Determination of tacrolimus crystalline fraction in the commercial immediate release amorphous solid dispersion products by a standardized X-ray powder diffraction method with chemometrics

Determination of tacrolimus crystalline fraction in the commercial immediate release amorphous solid dispersion products by a standardized X-ray powder diffraction method with chemometrics

International Journal of Pharmaceutics 475 (2014) 462–470 Contents lists available at ScienceDirect International Journal of Pharmaceutics journal h...

3MB Sizes 0 Downloads 8 Views

International Journal of Pharmaceutics 475 (2014) 462–470

Contents lists available at ScienceDirect

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

Determination of tacrolimus crystalline fraction in the commercial immediate release amorphous solid dispersion products by a standardized X-ray powder diffraction method with chemometrics$ Ziyaur Rahman, Akhtar Siddiqui, Srikant Bykadi, Mansoor A. Khan * Division of Product Quality and Research, Center for Drug Evaluation and Research, Food and Drug Administration, MD, USA

A R T I C L E I N F O

A B S T R A C T

Article history: Received 8 May 2014 Received in revised form 21 August 2014 Accepted 26 August 2014 Available online 27 August 2014

Clinical performance of an amorphous solid dispersion (ASD) drug product is related to the amorphous drug content because of the greater bioavailability of this form of the drug than its crystalline form. Therefore, it is paramount to monitor the amorphous and the crystalline fractions in the ASD products. The objective of the present investigation was to study the feasibility of using a standardized X-ray powder diffraction (XRPD) in conjunction with chemometric methods to quantitate the amorphous and crystalline fraction of the drug in several tacrolimus ASD products. Three ASD products were prepared in which drug to excipients ratios ranged from 1:19 to 1:49. The amorphous and crystalline drug products were mixed in various proportions so that amorphous/crystalline tacrolimus in the samples vary from 0 to 100%. XRPD of the samples of the drug products were collected, and PLSR and PCR chemometric methods were applied to the data. The R2 was greater than ‘0.987’ for all the models and bias in the models were statistically insignificant (p > 0.05). RMSEP and SEP values were smaller for PLSR models than PCR models. The models prediction capabilities were good and can predict as low as 10% when drug to excipient ratio is as high as 1:49. In summary, XRPD and chemometric provide powerful analytical tools to monitor the crystalline fractions of the drug in the ASD products. Published by Elsevier B.V.

Keywords: Tacrolimus Amorphous solid dispersion X-ray powder diffraction Partial least square regression Principle component regression

1. Introduction Solid forms of drugs are commonly delivered as crystalline forms, but amorphous forms are of increasing interest due to their better physicochemical and absorptionproperties (Yoo et al., 2009; Chokshi et al., 2007). However, the amorphous form of the drug is thermodynamically unstable, may revert to its stable crystalline form under various environmental conditions such as high humidity, temperature or both (Rumondor et al., 2011; Rahman et al., 2013; Sinclair et al., 2011). Due to stability reasons, most of the drugs present in the commercial products are crystalline and only a very few products contain the amorphous form of the drug (Janssens and Van den Mooter, 2009). Unlike a crystalline drug which has finite melting point, the amorphous drug is characterized by glass transition temperature, Tg. Tg is a temperature range over which the amorphous drug properties change from solid-like (glass) to liquid-like (supercooled liquid). Tg is often used as a benchmark to

$ The views and opinions expressed in this paper are only those of the authors, and do not necessarily reflect the views or policies of the FDA. * Corresponding author at: FDA/CDER/DPQR, White Oak LS Building 64, Room 1070, 10903 New Hampshire Ave, Silver Spring, MD 20993-002, USA. Tel.: +1 301 796 0016. E-mail address: [email protected] (M.A. Khan).

http://dx.doi.org/10.1016/j.ijpharm.2014.08.050 0378-5173/ Published by Elsevier B.V.

