Characterization of the Particle Size and Polydispersity of Dicumarol

Subscriber access provided by University of Newcastle, Australia

Article

Characterization of the particle size and polydispersity of dicumarol using solid-state NMR spectroscopy Kassibla Elodie Dempah, Joseph W. Lubach, and Eric J. Munson Mol. Pharmaceutics, Just Accepted Manuscript • DOI: 10.1021/acs.molpharmaceut.6b01073 • Publication Date (Web): 06 Feb 2017 Downloaded from http://pubs.acs.org on February 7, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Molecular Pharmaceutics is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 27

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Molecular Pharmaceutics

Characterization of the particle size and polydispersity of dicumarol using solid-state NMR spectroscopy Kassibla Elodie Dempah, Joseph W. Lubach, Eric J. Munson

ACS Paragon Plus Environment

1


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 27

Abstract A variety of particles sizes of a model compound, dicumarol, were prepared and characterized in order to investigate the correlation between particle size and solid-state NMR (SSNMR) proton spin-lattice relaxation (1H T1) times. Conventional laser diffraction and scanning electron microscopy were used as particle size measurement techniques, and showed crystalline dicumarol samples with sizes ranging from tens of micrometers to a few micrometers. Dicumarol samples were prepared using both bottom-up and top-down particle size control approaches, via antisolvent microprecipitation and cryogrinding. It was observed that smaller particles of dicumarol generally had shorter 1H T1 times than larger ones. Additionally, cryomilled particles had the shortest 1H T1 times encountered (8s). SSNMR 1H T1 times of all the samples were measured and showed as-received dicumarol to have a T1 of 1500 s, whereas the 1H T1 times of the precipitated samples ranged from 20 to 80 s, with no apparent change in the physical form of dicumarol. Physical mixtures of different sized particles were also analyzed to determine the effect of sample inhomogeneity on 1H T1 values. Mixtures of cryoground and as-received dicumarol were clearly inhomogeneous as they did not fit well to a one-component relaxation model, but could be fit much better to a two-component model with both fast-and slow-relaxing regimes. Results indicate that samples of crystalline dicumarol containing two significantly different particle size populations could be deconvoluted solely based on their differences in 1H T1 times. Relative populations of each particle size regime could also be approximated using two-component fitting models. Using NMR theory on spin diffusion as a reference, and taking into account the presence of crystal defects, a model for the correlation between the particle size of dicumarol and its 1H T1 time was proposed.


2

Page 3 of 27

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Keywords: solid-state NMR, relaxation, 1H T1 time, particle size, milling, cryogrinding, stability, spin diffusion

Introduction The size of the particles in a solid powder is a critical property to monitor and control in many industries, especially in the pharmaceutical industry. Particle size directly impacts properties such as dissolution rate, mixing uniformity, and compaction behavior

(1-4)

. Methods

commonly used to measure particle size include laser diffraction techniques and microscopy (5). Laser diffraction is typically used to measure particles nominally ranging from 0.2 to 2000 μm in size. The measurements typically require that the particles be suspended in a dispersing liquid non-solvent, often with a surfactant added. Sonication is often used during the measurement to break down possible agglomerates to measure primary particle size. Dry dispersion methods are also being increasingly used in the pharmaceutical industry. Laser diffraction measurements are volume-based and the results are typically presented in terms of percentile values, where the d10, d50, and d90 values are the numbers below which 10, 50, 90% of the volume of the sample lies (5)

. These values may or may not be representative of the actual dimensions of the particles,

depending on the particle morphology and aspect ratio. Contrary to laser diffraction, microscopy does not typically provide quantitative measurements (except with extensive image analysis), but it is a very useful particle sizing method, especially as an orthogonal technique to laser diffraction-based methods. It is the only method where the individual particles can be observed and true dimensions accurately measured, thus providing morphology information as well as information on agglomeration in addition to size. Optical microscopy can be used to image particles as small as 0.5μm. For particles in the nanometer size range, electron microscopy must be utilized. X-Ray Diffraction (XRD) has also been used, although sparingly so far, to measure the size of particles in the nanometer range

(6, 7)

. While XRD is typically used in the

pharmaceutical industry as a phase identification technique, it is commonly known that the particle size of the sample being analyzed can affect the width of the peaks in the powder X-Ray diffraction pattern. The Scherrer equation relates the width of the peak in the PXRD pattern to the size of the crystallite (6, 7). Solid-state NMR spectroscopy (SSNMR) is physical form characterization method that is also sensitive to the size of the particles in the sample being analyzed. Lubach and coworkers


