G Model

IJP 14293 1–10 International Journal of Pharmaceutics xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

International Journal of Pharmaceutics journal homepage: www.elsevier.com/locate/ijpharm

1

Pharmaceutical nanotechnology

2

Raman spectroscopy for in-line and off-line quantification of poorly soluble drugs in strip films

3

4 Q1 5 6 7 8 9 10

Jun Zhang a , Ye Ying c , Barbara Pielecha-Safira b , Ecevit Bilgili a , Rohit Ramachandran d, Rodolfo Romañach e , Rajesh Davé a, **, Zafar Iqbal c, * a

Department of Chemical, Biological, and Pharmaceutical Engineering, New Jersey Institute of Technology, Newark, NJ 07102, USA Wallington Public School, Pine Street, Wallington, NJ 07057, USA c Department of Chemistry and Environmental Science, New Jersey Institute of Technology, Newark, NJ 07102, USA d Department of Chemical and Biochemical Engineering, Rutgers University, Piscataway, NJ 08854, USA e Department of Chemistry, University of Puerto Rico, Mayagüez Campus, Puerto Rico, USA b

A R T I C L E I N F O

A B S T R A C T

Article history: Received 27 December 2013 Received in revised form 20 August 2014 Accepted 26 August 2014 Available online xxx

Raman spectroscopy was used as a process analytical technology (PAT) tool for in-line measurement of active pharmaceutical ingredient (API) content during continuous manufacturing of strip films containing nanoparticles of poorly water-soluble APIs. Fenofibrate and naproxen were used as model APIs, whose concentrations ranged from 3% to 26% (w/w) in the model calibration. For both in-line and off-line measurements, calibration models employed partial least square (PLS) analysis, yielding correlation coefficients (R2) greater than 0.9946 and root mean squared error of calibration (RMSEC) of about 0.44%, indicating the validity and accuracy of the calibration. The robustness of Raman spectroscopy as a PAT tool was established by considering three processing parameters after substrate interference correction: sensing location, substrate speed and film thickness. Calibration models for each API were validated using a separate batch of strip films by predicting the API concentrations to within 1.3%. Principal component analysis (PCA) was used to explain the interactions between processing variables and calibration models, which suggest that besides API concentration, film thickness could also be monitored using Raman spectroscopy. The results demonstrate the potential of Raman spectroscopy as an effective PAT tool for novel strip film manufacturing process, facilitating detection of drug form and concentration in real-time. ã 2014 Published by Elsevier B.V.

Keywords: Raman spectroscopy Calibration model Strip films In-line drug concentration monitoring Principal component analysis

11 12 13 14 15 16 17 18 19 20 21 22 23

1. Introduction This paper examines Raman spectroscopy as an in-line process analytical technology (PAT) tool for monitoring the production of poorly water-soluble drug particle loaded thin films, manufactured through continuous casting over a moving substrate. Biocompatible polymeric strip film technology has emerged as a promising platform for drug delivery due to its high potential for patient compliance, continuous processing and cost-effective manufacturing (Cilurzo et al., 2008; Dixit and Puthli, 2009). It has also been recently shown to be an excellent dosage form for enhancing the dissolution of poorly water-soluble drugs by incorporating them as crystalline nanoparticles or microparticles in contrast to conventional solvent-casting (Dixit and Puthli,

* Corresponding author. Tel.: +1 973 596 8571; fax: +1 973 596 3586. ** Corresponding author. Tel.: +1 973 596 5860; fax: +1 973 642 7088. E-mail addresses: [email protected] (R. Davé), [email protected] (Z. Iqbal).

2009; Sievens-Figueroa et al., 2012a; Susarla et al., 2013). Although there are many papers addressing various aspects of pharmaceutical strip film design, formulation and applications (see e.g., Dixit and Puthli, 2009), as well as processing and dissolution testing (Sievens-Figueroa et al., 2012a,b; Susarla et al., 2013) the topic of in-line monitoring for process control and quality assurance has not been explored. Due to their virtually two-dimensional format and their production in a continuous process, strip films offer easy incorporation of PAT tools, forging a pathway toward quality-by-design. In this regard, previous work has explored use of off-line near-IR chemical imaging to determine drug distribution within films (Sievens-Figueroa et al., 2012a; Susarla et al., 2013). On the other hand, use of inline Raman spectroscopy, which has not been explored for strip films, may offer additional benefits due to its ability to detect drug crystallinity and form. Therefore, the use of Raman spectroscopy is studied here for detecting the drug content, which can allow for in-line assessment of potency of the films containing nanoparticles of poorly water-soluble drugs.

http://dx.doi.org/10.1016/j.ijpharm.2014.08.051 0378-5173/ ã 2014 Published by Elsevier B.V.

Please cite this article in press as: Zhang, J., et al., Raman spectroscopy for in-line and off-line quantification of poorly soluble drugs in strip films. Int J Pharmaceut (2014), http://dx.doi.org/10.1016/j.ijpharm.2014.08.051

24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42

G Model

IJP 14293 1–10 2 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86

J. Zhang et al. / International Journal of Pharmaceutics xxx (2014) xxx–xxx

Particle size reduction is a popular method for improving the bioavailability of BCS (Biopharmaceutical Classification System) Class II drugs, which have poor water solubility (Brown et al., 2004; Liversidge and Cundy, 1995; Singh et al., 2012; Yu et al., 2002). Among various techniques of particle size reduction reported in the past few decades (Aulton and Wells, 2002), wet stirred media milling has become of great interest due to the efficiency by which drug nanoparticles and microparticles are produced (Afolabi et al., 2014; Merisko-Liversidge et al., 2003). Significant recent research on the production and characterization of drug nanosuspensions with various stabilizers such as polymers and surfactants will be leveraged in the present study (Beck et al., 2013; Bhakay et al., 2013, 2014; Bilgili and Afolabi, 2012; Dalvi et al., 2013; Heinz et al., 2009; Monteiro et al., 2013). Drug nanosuspensions, formed via media milling, have been used in diverse dosage forms where their small size and increased surface area may lead to an increased dissolution rate and bioavailability (Patravale and Kulkarni, 2004). Generally, a solid dosage form is preferred due to patient preference, ease of administration, and long-term physical stability issues associated with the nanosuspensions (Van Eerdenbrugh et al., 2008). However, preparation of solid dosage forms entails drying of nanosuspensions, which may lead to nanoparticle agglomeration, poor redispersion, and slow recovery of primary nanoparticles during redispersion and dissolution testing (Bhakay et al., 2013; Hu et al., 2011). In order to address these issues, in this work, drug nanoparticles are incorporated into a biocompatible, edible polymer strip film, taking advantage of such recently demonstrated phenomena as avoiding irreversible agglomeration, thereby retaining the large surface area of the nanoparticles (Sievens-Figueroa et al., 2012a; Susarla et al., 2013). Pharmaceutical products must meet strict specifications which often require time-consuming, expensive and inefficient off-line testing on randomly collected samples to evaluate the end product quality. In order to improve product quality through process design, multiple PAT tools have been developed to provide in-line and real-time process information that allows for monitoring of processing variables and prediction of product quality (FDA and CVM, 2004). Raman spectroscopy is one such PAT tool which can observe molecular vibrations and rotations, as well as lattice modes in solid molecular systems (Pelletier, 1999). It has been demonstrated as a useful non-destructive method to monitor pharmaceutical manufacturing because it can provide detailed inline processing information about the active pharmaceutical ingredient (API) (Michelet et al., 2013). It is also a promising tool