assess amorphous system stability and storage condition because chemical and physical stability of the amorphous system is affected if stored above its Tg (Wu et al., 2013; Yoshioka and Aso, 2007). In general, an amorphous solid dispersion (ASD) formulation (which contains only an amorphous drug and amorphous/crystalline excipient(s) or only amorphous/crystalline excipient(s) and crystalline drug or both drug and excipient(s) are in amorphous forms) is considered stable if stored at 50 K or greater below the Tg of the formulation (Hancock et al.,1995; Newman et al., 2012). However, Tg often fails to predict stability of the ASD especially in the multicomponent system formulation (Baird and Taylor, 2012). Commercially available formulations of tacrolimus are the ASD intended to improve its dissolution rate and bioavailability (Janssens and Van den Mooter, 2009). As with any amorphous drug form, it is intrinsically unstable and reverts to its stable forms (Newman et al., 2012). The conversions are even faster if not formulated (high drug loading and improper selection of excipients e.g., low Tg and/or immiscible polymer etc.) or processed properly (improper process and uncontrolled/unmonitored process parameters etc.) or exposed to high humidity/temperature conditions (Rumondor et al., 2011; Rahman et al., 2013; Sinclair et al., 2011).There were recalls of the tacrolimus ASD products due to their failure to meet critical quality attributes (CQAs) (FDA, 2014). The clinical performance of ASD can only be ensured if the

Z. Rahman et al. / International Journal of Pharmaceutics 475 (2014) 462–470

463

amorphous/crystalline ratio meet its release specification and maintain that ratio/fraction throughout product shelf life. The amorphous/crystalline ratio change in the ASD products can be measured by direct and indirect method. The indirect method involves performing discriminating dissolution methods. The product will show sign of crystalline reversion by not meeting dissolution specification if tested in discriminating dissolution conditions (Newman et al., 2012; Zidan et al., 2012). Direct methods, however, involve quantitating the crystalline drug fraction in the product by various analytical techniques such as DSC, NIR, FTIR, Raman, terahertz-pulsed, ssNMR and X-ray powder diffraction (XRPD) (Shah et al., 2006; Siddiqui et al., 2014). Vibration spectroscopies such as NIR, FTIR and Raman are not always sensitive to differentiate between amorphous and crystalline forms of the drug. Moreover, newer vibration technique, terahertz-pulsed spectroscopy probe lattice and low energy hydrogen bonding vibrations, and it showed more pronounced spectral changes for amorphous and crystalline forms compared to FTIR and NIR, however, this technique is still evolving (Shah et al., 2006). On the other hand, XRPD is the most definitive detection and quantification method for the crystalline/amorphous fraction due to distinctive diffractogram pattern of the two forms. XRPD have an advantage over DSC in terms of displaying multiple peaks and some of which are overlapping with the excipients of the formulation. In such cases non-overlapping peaks can be utilized in quantification of the crystalline fraction in the products (Siddiqui et al., 2014). Another technique that produced results similar to XRPD is ssNMR, however, it takes a very long time to collect good signal to noise spectra, especially for low dose drug products (Shah et al., 2006). This work was the extension of our previous work in which we have shown the use of XRPD in the quantification of the crystalline tacrolimus from the ASD formulations (Siddiqui et al., 2014). Our findings indicated that crystalline tacrolimus as low as 4% (when total drug content is 24.4%) can be quantitated with good precision in the ASD. However, detection and quantification limits depend upon the drug loading in the ASD formulation (Siddiqui et al., 2014). There is no literature report for quantification of crystalline fraction of tacrolimus in the commercial products by XRPD and chemometrics to the best of our knowledge. However, many investigator reported the use of XRPD in quantification of amorphous/crystalline or polymorphs fraction in the bulk drug substance (Chadha et al., 2012). Few investigators attempted to use XRPD univariate method to quantitate crystalline/amorphous fraction in the product for high dose drug or low drug to excipients ratio (higher drug percentage with respect to excipient) (Kommavarapu et al., 2013). Therefore, focus of present investigation was to study the feasibility of XRPD and chemometrics in quantification of the crystalline and amorphous fractions of tacrolimus in commercial ASD products that contained 2–5% drug. Furthermore, this manuscript provides novel methods to the sponsors of tacrolimus products that are otherwise unavailable in United States Pharmacopeia or literature.