3


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 27

have previously shown that a decrease in particle size of several compounds may be the cause of the decrease in their proton spin-lattice relaxation (1H T1) times, but a clear correlation between particle size and 1H T1 time was not established for crystalline compounds. In the first case reported by Lubach and coworkers, the 1H T1 time of crystalline lactose monohydrate decreased when lactose was cryoground and when it was compressed into a tablet

(8)

. While the reduction

in particle size caused by grinding was believed to be at least partially responsible for the decrease in 1H T1 time observed, lactose significantly converted to the amorphous state when milled for long periods of time, which created highly mobile domains of amorphous material, of which even small amounts could contribute to the decrease in 1H T1 time by acting as relaxation sinks for 1H magnetization. In another study, the 1H T1 time of gabapentin, a crystalline active pharmaceutical ingredient (API), also decreased when it was ground

(9)

. However, polymorphic

conversion occurred as a result of grinding, and crystal defects were also likely generated during grinding, as evidenced by the decrease in chemical stability of gabapentin. For that reason, a clear correlation between the particle size of gabapentin and its 1H T1 time could not be established. To the best of our knowledge no systematic study has been done to correlate the particle size of a crystalline compound to its 1H T1 time. However, NMR theory correlates domain sizes in amorphous polymers to the time it takes the magnetization to diffuse through those domains via the homonuclear 1H dipolar coupling network(10, 11). This diffusion of magnetization occurs through a process known as spin diffusion, which is the transfer of magnetization among protons from the source of magnetization to a relaxation sink. This magnetization transfer can occur through relatively long distances because protons are fully abundant and strongly coupled in the solid state. Therefore, domain sizes in copolymers ranging from 0.5 to 50 nm have been estimated through SSNMR 1H spin diffusion experiments (11, 12). In the study presented here, we investigate the correlation between the particle size of a model crystalline compound dicumarol and its SSNMR 1H T1 times in various preparations. Dicumarol is a pharmaceutical solid with only one known crystalline form. It was chosen as a model compound because it has no internal relaxation sinks in its structure, such as methyl groups. It was therefore expected to have a 1H T1 time on the order of hundreds to thousands of seconds. A long 1H T1 time in a starting material facilitates monitoring changes in its value as a function of different processing methods. In addition, dicumarol does not easily convert to the


4

Page 5 of 27

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


amorphous state, even upon high energy grinding, and amorphous dicumarol has never been reported. Particles of dicumarol ranging from a few hundred nanometers to tens of micrometers in size were prepared by precipitation and cryogrinding in bottom-up and top-down approaches to particle size control. The samples were then characterized by laser diffraction, SEM and SSNMR. The impact of the particle size inhomogeneity of a sample on its 1H T1 time was also evaluated by preparing mixtures of as-received and cryoground crystalline dicumarol and characterizing them by SSNMR. Finally, and most importantly, a clear correlation between the 1

H T1 time of dicumarol and its particle size was proposed.

Materials and Methods Sample preparation Cryogrinding About 1.5 g of dicumarol was placed in a vessel and ground while immersed in liquid nitrogen after a 15 min precooling period with alternating cycles of 2 min of cooling and 2 min of grinding (SPEX SamplePrep 6770 Freezer/Mill, SamplePrep, Inc., Metuchen, NJ). The rate of grinding was 10 counts per second. The total grinding time was 4 or 10 min.

Precipitation A suspension of dicumarol in dimethyl sulfoxide (DMSO) was stirred and heated to at least 100 ºC overnight to obtain a 1mg/mL solution. The solution was then injected into water under sonication (Fisher brand sonic dismembrator, Fisher Scientific, Pittsburgh, PA). The sonication amplitude was 20% for all the samples. A Fisher brand injector and a 5mL syringe with a 20gauge needle (Popper deflected non coring septum 20 x 8”, Popper & Sons, Hyde Park, NY) were used for the injections Table 1 summarizes the ratios of DMSO to water, as well as the injection speed employed for each lot prepared. 10-15 vials were prepared using each set of parameters. After injection, the suspensions were freeze-dried Virtis freeze- dryer (Advantage 2.0 Benchtop Freeze Dryer, SP Scientific, Gardiner, NY) using the following cycle: hold at -70 ºC for 120 min, extra freezing at -50 ºC for 60 min; hold at -5 ºC and 100 mTorr for 1320 min, ramp to 15 ºC in 300 min at 100 mTorr, hold at 15 ºC and 100 mTorr for 600 min, ramp to 30 ºC at 50 mTorr for 300 min, hold at 60 ºC and 10 mTorr for 60 min. The vials were held at 25 ºC and 200 mTorr until they were removed from the


5


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 27

lyophilizer.

