Chemometric Methods for the Quantification of Crystalline Tacrolimus in Solid Dispersion by Powder X‐Ray Diffractrometry

RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology

Chemometric Methods for the Quantification of Crystalline Tacrolimus in Solid Dispersion by Powder X-Ray Diffractrometry AKHTAR SIDDIQUI, ZIYAUR RAHMAN, SRIKANT BYKADI, MANSOOR A. KHAN Division of Product Quality Research, Office of Testing and Research, OPS, CDER, United States Food and Drug Administration Received 3 December 2013; revised 5 February 2014; accepted 7 February 2014 Published online in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.23912 ABSTRACT: The objective of this study was to develop powder X-ray diffraction (XRPD) chemometric model for quantifying crystalline tacrolimus from solid dispersion (SD). Three SDs (amorphous tacrolimus component) with varying drug to excipient ratios (24.4%, 6.7%, and 4.3% drug) were prepared. Placebo SDs were mixed with crystalline tacrolimus to make their composition equivalent to three SD (crystalline tacrolimus component). These two components were mixed to cover 0%–100% of crystalline drug. Uniformity of the sample mixtures was confirmed by near-infrared chemical imaging. XRPD showed three distinct peaks of crystalline drug at 8.5◦ , 10.3◦ , and 11.2◦ (2θ ), which were nonoverlapping with the excipients. Principal component regressions (PCR) and partial least square (PLS) regression used in model development showed high R2 (>0.99) for all the mixtures. Overall, the model showed low root mean square of standard error, standard error, and bias, which was smaller in PLS than PCR-based model. Furthermore, the model performance was evaluated on the formulations with known percentage of crystalline drug. Model-calculated crystalline drug percentage values were close to actual value. Therefore, these studies strongly suggest the application of chemometric-XRPD models as a quality control tool to quantitatively predict C 2014 Wiley Periodicals, Inc. and the American Pharmacists Association J Pharm Sci the crystalline drug in the formulation.  Keywords: tacrolimus; solid dispersion; powder X-ray diffraction; chemometric; principal component analysis; partial least square; multivariate analysis

INTRODUCTION Tacrolimus, a potent immunosuppressive agent, is the clinician’s drug of choice to prevent rejection of the transplanted solid organ in patients if cyclosporine A or steroid-based regimen fails.1,2 Despite its life saving potential in such patients, the drug has biopharmaceutical and solubility issues.3–5 Under Biopharmaceutical Classification System (BCS), it is placed in BCS class II drug exhibiting high permeability and low solubility. Even though a plethora of formulation and drug delivery modalities have been reported to improve bioavailability, which include self microemulsifying drug delivery system, prodrug, oily solution, complexation with cyclodextrin, micro or nanobased drug delivery system, and solid dispersion (SD)5,6 , only a few formulations have made it to the market. Commercially, it is available as injection (5 mg/mL), capsules (Prograf , 0.5, 1, and 5 mg, a SD formulation,4,7,8 and ointment8 (Protopic 0.03, and 1%). The main focus of formulating the poorly aqueous soluble drug is to present the drug in its molecular form to the absorbing surface, that is, gut mucosa. Any intrinsic or extrinsic factor triggering crystallization of the drug in the formulation can impact an overall absorption and thereby bioavailability of the drug, which is of particular concern to a drug like tacrolimus possessing a narrow therapeutic index (5– 15 :g/mL).9 Sub-potent dose of tacrolimus can increase the R

R

Correspondence to: Mansoor A. Khan (Telephone: +301-796-0016; Fax: +301796-9816; E-mail: [email protected]) Disclaimer: The findings and conclusions in this article have not been formally disseminated by the United States Food and Drug Administration and should not be construed to represent any Agency determination or policy. Journal of Pharmaceutical Sciences  C 2014 Wiley Periodicals, Inc. and the American Pharmacists Association

