Early stages of drug crystallization from amorphous solid dispersion via fractal analysis based on chemical imaging

Depósito de investigación de la Universidad de Sevilla https://idus.us.es/ “This is an Accepted Manuscript of an article published by Elsevier in: EUROPEAN JOURNAL OF PHARMACEUTICS AND BIOPHARMACEUTICS on 2018, available at: https://doi.org/10.1016/j.ejpb.2018.10.007” 1 Early stages of drug crystallization from amorphous solid dispersion via fractal analysis 1 based on chemical imaging 2 Renata Abreu-Villela 1,2, Monica Schönenberger 3, Isidoro Caraballo 1, Martin Kuentz 2 3 4 1Universidad de Sevilla, Department of Pharmacy and Pharmaceutical Technology, C/Profesor García Gon- 5 zález 2, 41012 Seville, Spain. 6 2University of Applied Sciences and Arts Northwestern Switzerland, Institute of Pharma Technology, 7 Hofackerstrasse 30, 4132 Muttenz, Switzerland 8 3University of Basel, Swiss Nanoscience Institute, Nano Imaging Lab, Klingelbergstrasse 82, 4056 Basel, 9 Switzerland. 10 11 12 13 Address correspondence to 14 Prof. Dr. Martin Kuentz 15 University of Applied Sciences and Arts Northwestern Switzerland 16 Institute of Pharma Technology 17 Hofackerstr. 30 18 4132 Muttenz / Switzerland 19 T +41 (61) - 228 56 42 20 [email protected] 21 22 23 24 2 Abstract 25 Early stages of crystallization from amorphous solid dispersion (ASD) are typically not detected by means 26 of standard methods like powder X-ray diffraction (XRPD). The aim of this study is therefore to evaluate 27 if fractal analysis based on energy dispersive x-ray imaging can provide the means to identify early signs 28 of physical instability. ASDs of the poorly water-soluble compound, felodipine (FEL) were prepared by 29 solvent evaporation using different grades of HPMCAS, at 50wt. % drug loading. Samples were stored at 30 accelerated conditions of 40˚C. Scanning electron microscopy equipped with an energy-dispersive X-ray 31 spectroscopy (SEM-EDS) was used for elemental mapping of tablet surfaces. Comparative data were gen- 32 erated with a standard XRPD and with more sensitive methods for detection of early instability, i.e. laser 33 scanning confocal microscopy (LSM) and atomic force microscopy (AFM). The SEM-EDS identified 34 changes of drug-rich domains that were confirmed by LSM and AFM. Early changes in drug clusters were 35 also revealed by a multifractal analysis that indicated a beginning phase separation and drug crystallization. 36 Therefore, the presented fractal cluster analysis based on chemical imaging bears much promise as a new 37 method to detect early signs of physical instability in ASD, which is of great relevance for pharmaceutical 38 development. 39 40 Keywords: solid dispersion; amorphous drug; phase separation; crystallization; chemical imaging; mul- 41 tifractal analysis 42 43 44 3 1. Introduction 45 Amorphous solid dispersion (ASD) has become an established oral delivery technique to formulate poorly 46 water-soluble drugs [1–3]. Given the large number of papers that are published on solid dispersions, it is 47 rather surprising that there are not many more of these products on the pharmaceutical market [4]. One of 48 the main reasons is that ASDs are metastable systems and bear the risk of physical instability during their 49 shelf life. Drug crystallization can occur depending on the history of an amorphous product and on given 50 conditions [5]. Particularly critical is storage at elevated temperature and moisture leading eventually to 51 phase separation and crystallization, which means a loss of biopharmaceutical advantages using ASD [6– 52 8]. The molecular processes of phase separation as well as crystallization from amorphous material are 53 complex and mobility in the glass state plays an important role. Relaxation of amorphous materials take 54 place on different time scales from primary diffusive (αrelaxation) to secondary local relaxation such as 55 Johari-Goldstein relaxations (β-relaxation). While αrelaxation becomes very slow in the glass state, it is 56 mostly the secondary relaxations that are relevant for crystallization from amorphous state [9–11]. Once 57 crystallization starts, it continues to reduce the system's free energy [12,13]. Thus, thermodynamics is the 58 driver for physical change and molecular mobility is a facilitator, while surface effects can act as modula- 59 tors, and heterogeneities as amplifiers of crystallization [14]. 