Sample characterization Differential scanning calorimetry (DSC) 1-4 mg of sample was weighed in a standard, hermetic or T-Zero normal TA aluminum pan. Samples were heated from -10ºC to 250 ºC or 300ºC using a 10ºC/min ramp (DSC Q200 or DSC Q2000, TA Instruments, New Castle, DE).

Laser diffraction A Malvern 2000 Mastersizer was used for particle size distribution measurements. The powders were suspended in hexane and sonicated for 2 min in the measuring chamber before each measurement. The refractive index of dicumarol was calculated using the ACD Labs software (Toronto, Ontario, Canada).

Scanning electron microscopy The samples were mounted on an SEM stage line with carbon tape, and coated with approximately 30 nm of gold or 15 nm of a gold-palladium mixture. The samples were imaged on a Hitachi-S3200-N scanning electron microscope (Hitachi High Technologies America, Inc., Schaumburg, IL).

Solid-State Nuclear Magnetic Resonance (SSNMR) spectroscopy All samples were packed under ambient conditions in 7 mm zirconia rotors (Revolution NMR, Fort Collins, CO).

13

C SSNMR spectra were collected using either a Chemagnetics CMX300

(Varian, Palo Alto, CA) or a Tecmag Apollo (Tecmag, Inc., Houston, TX) spectrometer both operating at a

13

C frequency of ~75 MHz. 3-methylglutaric acid was used as an external

standard, with the methyl peak referenced at 18.84 ppm(13). All spectra were acquired using cross-polarization and magic angle spinning (CP/MAS)(14-17). The magic angle spinning frequency was 4 kHz. A SPINAL64 decoupling pulse sequence was used at a 1H decoupling field of 70-80 kHz, and the spinning sidebands were suppressed using total sideband suppression (TOSS)(18). A contact time of 1 ms was employed for CP. The number of acquisition points was


6

Page 7 of 27

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


1024. 1H T1 relaxation times were measured by saturation recovery experiments with

13

C

detection. The entire spectrum was integrated to obtain each 1H T1 curve.

Results The primary objective of this study was to prepare crystalline dicumarol samples of various particle sizes, ideally monodispersed and of similar crystal quality or defect content. Particle size measurements via laser light diffraction are often challenging to correlate back to the actual size of individual crystallites, so direct measurement by electron microscopy was chosen as the primary technique to characterize particle size. The focus here was on preparing particles of dicumarol in the micrometer and nanometer size range using precipitation and grinding techniques.

Subsequently, 1H T1 relaxation times were measured to investigate

potential relationships between dicumarol particle size and solid-state NMR relaxation properties.

Characterization of precipitated dicumarol In the first part of this study six different samples of dicumarol were prepared. The samples include as-received dicumarol and five lots of dicumarol generated by anti-solvent precipitation. These five lots were labeled lots A through E and differ by the conditions used to prepare them as described in the method section. The samples were then characterized by laser diffraction, SEM and SSNMR. Figure 1 shows the laser diffraction particle size distributions of the six samples. The broadness of the laser diffraction curves and their modality provide information on the relative heterogeneity of the samples. As-received dicumarol has a monomodal particle size distribution that spans from approximately 5 to 74 μm. The curves of the precipitated samples span a lower size range than the as-received material, from about 0.2 to 56 μ m. These distributions are also generally broader, suggesting that those samples are significantly more heterogeneous in size than as-received dicumarol. Furthermore, there is significant overlap among the size distributions. The first four columns of Table 2 summarize the laser diffraction data of as-received dicumarol and of the precipitated lots of dicumarol. The d10, d50, and d90 values of as-received dicumarol are 11, 24, and 43 μm, respectively. For the precipitated samples, the d10 values range from about 1 to 2 μm, the d50 values from about 3 μm to 5 μm, and the d90 values


