1 Introduction Principles of Quality by Design Design intent Method development Design space Control strategy Current application of Quality by Design Design intent Method development Ion mobility spectrometry Principles of Ion mobility spectrometry Plasmagram Franz Diffusion Cell experiment Objectives Materials & methods Materials and reagents Skin permeation study HPLC Descriptive Method validation Linearity Accuracy Repeatability Sensitivity Specificity IMS Descriptive Operational qualification Noise check Carryover check Limit of detection check Detector linearity check... 21

5 Peak profile check Injection repeatability check Sample treatment Experimental design Method validation Results HPLC Validation of the HPLC reference method Linearity Detection limit Range Precision Accuracy Specificity IMS Operational qualification Design of experiments Maximum amplitude Relative standard deviation Validation of the IMS method Linearity Detection limit Range Precision Accuracy Specificity Sample treatment Application: transdermal diffusion experiment HPLC IMS Comparison of the two methods Discussion HPLC... 45

6 5.2 IMS Operational qualification Design of experiments Validation of the IMS method Sample treatment Application: transdermal diffusion experiment Conclusion List of references... 51

7 List of abbreviations APCI: atmospheric pressure chemical ionization CCC: central composite circumscribed design CCF: central composite face centered design cgmp: current Good Manufacturing Practices cqa: critical quality attributes CumAmp: Cumulative amplitude DAD: diode array detector DOE: Design of experiments Du: digital units FDA: Food and Drug Administration FDC: Franz diffusion cell FWHM: full-width half-maximum HPI: high performance injector HPLC: High performance liquid chromatography ICH: International Conference on Harmonisation IMS: Ion mobility spectrometer/-metry LoD: limit of detection LoQ: limit of quantification MALDI: Matrix-assisted laser desorption MaxAmp: Maximum amplitude MS: mean squares OECD: Organisation for economic cooperation and development

8 OFAT: One-factor-at-a-time OQ: operational qualification PAT: Process analytical technology PBS: phosphate buffered saline P-P Amp: peak-to-peak amplitude PRESS: prediction residual sum of squares QbD: Quality by design RMS Amp: root mean square amplitude RSD: relative standard deviation RSM: response surface model SEM: standard error on the mean SS: sum of squares SS REG : sum of squares by the regression SST: System suitability test TNT: 2,4,6-trinitrotoluene TOF: time of flight UV: ultraviolet

9 1 Introduction 1.1 Principles of Quality by Design Quality by design is a new concept on product development and on analytical analysis. Contrary to traditional empirically based control strategies (also called Quality by testing, where quality heavily relies on controls of batches of a process), QbD provides assurance of the quality throughout process-understanding (Borman et al., 2007). Quality by design is introduced in various guidelines from different organisations, such as the FDA (Pharmaceutical CGMPs for the 21st Century A Risk Based Approach) and the ICH (ICH Q8: Pharmaceutical development, Q9: Quality risk management and Q10: Pharmaceutical quality system). A proper definition (according to ICH Q8) for this concept is the following: A systematic approach to development that begins with predefined objectives and emphasizes product and process understanding and process control, based on sound science and quality risk management. Similar to this process-related QbD, analytical QbD can be applied to ensure that analytical tests methods and process analytical technology (PAT) are robust (ICH Q2) and rugged (USP). FIGURE 1.1.: Representation of Quality by design (Borman et al., 2007). 1

10 Quality by design consists of the following elements or phases: Design intent Method development using a design of experiment (DOE) approach Risk assessment: definition of design space Control strategy Design intent Before developing a new method, it is imperative to know what user requirements should be met e.g. what should be measured at which concentrations? During this phase, critical quality attributes (cqa) are identified by an understanding of the process monitoring and control requirements. cqa s are important to be controlled as they assure quality to the analytical method Method development Contrary to traditional method development, often based on one-factor-at-a-time (OFAT) processes, the method development in QbD uses a design of experiments (DOE) approach. DOE allows a multivariate approach, estimates all main factors as well as interaction factors, using a limited amount of experiments. The design consists of three stages, namely the screening, the optimization and the robustness stage. Screening studies are conducted to determine significant parameters in a process/analytical method, whereas optimization studies lead to finding optimal conditions. Finally, a robustness study is performed to evaluate the effect of small changes in method parameters on the analytical results (Dejaegher & Vander Heyden, 2009) Design space The design space can be defined as following: the multidimensional combination and interaction of input variables and process parameters that have been demonstrated to provide assurance of the quality (ICH Q8, 2009). The design space can be obtained out of fundamental theoretical knowledge and optimization studies. These studies acquire response surfaces, which are then used for finding the optimal conditions as well as setting limit 2

11 specifications to parameters. The surface within these limits represents the design space. within this surface, the quality of the analytical method is assured Control strategy As was stated in , working within the design space, the quality of the analytical method is assured. Therefore, measurements are needed to ensure working within this range. In analytical methods, these control strategies comprises different tools e.g. system suitability tests (SST), calibration and qualification. 1.2 Current application of Quality by Design Design intent This study comprises the development and evaluation of an ion mobility spectrometry analytical method, in comparison to a traditional HPLC analysis method. The method is intended to be applied in transdermal studies, i.e. the diffusion of ibuprofen throughout human skin using a Franz Diffusion cell (FDC) experiment. The samples will be analyzed to obtain skin penetration parameters (steady-state flux J ss, skin permeability coefficient K p, lag-time T lag ). Previously, a pilot FDC experiment on ibuprofen was conducted and gave knowledge to approximate concentrations at specific time points (table 1.1.) (Baert, 2009a). TABLE 1.1.: Diffusion concentration of ibuprofen in a pilot FDC experiment using HPLC. Time point (h) Cumulative concentration (µg/ml) 0 < LoD ± ± ± ± ± ± ± ± 9.73 LoD ibuprofen= 68 ng/µl (20 µl: 1.36 µg) 3