(Shin-Etsu Chemical Co., Ltd., Chiyoda-Ku, Tokyo, Japan), lactose monohydrate NF (LM) (Foremost Farms, Baraboo, WI, USA), magnesium stearate (MGS) (Sigma–Aldrich, St. Louis, MO, USA), colloidal silicon dioxide (CSD) (Aerosil 200, Evonik Industries AG, Hanau-Wolfgang, Germany), croscarmellose sodium (CC) (FMC Biopolymer, Philadelphia, PA), ethanol 200 proof (Decon Labs, Inc., King of Prussia, PA, USA) were purchased and used as received.

2. Materials and methods

3. Results and discussion

2.1. Materials

3.1. X-ray powder diffraction

NJ,

Tacrolimus monohydrate (Ria International LLC, East Hanover, USA), hydroxypropyl methyl cellulose USP (HPMC)

2.2. Methods 2.2.1. Preparation of sustained release ASD Three ASD products were selected that contained LM, HPMC, CC, MGS and CSD as excipients. Drug to excipients ratio of the products are given in Table 1. The ASDs were prepared by solvent evaporation method as described by Zidan et al. (2012) and resultant formulations were called the amorphous tacrolimus products. Similarly, placebo ASDs were prepared in the same way and have identical composition as that of the products but without the drug. Crystalline drug was added to placebo ASDs to make their composition equal to the amorphous tacrolimus products. The resultant formulations were called crystalline tacrolimus products. The drug physical forms in the products were confirmed by XRPD. The amorphous and crystalline tacrolimus products were mixed in various ratios to make sample matrices for product A (drug: excipients 1:19 and total drug content 5%), product B (drug: excipients 1:27 and total drug content 3.6%) and product C (drug:excipients 1:49 and total drug content 2%). Furthermore, the amorphous/crystalline tacrolimus in the sample matrices of the products vary from 0 to 100%. These sample matrices of each product were used to collect XRPD data and chemometric models development for the crystalline/amorphous fraction determination in the samples. 2.2.2. X-ray powder diffraction Tacrolimus forms in the amorphous product A, product B, product C, and their corresponding crystalline products were confirmed by XRPD. Diffractograms were collected using a Bruker D8 Advance with DaVinci design (Bruker AXS, Madison, Wisconsin) equipped with the LYNXEYE scintillation detector and Cu Ka radiation (l = 1.5405 Å) at a voltage of 40 KV and current of 40 mA. Before measurement, the instrument functionality was checked using corundum as an external standard. About 650 mg of sample was placed in the sample holder and six replicate diffractograms of each sample were collected over 2u range of 4–30 with an increment of 0.00915 at 1 s per step (3000 total steps). Sample holder was rotated during run to get the average diffractogram of the sample. The XRPD operation, data collection, and data analysis were achieved through Diffract. Suite (V2.2). 2.2.3. Data analysis XRPD data were analyzed by Unscrambler X software (version 10.1, Camo Process, Oslo, Norway) for the development of chemometric models.