60 61 A standard method to detect and characterize crystalline material is X-ray power diffraction (XRPD) but 62 like with other conventional methods such as differential scanning calorimetry, it is challenging to detect 63 small amounts of crystalline material. Such small amounts can be already present following manufacture 64 of an amorphous product or they are generated as initial instability but either ways, such crystallites can 65 negatively affect kinetic stability of ASDs [12,13]. 66 67 Interesting for detection of crystalline material are imaging techniques that can be based on different phys- 68 ical principles. Since amorphous formulations are typically multi-component systems, chemical image has 69 the advantage that an active pharmaceutical ingredient (API) is differentiated from excipients. Any 70 4 chemical imaging involves a sophisticated analytical technique for acquisition of images and spectra that 71 contain the chemical information [15], which typically enables spatial distribution of one or all formulation 72 components [16]. Images can be acquired at the surface and in the bulk by electron microscopy, such as, 73 transmission electron microscopy (TEM) (Yamada et al., 2017) and scanning electron microscopy (SEM) 74 with energy dispersive X-ray spectroscopy (EDS) [17,18]. Important are also vibrational spectroscopic 75 techniques with appropriate optics, such as Raman [15,19,20], near infrared (NIR), or terahertz spectros- 76 copy [14]. 77 78 It is critical for any imaging technique how the large amounts of data are evaluated. Suitable algorithms 79 such as modern chemometric methods can be applied to extract useful information from otherwise just large 80 and incomprehensible data sets [15]. An algorithmic topic in its own right is how clusters are analyzed in 81 images since a naked eye is not capable of detecting any subtle changes of an imaged microstructure. Prom- 82 ising for any such cluster analysis is fractal geometry. This approach was pioneered by Mandelbrot and 83 does not use classical geometry to describe physical objects [21]. Fractal geometry has been applied in 84 pharmaceutics, for example, to describe a solvent-mediated formation of a drug hydrate [22]. However, a 85 single fractal dimension is often not sufficient to adequately describe a complex heterogeneous system. A 86 more generalized mathematical concept is given by multifractal analysis, which decomposes the self-simi- 87 lar measures into intertwined fractal sets that describe the variations from the average in heterogeneous 88 systems [23]. 89 90 Recently, the multifractal formalism was introduced in solid dispersion technology to describe the spatial 91 distribution of an inorganic carrier [24]. In a subsequent work, the distribution of different drugs in ASDs 92 was revealed to have a multifractal character [25]. The mathematical formalism was found to model ade- 93 quately the heterogeneous nature of drug clusters in ASD and a next step would be to study changes over 94 time. The hypothesis of the present work is that multifractal analysis based on chemical imaging can proof 95 utility in analyzing early stages of physical instability. Thus, felodipine (FEL) was used as model drug in 96 solid dispersions with the polymer hydroxypropyl methylcellulose acetate succinate (HPMCAS). The 97 5 different grades LF and HF were used with 14-18% and 4-8% of succinoyl substitutions, respectively [26]. 