for controlling crystallization processes through Raman-signal feedback (Pataki et al., 2012). Recent implementations of Raman spectroscopy have demonstrated its use in identification of pharmaceutical ingredients during manufacturing (Cebeci Maltaş et al., 2013) and quantification of polymorphs in tablets during production (McGoverin et al., 2012). It has also been used in image analysis of API distribution and quantification of API loading in thin films loaded with 10 mm particles (Rozo et al., 2011). However, this technique required off-line processing that took minutes to hours depending on the desired accuracy. In other examples, API concentration in solid-dispersions was measured via on-line measurement from hot-melt extruded films (Tumuluri et al., 2008). However, their model validation used the same samples for both calibration and testing without examining the robustness of the technique by varying any processing variables. Overall, these studies indicate that Raman spectroscopy is a promising tool for the solid-state characterization of APIs. However, further investigation is required for developing this approach for in-line monitoring of the strip film manufacturing process, which is the focus of this work. A major advantage of Raman spectroscopy is that it is not sensitive to water, which makes it more suitable for monitoring API concentration in a wet strip film than NIR spectroscopy whose API spectra may be masked by broad water bands. Hence, it follows that Raman spectroscopy would be capable of monitoring the API concentration during the film drying process, starting immediately after film casting at the beginning of the drying process, which could facilitate the analysis of out-ofspecification events, deviations, and feedback control of strip film formation process. In what follows, an investigation is carried out to establish the feasibility of using Raman spectroscopy to monitor the concentration of two poorly water-soluble APIs, fenofibrate and naproxen, in a strip film during the drying process. API nanosuspensions prepared via wet stirred media milling were mixed with filmforming polymer solution in different mass ratios, and the resulting mixtures were wet-casted and dried to prepare strip films with different API content. The actual API content in the films was determined by assaying the drug in dissolved film samples via UV spectroscopy. In-line and off-line Raman calibration models were then established based on the actual concentrations and measured Raman spectra, and their robustness was analyzed by statistical indicators. Another batch of test strip films having different API concentrations from the films used for the calibration model development was produced with the objective of validating

Fig. 1. Chemical structures of BCS Class II API molecules: a) fenofibrate (FNB) and b) naproxen (NPX).

Please cite this article in press as: Zhang, J., et al., Raman spectroscopy for in-line and off-line quantification of poorly soluble drugs in strip films. Int J Pharmaceut (2014), http://dx.doi.org/10.1016/j.ijpharm.2014.08.051

87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130

G Model

IJP 14293 1–10 J. Zhang et al. / International Journal of Pharmaceutics xxx (2014) xxx–xxx Table 1 API concentrations of strip films containing FNB and NPX prepared from samples with different nanosuspension–polymer solution ratios (measured by direct assaying via UV spectroscopy). Sample

Group Group Group Group Group

131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157

A B C D E

Nanosuspension (g)

Polymer solution (g)

FNB reference concentration (w/w %)

NPX reference concentration (w/w %)

20 20 10 10 5

20 30 20 35 40

26.0 19.6 15.8 9.9 4.7

23.0 15.8 12.5 7.1 2.8

the established in-line calibration model. Subsequently, calibration models were also developed for different process parameters such as substrate speed, probe location and casting thickness. principal component analysis (PCA) was applied to explain the variation in Raman spectra as influenced by API concentration, process parameters and within-sample variation in the Raman spectrum. The study of different process parameters, along with the PCA interpretation of the data, allows us to demonstrate the robustness of in-line Raman spectroscopy for monitoring API concentration and possibly film thickness. 2. Materials and methods 2.1. Preparation of nanosuspension and polymer solution Two BCS Class II APIs, fenofibrate (FNB; Jai Radhe Sales) and naproxen (NPX; Medisia), were used in this study. Their molecular structures are shown in Fig. 1. Sodium dodecyl sulfate (SDS; Sigma– Aldrich) and hydroxypropyl methyl cellulose (HPMC; Methocel E15LV from Dow Chemical) were used as stabilizers during wet media milling. HPMC (E15LV) and polyvinylpyrrolidone (PVP; K90 from Sigma–Aldrich) were used as film formers, and glycerin (Sigma–Aldrich) was used as the plasticizer. Preparation of stable API nanosuspensions with a combination of cellulose-based polymers and anionic surfactants in a wet stirred media milling process has been discussed elsewhere (Bhakay et al., 2011, 2013; Bilgili and Afolabi, 2012; Monteiro et al., 2013; Sievens-Figueroa et al., 2012a). In the present study, two API nanosuspensions consisting of 8.85% (w/w) FNB and NPX, respectively, 2.21% (w/w) HPMC and 0.44% (w/w) SDS, were prepared. A

3

polymer solution consisting of 75% (w/w) water, 10% (w/w) HPMC, 10% (w/w) PVP and 5% (w/w) glycerin was prepared by mixing the components at 80–90  C, and the mixture was cooled down to room temperature (around 20  C) prior to strip film preparation. For each API, five mixtures of the respective nanosuspensions and polymer solution mentioned above were prepared as precursor film suspensions with the objective of developing a Raman spectroscopic calibration model for the films. The formulations for these five mixtures (Groups A–E) are presented in Table 1. The nanosuspension and polymer solution were mixed for 3 h by a dual-propeller mixer (McMaster) with a motor (VWR International), and the mixture was allowed to equilibrate at rest until bubbles completely disappeared. Although the same nanosuspension and polymer solution formulations were used for each API, the mass ratio of API nanosuspension-to-polymer solution was varied to prepare samples (mixtures) and resulting films with different API content.

158

2.2. Strip film formation process

175

The schematic of the strip film formation process is shown in Fig. 2. The entire process was conducted using precision tape casting equipment (Model TC-71LC, HED International Inc.), where the substrate roller (IX in Fig. 2) was attached to a motor (Baldor Electric), which allowed the strip film to be moved forward along with a 15 cm wide Mylar substrate. The doctor blade (Model 3700 from Elcometer Instruments) was used to hold the prepared mixture and extruded wet film on the surface of the Mylar substrate with a specific thickness controlled by the aperture. The drying began right after the wet film entered the drying chamber (I in Fig. 2). The 2-m long drying chamber consists of three zones as shown in Fig. 2: Zone 1 (II), Zone 2 (III) and Zone 3 (IV). The temperature of each zone can be manipulated independently to heat the bottom surface of the wet film, while counter-flowing air drawn in and out of the chamber by a blower (Model Dayton 1TDP3) and passed through an air heater (Model BHN717N10S24 from Watlow) heats the top surface of the film and removes moisture. Three windows (VI in Fig. 2) are available to monitor the drying process using specific sensors, for example, a fiber optic Raman probe. For the baseline process, wet film thickness, substrate speed, and drying temperatures were fixed and the API concentrations

176

Fig. 2. Schematic of the laboratory scale process used for the preparation of drug-loaded strip films.