XRPD is shown in Fig. 1A. HPMC and CSD showed diffuse and halo diffraction patterns which are the characteristics of an

Table 1 Composition of Tacrolimus Products. Components

Product A

Product B

Product C

Drug:excipients ratio Excipients

1:19 LM, HPMC, CC, MGS and CSD

1:27 LM, HPMC, CC, MGS and CSD

1:49 LM, HPMC, CC, MGS and CSD

464

Z. Rahman et al. / International Journal of Pharmaceutics 475 (2014) 462–470

amorphous material. On the other hand, tacrolimus, LM and MGS were crystalline, which were indicated by sharp diffraction peaks. Furthermore, CC showed a broad peak at 19.90 , which was indicative of its partly crystalline nature. Tacrolimus showed major diffraction peaks at 6.45, 8.55, 10.35, 11.25, 11.75, 12.65, 13.00, 13.75, 14.15, 15.30, 16.50, 16.90, 17.20, 17.40, 18.20, 19.05, 19.40, 19.75, 19.95, 20.35, 22.00, 23.05 and 23.55 . Additionally, peaks at 8.55, 10.35 and 11.25 were most intense. LM and MGS showed peaks at 8.2, 11.85, 12.55, 14.40, 16.45, 17.11, 18.05, 19.15, 19.60, 20.05, 20.85, 21.25, 22.80, 23.20, 23.80, 25.25, 25.60, 26.20, 26.55, 26.85, 27.50, 28.25, 28.55, 29.00, 30.25 and 31.05 , and 5.10, 7.10, 8.50, 8.90, 10.15, 12.50, 13.60, 14.90, 17.00, 18.65, 21.80, 23.45, 25.20 and 30.05 , respectively. The amorphous and crystalline tacrolimus products showed the peaks of formulation components. The amorphous products showed mostly the peaks of lactose which were confirmed by the absence of crystalline tacrolimus diffraction peaks at 8.55, 10.35 and 11.25 . On the other hand, the crystalline tacrolimus products showed peaks of both crystalline drug and lactose monohydrate. The univariate analysis of three peaks of crystalline drug at 8.55, 10.35 and 11.25 was tried for quantification of crystalline drug fraction in the products by calculating peak area but did not yield good calibration plot. 3.2. Chemometrics 3.2.1. Partial least square and principle component regressions Truncated XRPD data from 8.3 to 11.4 were selected for chemometric models development due to the absence of excipients peaks and presence of the most intense crystalline tacrolimus peaks at 8.55, 10.35 and 11.25 in the products. Despite two excipients of products namely LM and MGS are crystalline, LM did not show any peaks while MGS showed three peaks at 8.50, 8.90 and 10.15 in the selected data region and these peaks might interfere with crystalline tacrolimus peaks at 8.55 (Fig. 1B). However, MGS present in the formulation was very low compared to the drug. Furthermore, signal intensity of MGS peaks was very weak compared to intensity of three peaks of the drugs in the selected data region. Moreover, non-interference of MGS peaks was confirmed by comparing the diffraction patterns of the placebo products, the amorphous and crystalline tacrolimus products. The placebo and amorphous products showed no peaks of MGS or any other components of the formulations while the crystalline tacrolimus product showed only three peaks of the

Fig. 2. X-ray powder diffractograms of amorphous solid dispersion products.

crystalline drug in the selected data region (Fig. 2). Additionally, there was a gradual increase in the peak intensity with an increase in the crystalline tacrolimus fraction present in the products (Fig. 3). However, intensity count of peaks at 8.55, 10.35 and 11.25 varied among products depending upon the drug content. For example signal to noise ratio in product A and product C (contained 100% crystalline tacrolimus) were 27.70, 26.50 and 23.03, and 18.70, 14.41 and 14.05, respectively for peaks at 8.55, 10.35 and 11.25 . Multivariate methods used were partial least squares (PLSR) and principle component (PCR) regressions (Dunn et al., 1989). Both methods suppress spectral collinearity of the data. The most critical step in the development of multivariate chemometric models is selection of optimum number of latent variables (LVs) to prevent under or over fitting of the model. The optimum LV was selected based on the minimum value of predicted residual sums of squares (PRESS) and root square error of cross validation (RMSECV). Four LV resulted in the lowest values of these parameters; however, further increase in LV did not significantly change these parameters (Fig. 4). In general, PLSR models require fewer LV than PCR models due to its better efficiency. Both

Fig. 1. (A) Full and (B) truncated X-ray powder diffractograms of individual components of amorphous solid dispersion.