98 ASDs of FEL/HPMCAS were analyzed topographically by SEM-EDS as a chemical imaging technique, 99 which provided the basis for multifractal analysis. To have a comparison with other known sensitive meth- 100 ods, samples were also studied by laser scanning confocal microscopy (LSM) and atomic force microscopy 101 (AFM). 102 103 2. Materials and methods 104 2.1. Materials 105 Felodipine (FEL) was purchased from Kemprotec Ltd. (Smailthorn, Cumbria, UK) and the different grades 106 of HPMCAS (Shin-Etsu AQOAT®, Type LF and HF) were a generous gift from Shin-Etsu Chemical Co. 107 Ltd. (Tokyo, Japan). Dichloromethane and methanol (HPLC grade) were procured from Sigma-Aldrich (St. 108 Louis, Missouri, USA). The API chemical structure as well as monomer units of the polymer are given in 109 Fig. 1. Particularly highlighted are chloride atoms regarding their selective detection by energy dispersive 110 X-ray spectroscopy. 111 112 2.2. Methods 113 2.2.1. Preparation of physical mixtures and solid dispersions by rotary evaporation 114 Initial pretests of varying drug loads suggested that 50% FEL in polymer was rather challenging for amor- 115 phous stability, which therefore provided a suitable reference concentration in the ASDs of the main study. 116 Binary mixtures of 50% w/w FEL polymer were then prepared by dissolving drug and polymer in a solvent 117 mixture of 50: 50 (v/v) dichloromethane: methanol. All solvent mixtures were visually inspected to confirm 118 that the drug and the polymer were completely dissolved and the systems formed uniform one-phase solu- 119 tions. The solvent was then removed at 50˚C under reduced pressure using a rotavapor RE120 (Bünchi 120 Labortechnik AG, Flawil, Switzerland) with a vacuum controller CVC2 (Vacuubrand GMBH + CO, 121 Wertheim, Germany). The obtained solid mass was stored at room temperature overnight to remove any 122 6 residual solvents. The ASDs were freeze/milled with SPEX SamplePrep model 6770 (Metuchen, New jer- 123 sey, USA). For a comparison with amorphous formulations, physical mixtures (PM) were prepared by mix- 124 ing the powders of FEL and polymer for 5 minutes with a spatula. 125 126 Samples of the solid dispersions were compressed to tablets for subsequent surface and image analysis. 127 Thus, powders (100 mg) were manually fed into a hydraulic XP1 press (Korsch AG, Berlin, Germany) and 128 manually compacted. The compacts were flat-faced and round with a diameter of 7 mm. 129 130 2.2.2. X-ray powder diffraction 131 X-ray powder diffractogram (XRPD) were determined for PM as well as for ASD obtained from rota- 132 evaporation. A D2 Phaser diffractometer (Bruker AXS GmbH, Karlsruhe, Germany) was employed that 133 was equipped with a Co-2K KFL diffraction tube configured and a 1D- Lynxeye detector and with a Fe 134 filter. The applied voltage and current were 30kV and 10mA, respectively. Diffraction patterns were ob- 135 tained using a step width of 0.02˚ with a detector resolution in 2Ɵ between 6-40˚ and a scan speed of 2s/step 136 at room temperature. 137 138 2.2.3. Stability studies 139 Compacts were stored in hermetically closed glass vials at a temperature of 40˚C using a climate chamber 140 (Binder GmnH, Tuttlingen, Germany). The storage temperature was selected as a realistic but accelerating 141 condition to detect kinetic changes over time (e.g. phase separation and/or drug crystallization). Samples 142 were taken at specific time intervals and analyzed physically. 143 144 2.2.4. Scanning electron microscopy and energy-dispersive X-ray spectroscopy 145 7 The samples were placed on double-side adhesive carbon tabs and the surface of the compacts was coated 146 with gold under argon vacuum with a Sputter Coater SC7620 (Quorum Technologies Ltd., Kent, UK). 