Please cite this article in press as: Zhang, J., et al., Raman spectroscopy for in-line and off-line quantification of poorly soluble drugs in strip films. Int J Pharmaceut (2014), http://dx.doi.org/10.1016/j.ijpharm.2014.08.051

159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174

177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197

G Model

IJP 14293 1–10 4 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250

J. Zhang et al. / International Journal of Pharmaceutics xxx (2014) xxx–xxx

used are shown in Table 1. Five samples (i.e., Group A, B, C, D and E referred in Table 1) were cast at a wet thickness of 500 mm with a substrate speed of 0.17 cm/s and the temperature set at 50  C in all three zones (i.e., Zones I, II and III). Since the in-line Raman spectroscopy is ultimately intended for real-time monitoring of the API content in wet films during drying, the spectra were collected while the film was still drying. The typical amount of time required to partially dry the film was 19.6 min, which is the residence time in the drier. In the present work, the selection of drying temperature of 50  C and thinner wet film thickness permitted a relatively short residence time while assuring that possible heatinduced degradation of the physical and chemical properties of the strip film would be avoided. As a reference, a strip film with 1000 mm wet thickness loaded with drug nanoparticles required as long as 80 min to dry at 40  C to the desired moisture end-point (Susarla et al., 2013). All strip films together with the substrates were collected after coming out of the drying chamber, and were dried further at room temperature (approximately 20  C) until they could be peeled off the substrate. These strip films were labeled and sealed in polyethylene bags for further analysis. In order to validate the calibration models of FNB and NPX, another batch of strip films was produced. Using the aforementioned baseline process conditions, but with different mass ratios of nanosuspension-to-polymer solution from those reported in Table 1, allowed us to prepare films with API concentrations different from the baseline films. These strip films served as test samples for Raman spectroscopy in contrast to the previous set of films that were used for calibration. For each API, four samples were prepared with different concentrations than those used for the development of the calibration model. In the second part of the study, NPX in-line calibration models were established for three process scenarios different than the baseline process (see Table 2), while using the API formulations presented in Table 1. Substrate speed, probe location and casting or wet film thickness were considered because they are key process parameters that may impact how the drying process is monitored. For the three scenarios, the parameters were adapted from the baseline values based on the one-factor-at-a-time (OFAT) approach. The selection of these three process parameters can be rationalized as follows: all parameters affect the water content that impacts the composition of the films, which might be reflected in the Raman spectra. Raising the substrate speed will lead to higher water content in the film at a given location in the drier because the drying time is shorter (e.g., increasing the substrate speed from 0.17 to 0.37 cm/s reduced the drying time from 19.6 min to 9 min). Likewise, at any given position in the drier, a thick wet film will have more water content than a thin wet film under the same operating conditions. Obviously, more water is present in the wet film earlier in the drying process (e.g., the film is not as dry in Zone 2 as in Zone 3). While the water evaporation rate is mainly affected by the flow rate of hot air and the temperature of the three zones, these process parameters were not investigated and are outside the scope of the current study. As mentioned before, the

Table 2 Operation settings of 3 scenarios for naproxen. Substrate speed (cm/s) Scenario 1 Scenario 2 Scenario 3 a

Location of optic Casting thickness (mm)a probe

0.17

500

Zone 2

0.37

500

Zone 3

0.17

1000

Zone 3

Casting thickness indicates aperture of the doctor blade.

temperature was set at 50  C, while the hot air flow rate was set at the same value as in a previous study (Susarla et al., 2013). To investigate the impact of different process parameters on calibration model development, principal component analysis (PCA), was used to reduce the dimensionality of the data from the three process scenarios (Abdi and Williams, 2010). The percentage of difference between Raman spectra explained by changes in the three process parameters in the respective scenarios could allow for refinement of the baseline calibration model once the process parameters are varied.

251

2.3. Drug content measurement via direct assaying

261

The actual API concentrations in the films were determined based on drug assaying via off-line UV spectroscopy. These concentrations, hereafter referred to as reference concentrations, were subsequently used for the calibration model development and validation (Section 2.4). For FNB and NPX containing films, the drug assay measurement via UV spectroscopy was already developed and used as a standard procedure (Sievens-Figueroa et al., 2012a). In order to determine the drug assay, round strip film samples were punched out from a whole film by a hole punch (McMaster, US; I.D. 3/8 in.). The film thickness and weight were measured by a micrometer (McMaster, US) and an analytical balance (ML, Mettler Toledo, US), respectively. After these measurements, each sample (n = 3) was dissolved in 250 mL SDS solution (5.4 mg/mL) using a magnet stirring plate (RT5, IKA, US) for 6 h. Finally, an aliquot from this solution was taken to measure the absorption at a specific wavelength, i.e., 290 nm for FNB and 272 nm for NPX, by a UV spectrometer (EvolutionTM 300, Thermo Scientific). The absorbance was then converted into the API concentration via established calibration curves of absorbance vs. concentration.

262

252 253 254 255 256 257 258 259 260

263 264 265 266 267 268 269

Q2 270 271 272 273 274 275 276 277 278 279

2.4. In-line and off-line Raman spectroscopy and development of calibration model

280

For both in-line and off-line measurements, Raman spectroscopy was performed with an EZ Raman system from Enwave Optronics Inc. equipped with a fiber optic probe consisting of a 785 nm laser, emitting with a total power of 250 mW, capable of measuring Raman lines with frequencies ranging from 102 to 3300 cm 1 with 2 cm 1 resolution, and operating at a working distance of 7 mm from the top surface of the film. The equipment's software, EZ Raman Reader from Enwave Optronics Inc. was used for data acquisition. For in-line measurements, the fiber optic probe was located in Zone 3 (see Fig. 1). Spectral scanning involved three consecutive scans at 10 s exposure each of the moving strip film. Similar setting was also applied for the validation experiments. For off-line measurements, the strip film was cut into small rectangles and examined with the EZ Raman system coupled to an optical microscope (Leica BME) using a 10 objective with 10 mm spot size. The laser beam was focused on the samples prior to each measurement, and 10 s exposures each of the strip film at 5 randomly selected locations were used for the acquisition of Raman data. All spectral data were analyzed using the Unscrambler1 X version 10.3-software from CAMO. For Raman spectral analysis, the standard normal variate (SNV) method was used as the pre-treatment which has been widely accepted for eliminating the differences caused by variations of the baseline and spurious data generated by laser intensity fluctuations (Barnes et al., 1989). The partial least square (PLS) model was employed for regression analysis of the Raman spectra. In order to assess the calibration model, three statistical parameters, namely, the correlation coefficient (R2), root mean squared error of cross validation (RMSECV) and root mean squared error of calibration (RMSEC), were used. The RMSECV indicates the standard error generated by leave-one-out cross validation