Z. Rahman et al. / International Journal of Pharmaceutics 475 (2014) 462–470

465

Fig. 3. X-ray powder diffractograms of (A) product A, (B) product B and (C) product C.

methods construct new predictor variables called LV (PC or PLS) to explain data variability which are the linear combination of the original predictor variables. However, method of calculation is different in both the methods. PCR uses an algorithm to create PC without considering response variability. In contrast, PLSR does take into account response variability to create PLS. Furthermore, due to this reason, computed PC of PCR may or may not correlate with the studied physical property. In other words, there is a greater likelihood that the spectra may correlate well with the component and quantify its physical properties in the case of PLSR (Roggo et al., 2007). Internal validation or cross validation approach was used in which the same sample was used for validation of the calibration model. The amorphous/crystalline tacrolimus varies from 0 to 100% in the samples of products. Mathematical treatment of the data prior to PLSR or PCR analysis did not improve the quality of the data; attempted methods include standard normal variate, multiple scattering correction and derivatives. Mathematical treatment is usually done to enhance the searched information and reduce the influence of side information contained in the spectra. Outlier in the data was identified by Hotelling T2 test and leverage. Hotelling T2 test is used to test for outliers in the

multivariate, mean shift or other distributional deviation from the in-control distribution (Chou et al., 1999). The Hotelling T2 statistics measure the Mahalanobis distance of the sample from the mean of the distribution. Sample belongs to the distribution if sample T2 statistics values are lower than T2 statistics limit for a distribution. The samples values are lower than T2 statistics limit at 95% confidence upper limit for the entire factor selected for PLSR or PCR. Furthermore, the leverage or h-statistic of the samples was lower than 3A/I, where A is the number of LV and I is number of samples (3A/I value of product A, product B and product C were 0.129, 0.135 and 0.118, respectively) (Fig. 5). The sample will have a strong influence on the model if its leverage is greater than 3A/I. Four LV can explain more than 95% variance in the data of product A, product B, and product C for PCR and PLSR models. Furthermore, two LV accounted for 92% information in X related to about 98% information in Y for both models for all the products (Table 2). Additionally, in the PLS or PCA score plots, PCA1 and PCA2 or PLS1 and PLS2 values increased with an increase in the crystalline tacrolimus in the products. This can be explained by comparing the LVs spectra (loading plot) with the spectra of the amorphous/crystalline tacrolimus of the products. The selected data region has three peaks of crystalline tacrolimus at 8.55, 10.35, 11.25 . The first two LV of PCR and PLSR models of product A,

466

Z. Rahman et al. / International Journal of Pharmaceutics 475 (2014) 462–470

Fig. 4. Effect of number of number of factors (LVs) on RMSEP and PRESS of (A) product A, (B) product B and (C) product C.

product B and product C showed the diffractogram identical to crystalline tacrolimus. This indicated that first two LV was related to crystalline tacrolimus in the product for both models (Fig. 6). The other two LV of the PLSR and PCR models showed noisy diffractograms which may be related to the amorphous tacrolimus or other physical attributes of the products such as particle size and density etc. Calibration and validation plots (Fig. 7) were superimposable with the determination coefficient R2 > 0.987 for PLSR and PCR models for all products which further indicated the absence of

outlier in the data. Furthermore, the slopes were close to ‘1’ and intercepts were <0.716 for all the models (Table 3). Accuracy and precision of the models are measured by calculating root mean square error of prediction (RMSEP) and standard error of prediction (SEP). These parameters values should be low for high accuracy and high precision models and vice versa for low accurate and less precise models. The RMSEP and SEP values were lower for PLSR models when compared to PCR models. Furthermore, accuracy and precision of models to predict crystalline/amorphous tacrolimus fraction in product A to product C were decreasing as

Fig. 5. Leverage and residual of (A) product A, (B) product B and (C) product C.