potential of a graft rejection, whereas overdose can cause dosedependent neuro/nephrotoxicity. Therefore, a dosage form or delivery strategy is needed that can reliably deliver a calibrated concentration of the drug for the graft survival and minimization of the drug-induced toxicity. Because tacrolimus is recommended to be administered orally after initial intravenous infusion,10,11 SD of tacrolimus offers one of the several other options to improve bioavailability. SD is a formulation strategy to enhance drug solubility by molecularly dispersing the poorly aqueous soluble drug in hydrophilic polymer and converting crystalline to amorphous drug, enhancing wettability, and affecting carrier-mediated solubilization of the poorly aqueous soluble drug.12 Various dispersion techniques to prepare SD have been reported including solvent evaporation method, melting method, solvent wetting method, and surface attachment methods.9,13 Although SD approach improves the physicochemical properties of the drug, it is not a thermodynamically stable system leading to reversion of the molecules to its thermodynamically more stable form by crystallization over a period of time.14 Depending upon the manner the finished product is handled during the excursion of product from the industry to the pharmacy or end user, the onset of crystallization process or chances of drug falling out of specification may be sooner than later and thereby patients will not get the timely therapeutic benefit of the drug. There were instances of drug products recalled by United States Food and Drug Administration (US FDA) because of the crystallization of the amorphous drug in the product.15 Therefore, a universal evaluation technique that can be used to detect and quantitatively differentiate the crystallinity in SD or its product intermediate is required to monitor the state of drug (percentage crystalline/amorphous) in drug products. This determination could allow for the development of crystallinity-discrimination dissolution methods.

Siddiqui et al., JOURNAL OF PHARMACEUTICAL SCIENCES

1

2

RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology

Powder X-ray diffraction (XRPD) is one of the several techniques commonly used in evaluating solid-state property of the drug. Because polycrystallites in the powder always contain crystallite at an angle that satisfy Braggs law to produce diffraction, XRPD can produce constituent-specific diffraction at an angle (2θ ) whose position and intensities depend upon crystal lattice and unit cells contents, respectively [United States Pharmacopeia (USP) 35, Chapter 941]. In case of amorphous drug, absence of crystal lattice produces halo XRPD. Furthermore, proportionality of the peak intensity to the crystalline fraction present in the powder has been reported, which means fraction of crystalline constituent in the powder can be obtained by establishing correlation between drug and peak intensity.16 Because XRPD is a multivariate process, responses of the powder diffraction can be fit into a model by establishing multivariate calibration curve using samples of known crystallinity. Chemometrics, which is a method of relating data on chemical system to the state of the system using statistics,17 employs multivariate analysis techniques for model development. In this method, structures of the data are evaluated using principal component analysis (PCA) and assessments are performed whether data pretreatment that can improve the overall structure of the data. Multivariate model of the property of interest (percentage crystalline or amorphous) is then developed using principal component regression (PCR) or partial least square (PLS) regression. These regression models are then used to predict the property of interest in the unknown samples. Previously, chemometric models for estimating nimodipine polymorphs in the mixtures were built and their performances were evaluated using spectroscopic techniques, that is, FTIR, near infrared (NIR), and Raman.18 Therefore, the focus of this work was to build and validate chemometric-XRPD models for tacrolimus SD and evaluate their performance in predicting percentage crystalline (or amorphous) tacrolimus in tacrolimus product intermediates.

MATERIALS AND METHODS Materials Tacrolimus (Ria International LLC, East Hanover, New Jersey), hydroxypropyl methyl cellulose (HPMC, 6cps viscosity substitution type 2910 USP) (Shin Etsu Chemical Company, Tokyo, Japan), croscarmellose sodium (FMC Biopolymer, Philadelphia, Pennsylvania), lactose monohydrate (Sigma–Aldrich, St. Louis, Missouri), and ethyl alcohol (Decon Lab, Inc., King of Prussia, Pennsylvania) were purchased and used as received. All other chemicals and solvents used were of analytical grade. Methods Preparation of SD The commercial tacrolimus product (Prograf ) and its generic forms are the SD formulations, and contain HPMC (hypromellose), lactose monohydrate (LM), croscarmellose sodium (CC), and magnesium stearate as its inactive ingredients (US FDA Prograf capsule USP label). The capsule consists of two components: SD granules and extragranular ingredients.19,20 Tacrolimus SD formulations were prepared by solvent evaporation methods described by Yamashita et al.21 with some modification. Briefly, HPMC was added to the ethanolic drug solution. The mixtures were kept stirring for an hour to hydrate R

R

Siddiqui et al., JOURNAL OF PHARMACEUTICAL SCIENCES

Table 1.