282

Please cite this article in press as: Zhang, J., et al., Raman spectroscopy for in-line and off-line quantification of poorly soluble drugs in strip films. Int J Pharmaceut (2014), http://dx.doi.org/10.1016/j.ijpharm.2014.08.051

281

283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312

G Model

IJP 14293 1–10 J. Zhang et al. / International Journal of Pharmaceutics xxx (2014) xxx–xxx

5

Fig. 3. Raman spectra of the five FNB strip films with different drug loading, Mylar substrate and pure FNB powder. The left panel shows the Raman spectra in the wavenumber range of 102–3300 cm 1 that are offset along the intensity axis to enhance clarity. The right panel shows the Raman spectra in the wavenumber range of 1024–1550 cm 1 (see the boxed region in the left panel, but without offset along the intensity axis), which were actually used for the development of the calibration model. In the wavenumber range of 1024–1550 cm 1, notable Raman lines at 1092 cm 1 and 1148 cm 1 appeared from the FNB strip films and pure FNB powder due to presence of FNB, whereas no obvious Raman line appeared due to Mylar. 313

in poor film quality due to an insufficient amount of polymer for film formation. Future studies will consider higher drug loadings for the refinement of the calibration model.

333

315

(Rencher and Christensen, 2012), and RMSEC is a standard error calculated from the calibration data (Orzel et al., 2012; Staudenmayer et al., 2012).

316

3. Results and discussion

3.2. Interpretation of Raman spectra

336

317

3.1. Quantification of reference concentrations

Since the laser beam can easily penetrate the wet film during spectra collection, the Raman spectra would also include the contribution from the substrate, which is Mylar in this case. Figs. 3 and 4 illustrate the in-line spectra of the FNB and NPX films, respectively, along with those of Mylar and the as-received APIs. The different colors indicate the spectra of the strip films with different API concentrations. For Mylar, which is composed of polyethylene terephthalate (PET), two strong lines at 1602 and 1730 cm 1 assigned to the n(CQO) stretching mode of PET are observed. However, since carbonyl groups are present in both FNB and NPX (see Fig. 2), lines associated with the n(CQO) stretching mode also exist in the Raman spectra of FNB and NPX. To eliminate the interference due to Mylar, Raman spectra in the 1024–1550 cm 1 and 350–780 cm 1 regions for FNB and NPX, respectively, were studied. For FNB, two lines at 1092 and 1148 cm 1 assigned to the n(C O) mode (Saerens et al., 2011) were analyzed, whereas for NPX, three lines, which can be assigned

337

314

318 319 320 321 322 323 324 325 326 327 328 329 330 331 332

The API concentrations, also referred to as reference concentrations, in the strip films were measured using the procedure discussed in Section 2.3 by direct assaying via UV spectroscopy and are presented in Table 1. They were only used for the development of the calibration models. Table 1 indicates that the use of lower mass ratios of API nanosuspension-to-polymer solution in the mixture indeed yielded strip films with lower API concentrations, which was intended during the preparation step. The concentration difference between FNB and NPX for the same mass ratio of drug nanosuspension-to-polymer solution is most likely caused by the unknown amount of water lost during preparation of the nanosuspension and polymer solution as well as potential drug inhomogeneity in the mixtures. For the current investigation, the maximum API loading was kept at 26% (w/w) to assure good film quality because it was found that higher API concentration resulted

Fig. 4. Raman spectra of the five NPX strip films, Mylar substrate and pure NPX powder. The left panel shows the Raman spectra in the wavenumber range of 102–3300 cm 1 that are offset along the intensity axis to enhance clarity. The right panel shows the Raman spectra in the wavenumber range of 350–780 cm 1 (see the boxed region in the left panel, but without offset along the intensity axis), which were actually used for the development of the calibration model. In the wavenumber range of 350–780 cm 1, notable Raman lines at 410 cm 1, 524 cm 1 and 742 cm 1 appeared from the NPX strip films and pure NPX powder, whereas no obvious Raman line appeared due to Mylar.

Please cite this article in press as: Zhang, J., et al., Raman spectroscopy for in-line and off-line quantification of poorly soluble drugs in strip films. Int J Pharmaceut (2014), http://dx.doi.org/10.1016/j.ijpharm.2014.08.051

334 335

338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353

G Model

IJP 14293 1–10 6

J. Zhang et al. / International Journal of Pharmaceutics xxx (2014) xxx–xxx

Fig. 5. Second derivative Raman spectra of FNB (left panel) and NPX (right panel) strip films in the wavenumber ranges of 1100–1200 cm 1 and 700–800 cm 1, respectively. Details of the line at 1148 cm 1 emanating from the n(C O) mode and the line at 742 cm 1 emanating from the n(C C) mode, respectively, are depicted. For FNB strip films, the band locations of five samples are consistent with crystalline FNB, while a small shift of the peak position is observed in the amorphous form. For NPX strip films, the band locations of five samples are consistent with crystalline NPX, while a sizable shift of the peak position is observed for the amorphous form. The values of the 2nd derivative have been normalized to [0,1] for the sake of clarity. The Raman spectra of the amorphous phase were collected soon after formation of the metastable amorphous phase.

365

to planar and ring deformations associated with n(C C) motions (Jubert et al., 2006) appearing at 410, 524 and 742 cm 1, were analyzed. As expected, the intensities of these lines increased with an increase in the API concentration. Overall, interference due to Mylar was eliminated, and only the key molecular vibrational modes of the two APIs were included in the data set used to develop the calibration model. When a different substrate is used, a similar procedure may be used to eliminate spectral interference from that substrate. Off-line Raman measurements for FNB and NPX were made in the same spectral ranges as shown in Figs. 3 and 4, respectively, which were quite similar with in-line spectra and omitted for the sake of brevity.

366

3.3. Raman analysis of crystalline phase of the APIs

367

In Sievens-Figueroa et al. (2012a), Raman spectroscopy was used to analyze FNB and NPX nanoparticle carrying films. However, a systematic analysis to examine if the crystalline phase is preserved after drying of the strip film has not been reported. For this purpose, off-line Raman spectral data of the dry film was used for crystalline phase analysis. Accordingly, second derivative transformations of the Raman spectra, which are commonly used for spectral pre-treatment (Saerens et al., 2011), were performed for different concentrations of FNB and NPX as shown in Fig. 5. For the sake of brevity, only a single Raman wavenumber (spatial frequency) is shown in each case; i.e., 1148 cm 1 and 742 cm 1, for FNB and NPX, respectively.