147 These surfaces were then studied by means of scanning electron microscopy (SEM) TM3030 PLUS (Hita- 148 chi High-Technologies Corporation, Tokyo, Japan) and micrographs were collected in a mix mode. The 149 microscope was equipped with an energy-dispersive X-ray spectroscopy (EDS) device for elemental map- 150 ping. The analysis was based on a Quantax 70 system software (Bruker Nano GmbH, Berlin, Germany) 151 consisting of an X Flash Min SVE signal processing unit, a scan generator and Megalink interface together 152 with an X Flash silicon drift detector 410/30 H (Bruker Nano GmbH, Berlin, Germany). Samples were 153 scanned during 300 seconds at a voltage of 15 kV to map the drug distribution of chloride (Cl) atoms for 154 comparatively higher X-ray scattering intensity. This procedure was performed on all compacts at random 155 location in spot sizes of 113 × 85 µm2. 156 157 2.2.5. 3D-Laser scanning confocal microscopy 158 Laser scanning micrographs of the sample surfaces were collected by means of a 3D laser scanning confocal 159 microscopy (LSM) VK-X200 (Keyence, Osaka, Japan) using a violet laser (408nm) and a 150x objective 160 lens (Nikon Plan CF Apo, 150x/0.95, WD 0.2mm) . The surface is scanned at high speed in X, Y and Z, 161 allowing image capturing and height measurements with high lateral resolution. Reflected white light and 162 laser light emitted from the focal point are reflected back through the objective lens. The intensity of the 163 laser light that passes through a pinhole is determined by a very sensitive 16-bit photomultiplier. Since the 164 pinhole blocks most of the returning light (except the light from the focal point), confocal LSM delivers 165 much sharper images than conventional microscopy techniques. In addition, a true color image from the 166 integrated second light source is overlaid. 167 168 2.2.6. Atomic force microscopy 169 Atomic force microscopy (AFM) images of compacts were acquired at ambient conditions in dynamic AC 170 mode using a NanoWizard 4 AFM instrument (JPK Instruments AG, Berlin, Germany). Height and phase 171 8 images were collected simultaneously using a silicon PPP-NCHR cantilever (Nanosensors AG, Neuchâtel, 172 Switzerland) with a resonance frequency of approximately 320 kHz and 42 N m-1 spring constant. 173 174 2.2.7. Image processing 175 The Cl distribution images were exported to Image J software (National Institutes of Health, Bethesda, 176 Maryland, USA) and converted to maximum intensity projection. In a second step, the gray scale images 177 in the bmp format (with 1023 x 766 pixels and a resolution 72 dpi) were resized to the png format with 512 178 x 512 pixels while keeping the same resolution. The projected images were unsharpened with a radius of 179 12 pixels and they were binarized using MATLAB software package, version R2017b (The MathWorks, 180 Inc., Natick, USA). In a binary image, a signal pixel is defined as a digital element whose intensity is 1 181 after thresholding the images of 5 and this conversion to binary format (with a resolution of 96 dpi) was 182 conducted with 10 images per each sample analysis. 183 184 2.2.8. Multifractal and statistical data analysis 185 While classical fractals are mathematical objects with a single fractal dimension, multifractal formalism 186 decomposes self-similar measures into intertwined fractal sets (Gould et al., 2011; Li et al., 2012). A brief 187 review of basic multifractal theory is given in the Appendix. 188 189 Box-counting method algorithm was used to cover a 2-D image for determination of fractal dimensions. 190 Using the binary images, boxes (grids of 512 pixel sizes) were counted using at least one pixel of the 191 observed object. A multifractal spectrum was then determined with a customized MATLAB program as 192 proposed and described previously [27]. 