7


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 27

range from approximately 13 to 20 μm. All the precipitated samples have smaller d10, d50, and d90 values than as-received dicumarol, thus indicating that smaller sizes of dicumarol were successfully prepared by anti-solvent precipitation. The same samples were then imaged by SEM in order to compare the size of the particles observed by microscopy to the values obtained by laser diffraction. Figure 2 shows the SEM micrographs of lots A through E and of as-received dicumarol. Most of the particles in the five precipitated lots are significantly agglomerated. Using the scale on the images the size of the majority of the particles was estimated to be between 0.5 and 1 μm for each precipitate sample. Larger particles estimated to range from 2-8 μm can be seen in the micrographs of lot C and lot D, indicated in Table 2. The size of the particles in lot B could not be easily estimated due to relatively poor resolution of the SEM micrograph, but it is evident that the primary particles measure less than 5 μm. All these values are significantly smaller than the ones obtained by laser diffraction. One likely reason for the discrepancy between the two techniques is that laser diffraction measured the size of the agglomerates rather than the size of the individual particles. In order to examine the correlation between 1H T1 times and particle size of dicumarol, the true single crystal domain particle size estimated by SEM will be used rather than the values obtained by laser diffraction because they represent a more accurate description of the actual size of the primary crystallites in these materials. The sizes estimated using the SEM micrographs are also summarized in Table 2. Subsequently, the solid-state 1H T1 time of each sample was measured and the values obtained are summarized in Table 2. The 1H T1 time of as-received dicumarol is 1500 s. With values ranging from 20 s to 80 s, the 1H T1 times of all the precipitated lots are significantly shorter than the 1H T1 of as-received dicumarol. The relaxation data from lots C and D did not fit well with a mono-exponential equation typically used to fit 1H T1 data and will be revisited below. Interestingly, these are also the lots that show bimodal particle size distributions. The way that the size heterogeneity of a sample might impact its 1H T1 time will be addressed at a later point in this manuscript. In this first part of the study, we have seen that the smaller particles of dicumarol generated by precipitation have significantly shorter 1H T1 times than the larger particles of as-received dicumarol.

Characterization of cryoground dicumarol


8

Page 9 of 27

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


The first set of samples studied was generated using a bottom-up particle size control method. Subsequently samples were prepared using a top-down approach, namely cryogrinding, and characterized similarly to the precipitated samples. Figure 3 shows the SEM micrographs of dicumarol cryoground for 8 min and of as-received dicumarol. While irregular in shape and size, the particles of the cryoground sample are obviously smaller than the ones in as-received dicumarol. The surface of the cryoground particles is also fairly rough and heterogeneous, unlike the as-received particles that have smooth crystal surfaces with fewer obvious defects. This surface roughness is likely indicative of a larger concentration of crystal defects in these crystals. Since grinding has been known to induce form changes, SSNMR spectra of the samples were acquired in order to confirm that they remained crystalline upon grinding. Figures 4a through 4c show the 13C SSNMR spectra of as-received dicumarol and dicumarol cryoground for 4 and 10 min. The chemical shifts in the three spectra are the same, confirming that they contain the same crystalline form of dicumarol. However, the peaks in the spectra of the cryoground samples are slightly broader than the peaks of the as-received material. This is most evident with the loss of resolution at 115 ppm in both cryoground samples relative to the as-received material. Such linewidth increase in a cryoground sample has been previously observed with ibuprofen and is likely due to the reduction in particle size and introduction of defects(13). While grinding has been shown to generate amorphous material in some compounds, no evidence of amorphous dicumarol, such as a glass transition temperature, was observed in the differential scanning calorimetry thermograms (not shown) of the cryoground samples. The 1H T1 times of the cryoground samples were then measured. Dicumarol cryoground for 4 min had a 1H T1 of 14 s and dicumarol cryoground for 10 min had a 1H T1 of 8 s. Both of these values are significantly shorter than the 1500 s 1H T1 time of the as-received material. To summarize, dicumarol remained crystalline upon cryogrinding, but the cryoground samples had significantly shorter 1H T1 values than the unprocessed material. Mixture analysis In order to determine how the particle size uniformity of the dicumarol samples might impact their 1H T1 times, mixtures of as-received and cryoground dicumarol were prepared and characterized by SSNMR. The mixtures contained from 10 to 90% by weight of as-received dicumarol, the remainder being cryoground dicumarol. A SSNMR spectrum of the 25:75


9


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 27

physical mixture is presented in Figure 4d. Apart from a loss of resolution in the peaks around 115 ppm in the spectrum of the mixture, the spectrum looks identical to the as-received dicumarol spectrum. This was expected since both the as-received and the cryoground dicumarol are crystalline and physically mixing them was not expected to alter their physical state. In order to confirm that the mixing process did not significantly alter the size of the particles of as-received and cryoground, the 10:90 and the 50:50 physical mixtures were characterized by laser diffraction. Figure 5 shows their particle size distribution overlaid with the particle size distribution of as-received dicumarol. The particle size distribution of the 50:50 physical mixture is bimodal. Since the width and the height of the two peaks in the graph are approximately equal, one can conclude that the two particle size distributions in the sample represented by the two nodes are present in approximately equal amounts. The median of the population at the largest particle size, presumably due to the as-received dicumarol, overlaps with the distribution of the neat as-received dicumarol. This indicates that the as-received particles of dicumarol were not significantly altered during blending of the physical mixture. For the 10:90 physical mixture, instead of the expected bimodal distribution a trimodal distribution emerged. There are two humps in the larger particle size range, where the particles of the asreceived dicumarol are expected to show up. This could be due to the fact that the particles in the material are not completely uniform in size, and since as-received dicumarol is only present at a 10% amount, the non-uniformity might be more obvious. Alternatively, these could be agglomerates that did not disperse well in the measurement media. Nonetheless, the two humps mostly overlap with the particle size distribution of as-received dicumarol. The 1H T1 times of all the physical mixtures were then measured. NMR spin-lattice relaxation is an exponential process that can be described using the following equation: τ