354 355 356 357 358 359 360 361 362 363 364

368 369 370 371 372 373 374 375 376 377

For both FNB and NPX, the Raman frequencies at 1148 cm 1 and 742 cm 1, respectively, remained unchanged and consistent with the corresponding crystalline forms for different concentrations. It is noted that frequency shifts would be observed if there were a drug phase change such as conversion to the amorphous forms. This indicates that API–polymer interactions did not cause crystalline phase transformations during the strip film formation process, which include several steps, including mixing, casting and drying. It also indicates that the drying temperature of 50  C used did not lead to the temperature-induced formation of amorphous phases of the APIs, which could occur at high temperatures (Tumuluri et al., 2008). These results indicate that the film formation process is gentle and robust for processing of the two model BCS Class II drugs considered. The results also indicate that Raman spectroscopy could be used to monitor qualitative changes of crystallinity and crystal structure of the API during manufacturing, although here, it was done off-line. Doing this in-line would require integrating data collection steps with the processing algorithms, such as selecting a specific Raman band and applying second derivate treatment, and assessing the variance of possible Raman shift.

378

3.4. Development of the calibration model

398

For in-line measurements, the calibration models for the two APIs are presented in Fig. 6. For the FNB film, a correlation coefficient of 0.9976 was obtained. Three scans per concentration

399

Fig. 6. Plots of the in-line calibration models of API concentration (left, FNB; right, NPX) based on the corresponding in-line Raman spectra. R2, RMSEC and RMSECV are indicators used to assess the validity of the calibration models. The optimal number of PLS factors was selected based on the percentage of variance in the spectra and API concentrations explained by the calibration model, which was calculated by the Unscrambler1 software.

Please cite this article in press as: Zhang, J., et al., Raman spectroscopy for in-line and off-line quantification of poorly soluble drugs in strip films. Int J Pharmaceut (2014), http://dx.doi.org/10.1016/j.ijpharm.2014.08.051

379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397

400 401

G Model

IJP 14293 1–10 J. Zhang et al. / International Journal of Pharmaceutics xxx (2014) xxx–xxx

7

Fig. 7. Plots of the off-line calibration models of API concentration (left, FNB; right, NPX) based on the corresponding off-line Raman spectra. R2, RMSEC and RMSECV are indicators used to assess the validity of the calibration models. The optimal number of PLS factors was selected based on the percentage of variance in the spectra and API concentrations explained by the calibration model, which was calculated by the Unscrambler1 software. 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448

were used for the development of the calibration model, and the model was assessed with a RMSEC of 0.36% and a RMSECV of 0.69%. For the NPX film, three scans per concentration were used, and a correlation coefficient of 0.9961, RMSEC of 0.42% and RMSECV of 0.69% were obtained. Fig. 7 presents the calibration models that are based on off-line measurements for the two APIs. Five scanning results were used for the calibration model, so each group of points includes five measurements for each reference concentration of the APIs. For the FNB film, the RMSEC, RMSECV and correlation coefficient were 0.44%, 0.73% and 0.9946, respectively. For the NPX films, a strong linear relationship was observed by the correlation coefficient, RMSEC and RMSECV values of 0.9969, 0.36% and 1.1%, respectively. It is important to note that the Raman spectra of in-line and offline measurements were collected in different contexts: in-line spectra were collected from a moving strip film with partial residual water content and contained various background noise including some from Mylar substrate despite the minimization owing to the selection of a specific segment of the Raman spectra, whereas off-line spectra were collected from a static dry strip film. Hence, the quality of the off-line calibration model is expected to be better than that for the in-line. Nevertheless, based on the results shown in Figs. 6 and 7, the correlation coefficients and RMSECs generated from in-line and off-line measurements were similar, indicating good linearity of response throughout the calibration range. This suggests that the scattering cross-section variations caused by different residual water contents within inline and off-line samples will not significantly impact the calibration models. It also suggests that specific segments of the Raman spectra, i.e., from 1024 cm 1 to 1550 cm 1 and from 350 cm 1 and 780 cm 1 for FNB and NPX, respectively, can be used to calibrate the API concentration accurately. The optimal number of PLS factors, i.e., the number of latent variables, presented in Figs. 6 and 7, was selected based on the percentage of variance in the spectra and API concentrations explained by the calibration model, which was calculated by the Unscrambler1 software. The in-line PLS factors (3 for FNB and 2 for NPX) are similar to those utilized in the Raman calibration models used for monitoring API concentration during hot melt extrusion (HME) process (Saerens et al., 2011; Tumuluri et al., 2008). A difference between off-line and in-line calibration models was observed for the RMSECV values of NPX calibration models, i.e., 1.1% and 0.69%. This finding is counterintuitive as the off-line calibration model is expected to be better than the in-line calibration model due to the more stable environment of spectra collection in the former. The difference may be explained based on the scale of scrutiny used for the determination of the API

concentration by Raman spectroscopy. The off-line spectra represented the NPX concentration of five random 10 mm spots, whereas the in-line spectra represented the averaged NPX concentration within the area of 10 mm by 17 mm, which is obviously much larger than the 10 mm spots used for off-line analysis. A larger scale of scrutiny may have led to lower RMSECV of the in-line calibration model despite the non-stationary nature of the film. RMSECV of the off-line calibration models suggests that the relative standard deviation of FNB is better than that of NPX, whereas the in-line models suggest identical RMSECV for both drugs. This difference between the in-line and off-line models may also originate from the different scales of scrutiny and moving vs. stationary nature of the films during the spectra collection.

449

3.5. Validation of calibration models

462

The validation results are given in Table 3, where the reference concentration is that of the dry film quantified by direct assaying via UV spectroscopy (Section 2.3), the predicted concentration is the one calculated by projection of the in-line Raman spectra collected during the strip film formation onto the corresponding in-line calibration models. The optimal number of PLS factors were 3 and 2 for FNB and NPX, respectively. The prediction involved three samples and corresponding standard deviation values are provided in Table 3. The validation results suggest that the calibration models developed under the baseline process conditions can accurately monitor the API concentration in-line during the strip film formation process. Since accurate predictions were observed, it

463

Table 3 Validation of FNB and NPX in-line Raman calibration models using a different batch of strip films. API

Reference concentrationa (w/w %)

Predicted concentrationb (w/w %), n = 3

Standard deviation (%), n = 3

FNB (PLS factor:3)

14.1 19.2 7.5 17.4

15.0 19.3 8.5 18.7

0.71 1.15 0.53 0.62

NPX (PLS factor:2)

14.6 19.1 7.3 17.5

15.6 19.7 7.7 18.7

0.76 1.67 0.86 0.92

a Determined from drug assaying of the films, which were not used in the Raman calibration, via UV spectroscopy. b Predicted using established in-line Raman calibration model.