193 194 15 smoother (Figure 6c/d). From phase imaging, it is suggested that initially (Fig. 7a/b) the drug and the pol- 349 ymer were remaining both in the amorphous state and showed more or less homogenous contrast in the 350 phase signal. The appearance of a brighter domain in the AFM phase image (Fig. 7c) give evidence of the 351 existence of crystalline domains among partially amorphous and highly dissipating polymer regions, indi- 352 cating a heterogeneous surface due to the phase separation. These findings seems contradictory at first sight, 353 but under assumption that amorphous regions of the drug are re-crystallizing with accelerating temperature, 354 the polymer needs to reorganize as well and starts to flatten out. This is in agreement with the relaxation 355 phenomenon and mobility in glass state [14]. 356 357 The orthogonal techniques LSM and AFM would be in line with the assumption that crystalline-amorphous 358 phase separation may have occurred [32]. Based on this mechanism, a larger amount of drug can be uni- 359 formly de-mix and segregate in a short amount of time, while nucleation and growth act only locally [39]. 360 The de-mixing was likely to accelerate re-crystallization of drug. 361 362 In summary, the finding of the orthogonal methods of SEM-EDS, LSM and AFM suggest that even the 363 freshly prepared solid dispersions had drug-rich and polymer-rich clusters and this heterogeneity was re- 364 flected in the binary images obtained from SEM-EDS. The subsequent dynamics of de-mixing and re- 365 crystallization was captured as complex changes in cluster dynamics of the binary images leading to meas- 366 urable changes in multifractal dimension, LSM and AFM images, whereas in the classical XRPD analysis 367 no changes were observed throughout the same time period. 368 369 5. Conclusions 370 The present work addressed the need for novel tools in early identification of physical instability of solid 371 dispersions. Multifractal analysis was introduced successfully to early stages of stability testing using amor- 372 phous solid dispersions. Changes in the fractal dimensions were noted early in stability testing, when no 373 16 changes were appreciated based on XRPD analysis. The orthogonal techniques of SEM-EDS, LSM and 374 AFM that are known to be sensitive for microstructural change, suggested that the initial solid dispersions 375 already displayed heterogeneity in terms of drug-rich and polymer-rich domains and de-mixing of the com- 376 ponents was likely to precede nucleation of crystalline material. 377 378 A decrease of the fractal dimensions D0, D1 and D2 was statistically significant after four weeks of stability 379 testing, while a possible effect of the HPMCAS grade was not revealed. Although the use of multifractals 380 was successful for early instability detection, care is needed to expect the same cluster dynamics in other 381 solid dispersions too. We discussed that different mechanisms of microstructural change can affect clusters 382 and therefore fractal dimensions. However, a clear strength of the presented data evaluation is that this 383 cluster analysis can be based even on more than one physical method of imaging. It could be, for example, 384 also used for imaging based on Raman or near infrared spectroscopy. Moreover, it would be interesting to 385 study different types of solid dispersions. The present work holds much promise but further research is 386 needed to better assess the capability of multifractals to act as early warning tool for physical changes in 387 metastable drug formulations. 