Composition of Tacrolimus Solid Dispersion Formulations Drug:Excipients

Formulation SD1 SD2 SD3

Tacrolimus

Excipients (HPMC, CC, LM)

1 1 1

3 14 22

the polymer, then croscarmellose sodium and lactose monohydrate were added under stirring. Ethyl alcohol was allowed to evaporate under stirring. Residual alcohol was removed by drying under reduced pressure at 25◦ C for 24 h. The dried mass was then pulverized using glass mortar and pestle and passed through USP sieve no 80/120. The formulations proportions that passed through USP sieve no 80 but retained on 120 were used in analysis. Because SD can be prepared in different drug to excipients ratios, drug to excipients ratios (Table1) 1:3.10 (SD-1), 1:14 (SD-2), and 1:22 (SD-3), which represent 24.4%, 6.7%, and 4.3% of tacrolimus with respect to excipients, were chosen in the present work to prepare three tacrolimus SD granules and called amorphous tacrolimus component (SD-1, SD-2, and SD-3). Their amorphous state was confirmed by differential scanning calorimetry and XRPD. Furthermore, three placebo formulations were prepared in a similar manner having identical qualitative and quantitative excipients composition as amorphous SD components (SD-1, SD-2, and SD-3). Crystalline tacrolimus equivalent to SD-1, SD-2, and SD-3 was added to placebo and labeled as crystalline tacrolimus component (PM-1, PM-2, and PM-3). Preparation of Tacrolimus Samples SD-1 and PM-1, SD-2 and PM-2, and SD-3 and PM-3 were mixed in various ratios to make samples that have 0%–100% of added crystalline tacrolimus of the formulations. These sets of mixtures were labeled as FD-1, FD-2, and FD-3. These mixtures were then swirled 100 times in horizontal and vertical direction to allow mixing of the powder components. NIR Chemical Imaging (NIR-CI) The components distribution was tested by NIR chemical imaging (NIR-CI) technique. NIR-CI of FD-1, FD-2, and FD-3 samples sets was acquired by a Sapphire imaging system using SapphireGo software (Malvern, Worcestershire, UK). The instrument is equipped with indium-gallium-arsenide focal-plane array detector and liquid-crystal tunable filter to allow diffuse light from the sample and produces 320 × 256 pixel images. Before capturing images, the background reference and dark response were collected using Spectralon 99 (Labsphere, North Sutton, New Hampshire) and a clean stainless steel mirror, respectively. In the NIR range starting from 1400 to 2450 nm, the data were obtained with an increment of 10 nm, and eight scans were coadded to produce an average spectrum. The obtained data were analyzed by ISys chemical imaging software (Malvern). Data were treated before analysis, which include transformation of reflectance to absorbance, masked, truncated, and normalized by mean centering and scaling to unit variance by spectrum. The libraries for each set of sample (FD1, FD-2, or FD-3) consisting of two components namely amorphous tacrolimus component (added 100% amorphous drug, SD-1, SD-2, or SD-3) and crystalline tacrolimus component DOI 10.1002/jps.23912

RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology

(added 100% crystalline drug, PM-1, PM-2, or PM-3) were built, and PLS2 (partial least squares 2) fitting was employed to obtain PLS concentration scores and images. Powder X-Ray Diffraction (XRPD) Powder X-ray diffraction patterns for individual components of the formulation (HPMC, croscarmellose sodium, lactose monohydrate), placebos formulations, and FD formulations sets were collected using a Bruker D8 Advance with DaVinci design (Bruker AXS, Madison, Wisconsin) using Cu K" radiation (8 = ˚ at a voltage of 40 kV and current of 40 mA and 1.5405 A) equipped with the LYNXEYE scintillation detector. Corundum was used as an external standard to calibrate the XRPD instrument. Powder sample equivalent to 500 mg in case of FD-1 and 550 mg for FD-2 and FD-3 samples sets was used for data collection. Six-replicate diffractograms were obtained for each set of samples. Weighed powder was placed in the sample holder, pressed using zero diffraction plate, and scanned over 2θ range of 4◦ –30◦ with a step size of 0.00915◦ at 1 s per step (3000 total steps). Samples were rotated at 15/min during measurements to get average diffractogram of the sample. The XRPD data collection and data analyses were achieved through Diffract.Suite (V2.2). Statistical Analysis Multivariate data analysis of the data was carried out by UnscramblerX chemometrics software (version 10.1; Camo Process, Oslo, Norway). For PCR and PLS regression model development, the numbers of latent variables (LVs) or factors are very critical. The optimum number of latent factors was determined by the lowest value of RMSECV (root mean squared error

3

of cross validation) from “leave one batch out” cross-validation. Model performances were evaluated by R, R2 , RMSEC, RMSEP (root mean square of calibration and prediction), SEP, and SEC (standard error or prediction and calibration). Their predictive capabilities were assessed by RMSEP and bias.