12 1.2.2 Method development The IMS method will be developed using an experimental design; the design chosen is the Doehlert design, which is a traditional response surface modelling (RSM) design. The RSM methodology allows exploring the relationships between several explanatory variables and one or more response variables. The Doehlert designs are quadratic RSM designs with some special properties (buildable and extendable to other factor intervals). They allow the estimation of all main effects, all first order interactions, and all quadratic effects without confounding. They are saturated designs with similar properties to the CCF and CCC designs. Geometrically they are polyhedrons based on hyper-triangles (simplexes), with a hexagon in the simplest two-factor case and a centered dodecahedron in the three-factor case (figure 1.2.) (Bezerra et al., 2008). FIGURE 1.2.: Representation of Doehlert design in a 3 dimensional space (Dejaegher & Vander Heyden, 2009). The Doehlert design contains k² + k + 1 experimental points for k variables. In this case, for three variables, a minimum set of 13 experiments is required, characterized by the uniform distribution of the experiments in the 3-dimensional variable space. These comprises of 12 experiments that are equidistant from a central point, which is repeated several times to estimate the variability (Eriksson et al., 2000). 4

13 In a three-factor Doehlert design, one variable is represented at seven levels, the second variable at five levels and the third variable at three levels (table 1.2.). TABLE 1.2.: Experimental matrix for a three-factor Doehlert design. Factors Experiment A B C etc. * * additional centre points are added to the design Ion mobility spectrometry Ion mobility spectrometry is an electrophoretic technique used to characterize chemical substances on the basis of velocity (v d ) of gas-phase ions in an electric field. IMS is comparable to the time-of-flight (TOF) method in mass spectrometry, with the exception that IMS is conducted at ambient pressure, while TOF in mass spectrometry is run in vacuum conditions (Collins & Lee, 2002). The principle of ion mobility was already discovered in the late 19 th century. However, it was first described as an analytical tool in 1970 by F.W. Karasek, which used the term plasma chromagraphy. Since then, IMS has mainly been developed in security and military venues, specifically for detection of explosives, narcotics and war gases (Borsdorf & Eiceman, 2006). Now, IMS is also increasingly being used in the pharmaceutical field for cleaning validation as well as quantitative analysis (Walia et al., 2002). Ion mobility spectrometry allows several advantages (Smiths Detection, 2007): Fast response (analysis time of seconds) 5

14 High sub-nanogram sensitivity Low cost per sample: no use of mobile phases... No need for vacuum conditions, purified air is often used a drift gas. FIGURE 1.3.: Representation of a typical ion mobility spectrometer (Smiths Detection, 2006) Principles of Ion mobility spectrometry As stated in , ion mobility spectrometry determines the drift velocity of gasphase ions derived from constituents in a sample under the influence of an electric field. This already shows the major limitations of the technique, i.e. a substance must be volatile and it must be ionisable. The drift velocity (v d ) is proportional to the strength of the electric field: the proportionality constant equals the mobility (K) of the ions (equation 1.1.): v d = K x E = d/t (cm/s) (equation 1.1.) With: K: ion mobility coefficient (cm²/vs) E: electric field strength (V/cm) d: drift length (cm) t: drift time (s) 6

15 The ion mobility coefficient K depends on the size, shape and mass of the ion and the drift gas molecule (the collision cross section). It also depends on environmental factors. Therefore, this parameter is usually normalized to temperature and pressure and is represented as the reduced mobility (K 0 ) (Smiths Detection, 2005a): K 0 = (v d x E) x (p/760) x (273/T) (cm²/vs) (equation 1.2.) With: v d = d/t = drift velocity (cm/s) E = field strength (V/cm) p = pressure (torr) T = temperature (K) The processes during ion mobility spectrometry can be summarized in the following (Borsdorf & Eiceman, 2005): Transfer of sample as a vapour into the ion source Formation of ions from neutral gaseous sample molecules at atmospheric pressure Injection of ion swarm into the drift region Determination of drift velocities of the swarms under the influence of the electric field of the drift region and in a supporting atmosphere, the drift gas Detection of ions and storage or display of the electrical signal with or without automated analysis of the result Sample introduction The ion mobility spectrometer used in this study (IONSCAN -LS from Smiths Detection) provides two methods of sample introduction: - Teflon substrate - High performance injector (HPI) In both methods, the analyte sample is deposited onto the substrate or injected into the HPI. The solvent is evaporated and the analyte undergoes thermal desorption. A carrier gas (i.e. purified air) transfers the gas-phase analyte molecules through the inlet region and to the ionisation region. 7

16 Ionization In traditional IMS, ionization occurs through atmospheric pressure chemical ionization (APCI). APCI allows ionization of analyte molecules without fragmentation in the positive and negative mode: analyte molecules with a high proton affinity will be positively ionized, while analyte molecules with a high electronegativity will be ionized in the negative mode. A radioactive source, mostly 63 Ni, is used as ionization source (Smiths Detection, 2005a). The IONSCAN -LS from Smiths Detection contains a 555MBq 63 Ni-source (equivalent to 15 mci). 63 Ni is a β-emitter with an approximate half-life of 100 years. The emission of highly energized electrons (a maximum energy of 67 kev is obtained) occurs using the following equation: (equation 1.3.) The β - -particles react with molecules contained in the drift gas, i.e. nitrogen and oxygen, hence rendering the following positive and negative ions: N 2 + β - N β + e - (equation 1.4.) O 2 + e - - O 2 (equation 1.5.) As next reaction in the ionization process, water (present in traces in the drift gas) reacts with the positive nitrogen or with the negative oxygen molecules, forming (H 2 O) n H + and (H 2 O) n O - 2 respectively, whereas n is dependent of the gas temperature and the moisture level in the ionization region (Creaser et al., 2004). Positive ionization mode Analyte molecules with a high proton affinity such as ketones (aliphatic and aromatic), amines and sulfoxides undergo positive ionization through reaction with the positively charged water clusters, (H 2 O) n H +. 8