Please cite this article in press as: Zhang, J., et al., Raman spectroscopy for in-line and off-line quantification of poorly soluble drugs in strip films. Int J Pharmaceut (2014), http://dx.doi.org/10.1016/j.ijpharm.2014.08.051

450 451 452 453 454 455 456 457 458 459 460 461

464 465 466 467 468 469 470 471 472 473 474 475

G Model

IJP 14293 1–10 8 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537

J. Zhang et al. / International Journal of Pharmaceutics xxx (2014) xxx–xxx

is reasonable that any sample with concentration ranging from 7.2 to 19.2% (w/w) can be accurately predicted by the calibration models, and it is expected that to a first approximation, the calibration models would also be valid for the whole reference concentration range, i.e., 2.6 to 26% (w/w). 3.6. Impact of process parameters on Raman spectroscopic monitoring of API concentration In order to examine how changes in the processing conditions can impact the calibration models, the in-line calibration models for NPX in three process scenarios, which are different from the baseline process, were considered (refer to Table 2). These calibration models are used not only to confirm the capability of Raman spectroscopy for quantifying the API concentration under different process conditions, but also to help better understand the strip film formation process, especially that of drying. At first, the same NPX calibration model that was obtained from the baseline process was utilized to predict the API concentration from each of the Raman spectra collected from the three process scenarios (Fig. 6). Significant errors in the prediction of API concentrations were observed because the water contents in these different process scenarios are different. Different wet film thicknesses and working distances from the laser source may impact the penetration and scattering of the light beam at the same setting, ultimately causing a variation of the Raman spectral intensities. To alleviate this problem, a new calibration model for each of the three process scenarios was developed in the same way as for the baseline process by collecting the Raman spectra and using the measured reference concentrations of the API in the films via drug assay using UV spectroscopy. For these new models, the correlation coefficients (R2) were all greater than 0.9937 for both in-line and off-line data, and the RMSECs were all less than 0.54%, demonstrating predictive capability. Thus overall, these results indicate the validity of the proposed methodology and that a new in-line calibration model should be established and re-validated once the operating conditions have been changed, which is also required as per PAT guidance (FDA and CVM, 2004). 3.7. PCA analysis of the impact of process parameters on Raman spectra In order to interpret the Raman spectra, three components were considered: API concentration, process parameters and withinsample variation in the Raman spectrum. The first two are highly relevant to the use of Raman for the film process monitoring, whereas the last one gives a measure of relative error in spectral data. In order to sort out the weight of each of the three components with regard to the interpretation of the variations of Raman spectra, PCA analysis was implemented via transitive comparison. Within-sample variations of the Raman spectrum and process parameters were analyzed first to determine their weight in explaining the Raman spectrum under identical API concentration. Then, the variations of the Raman spectra were evaluated when both the API concentration, and one of three process parameters were varied. Based on these results, the impact of the three processing parameters and API concentration on the calibration models can be quantified. The PCA results are given in Fig. 8 where the ellipse shown in all figures is indicative of the multivariate 95% confidence interval for the spectral variance, and the percentages of the first two principal components (factors), i.e., PC-1 and PC-2, were computed by Unscrambler1 X, and their definitions in each PCA result are explained below. The notation employed for the panels within Fig. 8, is as follows; 1, 2, and 3 represent variation in the process parameters, specifically, probe location, substrate speed, and wet film thickness, respectively;

whereas, within each scenario 1, 2, and 3, “a” represents fixed API concentration, while “b” represents varying API concentration. First, the influence of three process parameters is examined for fixed API concentration (Group A). PCA results for different locations of the fiber optic probe are shown in Fig. 8, panel 1a. Here, 61% of the Raman spectral variance could be explained by the first major component, i.e., PC-1. We can infer that in this case, PC-1 mainly represents different probe locations, while PC-2 mainly represents within-sample spectral variation, accounting for 22%. This indicates that the spectral variation is greater between Zone 2 and Zone 3 (different probe locations) than for the samples obtained within one of the zones. This means that the differences between locations are highly evident. The PCA score plot (Fig. 8 panel 1a) also shows that scores from Zone 3 are very close together which is indicative of similarity in spectra, indicating that intrinsic error is low compared to the variation in the process parameter, probe location. Similar results were obtained from the scenarios for the remaining two process parameters; substrate speeds and wet film thicknesses and are shown in Fig. 8, panels 2a and 3a, respectively. For example, panel 3a shows that scores from the 500 mm and 1000 mm lie on different sides of PC-1, indicating spectral differences that are from thickness variation. The PCA score plot also shows that Raman spectra are affected by film thickness, indicating the feasibility of predicting film thickness, which will be discussed further through the results from panel 3b. Overall, the results from panels 1a, 2a, and 3a, suggest that the process parameters would have greater influence in interpreting the Raman spectra than within-sample spectral variation, as long as the API concentration is fixed. On the other hand, when the process conditions are the same, one would expect that the API concentration would influence a greater proportion of the Raman spectra than within-sample spectrum variation (Saerens et al., 2011). Therefore, next, the API concentrations are also changed. PCA was used to evaluate spectral changes when both concentration and processing parameters are varied, and the results are shown in the b panels of Fig. 8. For the spectra obtained from Zones 2 and 3 (Fig. 8, panel 1b), PC-1 accounts for 79% of the variance in the Raman spectra, whereas PC-2 accounts for only 6%. Here, it may be inferred that PC-1 mainly represents the different API concentrations, and PC-2 mainly represents different locations. For the same API concentration, different locations would have varying water content but similar concentrations of API particles per cross-sectional segment of a film. Since Raman spectroscopy would be only weakly sensitive to the n(O H) stretching vibration of water (McGoverin et al., 2012; Shi et al., 2012), it is expected that the spectra would arise mainly from the API particles, but still would be impacted by residual water content, e.g., 6% of Raman spectral variations. Similar results were obtained from the scenario of different substrate speeds (Fig. 8, panel 2b). The PC-1, representing API concentration, accounts for 77% of the variance in the Raman spectra, while PC-2, representing different substrate speeds accounts for only 8%. Interestingly, the trends seen in the results for different thicknesses do not follow that observed in the previous two cases. That is because the number of API particles per cross-sectional segment is different from that reflected in the spectra (Fig. 8, panel 3b), where PC-1, which here represents mainly the thickness, determines 51% of Raman spectral variations, and PC-2, representing the API concentration, determines 32%. This is an interesting and important outcome that suggests that Raman spectroscopy could also be used to monitor the film thickness since the thickness of API loaded film contributed more significantly to the Raman spectral variation than the API concentration did. These results have also been validated using the K-mean cluster for group analysis. For example, the spectra from samples of different thicknesses (Fig. 8, panel 3b) can be isolated into two groups, each representing one thickness, whereas the spectra from different locations and substrate speeds cannot be similarly isolated. These

Please cite this article in press as: Zhang, J., et al., Raman spectroscopy for in-line and off-line quantification of poorly soluble drugs in strip films. Int J Pharmaceut (2014), http://dx.doi.org/10.1016/j.ijpharm.2014.08.051

538 539 540 541 542 543 544 545 546 547 548 549

Q3 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603

G Model

IJP 14293 1–10 J. Zhang et al. / International Journal of Pharmaceutics xxx (2014) xxx–xxx

9

Fig. 8. PCA of the in-line calibration spectra of NPX strip films obtained from three process scenarios involving different probe locations (1a, 1b), different substrate speeds (2a, 2b), and different wet film thicknesses (3a, 3b). Panels “a” are for fixed API concentration (Group A), while “b” for varying API concentration (Groups A–E in Table 1). Each data point corresponds to a different Raman spectrum. The ellipse is indicative of the multivariate 95% confidence interval for the spectral variance. The percentage of spectral variance explained by the first two principal components, i.e., PC-1 and PC-2, were computed by Unscrambler1 X. Capital and lowercase letters for a group label denote the same formulation and API concentration, but are designated for the two different values of each process parameter. 604 605 606

results also address the explanation in Section 3.6 that calibration model should be revalidated once the operating conditions have been changed, which is in line with FDA recommendations.