RESULTS AND DISCUSSION NIR Chemical Imaging (NIR-CI) NIR-CI is a powerful technique used widely in pharmaceutical industry for counterfeit detection, quantification of polymorphs, and distribution of the components of complex dosage forms, and so on. Each pixel in the images contains spatial and spectral information. The color and intensity in the image, which are represented by a color bar, demonstrate the relative distribution of the components in the mixtures. After mathematically treating the data (inverse log, masking, Savitzky–Golay), partial least square 2 (PLS2) was applied considering scanned wavelengths as variables and intensity as response. NIR-CI images of FD-1 and FD-2 (not shown FD-3 set) samples sets are shown in Figure 1. The images showed uniform distribution of amorphous and crystalline tacrolimus components in the sample sets except a few regions rich or poor in amorphous and/or crystalline tacrolimus, which was indicated by dark red or blue spots in the images, respectively. Furthermore, homogeneity of the sample mixture was supported by narrow values of the skewness, kurtosis and bell-shaped histogram (not shown). The images, when arranged according to the increased amorphous or crystalline component, showed increased mean absorbance of the image, which matched qualitatively with the actual percentage of the respective components in the mixtures.

Figure 1. PLS concentration image with respect to amorphous and crystalline tacrolimus component of (a) FD-1 and (b) FD-2 data sets. DOI 10.1002/jps.23912

Siddiqui et al., JOURNAL OF PHARMACEUTICAL SCIENCES

4

RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology

Figure 2. XRPD of individual components, amorphous and crystalline tacrolimus components, and their placebos.

Powder X-Ray Diffraction (XRPD) The XRPD results are shown in Figure 2. Tacrolimus monohydrate and lactose monohydrate showed numerous diffraction peaks, which indicated their crystalline nature. On the other hand, HPMC and croscarmellose sodium yielded halo diffractogram confirming their amorphous nature. The SD formulations SD-1, SD-2, and SD-3 showed only the diffraction peaks of the lactose monohydrate but no peaks diagnostic of tacrolimus monohydrate, which indicated the conversion of crystalline to amorphous form of drug during their formulation. This was further supported by their placebo formulations diffractrogram, which look like their SD formulations. Furthermore, amorphous tacrolimus component and placebos formulations showed only diffraction peaks of the lactose monohydrate. Moreover, crystalline tacrolimus component showed the diffraction peaks of the drug and lactose monohydrate. All the peaks of the drug were overlapping with the lactose monohydrate except peaks in the 2θ region of 8.2◦ –11.4◦ , which contain three drug peaks. These peaks were distinct, nonoverlapping with any of the formulation excipients peaks, and could be used for chemometric model development. The three peaks are located at 8.5◦ , 10.3◦ , and 11.2◦ and their intensity increased or decreased with increased crystalline or amorphous tacrolimus in the formulations. Additionally, XRPD was able to detect 2.5%, 5%, and 10% crystalline tacrolimus component in FD-1, FD-2, and FD-3, respectively (Fig. 3). These formulations contained total drug of 24.4%, 6.7%, and 4.3% with respect to excipients. It means XRPD was capable of detecting 0.61%, 0.34%, and 0.43% (w/w) of crystalline tacrolimus in FD-1, FD-2, and FD-3 samples sets. Chemometric Models Truncated data from 8.2◦ to 11.4◦ 2θ were used in chemometric model development, which contained three distinct peaks of tacrolimus. The pretreatment of the data with various mathematical methods did not improve the data as the primary source of data variation in XRPD is preferred crystal orientation and sample heterogeneity. The preferential crystal orientation Siddiqui et al., JOURNAL OF PHARMACEUTICAL SCIENCES