17 Collisions between molecules (M) from a sample and the reactant ions can lead to the formation of an adduct ion (MH + (H 2 O) n ) (equation 1.6.) which leads to the formation of product ions (MH + ), by loss of water. Additional product ions can also be formed, depending on the physicochemical properties of the substances and their concentration, as well as on the experimental conditions. These ions can be formed by cluster formation, dimerization or by a combination of both (respectively, equation 1.7., 1.8., 1.9.) (Borsdorf & Eiceman, 2005). M + H + (H 2 O) n MH + (H 2 O) n-x + x H 2 O (equation 1.6.) MH + + nl MH +.L n (equation 1.7.) MH + + M M 2 H + (equation 1.8.) M 2 H + + nl M 2 H +.L n (equation 1.9.) Negative ionization mode Analyte molecules with a high electronegativity such as aromatic hydrocarbons, carboxylic acids and nitro compounds react with (H 2 O) n O - 2 leading to negative ionization of the analyte. The ions are formed through charge transfer reactions, resulting in dissociative and associative electron attachment as well as proton abstraction of the analyte molecule (equation 1.10., & 1.12.). MX + (H 2 O) n O 2 - MX + (H 2 O) n O 2 - M + (H 2 O) n O 2 - M + X - + nh 2 O + O 2 (equation 1.10.) (MX)O nh 2 O (equation 1.11.) (M H) - + neutrals (equation 1.12.) Besides 63 Ni, 241 Am or 3 T (tritium) can be used as β-emitters. Newer ionization methods use non-radioactive sources to avoid safety issues concerning radioactivity, i.e. photoionization, corona discharge ionization, MALDI-ionization and electrospray ionization. However, β-emitters provide stable production of the ionization source (high energy electrons), have low weight and low power requirements and are simple to use (Borsdorf & Eiceman, 2005). Moreover, safety issues are somewhat obsolete: the radioactive source is fully sealed resulting in no direct radiation hazard. Internal calibrant 9

18 IMS uses an internal calibrant that is injected with each analysis. The internal calibrant is a substance with a known and fixed mobility. This value, in addition with the experimentally determined drift times, is used to calculate the K 0 of the analyte using the following equation, derived from equation 1.1. and 1.2. (Smiths Detection, 2005a): d/e = K 0 analyte x t drift of analyte = K 0 cal x t drift of cal (equation 1.13.) The standard internal calibrant on the IMS are the following: positive ion mode: nicotinamide negative ion mode: 4-nitrobenzonitrile Reactant/dopant In order to obtain an accurate identification of the analyte molecules, a dopant/reactant is used in the positive/negative mode, respectively. These molecules form ions that interact with the analyte molecules, leading to the formation of ion clusters. In overall, a reactant/dopant generally results into a more efficient ionization of the analyte molecules (Puton et al., 2008). In the negative ion mode, hexachloroethane is a commonly used reactant; in the positive ion mode, nicotinamide is used as dopant, which also acts as internal calibrant. Injection of ion swarm into the drift region Once the analyte molecule is ionized, it is injected into the drift region, where it migrates to the detector, along the electric field. Between the ionization region and the drift tube, a shutter is situated that allows pulse injections of the ions. Two designs of the shutter exists, i.e. the Bradbury-Nielsen design and the Tyndall design. Only the Bradbury-Nielsen gate, which is present in the IONSCAN -LS, will be explained. The Bradbury-Nielsen gate (figure 1.4.) comprises of parallel wires, with opposite voltages between neighbouring wires, resulting in a strong electric field between the wires (due to a close spacing of the wires). This electric field causes deflection of ions, sweeping them away out of the drift tube. When the electric field is momentarily removed (= injection of analyte ions), ions pass through the gate under the influence of the electric field in the drift 10

19 tube (a). An injection is completed when the voltage difference between the wires is reestablished, blocking further flow through the ion shutter (b). The pulse width on the IONSCAN -LS lasts for 200 µs, resulting into a finite pulse of products into the drift tube (Borsdorf & Eiceman, 2005). FIGURE 1.4.: Representation of a Bradbury-Nielsen design (Zuleta et al., 2007). Detection of the ion swarm/determination of the drift velocity In the drift tube, the ions move toward a detector down a linear voltage gradient through a purified air atmosphere, that flows in an opposite direction. The detector consists of a Faraday plate: a circular metal plate which allows capture of the analyte ions. Collision and annihilation of ions on the detector cause a current flow of nano- and pico-amps, which is transformed to a voltage of 5 to 10 V, using current to voltage amplifiers. An aperture grid is placed at a short distance from the Faraday plate, to prevent a current flow before the ions are striking the detector, caused by an approaching ion swarm. This prevents fronting of the peaks and results in an enhanced resolution in the mobility spectrum Plasmagram A plasmagram represents the results of an IMS analysis. The peak height, in digital units (du) is set out in function of the drift time (in milliseconds). On the plasmagram, the 11

20 drift time of the analyte molecule and the internal calibrant is shown. The internal calibrant is used to determine the reduced mobility of the analyte, as stated in FIGURE 1.5.: Plasmagram of Ibuprofen (concentration: 1 µg/ml). A plasmagram can be represented in two dimensions or in three dimensions (figure 1.5.), where the third dimension represents the total analysis time. Each time the shutter gate opens, a new scan is recorded. Several scans are combined into one segment, in order to obtain a better signal-to-noise response. The 3D representation consists of the several segments during the total analysis, giving rise to a desorption profile of an analyte. Quantification of the peaks can be done by taking the peak height of the highest peak (Maximum amplitude or MaxAmp) or by adding up all the peak heights of all the segments (Cumulative amplitude or CumAmp) (Smiths Detection, 2005a) Franz Diffusion Cell experiment The diffusion of ibuprofen through human skin is assessed using a static Franz diffusion cell experiment, in accordance with OECD guideline 28 & 428. A diffusion cell consists of a donor chamber, where the analyte solution is deposited and a receptor compartment, which is continuously stirred to assure homogeneity of the receptor fluid, in this case a PBS solution. Between the two chambers, excised human skin is positioned that acts as 12