388 389 Acknowledgements 390 391 Financial support to RAV from Brazilian Federal Agency for Support and Evaluation of Graduate Educa- 392 tion within the Ministry of Education of Brazil (CAPES) grant number 99999.001486 is greatly acknowl- 393 edged. We also thank Spanish Ministry of Economy and Competitiveness and FEDER Funds (European 394 Union) for the support of the project MAT2016-77345-C3-3-P 395 396 Appendix: Multifractal theory 397 17 398 The fractal dimension is measured by overlaying the binary image with grid of boxes and counting the 399 number of boxes, N (  ), this is expressed as [23,27] 400 𝑁(𝜀)~𝜀−𝐷0 (1) 401 where D0 is the fractal dimension, calculated from the following equation: 402 𝐷0= 𝜀→0 𝑙𝑖𝑚 log𝑁(𝜀) log1 𝜀 (2) 403 D0 is derived by counting the number of boxes with various sizes to cover the image and then estimating 404 the linear region in the log-log plot. However, complex structures may not entirely be described by single 405 fractal dimension, but by multifractal analysis, which considers the amount of mass inside each box, in this 406 way characterize these complex structures. The probability Pi of finding the object pixel in the ith box is 407 determined by 408 𝑃𝑖(𝜀)~𝜀𝛼𝑖 (3) 409 where αi is the singularity strength which corresponds to the density in the ith box. 410 The probability distribution for multifractal measurements is 411 ∑[𝑃𝑖(𝜀)𝑞] 𝑖 ~𝜀𝜏(𝑞) (4) 412 Where q is the exponent expressing the fractal properties in different scales of the object. 𝜏(q) can be defined 413 as: 414 𝜏(𝑞)= lim 𝑟→0[ln (∑𝑃𝑖𝑖 (𝜀)𝑞)]ln(1/𝜀) ⁄ (5) 415 The full plot of Dq versus q is representative of the strength of the multifractality of finite measure, and the 416 generalized dimension Dq which is related with q can be expressed as 417 𝐷𝑞= 𝜏(𝑞) (𝑞−1) ⁄ (6) 418 Also, the relationship between parameters of f(α) versus α are used to calculate the multifractal spectra: 419 𝑁(𝛼)~ 𝜀−𝑓(𝛼) (7) 420 18 where the number of boxes N(α) for each probability Pi (Ɛ) has singularity strengths between α and α + dα 421 is found to scale. f (α) against α, in general way it gives the “fractal dimension” f (α) of sets where the 422 measure scales locally with the same exponent α. The multifractal spectrum gives one dimension for each 423 set where the data scales similarly. The variable f (α(q)) gives the local fractal dimension at resolution q. 424 f(α) has the same information of generalized information Dq and can be defined as [23,27,28]: 425 𝑓(𝛼(𝑞))=𝑞𝛼(𝑞)− 𝜏(𝑞) (8) 426 where α(q) can be defined as: 427 𝛼(𝑞)=𝑑𝜏(𝑞) 𝑑𝑞 ⁄ (9) 428 In case of monofractal, D0= D1= D2, whereas different values D0 ≥ D1≥ D2 indicate a multifractal system 429 [28]. 430 431 References 432 [1] A. Newman, G. Knipp, G. Zografi, Assessing the Performance of Amorphous Solid Dispersions, 433 IJOURNAL Pharm. Sci. 101 (2012) 1355–1377. 434 [2] Y. Huang, W.-G. Dai, Fundamental aspects of solid dispersion technology for poorly soluble 435 drugs, Acta Pharm. Sin. B. 4 (2014) 18–25. 436 [3] S. Baghel, H. Cathcart, N.J. O’Reilly, Polymeric Amorphous Solid Dispersions: A Review of 437 Amorphization, Crystallization, Stabilization, Solid-State Characterization, and Aqueous 438 Solubilization of Biopharmaceutical Classification System Class II Drugs, J. Pharm. Sci. 105 439 (2016) 2527–2544. 440 [4] N. Wyttenbach, M. Kuentz, Glass-forming ability of compounds in marketed amorphous drug 441 products, Eur. J. Pharm. Biopharm. 112 (2017) 204–208. 442 [5] J.B. Nanubolu, J.C. Burley, Investigating the recrystallisation behaviour of amorphous 443 paracetamol by variable temperature Raman studies and surface Raman mapping ., Mol. Pharm. 9 444 (2012) 1544–1558. 445 19 [6] P. Mistry, K.K. Amponsah-Efah, R. Suryanarayanan, Rapid Assessment of the Physical Stability 446 of Amorphous Solid Dispersions, Cryst. Growth Des. 17 (2017) 2478–2485. 447 [7] M. Milne, W. Liebenberg, M. Aucamp, The Stabilization of Amorphous Zopiclone in an 448 Amorphous Solid Dispersion, AAPS PharmSciTech. 16 (2015) 1190–1202. 449 [8] Z. Yang, K. Nollenberger, J. Albers, J. Moffat, D. Craig, S. Qi, The effect of processing on the 450 surface physical stability of amorphous solid dispersions, Eur. J. Pharm. Biopharm. 88 (2014) 451 897–908. 452 [9] E.O. Kissi, H. Grohganz, K. Löbmann, M.T. Ruggiero, J.A. Zeitler, T. Rades, Glass-Transition 453 Temperature of the β-Relaxation as the Major Predictive Parameter for Recrystallization of Neat 454 Amorphous Drugs, J. Phys. Chem. B. (2018) acs.jpcb.7b10105. 