could be prevented by reducing the crystal size and pressing the powder sample in the sample holder before measurement.22 Principal Component Analysis Under this multivariate analysis approach, the original variables are transformed into a linear combination of data for the investigation of the data structure with reduced dimensionality 23 . The linearly combined new variables are called LVs or principal components (PCs)24 . Usually, PCs are ordered with the first one explaining the maximum variance in the data and remaining variances described by successive PCs, which are orthogonal to the previous ones. Each PC consists of score and loading values representing a scalar and vector quantity as score represents the distance of each sample from the PC axis and loading demonstrates the relative orientation of PC to the original variables. PCA of the XRPD raw data was performed without any data pretreatment. As shown the PCA score plot in Figure 4, PC1 and PC2 could explain 95% and 3%, 84% and 12%, and 75% and 15% data variability of FD-1, FD-2, and FD-3 data set, respectively. Additionally, data variation explained by PC1 decreased, whereas it increased in case of PC2 from FD-1 to FD-3 formulations, which might be explained by an increased excipients proportion in the formulations. Furthermore, two PCs were able to explain more than 90% variability in all the data sets. Interestingly, one peculiar finding was observed in the PCA data sets. Total variability explained by two PCs decreased as the drug percentage decreased in the formulations. This could be explained by a decrease in the drug signal and an increase in the noise in the data with subsequent decrease in the drug percentage in the formulations. PCA score plot further demonstrated clustering of the same samples in all the data sets at added crystalline tacrolimus level from 0% to 100%. PC1 score number was increasing with increased crystalline tacrolimus component in the samples. Therefore, variation of crystallinity in the formulation could be assigned to PC1. This was further supported by loading plot of the PC1 of all the data sets (Fig. 5). The PC1 loading plots were superimposable to crystalline tacrolimus with three distinct peaks at 8.5◦ , 10.3◦ , DOI 10.1002/jps.23912

RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology

5

Figure 3. Truncated XRPD of (a) FD-1, (b) FD-2, and FD-3 data sets.

Figure 4. PCA score plot of (a) FD-1, (b) FD-2 and FD-3 data sets. DOI 10.1002/jps.23912

Siddiqui et al., JOURNAL OF PHARMACEUTICAL SCIENCES

6

RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology

and 11.2◦ 2θ. On the other hand, PC2, which accounted for minor variation in data sets, did not show pattern similar to PC1 rather it showed randomness on the PCA score plot. It might be related to physical attributes of the formulations such as particle size and density, and so on. Calibration Model Partial least square (PLS) and Principal compnent regression (PCR) algorithm were used for calibration model development. For PLS and PCR model development, the number of LVs was determined by low value of RMSECV and RMSEP.23 It was found that two and three LVs were giving lowest value of RMSECV and PMSEP for FD-1 and FD-2, and FD-3 data sets, respectively, and were selected for further model development. Two and three LVs of PLS could explain 99%, 96%, and 95% variability in the data of FD-1, FD-2, and FD-3 samples, respectively. First LV accounted for most of variability in the data, followed by second and third LVs, and similar trend was

shown by PCR analysis. Furthermore, first LVs of PLS and PCR are related to percentage crystalline tacrolimus added to the formulation as indicated by plot with three peaks of the crystalline tacrolimus (not shown). The built calibration model after mean centering of the data was validated by test set approach in which a subset of samples are selected, which were not included in the computation of the model parameters. The calibration (blue) and validation (red) plots are shown in Figure 6, and these plots were almost superimposable. This figure shows linear relationship between theoretical and predicted values of amorphous/crystalline tacrolimus. Residual values (differences between actual and model predicated values of amorphous/crystalline tacrolimus) were small and showed random distribution on the plot, which indicated good fitting of the data (Fig. 7). Furthermore, slope was almost equal to one, which indicated the absence of systemic and constant error in the data. The ideal line passing through origin represents the best fitting line with determination coefficient (R2 ) of one and zero

Figure 5. PCA loading plot of (a) FD-1, (b) FD-2 and FD-3 data sets. Siddiqui et al., JOURNAL OF PHARMACEUTICAL SCIENCES

DOI 10.1002/jps.23912

RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology

7

Figure 6. PLS model calibration and validation plot of (a) FD-1, (b) FD-2 and FD-3 data sets.

Figure 7. Residual plot (a) FD-1, (b) FD-2 and FD-3 data sets.

DOI 10.1002/jps.23912

Siddiqui et al., JOURNAL OF PHARMACEUTICAL SCIENCES

8

RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology

Table 2.

PLS and PCR Models Statistical Parameters of Data Sets

Sample Set

Model

Component

Plot

Sample Number

Slope

Offset

FD-1

PLS

Crystalline

Calibration Validation Calibration Validation Calibration Validation Calibration Validation

54 36 54 36 54 36 54 36

0.998 1.008 0.998 1.008 0.998 1.008 0.998 1.008

0.081 0.624 0.104 − 1.401 0.086 0.597 0.110 − 1.391

0.999 0.999 0.999 0.999 0.999 0.999 0.999 0.999

0.998 0.999 0.998 0.999 0.998 0.999 0.998 0.999

1.399 1.092 1.399 1.092 1.437 1.060 1.437 1.060

1.412 0.733 1.412 0.733 1.437 0.712 1.437 0.712

4.945e−07 0.818 −5651e−07 − 0.818 1.201e−06 0.795 − 7.771e−07 − 0.795

Calibration Validation Calibration Validation Calibration Validation Calibration Validation