(1) Integral = m (1 − e ) where the integral is the area under resonances in the NMR spectrum, τ is a time delay interval used during the NMR relaxation measurement. Figures 6a and 6c are the mono-exponential 1H T1 fits for the 50:50 and the 90:10 physical mixtures, respectively. In both cases the mathematically determined line does not fit well to all the data points obtained during the saturation recovery 1H T1 time experiment. In both cases, some data points are clearly left out of the fit. On the other hand, when the same data is fitted with a bi-exponential fit, the agreement between data and fit is significantly better for both mixtures as shown in Figures 4b and 4d.


10

Page 11 of 27

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Similarly for all the other physical mixtures, relaxation time curves were best fitted with a biexponential equation. This implies that two spin-lattice relaxation processes with different relaxation rates were occurring in the samples. The equation used to fit a two-component biexponential spin-lattice relaxation process is the following: (2) = 1 −

()

+ " (1 −

(#)

)

where T1(1) and T1(2) are the relaxation times of the two components detected in the mixture. The pre-exponential coefficients m1 and m2 can be used to calculate the ratio of each of the components in the mixture using the following equation: (3) (% %&'& 1) =

(

( )(#

(4) (% %&'& 2) = (

(#

)(#

Table 3 summarizes the 1H T1 times of the five physical mixtures of as-received and cryoground dicumarol prepared. When the 1H T1 times of the physical mixtures were fitted to twocomponent models, the values ranged from 890-2000 s for the slow-relaxing component and from 5.2-11 s for the fast-relaxing component. The as-received dicumarol was most likely at the origin of the long 1H T1 time and the cryoground material at the origin of the short one. There is a notable difference between the 1H T1 times measured for the as-received and cryoground dicumarol when they are in a mixture compared to when they are pure components. In all physical mixtures containing 25% or more of as-received dicumarol the 1H T1 of as-received dicumarol was longer than the one of the pure material, with values ranging from 1700-2000 s; the 1H T1 time of pure as-received dicumarol was 1500 s. In the mixture containing 10% of asreceived dicumarol, the 1H T1 time of this component was 890 s. Table 3 also contains the theoretical weight-based ratios of as-received and cryoground dicumarol in the mixtures calculated using equations (3) and (4). In all the cases, the calculated amounts were within 2.5% of the theoretical values. These results demonstrate that heterogeneous mixtures of particle sizes can be distinguished based on their difference in 1H T1 times, and relatively quantitated using the m1/m2 constants from equations (3) and (4). Based on these results, lots C and D of precipitated dicumarol were again fit using a biexponential equation. The resulting fits were a better description of the data, thus indicating that these two lots contain significant particle size heterogeneity. This is confirmed by looking at the


11


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 27

SEM micrographs of those two lots (Figure 2). Both lots contain larger chunks in addition to many smaller particles. The calculated 1H T1 times are 68 s and 16 s for lot C, and 49 s and 520 s for lot D.

Discussion In the first part of the study we have seen that the 1-10 μm particles of dicumarol prepared by precipitation had shorter 1H T1 times than the larger as-received particles. The smaller particles of cryoground dicumarol also had shorter 1H T1 times than as-received dicumarol, as shown in the second part of this study. While processes such as grinding and precipitation can generate amorphous material, dicumarol was chosen as a model compound for this study because amorphous dicumarol has never been reported in the literature. In this section we will discuss the potential correlation between dicumarol particle size and respective 1H T1 relaxation times. NMR theory relates domain sizes to the time that it takes the magnetization to diffuse through that domain via spin diffusion with the following equation (5)

(11, 19)

:

+" = 6-

where L is the domain length scale, and D is the diffusion coefficient (ref). Spin diffusion experiments have been used to structurally characterize heteropolymers

(10, 20, 21)

. Spin diffusion,

enabled by the dense network of 1H dipolar coupling, describes the transfer of magnetization from rigid moieties to more mobile relaxation sinks, allowing the nuclear spins to return to equilibrium. As a result, equation (5) can be rewritten as: (6)

+" = 6- ( . / )