21 diffusion membrane. A sampling port allows sample taking at specific time intervals. Finally, a water jacket around the receptor chamber is used to allow accurate temperature control (figure 1.6.). FIGURE 1.6.: Schematic of a static Franz Diffusion Cell (http// Diffusion across membranes can be explained by Fick s law, which represents the amount of mass transferred through a surface (equation 1.14.). J = - D x dc/dx (equation 1.14.) Where J = diffusion flux (µg/cm²s) D = diffusion coefficient (cm/s) dc/dx = concentration gradiant (µg/cm³) A Franz diffusion cell experiment is used as in-vitro alternative to determine several skin parameters. During this experiment, a condition of steady-state flux is achieved, i.e. the flux is time-independent. The skin parameters that are calculated are the following: the steady-state flux J ss (µg/cm²): the lag-time T lag (h): the slope of the steady-state linear part, divided by the diffusion surface (0.64 cm²) the X-intercept, obtained from the linear regression curve the permeability coefficient K p (cm/s): K p = J ss /C dose, (equation 1.15.) where C dose = applied dose concentration in µg/ml In this study, ibuprofen [(±)-(R,S)-2-(4-isobutylphenyl)-propionic acid] will be used as compound. IMS necessitates organic solutions, since the solvent needs to be readily 13

22 evaporated during analysis. However, in FDC experiments, aqueous solvents are used, which could cause problems during the IMS analysis. This can easily be solved by using a liquidliquid extraction procedure. Underlying figure shows the molecular structure of ibuprofen, which contains a carboxyl group, resulting in an acidic structure (pka = 4.4). FIGURE 1.7.: structure of ibuprofen. According to the Hendersson-Hasselbalch equation (equation 1.16.), the distribution between ionized and un-ionized forms can be modified by adjusting the ph. For ibuprofen, the solubility of ibuprofen can be decreased by lowering the ph, i.e. in a acidic ph, the carboxylgroup is protonated, thus un-ionized. ph = pka + log (base/acid) (equation 1.16.) 14

23 2 Objectives The main objective of this study is to develop a new ion mobility spectrometer method, validate and compare it with the traditional HPLC, using a Franz diffusion cell experiment as application. In order to achieve this goal, several objectives should be accomplished: 1) An IMS method should be developed that allows quantification of aqueous ibuprofen samples, using a liquid-liquid extraction method and IMS-substrate sample introduction: 1.1) Basic qualification of the IMS-instrumentation An operational qualification (OQ) is conducted to assure adequate performance of the IONSCAN -LS instrumentation. 1.2) Design of experiments (DOE) on the IMS-analysis of ibuprofen According to a QbD approach, a new IMS-method for the assay of ibuprofen is developed using DOE. A Doehlert design is used, which is evaluated using statistical analysis, and optimal conditions are determined using a response surface. 1.3) Development of sample treatment The ion mobility spectrometry method is based on organic solutions of ibuprofen; however, the samples acquired from the FDC experiment are aqueous, in order to avoid skin irritation by organic solvents. A rapid, simple and inexpensive liquid-liquid extraction is developed. Moreover, IMS methods have small working ranges; therefore, dilution of the sample if required will also be part of the sample treatment. 1.4) Validation of the method Once the whole method development is completed, it will be appropriately validated to proof its performance for the intended application. 2) An existing HPLC method for this analysis is slightly modified, i.e. a new column will be used (a HALO hexyl-phenyl column). This method should also be validated for its intended use. 3) Finally, the FDC experiment is run during 24 hours. Samples are taken at specific time points and analyzed on both HPLC and IMS. A comparison of the results obtained with both methods is made. 15

24 3 Materials & methods 3.1 Materials and reagents Ibuprofen (Ph.Eur grade) was obtained from ABC Chemicals (Vemedia, Wouters- Brakel, Belgium). Phosphate buffered saline (PBS), TNT, hydrochloric acid and LC-MS grade formic acid were bought from Sigma-Aldrich (Buchs, Switserland). GC grade n-hexane came from Fluka (St. Louis, MO, USA). HPLC grade methanol was obtained from Fisher Scientific (Leicestershire, UK). Water was purified using an Arium 611 purification system (Sartorius, Göttingen, Germany) resulting in ultrapure water of 18.2 M.cm quality. Whatman 2 µm PTFE 46.2 mm filter (Teflon membrane) were purchased from VWR (Leuven, Belgium). HPLC vials were purchased from Waters (Milford, MA, USA), the PTFE/silicon vial tops were obtained from Alltech (Deerfield, IL, USA). 3.2 Skin permeation study The penetration of ibuprofen through human skin was determined using static Franz diffusion cells (Logan Instruments Corp., New Jersey, USA) with a receptor compartment of 5 ml. Excised human skin from a healthy patient that had undergone an abdominoplasty procedure was used. After cleaning the skin with 0.01 M PBS ph 7.4 and removal of the subcutaneous fat, the skin samples were wrapped in aluminium foil and stored at -20 C for not longer than 3 months. Just before the start of the experiments, the skin samples were thawed and dermatomed to a pre-set thickness of 510 µm. The experimentally obtained thickness was determined using a micrometer (Mitutoyo, Tokyo, Japan). Skin samples were sandwiched between the donor and the receptor chambers of the diffusion cells. The receptor compartment was filled with PBS, making sure all air under the skin/membrane was removed. The whole assembly was fixed on a magnetic stirrer and the solution in the receptor was continuously stirred using a Teflon coated magnetic stirring bar. Before starting the skin experiments, skin impedance was measured using an automatic micro-processor controlled LCD Impedance Bridge (Tinsley, Croydon, UK) to ensure that there was no skin damage. Skin pieces with an impedance value below 20 k, a validated system suitability cut-off value, were discarded and replaced. Ibuprofen was topically applied to the surface of the skin as 500 µl of an ethanol/water (50:50, V/V) solution. The donor compartment was covered with parafilm (American National Can, Chicago, USA). The temperature of receptor compartment was kept at 32 ± 3 C by a water jacket. The available diffusion area was