36 36 36 36 36 36 36 36

0.994 0.978 0.994 0.978 0.992 0.970 0.992 0.970

0.255 0.477 0.370 1.716 0.309 0.620 0.448 2.333

0.997 0.998 0.997 0.998 0.996 0.997 0.996 0.997

0.993 0.996 0.993 0.996 0.992 0.995 0.992 0.995

2.515 1.983 2.515 1.983 2.769 2.272 2.769 2.272

2.550 1.995 2.550 1.995 2.808 2.274 2.808 2.274

− 1.589e−06 − 0.254 1.272e−06 0.254 − 1.113e−06 − 0.365 1.272e−06 0.365

Calibration Validation Calibration Validation Calibration Validation Calibration Validation

42 30 42 30 42 30 42 30

0.998 0.969 0.998 0.969 0.998 0.970 0.998 0.970

0.079 − 0.158 0.075 3.256 0.098 − 0.222 0.094 3.23

0.999 0.999 0.999 0.999 0.999 0.999 0.999 0.999

0.998 0.996 0.998 0.996 0.998 0.996 0.998 0.996

1.183 1.830 1.183 1.830 1.319 1.856 1.319 1.856

1.198 1.413 1.198 1.413 1.336 1.418 1.336 1.418

− 3.815e−07 − 1.195 1.431e−07 1.195 − 5.722e−07 − 1.229 2.384e−07 1.229

Amorphous PCR

Crystalline Amorphous

FD-2

PLS

Crystalline Amorphous

PCR

Crystalline Amorphous

FD-3

PLS

Crystalline Amorphous

PCR

Crystalline Amorphous

intercept. The correlation coefficient (R) and R2 were >0.99 for all the models for amorphous and crystalline drug prediction in the formulations. The errors in the models are indicated by RMSEC, PMSEP, SEC, and SEP and these values were low. Furthermore, these values were relatively low in the FD-1 in comparison with FD-2 and FD-3 samples sets, which were related to percentage drug present in the formulations. Overall, RMSEP, which represents error in the prediction using model, for both the models were low, which reflects their ability to predict the level of crystalline drug in an unknown formulation with minimum margin of error (Table 2). When comparison was made between two model algorithms, RMSEP and SEP values for PCR were higher than for PLS models, which indicated improved PLS prediction ability than PCR. Bias, which represents difference between population means and true value, for the PLS was found to be lower than PCR, which further indicates lower prediction variance.

Prediction of Crystalline Tacrolimus Using the PCR and PLS model, prediction of crystalline/ amorphous tacrolimus in the formulations having crystalline/ amorphous level was performed to evaluate performance and accuracy of the model. Four independent samples of known crystalline tacrolimus percentage of the FD-1, FD-2, and FD-3 sets were prepared. These samples were not used in the model development. The results of model predictions are given in Table 3. In case of the FD-1, samples of known crystallinity yielded results that were very close to actual crystalline tacrolimus percentage value in the formulation. Furthermore, error was also low as indicated by standard deviation. The results of the PCR model of the FD-1 samples have less error than the PLS model. This indicated superior results for PCR Siddiqui et al., JOURNAL OF PHARMACEUTICAL SCIENCES

Correlation

R2

RMSE

SEP

Bias

in comparison with PLS model for FD-1 formulations samples that has 24.4% drug. On the other hand, results of known crystalline tacrolimus samples set of the FD-2 and FD-3 yielded large variability in the results. These formulations have high excipients proportion in comparison with drug (6.7% and 4.3% drug, respectively). The intensity of the three peaks of crystalline tacrolimus would decrease as the percentage of the drug decrease in the formulations. This would decrease the signal to noise ratio of the data. On the contrary to FD-1 models, the PLS models of these samples were better than PCR. In summary, both PCR and PLS models of the entire formulation samples were able to predict low crystalline tacrolimus drug in the formulations with acceptable error.