Z. Rahman et al. / International Journal of Pharmaceutics 475 (2014) 462–470

467

Table 2 Amount of information explained by PLSR and PCR models. Product

Regression Model

No. of factor

A

PLSR

1 2 3 4 1 2 3 4

PCR

B

PLSR

PCR

C

PLSR

PCR

Variance explained by X-block

Variance explained by Y-block

Factor

Accumulated

Factor

77 18 2 1 79 16 2 1

77 95 97 98 79 95 97 98

76 23 0 0 61 38 0 0

76 99 99 99 61 99 99 99

1 2 3 4 1 2 3 4

43 50 2 0 52 43 1 1

43 93 95 95 52 95 96 97

99 0 1 0 0 99 0 0

99 99 100 100 0 99 99 99

1 2 3 4 1 2 3 4

35 57 1 3 76 16 3 1

35 92 93 96 76 92 95 96

79 19 1 0 2 96 0 1

79 98 99 99 2 98 98 99

Fig. 6. Loading plots PLS and PC factors of (A) product A, (B) product B and (C) product C.

Accumulated

468

Z. Rahman et al. / International Journal of Pharmaceutics 475 (2014) 462–470

Fig. 7. Calibration and validation plots of PLSR models for crystalline tacrolimus of (A) product A, (B) product B and (C) product C.

the drug level decreased from product A to product C. This was probably due to decrease in signal to noise ratio. Bias which is the systemic error in the model or trueness of the models is related to RMSEP and SEP by following equation (Davies and Fearn, 2006) 2

RMSEP2 ¼ Bias þ SEP2 Bias will be very low when RMSEP  SEP. Accuracy encompasses both precision and bias (trueness) according to ISO definition of accuracy (ISO, 1993). Bias in the models were low and their significance was checked by a student t-test at 95% confidence and (n  1) degree of freedom, where n is the total number of the validation sample

tbias ¼

jbiasj  SEP

pffiffiffi n

The bias was statistically insignificant (p < 0.05) for all the models for both the crystalline and amorphous tacrolimus estimation in the products. 3.2.2. Model prediction The prediction capability of the developed models was determined by estimating the crystalline tacrolimus and amorphous fractions from the products and comparing with actual values. Three samples of 10, 20 and 90% crystalline tacrolimus of product A, product B, and product C (Table 1) were prepared by the

Table 3 Statistical parameters of chemometric models. Slope

Offset

Correlation

R2

RMSEP

SEP

Bias

93 93 93 93 93 93 93 93

0.995 0.994 0.995 0.994 0.992 0.989 0.992 0.989

0.281 0.363 0.188 0.282 0.474 0.595 0.318 0.455

0.998 0.997 0.998 0.997 0.996 0.995 0.996 0.995

0.995 0.994 0.995 0.994 0.992 0.991 0.992 0.991

1.736 1.966 1.736 1.966 2.256 2.413 2.256 2.413

1.745 1.976 1.745 1.976 2.268 2.426 2.268 2.426

– 0.023 – 0.023 – 0.033 – 0.034

Calibration Validation Calibration Validation Calibration Validation Calibration Validation

89 89 89 89 89 89 89 89

0.998 0.996 0.998 0.996 0.994 0.991 0.994 0.991

0.092 0.244 0.095 0.160 0.297 0.406 0.305 0.434

0.999 0.998 0.999 0.998 0.997 0.997 0.997 0.997

0.998 0.997 0.998 0.997 0.994 0.993 0.994 0.993

1.330 1.835 1.330 1.835 2.384 2.541 2.384 2.541

1.338 1.845 1.338 1.845 2.397 2.556 2.397 2.556



Calibration Validation Calibration Validation Calibration Validation Calibration Validation