607

4. Conclusions

608

This study demonstrates that Raman spectroscopy is an attractive PAT tool for monitoring the strip film manufacturing process with respect to in-line quantification of API concentration. All the RMSECs obtained were less than 0.6%, demonstrating good accuracy of the calibration models. The off-line results supported the in-line measurements, which indicated that the partial segments of the Raman spectra used to avoid interference from the substrate spectrum and residual water content provide accurate quantification of the API concentrations in the strip films. The results also suggest that in-line Raman spectroscopy is

609 610 611 612 613 614 615 616 617

an efficient PAT approach for monitoring API concentration since it allows the user to avoid repetitious sample preparation and to make measurements in real-time. The evaluations conducted on a separate batch of strip films enabled us to validate the calibration model and to demonstrate that the API concentration can be accurately predicted. These findings suggest that Raman spectroscopy would be a suitable PAT tool to implement feedback control strategy in continuous manufacturing of strip films through monitoring of API concentration. In addition, Raman spectroscopy is also able to detect possible changes in the solid-state of the API. The calibration models established by varying three different operating conditions confirmed the applicability of the methodology and the capability of Raman spectroscopy for API concentration monitoring during the drying process. In addition, this exercise demonstrated that the in-line calibration model must be revalidated once the operating conditions have been changed; a

Please cite this article in press as: Zhang, J., et al., Raman spectroscopy for in-line and off-line quantification of poorly soluble drugs in strip films. Int J Pharmaceut (2014), http://dx.doi.org/10.1016/j.ijpharm.2014.08.051

618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633

G Model

IJP 14293 1–10 10 634

J. Zhang et al. / International Journal of Pharmaceutics xxx (2014) xxx–xxx

647

protocol that is in line with the FDA guidelines. The PCA analysis of three process scenarios demonstrated that Raman spectroscopy is not only a flexible tool for monitoring the API concentration in-line under different operating conditions, but also may be used to gain process understanding. From the process control perspective, the PCA results indicated that the film thickness could also be monitored by Raman spectroscopy, although the substrate speed would require an extra sensor for control implementation. It was evident that the Raman spectrometer can be installed at any location of the strip film formation process for the purpose of quantifying the API concentration in-line. Thus overall, this work demonstrates the potential of Raman spectroscopy as an effective PAT tool for strip film manufacturing process through facilitating detection of drug concentration and form in real-time.

648

Acknowledgments

649

653

The authors acknowledge financial support of this work through an NSF-Engineering Research Center award (EEC0540855), as well as support for BPS through an NSF-RET award (EEC-0908889). The authors are grateful to Scott Krull for his meticulous editing of the manuscript.

654

References

635 636 637 638 639 640 641 642 643 644 645 646

650 Q4 651 652

655 656 657 658 659

Abdi, H., Williams, L.J., 2010. Principal component analysis. Wiley Interdiscip. Rev. Comput. Stat. 2, 433–459. Afolabi, A., Akinlabi, O., Bilgili, E., 2014. Impact of process parameters on the breakage kinetics of poorly water-soluble drugs during wet stirred media milling: a microhydrodynamic view. Eur. J. Pharm. Sci. 51, 75–86. Q5 Aulton, M.E., Wells, T., 2002. Pharmaceutics: The Science of Dosage Form Design. 660 Barnes, R., Dhanoa, M., Lister, S.J., 1989. Standard normal variate transformation and 661 de-trending of near-infrared diffuse reflectance spectra. Appl. Spectrosc. 43, 662 772–777. 663 Beck, C., Sievens-Figueroa, L., Gärtner, K., Jerez-Rozo, J.I., Romañach, R.J., Bilgili, E., 664 Davé, R.N., 2013. Effects of stabilizers on particle redispersion and dissolution 665 from polymer strip films containing liquid antisolvent precipitated griseofulvin 666 particles. Powder Technol. 236, 37–51. 667 Bhakay, A., Merwade, M., Bilgili, E., Dave, R.N., 2011. Novel aspects of wet milling for 668 the production of microsuspensions and nanosuspensions of poorly water669 soluble drugs. Drug Dev. Ind. Pharm. 37, 963–976. 670 Bhakay, A., Davé, R., Bilgili, E., 2013. Recovery of BCS Class II drugs during aqueous 671 redispersion of core–shell type nanocomposite particles produced via fluidized 672 bed coating. Powder Technol. 236, 221–234. 673 Bhakay, A., Azad, M., Bilgili, E., Dave, R., 2014. Redispersible fast dissolving 674 nanocomposite microparticles of poorly water-soluble drugs. Int. J. Pharm. 461, 675 367–379. 676 Bilgili, E., Afolabi, A., 2012. A combined microhydrodynamics–polymer adsorption 677 analysis for elucidation of the roles of stabilizers in wet stirred media milling. 678 Int. J. Pharm. 439, 193–206. 679 Brown, C.K., Chokshi, H.P., Nickerson, B., Reed, R.A., Rohrs, B.R., Shah, P.A., 2004. 680 Q6 Dissolution testing of poorly soluble compounds. Pharm. Technol. 681 Cebeci Maltaş, D., Kwok, K., Wang, P., Taylor, L.S., Ben-Amotz, D., 2013. Rapid 682 classification of pharmaceutical ingredients with Raman spectroscopy using 683 compressive detection strategy with PLS-DA multivariate filters. J. Pharm. 684 Biomed. Anal. 80, 63–68. 685 Cilurzo, F., Cupone, I.E., Minghetti, P., Selmin, F., Montanari, L., 2008. Fast dissolving 686 films made of maltodextrins. Eur. J. Pharm. Biopharm. 70, 895–900. 687 Dalvi, S.V., Azad, M.A., Dave, R., 2013. Precipitation and stabilization of ultrafine 688 particles of fenofibrate in aqueous suspensions by RESOLV. Powder Technol. 689 236, 75–84. 690 Dixit, R., Puthli, S., 2009. Oral strip technology: overview and future potential. J. 691 Control. Release 139, 94–107. Q7 FDA, U., CVM, O., 2004. Guidance for Industry: Pat-A Framework for Innovative 692 Pharmaceutical Development, Manufacturing, and Quality Assurance, Rock693 ville, MD.