455 [10] J. Sibik, J.A. Zeitler, Direct measurement of molecular mobility and crystallisation of amorphous 456 pharmaceuticals using terahertz spectroscopy, Adv. Drug Deliv. Rev. 100 (2016) 147–157. 457 [11] G. Zografi, A. Newman, Interrelationships Between Structure and the Properties of Amorphous 458 Solids of Pharmaceutical Interest, J. Pharm. Sci. 106 (2017) 5–27. 459 [12] C. Correa-Soto, N.S. Trasi, P.D. Schmitt, Y. Su, Z. Liu, E. Miller, N. Variankaval, P.J. Marsac, 460 G.J. Simpson, L.S. Taylor, Second harmonic generation microscopy as a tool for the early 461 detection of crystallization in spray dried dispersions, J. Pharm. Biomed. Anal. 146 (2017) 86–95. 462 [13] N.K. Thakral, S. Mohapatra, G.A. Stephenson, R. Suryanarayanan, Compression-Induced 463 Crystallization of Amorphous Indomethacin in Tablets: Characterization of Spatial Heterogeneity 464 by Two- Dimensional X ‑ ray Di ff ractometry, Mol. Pharm. 12 (2015) 253–263. 465 [14] M. Descamps, E. Dudognon, Crystallization from the amorphous state: Nucleation-growth 466 decoupling, polymorphism interplay, and the role of interfaces, J. Pharm. Sci. 103 (2014) 2615– 467 2628. 468 [15] N.L. Calvo, R.M. Maggio, T.S. Kaufman, Characterization of pharmaceutically relevant materials 469 at the solid state employing chemometrics methods, J. Pharm. Biomed. Anal. 147 (2018) 538–564. 470 20 [16] B. Vajna, H. Pataki, Z. Nagy, I. Farkas, G. Marosi, Characterization of melt extruded and 471 conventional Isoptin formulations using Raman chemical imaging and chemometrics, Int. J. 472 Pharm. 419 (2011) 107–113. 473 [17] S. Rades, V.-D. Hodoroaba, T. Salge, T. Wirth, M.P. Lobera, R.H. Labrador, K. Natte, T. Behnke, 474 T. Gross, W.E.S. Unger, High-resolution imaging with SEM/T-SEM, EDX and SAM as a 475 combined methodical approach for morphological and elemental analyses of single engineered 476 nanoparticles, RSC Adv. 4 (2014) 49577–49587. 477 [18] N. Scoutaris, K. Vithani, I. Slipper, B. Chowdhry, D. Douroumis, SEM/EDX and confocal Raman 478 microscopy as complementary tools for the characterization of pharmaceutical tablets, Int. J. 479 Pharm. 470 (2014) 88–98. 480 [19] S. Hellstén, H. Qu, T. Heikkilä, J. Kohonen, S.-P. Reinikainen, M. Louhi-Kultanen, Raman 481 spectroscopic imaging of indomethacin loaded in porous silica, CrystEngComm. 14 (2012) 1582– 482 1587. 483 [20] H. Ueda, Y. Ida, K. Kadota, Y. Tozuka, Raman mapping for kinetic analysis of crystallization of 484 amorphous drug based on distributional images, Int. J. Pharm. 462 (2014) 115–122. 485 [21] C. Demetzos, N. Pippa, Fractal analysis as a complementary approach to predict the stability of 486 drug delivery nano systems in aqueous and biological media: A regulatory proposal or a dream?, 487 Int. J. Pharm. 473 (2014) 213–218. 488 [22] W. Kirchmeyer, N. Wyttenbach, J. Alsenz, M. Kuentz, Influence of excipients on solvent- 489 mediated hydrate formation of piroxicam studied by dynamic imaging and fractal analysis, Cryst. 490 Growth Des. 15 (2015) 5002–5010. 491 [23] K. Liu, M. Ostadhassan, Quantification of the microstructures of Bakken shale reservoirs using 492 multi-fractal and lacunarity analysis, J. Nat. Gas Sci. Eng. 39 (2017) 62–71. 493 [24] C. Adler, A. Teleki, M. Kuentz, Multifractal Characterization of Pharmaceutical Hot-Melt 494 Extrudates, Pharm. Res. 34 (2017) 321–332. 495 21 [25] R. Abreu-Villela, C. Adler, I. Caraballo, M. Kuentz, Electron microscopy/energy dispersive X-ray 496 spectroscopy of drug distribution in solid dispersions and interpretation by multifractal geometry, 497 J. Pharm. Biomed. Anal. 150 (2018) 241–247. 498 [26] Shin Etsu, Hypromellose Acetate Succinate NF Technical Information, Usp Nf,. (2005) 1–20. 499 [27] D.J. Gould, T.J. Vadakkan, R.A. Poché, Dj.M.E. Dickinson, Multifractal and Lacunarity Analysis 500 of Microvascular Morphology and Remodeling, (2011) 136–151. 501 [28] H. Salat, R. Murcio, E. Arcaute, Multifractal methodology, Physica A. 473 (2017) 467–487. 