102 102 102 102 102 102 102 102

0.992 0.992 0.992 0.992 0.988 0.988 0.988 0.988

0.436 0.540 0.361 0.295 0.650 0.716 0.538 0.516

0.996 0.995 0.996 0.995 0.994 0.993 0.994 0.993

0.992 0.990 0.992 0.990 0.988 0.987 0.988 0.987

2.818 3.206 2.818 3.206 3.441 3.635 3.441 3.635

2.832 3.221 2.832 3.221 3.458 3.652 3.458 3.652



Product

Regression Model

Component

Model

A

PLS

Crystalline

Calibration Validation Calibration Validation Calibration Validation Calibration Validation

Amorphous PCR

Crystalline Amorphous

B

PLS

Crystalline Amorphous

PCR

Crystalline Amorphous

C

PLS

Crystalline Amorphous

PCR

Crystalline Amorphous

Sample no.

0.045 – 0.045 – 0.008 – 0.008

0.083 – 0.083 – 0.042 – 0.042

Z. Rahman et al. / International Journal of Pharmaceutics 475 (2014) 462–470

469

Table 4 Crystalline and amorphous tacrolimus fraction prediction results from the products. Product

A

Regression Model

PLSR

PCR

B

PLSR

PCR

C

PLSR

PCR

Theoretical value

Predicted value

Crystalline (%)

Amorphous (%)

Crystalline (%)

Amorphous (%)

10 20 90 10 20 90

90 80 10 90 80 10

15.37  1.63 19.12  1.85 88.01  1.85 15.05  2.06 20.63  2.48 87.96  2.24

84.63  1.63 80.88  1.85 11.99  1.85 84.95  2.06 79.37  2.48 12.04  2.24

10 20 90 10 20 90

90 80 10 90 80 10

8.07  1.71 24.22  1.65 87.62  2.83 6.98  2.39 22.33  2.40 78.42  4.67

91.93  1.71 75.78  1.65 12.38  2.83 93.02  2.39 77.67  2.40 21.58  4.67

10 20 90 10 20 90

90 80 10 90 80 10

13.51  2.86 19.98  2.49 91.74  3.54 14.68  3.50 20.62  2.98 91.22  4.19

86.49  2.86 80.02  2.49 8.26  3.54 85.32  3.50 79.38  2.98 8.78  4.19

same procedure as described in Section 2.2. These samples were independent in the sense that they were not used to develop the chemometric models. The models predicted values are given in the Table 4 and the predicted values were in close agreement with actual values especially at 20% and 90% crystalline tacrolimus. However, prediction values were not good for 10% crystalline fraction samples possibly due to low signal to noise ratio. Additionally, predicted values by PLSR or PCR model were very similar with the exception of higher error in the PCR predicted values which were in-line with the higher values of RMSEP and SEP in those models. Furthermore, error in the results increased as the drug proportion in the sample decrease. For example, the sample of product A has less error compared to product B and product C. This is in fact related to higher signal to noise ratio in product A samples due to higher drug content. 4. Conclusion It is very important to have good in-vitro quality control tests in place to control CQAs of the drug products so that consistent therapeutic response especially for low dose and narrow therapeutic drugs such as tacrolimus can be achieved. Importance of in-vitro tests (especially, discriminating dissolution and quantitative method for crystalline fraction determination of the drug in the ASD products) further increase for thermodynamically unstable formulations such as ASD as these products change the CQAs of the product during usage or storage which are reflected in the clinical outcome of the product (Newman et al., 2012). The change in CQAs of the ASD can be quantitated by measuring the fraction of amorphous to crystalline drug in the product. The PLSR and PCR models prediction of crystalline/amorphous in the drug products were similar except high error in the PCR models predicted values. Furthermore, the developed chemometric models can quantitate up to 10% of crystalline tacrolimus when drug to excipients ratio is as high as 1:49. The developed models have reasonable accuracy and low error. The clinical performance of the tacrolimus products can be predicted by knowing the amorphous to crystalline fraction, comparing it to the initial value when the product was manufactured or the value of the product used in the bioequivalence studies.