Using equation (6) the level of miscibility between drug and polymers for amorphous solid dispersions has been evaluated (22-24). Equation 6 can be linearized by taking the logarithm of each term, thus giving equation (7): (7) log L = log √(6D) + log √(1H T1) In order to determine how the correlation between 1H T1 and particle size of dicumarol compares to the theory, a log-log plot of the square root of 1H T1 versus particle size was constructed for the five precipitated lots of dicumarol and as-received dicumarol (Figure 7). For the two precipitated lots of dicumarol that fit best to a two-component T1 model, the large particles seen in the SEM micrographs were assigned to the longer 1H T1 time and the smaller particles to the


12

Page 13 of 27

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


shorter 1H T1 time. The resulting plot can be described with a linear trend line whose slope is 0.46 and y-intercept is 0.87. In order to visualize how the experimental relaxation data from dicumarol compare with theory, the purely theoretical relationship between 1H T1 time and domain size described in equation (7) was also plotted using a spin diffusion coefficient of 0.2 nm2/s. Spin diffusion coefficients have only been measured for some polymers and that value is on the low end of the range of values obtained

(20, 25, 26)

. Had a different value of the spin

diffusion coefficient been used to plot the theoretical relaxation relationship in Figure 7, the yintercept would be altered but the slope would remain the same. As can be clearly observed in Figure 7, the theoretical line lies above the line constructed with the dicumarol data, indicating that the 1H T1 times of the dicumarol samples are shorter than would be predicted based on the theory. The slope of the dicumarol trend line is also smaller than the slope of the theoretical line, indicating that the 1H T1 time of dicumarol increases more slowly with particle size than predicted by the model. Also plotted in Figure 7 is a data point corresponding to a sample of cryoground dicumarol, of which the average particle size was estimated to be about 1 μm. That data point does not overlap with any of the data points of precipitated dicumarol of a similar size, suggesting that particles of dicumarol prepared by grinding possess an underlying relaxation mechanism that results in shorter 1H T1 times, relative to particles of a similar size prepared by precipitation. The data point to a definite correlation between particle size and spin-lattice relaxation, but also that other factors must also impact 1H T1. We hypothesize that this difference is due to the extent of crystal defects that a particle contains. Grinding is a top-down particle size control method, which can impart a significant amount of defect sites throughout particles, while precipitation is a bottom-up particle size control technique which is not expected to generate as many crystal defects. Defects facilitate higher mobility in molecules at the defect site, as they are not tightly locked into the lattice in all three dimensions. These more mobile sites can serve as relaxation sinks, and the more relaxation sinks a sample contains, the shorter its 1H T1 time is due to greater spin diffusion efficiency. The extent of crystal defects present in the samples could also explain the discrepancy between the theory and the experimental data that is evident in Figure 7. The spin diffusion theory assumes perfect crystals and that diffusion rate is uniform throughout the particles, which is certainly not the case with the dicumarol samples. As crystals get bigger they often contain more defects due


13


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 27

to imperfections sustained during crystal growth, therefore larger particles of dicumarol, including the precipitated samples, are likely to contain more defects and thus have shorter 1H T1 times than predicted solely based on theory. These observations on the impact of crystal defects on the magnitude of 1H T1 can also be applied to the results of the milling study of gabapentin mentioned in the introduction paragraph(9). In that study it was shown that milled gabapentin samples with shorter 1H T1 times also had lower chemical stability. It was hypothesized in that study that these samples contained a greater extent of crystal defect sites than the other samples, and that the chemical degradation reaction initiated primarily at these defect sites due to greater molecular mobility there. The findings in the dicumarol study presented here support that hypothesis. The findings in the study presented here also have implications in terms of characterization of the drug substance in the drug formulation. Because of its selectivity, solidstate NMR spectroscopy (SSNMR) can be used to characterize the API in the presence of excipients, including measuring relaxation time of only the API as it exists in a solid dosage form. Therefore, based on the results presented here, the impact of blending, granulation, compression, or other manufacturing processes on the particle size of the API and the amount of crystal defects can potentially be quantitatively evaluated.