CONCLUSIONS The XRPD of the crystalline drug formulations showed three distinct peaks of the crystalline drug. These peaks were nonoverlapping with the excipients in the formulation and were used for building chemometric models. NIR-CI indicated the homogeneity of the prepared formulations containing amorphous and crystalline tacrolimus. PCR and PLS models for crystalline and amorphous tacrolimus formulations were built and validated. These models were applied on formulations with known drug crystallinity to assess the model capabilities and accuracy. Accuracy to predict crystalline and amorphous tacrolimus in the formulations depends on the drug to excipients ratios. The models based on formulation having higher drug to excipients ratio have better model prediction capability as indicated by their statistical parameters. On the other hand, error increased and prediction capability decreased with a decrease in drug to excipient ratio. Moreover, models were able to predict formulation containing 4%, 7.5%, and 12.5% crystalline DOI 10.1002/jps.23912

RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology

Table 3.

9

Results of PLS and PCR Prediction of Crystalline and Amorphous Tacrolimus Component from Data Sets Actual

Sample Set FD-1

Model PLS

PCR

FD-2

PLS

PCR

FD-3

PLS

PCR

Crystalline Tacrolimus Component (%)

Amorphous Tacrolimus Component (%)

4 7.5 12.5 75 4 7.5 12.5 75 7.5 12.5 17.5 75 7.5 12.5 17.5 75 12.5 17.5 30 75 12.5 17.5 30 75

96 92.5 88.5 25 96 92.5 88.5 25 92.5 88.5 82.5 25 92.5 88.5 82.5 25 87.5 82.5 70 25 87.5 82.5 70 25

tacrolimus, which represent 24.4%, 6.7%, and 4.3% of the total drug in the formulation (normalized crystalline tacrolimus value with respect to total percentage would be 0.98%, 0.50%, and 0.54%) with low margin of error. Literature reported crystalline drug determination in the mixture of crystalline and amorphous drug by XRPD is 5%. Combining XRPD with chemometric methods extended the capability of XRPD. Furthermore, these models can be applied to a variety of tacrolimus SD granules to monitor the crystalline reversion of the drug.

ACKNOWLEDGMENTS This work was supported by funding from the NIH and the Medical Counter Measure initiative. The authors would like to thank Oak Ridge Institute for Science and Education (ORISE) for supporting post-doctoral research program.

REFERENCES 1. Borhade V, Nair H, Hegde D. 2008. Design and evaluation of selfmicroemulsifying drug delivery system (SMEDDS) of tacrolimus. AAPS Pharm Sci Tech 9:13–21. 2. Letko E, Bhol K, Pinar V, Foster CS, Ahmed AR. 1999. Tacrolimus (FK 506). Ann Allergy Asthma Immunol 83:179–190. 3. Yoshida T, Kurimoto I, Yoshihara K, Umejima H, Ito N, Watanabe S, Sako K, Kikuchi A. 2012. Aminoalkyl methacrylate copolymers for improving the solubility of tacrolimus. I: Evaluation of solid dispersion formulations. Int J Pharm 428:18–24. 4. Arima H, Yunomae K, Miyake K, Irie T, Hirayama F, Uekama K. 2001. Comparative studies of the enhancing effects of cyclodextrins on the solubility and oral bioavailability of tacrolimus in rats. J Pharm Sci 90:690–701.

DOI 10.1002/jps.23912

Model Predicted Crystalline Tacrolimus Component (%) 3.89 7.20 14.43 74.22 3.89 7.28 14.63 73.68 6.58 13.26 21.32 84.52 7.00 13.67 21.83 85.56 15.42 18.36 26.98 72.96 15.41 18.41 27.05 72.97

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

0.86 0.98 1.62 3.98 0.83 0.94 1.55 3.88 1.45 1.51 1.89 3.30 1.69 1.75 2.18 3.79 1.52 1.64 2.24 2.24 1.54 1.65 2.32 2.25

Amorphous Tacrolimus Component (%) 96.11 92.80 85.57 25.78 96.11 92.72 85.37 26.32 93.42 86.74 78.68 15.48 93.00 86.33 78.17 14.44 84.58 81.64 73.02 27.04 84.59 81.59 72.95 27.03

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

0.86 0.98 1.62 3.98 0.83 0.94 1.55 3.88 1.45 1.51 1.89 3.30 1.69 1.75 2.18 3.79 1.52 1.64 2.24 2.24 1.54 1.65 2.32 2.25