25 cm². Samples (200 µl) were drawn at regular time intervals from the sample port (0, 2, 4, 6, 8, 12, 17, 22 and 24h) and were immediately replaced by 200 µl fresh solution. The analytically determined assay values for the model compounds were correspondingly corrected for the replenishments. 3.3 HPLC Descriptive A HPLC method was used for assaying ibuprofen in the receptor fluid. The apparatus consisted of a Waters Alliance 2695 separation module coupled to a Waters 2996 photodiode array detector (DAD) and controlled by Empower 2 software (all Watters, Millford, USA). The analytical column used was a HALO Phenyl-hexyl (50 x 4.6 mm, 2.7 µm particle size) column, combined to a guard column, both from Advanced Material Technology (Wilmington, DE, USA). Samples (20 µl) were injected onto the column, with a temperature maintained at 30 C. Isocratic elution was achieved at a flow rate of 1 ml/min by a degassed mobile phase consisting of a mixture (35/65, V/V) of 0.1% m/v formic acid in water and 0.1% m/v formic acid in methanol. UV detection was done at 220 nm, with a typical elution time of ibuprofen at 3.7 minutes Method validation The HPLC method was validated for linearity, accuracy, repeatability, detection limit (LoD & LoQ) and specificity. These parameters were validated using the ICH Q2 guideline Linearity A linear relationship should be evaluated across the range of the analytical procedure by visual inspection of a plot of signals as a function of analyte concentration. If there is a linear relationship, test results should be evaluated by appropriate statistical methods, by calculation of a regression line by the method of least squares. The correlation coefficient, y-intercept, slope of the regression line and residual sum of squares should be submitted. For the establishment of linearity, a minimum of 5 concentrations is recommended. 17

26 Accuracy The accuracy of an analytical procedure expresses the closeness of agreement between the reference value and the value found. Accuracy should be established across the specified range of the analytical procedure. It should be assessed using a minimum of 9 determinations over a minimum of 3 concentration levels covering the specified range (3 concentrations/3 replicates each). Accuracy should be reported as percent recovery by the assay of known added amount if analyte in the sample, compared to a reference of known concentration Repeatability The precision of an analytical procedure expresses the closeness of agreement between a series of measurements obtained from multiple sampling of the same homogeneous sample under the prescribed conditions. Precision may be considered at three levels: repeatability, intermediate precision and reproducibility. The precision of an analytical procedure is usually expressed as the variance, standard deviation or coefficient of variation of a series of measurements. Repeatability should be assessed using a minimum of 9 determination covering the specified range for the procedure (3 concentrations/3 replicates each). The standard deviation, and relative standard deviation should be reported Sensitivity The detection limit of an individual analytical procedure is the lowest amount of analyte in a sample which can be detected but not necessarily quantitated as an exact value. It may be expressed as: LoD: 3.3 x S (equation 3.1.) 18

27 The standard deviation may be determined based on the calibration curve, using the standard deviation of the y-intercept of the regression line. The slope S may be estimated from the calibration curve of the analyte, which is constructed using a minimum of five samples. The quantitation limit of an individual analytical procedure is the lowest amount of analyte in a sample which can be quantitatively determined with suitable precision and accuracy. It may be expressed as: LoQ: 10 x S (equation 3.2.) The limit of quantification value is also determined using the traditional visual analysis, i.e. a chromatogram from the determined LoQ will be run and inspected for a signalto-noise ratio of Specificity Specificity is the ability to assess unequivocally the analyte in the presence of components which may be expected to be present, e.g. matrix components. The specificity of a procedure may be confirmed by obtaining positive results from samples containing the analyte, coupled with negative results from samples which do not contain the analyte. 3.4 IMS Descriptive IMS analyses were performed using an IONSCAN -LS (Smiths Detection, Warren, NJ, USA), while IM station software (version 5.389) was used for data acquisition and processing. Ibuprofen was analyzed in the negative ionization method. The system was equipped with an internal 4-nitrobenzonitrile calibrant. A 1 µl sample was deposited onto a Teflon membrane using an autosampler and the volatile solvent was allowed to evaporate during a post-dispense delay of 3 s. The substrate was then introduced into the IMS system and placed on the desorber heater, with a temperature of 100 C. Analyte molecules were 19

28 vaporized and carried from the heated inlet (temperature of 100 C) to the ionization chamber in a flow of dry air. As the vapors enter the ionization chamber, a 555 MBq 63 Ni radioactive source emits low energy β-particles initiating ionization of the analyte. The negative ions were gated into the drift tube every 25 ms with a pulse width of 0.2 ms, where they were accelerated under an applied electric field toward a collector electrode against a counterflow of 1 atmosphere of dry air. The total analysis duration was 20 seconds. Ibuprofen has a reduced mobility value of K 0 = , a drift time of ms and a FWHM-value of 357 µs Operational qualification The operational qualification was performed based on a document provided by Smiths Detection: Installation qualification and operational qualification (Smiths Detection, 2005b) Noise check Noise is an important characteristic that specifies a detector s performance. Low noise level is especially important when analyzing compounds at low trace levels. IONSCAN detector noise primarily originates from micro-phonic noise (motion of guard grid relative to collector plate) and preamp noise (resistance noise). The root mean square noise is determined by using a built-im-station software function to measure the root mean square noise in a defined drift time interval. RMS Amp RMS (root mean square) deviation from the mean signal amplitude over all segments of the Seg Range, in the drift time interval specified. P-P Amp Displays the peak-to-peak amplitude of all segments of the range (difference between maximum and minimum of all segments) Carryover check When a high-concentration sample is analyzed, it has the potential to persist in the instrument and contribute to the instrument response observed for the next sample. This is 20