Conclusion A model based on NMR spin diffusion theory was proposed that quantitatively relates the 1

H T1 time of dicumarol to its particle size in similarly processed samples. The relative ratio of

large and small particle size components of a heterogeneous mixture of dicumarol crystallites was also evaluated. As SSNMR can be used to characterize APIs in the presence of the excipients in a formulation, such evaluation of the particle size polydispersity of the API could be conducted on formulated drug products during development. The potential implications of the findings presented here include the ability to characterize the effect of tableting on the particle size of crystalline APIs and excipients by SSNMR, something previous inaccessible using classical techniques. Depending on the magnitude of the reduction in 1H T1 times observed upon tableting, if any, conclusions could be drawn on the impact of the compaction process on the particle size of the API or excipients, and subsequently other endpoints such as drug stability. An investigation of the effect of compression force on 1H T1 times is ongoing and will be presented


14

Page 15 of 27

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


in a follow up manuscript. Additionally, other model compounds should be studied in the same manner as dicumarol in order to support the universality of the relaxation/particle size model. Finally, another possible implication of the results presented here is the ability to assess the relative amount of crystal defects present in a given crystalline material. If the decrease in 1H T1 time upon milling a crystalline solid is larger than expected for the particle size obtained, this could indicate high defect density, which in turn could lead to a reduction in chemical stability. Knowing this ahead of time and designing the drug product formulation and process to minimize defect formation could save significant time and resources throughout the development process. Nonetheless, one of the biggest challenges in establishing 1H T1 time measurements as a tool for particle size evaluation is the difficulty in attributing only one factor as being at the origin of a change in relaxation time observed. In the study presented here, we isolated to the best of our ability the effect of particle size on 1H T1 time from any other factor that could decrease the 1H T1 time. However, defect-free uniformly sized crystalline particles would need to be generated for that purpose, and efforts to achieve this are still ongoing.

-----------

Notes The results presented here are from academic work at University of Kentucky, and no data from Kansas Analytical Services are presented. The authors declare the following competing financial interest(s): EJM is a partial owner of Kansas Analytical Services, a company that provides solid-state NMR services to the pharmaceutical industry. ■ ACKNOWLEDGMENTS The authors would like to thank NSF I/UCRC Center for pharmaceutical Development (IIP-1063879 and industrial contributions) for financial support.


15


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 27

References 1. Sun J, Wang F, Sui Y, She Z, Zhai W, Wang C, et al. Effect of particle size on solubility, dissolution rate, and oral bioavailability: evaluation using coenzyme Q(1)(0) as naked nanocrystals. Int J Nanomedicine. 2012;7:5733-44. 2. Swaminathan V, Kildsig D. Polydisperse Powder Mixtures: Effect of Particle Size and Shape on Mixture Stability. Drug Development & Industrial Pharmacy. [Article]. 2002;28(1):418. 3. Yalkowsky SH, Bolton S. Particle size and content uniformity. Pharm Res. 1990;7(Copyright (C) 2012 U.S. National Library of Medicine.):962-6. 4. Zhang Y, Johnson KC. Effect of drug particle size on content uniformity of low-dose solid dosage forms. International Journal of Pharmaceutics. 1997;154(2):179-83. 5. Shekunov BY, Chattopadhyay P, Tong HHY, Chow AHL. Particle Size Analysis in Pharmaceutics: Principles, Methods and Applications. Pharm Res. [10.1007/s11095-006-91467]. 2007;24(Copyright (C) 2012 American Chemical Society (ACS). All Rights Reserved.):20327. 6. Borchert H, Shevchenko EV, Robert A, Mekis I, Kornowski A, Grübel G, et al. Determination of Nanocrystal Sizes: A Comparison of TEM, SAXS, and XRD Studies of Highly Monodisperse CoPt3 Particles. Langmuir. 2005 2005/03/01;21(5):1931-6. 7. Weibel A, Bouchet R, Boulc F, Knauth P. The Big Problem of Small Particles:A Comparison of Methods for Determination of Particle Size in Nanocrystalline Anatase Powders. Chemistry of Materials. 2005 2005/05/01;17(9):2378-85. 8. Lubach JW, Xu D, Segmuller BE, Munson EJ. Investigation of the effects of pharmaceutical processing upon solid-state NMR relaxation times and implications to solid-state formulation stability. Journal of Pharmaceutical Sciences. 2007;96(4):777-87. 9. Dempah KE, Barich DH, Kaushal AM, Zong Z, Desai SD, Suryanarayanan R, et al. Investigating Gabapentin Polymorphism Using Solid-State NMR Spectroscopy. AAPS PharmSciTech. 2012 Nov 22. 10. Clauss J, Schmidt-Rohr K, Spiess HW. Determination of domain sizes in heterogeneous polymers by solid-state NMR. Acta Polymerica. 1993;44(1):1-17. 11. Henrichs PM, Tribone J, Massa DJ, Hewitt JM. Blend miscibility of bisphenol A polycarbonate and poly(ethylene terephthalate) as studied by solid-state high-resolution carbon13 NMR spectroscopy. Macromolecules. 1988 1988/05/01;21(5):1282-91. 12. Buda A, Demco DE, Bertmer M, Blümich B, Reining B, Keul H, et al. Domain sizes in heterogeneous polymers by spin diffusion using single-quantum and double-quantum dipolar filters. Solid State Nuclear Magnetic Resonance. 2003;24(1):39-67. 13. Barich DH, Davis JM, Schieber LJ, Zell MT, Munson EJ. Investigation of solid-state NMR line widths of ibuprofen in drug formulations. Journal of Pharmaceutical Sciences. 2006;95(7):1586-94. 14. Pines A. Proton-enhanced NMR of dilute spins in solids. J Chem Phys. [10.1063/1.1680061]. 1973;59(2):569. 15. Pines A, Gibby MG, Waugh JS. Proton-enhanced nuclear induction spectroscopy 13C chemical shielding anisotropy in some organic solids. Chemical Physics Letters. 1972;15(3):3736.