5. Wang Y, Zhang T, Liu H, He F, Hegde D. 2011. Enhanced oral bioavailability of tacrolimus in rats by self-emulsifying drug delivery systems. Drug Dev Ind Pharm 37:1225–1230. 6. Overhoff K, McConville J, Yang W, Johnston K, Peters J, Williams R. 2008. Effect of stabilizer on the maximum degree and extent of supersaturation and oral absorption of tacrolimus made by ultra-rapid freezing. Pharm Res 25:167–175. 7. Tamura S, Ohike A, Ibuki R, Amidon GL, Yamashita S. 2002. Tacrolimus is a class II low-solubility high-permeability drug: The effect of P-glycoprotein efflux on regional permeability of tacrolimus in rats. J Pharm Sci 91:719–729. 8. Pople PV, Singh KK. 2010. Targeting tacrolimus to deeper layers of skin with improved safety for treatment of atopic dermatitis. Int J Pharm 398:165–178. 9. Patel P, Patel H, Panchal S, Mehta T. 2012. Formulation Strategies for drug delivery of tacrolimus: An overview. Int J Pharm invest 2:169– 175. 10. Staatz C, Tett S. 2004. Clinical pharmacokinetics and pharmacodynamics of tacrolimus in solid organ transplantation. Clin Pharmacokinet 43:623–653. 11. Borhade VB, Nair HA, Hegde DD. 2009. Development and characterization of self-microemulsifying drug delivery system of tacrolimus for intravenous administration. Drug Dev Ind Pharm 35:619–630. 12. Yamashita K, Nakate T, Okimoto K, Ohike A, Tokunaga Y, Ibuki R, Higaki K, Kimura T. 2003. Establishment of new preparation method for solid dispersion formulation of tacrolimus. Int J Pharm 267:79–91. 13. Joe JH, Lee WM, Park YJ, Joe KH, Oh DH, Seo YG, Woo JS, Yong CS, Choi HG. 2010. Effect of the solid-dispersion method on the solubility and crystalline property of tacrolimus. Int J Pharm 395:161–166. 14. Serajuddin ATM. 1999. Solid dispersion of poorly water-soluble drugs: Early promises, subsequent problems, and recent breakthroughs. J Pharm Sci 88:1058–1066. 15. Guo Y, Shalaev E, Smith S. 2013. Physical stability of pharmaceutical formulations: Solid-state characterization of amorphous dispersions. TrAC Trends Anal Chem 49:137–144.

Siddiqui et al., JOURNAL OF PHARMACEUTICAL SCIENCES

10

RESEARCH ARTICLE – Pharmaceutics, Drug Delivery and Pharmaceutical Technology

16. Liu J, Nagapudi K, Kiang Y-H, Martinez E, Jona J. 2009. Quantification of compaction-induced crystallinity reduction of a pharmaceutical solid using 19F solid-state NMR and powder X-ray diffraction. Drug Dev Ind Pharm 35:969–975. 17. Wise BM, Gallagher NB. 1996. The process chemometrics approach to process monitoring and fault detection. J Process Control 6:329–348. 18. Siddiqui A, Rahman Z, Sayeed VA, Khan MA. 2013. Chemometric evaluation of near infrared, Fourier transform infrared, and raman spectroscopic models for the prediction of nimodipine polymorphs. J Pharm Sci 102:4024–4035. 19. Medicines Evaluation Board (MEB) 2010. Public assessment report of the medicine evaluation board in the Netherlands [tacrolimus Sandoz 0.5 mg, 1 mg, 5 mg, capsules, Hard Sandoz B.V., The Netherlands tacrolimus (as monohydrate)].

Siddiqui et al., JOURNAL OF PHARMACEUTICAL SCIENCES

20. Medicines Evaluation Board (MEB) 2010. Public assessment report of the medicines evaluation board in the Netherlands [(tacrolimus Accord 0.5 mg and 1mg capsules, hard Accord Healthcare Ltd., United Kingdom tacrolimus (as monohydrate)]. 21. Yamashita K, Hashimoto E, Nomura Y, Shimojo F, Tamura S, Hirose T, Ueda S, Saitoh T, Ibuki R, Ideno T. 2005. Google Patents US Patent No; US6884433 Sustained release formulation containing tacrolimus. 22. SUDA M, Takayama K, Otsuka M. 2008. An accurate quantitative analysis of polymorphic content by chemometric X-ray powder diffraction. Anal Sci 24:451–457. 23. Wu Z, Sui C, Xu B, Ai L, Ma Q, Shi X, Qiao Y. 2013. Multivariate detection limits of on-line NIR model for extraction process of chlorogenic acid from Lonicera japonica. J Pharm Biomed Anal 77:16–20.

DOI 10.1002/jps.23912