502 [29] A. Giri, S. Tarafdar, P. Gouze, T. Dutta, Multifractal analysis of the pore space of real and 503 simulated sedimentary rocks, Geophys. J. Int. 200 (2015) 1106–1115. 504 [30] T. Kano, Y. Yoshihashi, E. Yonemochi, K. Terada, Clarifying the mechanism of aggregation of 505 particles in high-shear granulation based on their surface properties by using micro-spectroscopy, 506 Int. J. Pharm. 461 (2014) 495–504. 507 [31] Ž. Mitić, A. Stolić, S. Stojanović, S. Najman, N. Ignjatović, G. Nikolić, M. Trajanović, 508 Instrumental methods and techniques for structural and physicochemical characterization of 509 biomaterials and bone tissue: A review, Mater. Sci. Eng. C. 79 (2017) 930–949. 510 [32] D.J. Van Drooge, K. Braeckmans, W.L.J. Hinrichs, K. Remaut, S.C. De Smedt, H.W. Frijlink, 511 Characterization of the mode of incorporation of lipophilic compounds in solid dispersions at the 512 nanoscale using Fluorescence Resonance Energy Transfer (FRET), Macromol. Rapid Commun. 513 27 (2006) 1149–1155. 514 [33] N. Pippa, S. Pispas, C. Demetzos, The fractal hologram and elucidation of the structure of 515 liposomal carriers in aqueous and biological media, Int. J. Pharm. 430 (2012) 65–73. 516 [34] Y. Chen, J. Wang, J. Feng, Understanding the fractal dimensions of urban forms through spatial 517 entropy, Entropy. 19 (2017) 1–18. doi:10.3390/e19110600. 518 [35] K. Izawa, N. Hattori, H. Okabayashi, Growth process for fractal polymer aggregates formed by 519 22 perfluorooctyltrimethoxysilane . Time-resolved small-angle X-ray scattering study, Colloid 520 Polym. Sci. 278 (2000) 293–300. 521 [36] J. Ofner, F. Brenner, K. Wieland, E. Eitenberger, J. Kirschner, C. Eisenmenger-Sittner, S. Török, 522 B. Döme, T. Konegger, A. Kasper-Giebl, H. Hutter, G. Friedbacher, B. Lendl, H. Lohninger, 523 Image-Based Chemical Structure Determination, Sci. Rep. 7 (2017) 1–11. 524 [37] Y. Chen, Data fusion for accurate microscopic rough surface metrology, Ultramicroscopy. 165 525 (2016) 15–25. 526 [38] H.X. Guo, J. Heinämäki, J. Yliruusi, Characterization of particle deformation during compression 527 measured by confocal laser scanning microscopy, Int. J. Pharm. 186 (1999) 99–108. 528 [39] M.E. Lauer, M. Siam, J. Tardio, S. Page, J.H. Kindt, O. Grassmann, Rapid Assessment of 529 Homogeneity and Stability of Amorphous Solid Dispersions by Atomic Force Microscopy—From 530 Bench to Batch, Pharm. Res. 30 (2013) 2010–2022. 531 532 533 23 Figure captions 534 535 Fig. 1. Chemical structures of felodipine (FEL) and HPMCAS. 536 537 Fig. 2. Powder X-ray diffraction of FEL in physical mixture (PM) with HPMCAS of different polymer 538 grades and formulated as amorphous solid dispersion (ASD). From bottom to top: ASDs: FEL/HPMCAS- 539 LF, FEL/HPMCAS-HF after 4 weeks stored at 40˚C; PM: FEL/HPMCAS-LF, FEL/HPMCAS-HF (all sam- 540 ples at a drug load of 50 wt% of FEL). 541 542 Fig. 3. Results of energy dispersive X-ray spectroscopy (EDS) to obtain two-dimensional binary images of 543 ASDs FEL/HPMCAS-LF, FEL/HPMCAS-HF (50 wt% of FEL), following storage at 40˚C. Drug-rich 544 phase is shown as white domains. 545 546 Fig. 4. Statistical means plot of FEL/HPMCAS ASD formulations (50 wt% of FEL) based on a two-factor 547 ANOVA of how D0 (a) and D2 (b) are affected by storage time (at 40˚C), and intervals of Fisher's Least 548 Significant Difference (LSD, 95%) are shown. 549 550 Fig. 5. Confocal laser microscopy of ASDs before storage (a, b) and after storage (c, d) for 4 weeks at 40˚C. 551 FEL/HPMCAS-LF (a, c), FEL/HPMCAS-HF (b, d), at 50 wt% of FEL. The scale bar is 10 µm. 552 553 Fig. 6. AFM topographical images of FEL/HPMCAS ASDs before storage (a, b) and after storage (c, d) 554 for 4 weeks at 40˚C. FEL/HPMCAS-LF (a, c) and FEL/HPMCAS-HF (b, d). 555 556 24 Fig. 7. AFM phase images of FEL/HPMCAS ASDs; before (a, b) and after storage (c, d) for 4 weeks at 557 40˚C. FEL/HPMCAS-LF (a, c) and FEL/HPMCAS-HF (b, d). 558 559