16

Page 17 of 27

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


16. Andrew ER, Bradbury A, Eades RG. Removal of Dipolar Broadening of Nuclear Magnetic Resonance Spectra of Solids by Specimen Rotation. Nature. [10.1038/1831802a0]. 1959;183(4678):1802-3. 17. Lowe IJ. Free Induction Decays of Rotating Solids. Physical Review Letters. 1959;2(7):285-7. 18. Dixon WT, Schaefer J, Sefcik MD, Stejskal EO, McKay RA. Total suppression of sidebands in CPMAS C-13 NMR. Journal of Magnetic Resonance (1969). 1982;49(2):341-5. 19. Schantz S, Ljungqvist N. Structure and dynamics in polymer blends: a carbon-13 CPMAS NMR study of poly(3-octylthiophene)/poly(phenylene oxide). Macromolecules. 1993 1993/11/01;26(24):6517-24. 20. Schmidt-Rohr K, Clauss J, Blümich B, Spiess HW. Miscibility of polymer blends investigated by 1H spin diffusion and 13C NMR detection. Magnetic Resonance in Chemistry. 1990;28(13):S3-S9. 21. Schmidt-Rohr K, Spiess H.W. Multidimensional solid-state NMR and polymers: Academic Press; 1994. 22. Pham TN, Watson SA, Edwards AJ, Chavda M, Clawson JS, Strohmeier M, et al. Analysis of Amorphous Solid Dispersions Using 2D Solid-State NMR and 1H T1 Relaxation Measurements. Molecular Pharmaceutics. 2010 2010/10/04;7(5):1667-91. 23. Yuan X, Sperger D, Munson EJ. Investigating Miscibility and Molecular Mobility of Nifedipine-PVP Amorphous Solid Dispersions Using Solid-State NMR Spectroscopy. Molecular Pharmaceutics. 2014 2014/01/06;11(1):329-37. 24. Chiang P-C, Cui Y, Ran Y, Lubach J, Chou K-J, Bao L, et al. In Vitro and In Vivo Evaluation of Amorphous Solid Dispersions Generated by Different Bench-Scale Processes, Using Griseofulvin as a Model Compound. The AAPS Journal. 2013;15(2):608-17. 25. Clauss J, Schmidt-Rohr K, Spiess HW. Determination of domain sizes in heterogeneous polymers by solid-state NMR. Acta Polym. 1993;44(Copyright (C) 2012 American Chemical Society (ACS). All Rights Reserved.):1-17. 26. Spiegel S, Schmidt-Rohr K, Boeffel C, Spiess HW. 1H spin diffusion coefficients of highly mobile polymers. Polymer. 1993;34(21):4566-9.


17


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Table 1. Conditions used to generate different lots of dicumarol by anti-solvent precipitation of a solution of dicumarol in DMSO into water

Lot

Ratio DMSO:H2O

Injection speed

A

1:9

60 mL/h

B

1:9

30 mL/h

C

1:1

40 mL/h

D

1:1

60 mL/h

E

2:1

40 mL/h


Page 18 of 27

Page 19 of 27

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 1. Laser diffraction particle size distribution curves of the precipitated dicumarol lots 255x182mm (300 x 300 DPI)



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 20 of 27

Table 2. Laser diffraction particle size of as-received dicumarol and dicumarol samples prepared by precipitation and 1H T1 times of dicumarol. SEM sizes are reported with approximate sizes of representative particles.

Sample

d10 (μm)

d50 (μm)

Lot A

1.6

4.9

18.0

Lot B

0.9

4.8

19.0

Lot C

0.9

2.8

20.0

Lot D

1.2

3.8

16.5

Lot E

1.0

3.3

13.3

Asreceived

10.9

23.8

43.3

H T1 (s)

√1H T1

0.5

20 ± 2

4.5

Characterization of the Particle Size and Polydispersity of Dicumarol

Recommend Documents