Heinz, A., Gordon, K.C., McGoverin, C.M., Rades, T., Strachan, C.J., 2009. Understanding the solid-state forms of fenofibrate–a spectroscopic and computational study. Eur. J. Pharm. Biopharm. 71, 100–108. Hu, J., Ng, W.K., Dong, Y., Shen, S., Tan, R.B., 2011. Continuous and scalable process for water-redispersible nanoformulation of poorly aqueous soluble APIs by antisolvent precipitation and spray-drying. Int. J. Pharm. 404, 198–204. Jubert, A., Legarto, M.L., Massa, N.E., Tévez, L.L., Okulik, N.B., 2006. Vibrational and theoretical studies of non-steroidal anti-inflammatory drugs ibuprofen [2-(4isobutylphenyl)propionic acid]; naproxen [6-methoxy-a-methyl-2-naphthalene acetic acid] and tolmetin acids [1-methyl-5-(4-methylbenzoyl)-1Hpyrrole-2-acetic acid]. J. Mol. Struct. 783, 34–51. Liversidge, G.G., Cundy, K.C., 1995. Particle size reduction for improvement of oral bioavailability of hydrophobic drugs: I. Absolute oral bioavailability of nanocrystalline danazol in beagle dogs. Int. J. Pharm. 125, 91–97. McGoverin, C.M., Hargreaves, M.D., Matousek, P., Gordon, K.C., 2012. Pharmaceutical polymorphs quantified with transmission Raman spectroscopy. J. Raman Spectrosc. 43, 280–285. Merisko-Liversidge, E., Liversidge, G.G., Cooper, E.R., 2003. Nanosizing: a formulation approach for poorly-water-soluble compounds. Eur. J. Pharm. Sci. 18, 113–120. Michelet, A., Boiret, M., Lemhachheche, F., Malec, L., Tfayli, A., Ziemons, E., 2013. Use of Raman spectrometry in the pharmaceutical field. STP Pharma Pratiques 23. Monteiro, A., Afolabi, A., Bilgili, E., 2013. Continuous production of drug nanoparticle suspensions via wet stirred media milling: a fresh look at the Rehbinder effect. Drug Dev. Ind. Pharm. 39, 266–283. Orzel, J., Daszykowski, M., Walczak, B., 2012. Controlling sugar quality on the basis of fluorescence fingerprints using robust calibration. Chemometr. Intell. Lab. Syst. 110, 89–96. Pataki, H., Csontos, I., Nagy, Z.K., Vajna, B., Molnar, M., Katona, L., Marosi, G., 2012. Implementation of Raman signal feedback to perform controlled crystallization of carvedilol. Org. Process Res. Dev. 17, 493–499. Patravale, V., Kulkarni, R., 2004. Nanosuspensions: a promising drug delivery strategy. J. Pharm. Pharmacol. 56, 827–840. Pelletier, M.J., 1999. Analytical Applications of Raman Spectroscopy. WileyBlackwell. Rencher, A.C., Christensen, W.F., 2012. Methods of Multivariate Analysis. John Wiley & Sons. Rozo, J.I., Zarow, A., Zhou, B., Pinal, R., Iqbal, Z., Romanach, R.J., 2011. Complementary near-infrared and Raman chemical imaging of pharmaceutical thin films. J. Pharm. Sci. 100, 4888–4895. Saerens, L., Dierickx, L., Lenain, B., Vervaet, C., Remon, J.P., De Beer, T., 2011. Raman spectroscopy for the in-line polymer–drug quantification and solid state characterization during a pharmaceutical hot-melt extrusion process. Eur. J. Pharm. Biopharm. 77, 158–163. Shi, L., Gruenbaum, S.M., Skinner, J.L., 2012. Interpretation of IR and Raman line shapes for H2O and D2O ice Ih. J. Phys. Chem. B 116, 13821–13830. Sievens-Figueroa, L., Bhakay, A., Jerez-Rozo, J.I., Pandya, N., Romanach, R.J., Michniak-Kohn, B., Iqbal, Z., Bilgili, E., Dave, R.N., 2012a. Preparation and characterization of hydroxypropyl methyl cellulose films containing stable BCS Class II drug nanoparticles for pharmaceutical applications. Int. J. Pharm. 423, 496–508. Sievens-Figueroa, L., Pandya, N., Bhakay, A., Keyvan, G., Michniak-Kohn, B., Bilgili, E., Davé, R.N., 2012b. Using USP I and USP IV for discriminating dissolution rates of nano-and microparticle-loaded pharmaceutical strip-films. AAPS PharmSciTech 13, 1473–1482. Singh, R., Ierapetritou, M., Ramachandran, R., 2012. An engineering study on the enhanced control and operation of continuous manufacturing of pharmaceutical tablets via roller compaction. Int. J. Pharm. 438, 307–326. Staudenmayer, J., Zhu, W., Catellier, D.J., 2012. Statistical considerations in the analysis of accelerometry-based activity monitor data. Med. Sci. Sports Exerc. 44, S61–67. Susarla, R., Sievens-Figueroa, L., Bhakay, A., Shen, Y., Jerez-Rozo, J.I., Engen, W., Khusid, B., Bilgili, E., Romañach, R.J., Morris, K.R., Michniak-Kohn, B., Davé, R.N., 2013. Fast drying of biocompatible polymer films loaded with poorly watersoluble drug nano-particles via low temperature forced convection. Int. J. Pharm. 455, 93–103. Tumuluri, V.S., Kemper, M.S., Lewis, I.R., Prodduturi, S., Majumdar, S., Avery, B.A., Repka, M.A., 2008. Off-line and on-line measurements of drug-loaded hot-melt extruded films using Raman spectroscopy. Int. J. Pharm. 357, 77–84. Van Eerdenbrugh, B., Van den Mooter, G., Augustijns, P., 2008. Top–down production of drug nanocrystals: nanosuspension stabilization, miniaturization and transformation into solid products. Int. J. Pharm. 364, 64–75. Yu, L.X., Amidon, G.L., Polli, J.E., Zhao, H., Mehta, M.U., Conner, D.P., Shah, V.P., Lesko, L.J., Chen, M.-L., Lee, V.H., 2002. Biopharmaceutics classification system: the scientific basis for biowaiver extensions. Pharm. Res. 19, 921–925.

Please cite this article in press as: Zhang, J., et al., Raman spectroscopy for in-line and off-line quantification of poorly soluble drugs in strip films. Int J Pharmaceut (2014), http://dx.doi.org/10.1016/j.ijpharm.2014.08.051

694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 Q8 714 715 716 717 718 Q9 719 720 721 722 723 724 725 Q10 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765

Raman spectroscopy for in-line and off-line quantification of poorly soluble drugs in strip films.

Raman spectroscopy was used as a process analytical technology (PAT) tool for in-line measurement of active pharmaceutical ingredient (API) content du...
2MB Sizes 6 Downloads 6 Views