29 termed carryover. To test for carryover, an analysis of a high-concentration standard is performed, followed by an analysis of a solvent blank. The standard should not be detected when the blank is analyzed Limit of detection check This test verifies the ability of the IONSCAN-LS to detect low levels of a standard. This is accomplished by demonstrating the detection of 25 pg on TNT on Teflon in negative ion mode Detector linearity check Detectors become non-linear when the amount of the test compound increases above a certain limit. The linearity test is based on a sequence of runs where a TNT solution is injected several times from vials containing different concentration using the same injection volume. For the appropriate concentration range, when the instrument response is plotted against the concentrations, the points should lie on a straight line. The square of the correlation coefficient, r², of this calibration curve is used to assess detector linearity. Three injections of five standards will be analyzed Peak profile check The peak profile of a standard is an important indicator of optimal performance of the IONSCAN -LS/HPI system. A peak at the characteristic ion drift time for the standard should be detected a certain length of time after the start of the IONSCAN analysis, and the time profile of the peak should have a certain width. This is evaluated by analyzing a 200 pg/µl standard of TNT on Teflon in the negative ion mode. The desorption time at maximum amplitude and number of segments in which the peak is detected are recorded and compared to operational specifications Injection repeatability check The injection volume precision is based on a sequence of runs where a TNT standard is used in negative ion mode on Teflon. Each standard is injected 5 times, always with the same volume. The average amplitude and relative standard deviation (RSD) are determined for each standard, and the RSDs are compared to acceptance criteria. The test uses cumulative amplitude response for Teflon sample introduction. 21

30 3.4.3 Sample treatment 200 µl of each sample obtained from the Franz cell experiment was transferred into 1 ml glass vial, acidified with 20 µl of 0.1M HCl and an equal volume of n-hexane (220 µl) was added to the vial. The mixture was vortexed for 5 minutes, subjected to centrifugation at g (ambient temperature) and 100 µl of the upper organic phase was removed for IMS analysis. If needed, the organic phase was diluted to obtain a sample in the working range of the IMS Experimental design A Doehlert design was constructed using three different variables, each with a specific experimental range: Inlet temperature (70 C 150 C) Desorber temperature (70 C 150 C) Post-dispense delay (3s 25s) The model was constructed using MODDE ; throughout the experiment, five replicates of a centre point were run to calculate variability on the model. Once the experiments run, the software was used for evaluation of the model, as well as statistical analysis of the effects and the visualisation of the response surface. The evaluation of the model is based on the calculation of four parameters (Eriksson et al., 2000): R² The first parameter is R 2, which represents the goodness of fit. This is the fraction of the variation of the response explained by the model: R² = SS REG (equation 3.3.) SS SS REG = the sum of squares of Y corrected for the mean, explained by the model. SS = the total sum of squares of Y corrected for the mean. The R 2 value is always between 0 and 1. Values close to 1 for both R 2 and Q 2 indicate very good model with excellent predictive power. 22

31 Q² The second parameter is Q 2, which represents the goodness of prediction. This is the fraction of the variation of the response predicted by the model according to cross validation and expressed in the same units as R 2. Q² = 1 - PRESS SS PRESS = the prediction residual sum of squares. SS = the total sum of squares of Y corrected for the mean. (equation 3.4.) The Q 2 is usually between 0 and 1. Q 2 can be negative for very poor models. Model validity model. The third parameter is the Model Validity and is a measure of the validity of the When the model validity bar is larger than 0.25, there is no lack of fit of the model. This means that the model error is in the same range as the pure error (reproducibility). Reproducibility The forth parameter is the Reproducibility which is the variation of the response under the same conditions (pure error), often at the center points, compared to the total variation of the response. Reproducibility = 1 - (MS(Pure error)/ms(total SS corrected)) (equation 3.5.) MS = Mean squares, or Variance. According to Eriksson et al. (2000), the conditions for a good model are the following: Difference R² - Q² < Q² > 0.5 Model validity > 0.25 Reproducibility >

32 In total, four different Doehlert designs were constructed (four blocks); in this study, the model with the best parameters was used for further analysis. However, the significant effects were similar on all four blocks. Thus, the used model is confirmed by three other blocks. The overall data can be found in a DruQuaR internal document (Baert, 2009b). Underlying table represents the construct of the Doehlert design. TABLE 3.1.: Experiments on the Doehlert design. Exp No Inlet temperature ( C) Desorber temperature ( C) Post-dispense delay (s) Method validation The IMS method was validated for linearity, accuracy, repeatability, for detection limit (LoD & LoQ) and specificity. These parameters are described in

33 4 Results 4.1 HPLC The HPLC method is validated for linearity, sensitivity, range, precision and accuracy as described in A typical chromatogram on the analysis of ibuprofen is shown in figure 4.1. FIGURE 4.1.: chromatogram of ibuprofen Validation of the HPLC reference method Linearity Linearity is assessed using concentrations from 1 µg/ml up to 500 µg/ml. The upper linear range corresponds to a concentration of 200 µg/ml. Following table and graph (table 4.1. and graph 4.1.) represents a summary and visualisation of the regression analysis. TABLE 4.1.: Summary of linear regression of ibuprofen on the HPLC. Slope Y-intercept Correlation coefficient Residual sum of squares 1.02E+11 Standard error on intercept ( ) F- value of the regression Confidence interval (95%): y = ax + b a: b:

34 GRAPH 4.1.: Linearity of Ibuprofen on the HPLC Area (µv x sec) Concentration (µg/ml) y = 47421x R² = Detection limit Using the data from the linearity experiment, the LoD and the LoQ can be calculated using the following formula: LoD: 3.3 x S (equation 4.1) The limit of detection corresponds to a concentration of 87.8 ng/ml (for an injection volume of 20 µl, this corresponds to a mass of ng). The limit of quantification can be calculated from the following formula: LoQ: 10 x S (equation 4.2) The limit of quantification corresponds to a concentration of ng/ml (for an injection volume of 20 µl, this corresponds to a mass of ng). Underlying figure is a chromatograph obtained from the injection of an ibuprofen solution with concentration = 250 ng/ml. 26

35 FIGURE 4.2.: Chromatogram of an ibuprofen sample at LoQ concentrations (250 ng/ml) Range The range for the analysis of ibuprofen is determined from the linearity and the sensitivity of the method; the lower range corresponds to the LoQ, i.e. 250 ng/ml, while the upper range corresponds with the upper linear part, i.e. 200 µg/ml Precision Repeatability As described in , the repeatability is assessed using 3 concentrations in triplicate. The solutions used, corresponded to a concentration of 1, 25 and 100 µg/ml, values that are spread throughout the linear range. 27

36 TABLE 4.2.: Summary repeatability results. Concentration (µg/ml) Average (µg/ml) Standard Deviation (µg/ml) RSD (%) Accuracy As with the linearity experiment, three concentrations are used for the accuracy experiment: 1, 25 and 100 µg/ml. Concentrations are calculated by the areas of the analyte, obtained from the chromatograms. These values are compared to the theoretical concentrations, yielding the recovery of the analyte. TABLE 4.3.: Summary of accuracy experiment. Sample Theoretical concentration (µg/ml) Calculated Concentration (µg/ml) % Calculated/Theoretical concentration Average % Standard deviation 1 µg/ml µg/ml µg/ml

37 Specificity Specificity of the method is assessed in order to ensure no interference with matrix components, originating from the FDC experiment. A skin sample is incubated in a 0.01M PBS solution for a period of 24 hours and is analyzed through HPLC. Following figure represents the chromatogram: no interfering compounds are detected. FIGURE 4.3: Placebo sample of FDC cell experiment. 4.2 IMS Operational qualification The operational qualification is executed to ensure performance of the IMS instrumentation. The tests are described in The noise check is executed in the negative ion mode and should comply with the specifications, i.e. the RMS Amp should not exceed 3 du, the P-P Amp should not exceed the 15 du. Both values are measured in a time interval of 5.0 to 5.3 msec. The obtained values are 1.2 du for the RMS Amp and 7.6 du for the P-P Amp; both values are in compliance with the specifications. The carryover check is run using three highly concentrated TNT samples. Carryover should not exceed a 2% value, compared to the cumulative amplitude of the sample. 29

38 As calculated in the following table, no carryover was detected. TABLE 4.4.: Carryover check. Sample 500 pg TNT (CumAmp; du) A Blanks (TNT CumAmp; du) Blank 1 Blank 2 Blank 3 Sum of Blanks (CumAmp; du) B % Carry Over (B/A) x Average The limit of detection check consisted of three runs of a lowly concentrated 25 pg/µl TNT solution. These are preceded by a blank solution; all three runs should result in a CumAmp signal. TABLE 4.5.: Limit of detection check. Run order TNT signal (CumAmp; du) Table 4.5. concludes a valid limit of detection check: all three runs were detected. As following OQ test, a linear curve is constructed out of 5 concentrations, varying from 50 pg/µl to 250 pg/µl. Each concentration is run in triplicate. The linear curve should have a minimal R²- value of In this case, a suitable R² value of is obtained (table 4.6.). 30

39 TABLE 4.6.: Detector linearity check. TNT on Teflon (CumAmp; du) Sample Conc. (pg/µl) Run 1 Run 2 Run 3 Avg. Sol Sol Sol Sol Sol Slope = Intercept = r² = The peak profile check allows determination of the desorption profile of a TNT sample. A 200 pg TNT sample should comply to the following specifications: A maximum amplitude within the first 1.5 s A number of detected segments, situated between 5 and 26. The analysis of the sample results in a maxamp desorption time at 0.5 s and detection of 19 segments. Both specifications are fulfilled. TABLE 4.7.: Injection repeatability check. Sample 250 pg TNT Teflon membrane (CumAmp; du) Average Std Dev Relative Standard Deviation (RSD) 4.67 % As final OQ test, the injection repeatability is assessed by determining the relative standard deviation of 5 equal samples, which should not exceed 8%. A RSD value of 4.67% is obtained; thus, this test complies with the specification. 31

40 4.2.2 Design of experiments The Doehlert design was run four times, rendering four blocks. Underlying table represents the best model. Each experiment in this design was run three times. Both MaxAmp as the relative standard deviation are statistically analyzed using the computer software. TABLE 4.8.: Results on the Doehlert design. Exp No Inlet temperature ( C) Desorber temperature ( C) Postdispense delay (s) Average MaxAmp (du) (n = 3) Standard deviation MaxAmp (du) RSD MaxAmp (%) Maximum amplitude The maximum amplitude was fit into a model and resulted into the following findings: All main factors have a significant negative influence on the MaxAmp value: inlet temperature (p = 0.005), desorber temperature (p = 0.013) and post-dispense delay (p = 0.027). A negative significant influence is observed in the quadratic term desorber temperature x desorber temperature (p = 0.027). 32

41 TABLE 4.9.: Analysis of the model of maximum amplitude. Factor Coefficient (scaled & centered) Standard error P-value Confidence interval (±) Constant E Inlet ( C) Desorber ( C) Post-dispense delay (s) Inlet x Inlet ( C x C) Desorber x Desorber ( C x C) Post-dispense delay x Post-dispense delay (s x s) Inlet x Desorber ( C x C) Inlet x Post-dispense delay ( C x s) Desorber x Post-dispense delay ( C x s) This model is corrected for the significant factors only, with retention of the quadratic postdispense delay x post-dispense delay, which turns out to have a significant positive influence after the modifications (figure 4.4. and table 4.10.). FIGURE 4.4.: Coefficient plot of the adjusted model. 33

42 TABLE 4.10.: Coefficient list of the adjusted model. Factor Coefficient (scaled & centered) Standard error P-value Confidence interval (±) Constant E Inlet ( C) Desorber ( C) Post-dispense delay (s) Inlet x Inlet ( C x C) Post-dispense delay x Post-dispense delay (s x s) The adjusted model is evaluated using the four parameters R², Q², model validity and R², as described in Underlying graph states the summary of the evaluation: all parameters are compliant to the specifications. GRAPH 4.2.: Summary plot of adjusted model fit. R² 0.85 Q² 0.70 Model validity 0.92 Reproducibility 0.72 R² - Q²

43 The model was used to set up a response surface, which would allow prediction of the optimal conditions for the analysis of ibuprofen (figure 4.5.). FIGURE 4.5.: Contour plot of model. Optimal conditions could be found with following parameters: Inlet temperature: 85 C Desorber temperature: 100 C Post-dispense delay: 3s However, due to mechanical constraints, the inlet temperature will be set to 100 C. This is further explained in discussion section Relative standard deviation As second result, the relative standard deviation is statistically evaluated. Analysis of the coefficient plot (table 4.11.) showed no significant effect of the factors, with the exception of the post-dispense delay (p = 0.009). However, by comparing the three other blocks, postdispense delay doesn t have any significant effect on the relative standard deviation. The relative standard deviation isn t further evaluated, since it shows to be nondependent of the parameters. 35

44 TABLE 4.11.: Analysis of the model of relative standard deviation. Factor Coefficient (scaled & centered) Standard error P-value Confidence interval (±) Constant E Inlet ( C) Desorber ( C) Post-dispense delay (s) Inlet x Inlet ( C x C) Desorber x Desorber ( C x C) Post-dispense delay x Post-dispense delay (s x s) Inlet x Desorber ( C x C) Inlet x Post-dispense delay ( C x s) Desorber x Post-dispense delay ( C x s) Validation of the IMS method A typical plasmagram of ibuprofen in hexane can be found in figure Linearity Linearity is assessed using concentrations from 0.25 µg/ml up to 5 µg/ml. Each concentration is analyzed in quintuplicate The upper linear range corresponds to a concentration of 1 µg/ml. Following table and graph (table and graph 4.3.) represent a summary and visualisation of the linearity experiment. TABLE 4.12.: Summary of linearity experiment. Concentration (µg/ml) Avg MaxAmp (du) (n = 5) Standard deviation (du) RSD (%)

45 GRAPH 4.3.: Linearity of Ibuprofen on the IONSCAN-LS. 600,00 500,00 Maximal Amplitude (du) 400,00 300,00 200,00 100,00 0,00 0,00 0,20 0,40 0,60 0,80 1,00 1,20 Concentration (µg/ml) y = x R² = Detection limit Regression analysis of the linearity experiment is used to determine the LoD & the LoQ on the analysis of ibuprofen (table 4.13.). TABLE 4.13.: Summary of regression analysis. Slope 566 Y-intercept -96 Correlation coefficient Residual sum of squares Standard error on intercept ( ) F- value of the regression Confidence interval 95%: y = ax + b a: b: The limit of detection is calculated from the following formula: LoD: 3.3 x S The limit of detection in a sample corresponds to a concentration of 73 ng/ml (for a volume of 1 µl, this corresponds to a mass of 73 pg). 37

46 The limit of quantification is calculated from the following formula: LoQ: 10 x S The limit of quantification in a sample corresponds to a concentration of 222 ng/ml (for a volume of 1 µl, this corresponds to a mass of 222 pg). Underlying figure represents a 2D-plasmagram of an ibuprofen sample with a concentration in proximity of the LoQ (250 ng/ml). The segment with the maximum amplitude is represented. A maximum amplitude of 47 du is recorded. FIGURE 4.6.: 2D Plasmagram of ibuprofen in LoQ concentration Range The range for the analysis of ibuprofen is determined from the linearity and the sensitivity of the method; the lower range corresponds to the LoQ, i.e. 222 ng/ml, while the upper range corresponds with the upper linear part, i.e. 1 µg/ml Precision The repeatability is determined during the linearity experiment. The relative standard deviation varies from 3.03% to 9.48% at lower concentrations. 38

47 Accuracy Accuracy is determined according to : simulated FDC-samples are submitted to an extraction procedure and are compared to the theoretical concentrations. Table represents the summary of the accuracy experiment. TABLE 4.14.: Summary of accuracy experiment. Samples Theoretical concentrati on (µg/ml) Calculated Concentrati on (µg/ml) % Calculated/Theoretical concentration Average % Standard deviation 0.25 µg/ml 0.5 µg/ml µg/ml Specificity Similar to , the specificity of the IMS method is assessed by incubation of skin samples. Skin samples in PBS, ethanol and ethanol/pbs (50:50, V/V) are used throughout the experiment. Both the ethanol and the ethanol/pbs sample yielded positive interference: a compound is falsely mistaken for ibuprofen, which is not present in the placebo sample. Underlying figure shows a positive interference of the ethanol sample: a maximum amplitude of 81 du is obtained during IMS analysis. FIGURE 4.7.: plasmagram of placebo sample in ethanol. 39