References Baird, J.A., Taylor, L.S., 2012. Evaluation of amorphous solid dispersion properties using thermal analysis techniques. Adv. Drug Deliv. Rev. 64, 396–421. Chadha, R., Arora, P., Bhandari, S., 2012. Polymorphic forms of lamivudine: characterization, estimation of transition temperature, and stability studies by thermodynamic and spectroscopic studies. ISRN Thermodyn Article ID 671027, 8 pages. Chokshi, R.J., Zia, H., Sandhu, H.K., Shah, N.H., Malick, W.A., 2007. Improving the dissolution rate of poorly water soluble drug by solid dispersion and solid solution: pros and cons. Drug Deliv. 14, 33–45. Chou, Y.M., Mason, R.L., Young, J.C., 1999. Power comparisons for a Hotelling’s T2 statistic. Commun. Stat. Simul. Comput. 28, 1031–1050. Davies, A.M.C., Fearn, T., 2006. Back to basics: calibration statistics. Spectrosc. Eur. 18, 31–32. Dunn III, W.J., Scott, D.R., Glen, W.G., 1989. Principal components analysis and partial least squares regression. Tetrahedron Comput. Methodol. 2, 349–376. FDA, 2014. Recalls, market withdrawals and safety alerts http://www.fda.gov/ safety/recalls/enforcementreports/ucm310739.htm (accessed 08.05.14). Hancock, B.C., Shamblin, S.L., Zografi, G., 1995. Molecular mobility of amorphous pharmaceutical solids below their glass-transition temperatures. Pharm. Res. 12, 799–806. ISO 3534-1, 1993. Statistics – Vocabulary and Symbols – Part 1: Probability and General Statistical Terms, International Organization for Standardization, Geneva. Janssens, S., Van den Mooter, G., 2009. Review: physical chemistry of solid dispersions. J. Pharmacol. 61, 1571–1586. Kommavarapu, P., Maruthapillai, A., Allada, R., Palanisamy, K., Chappa, P., 2013. Simultaneous estimation of degree of crystallinity in combination drug product of abacavir, lamivudine and neverapine using X-ray powder diffraction technique. J. Young Pharm. 5, 127–132. Newman, A., Knipp, G., Zografi, G., 2012. Assessing the performance of amorphous solid dispersions. J. Pharm. Sci. 101, 1355–1377. Rahman, Z., Siddiqui, A., Khan, M.A., 2013. Characterization of a nonribosomal peptide antibiotic solid dispersion formulation by process analytical technologies sensors. J. Pharm. Sci. 102, 4337–4346. Roggo, Y., Chalus, P., Maurer, L., Lema-Martinez, C., Edmond, A., Jent, N., 2007. A review of near infrared spectroscopy and chemometrics in pharmaceutical technologies. J. Pharm. Biomed. Anal. 44, 683–700. Rumondor, A.C., Wikström, H., Van Eerdenbrugh, B., Taylor, L.S., 2011. Understanding the tendency of amorphous solid dispersions to undergo amorphous– amorphous phase separation in the presence of absorbed moisture. AAPS PharmSciTech 12, 1209–1219. Shah, B., Kakumanu, V.K., Bansal, A.K., 2006. Analytical techniques for quantification of amorphous/crystalline phases in pharmaceutical solids. J. Pharm. Sci. 95, 1641–1665. Siddiqui, A., Rahman, Z., Bykadi, S., Khan, M.A., 2014. Chemometric methods for the quantification of crystalline tacrolimus in solid dispersion by powder X-ray diffractrometry. J. Pharm. Sci. Sinclair, W., Leane, M., Clarke, G., Dennis, A., Tobyn, M., Timmins, P., 2011. Physical stability and recrystallization kinetics of amorphous ibipinabant drug product by fourier transform Raman spectroscopy. J. Pharm. Sci. 100, 4687–4699.

470

Z. Rahman et al. / International Journal of Pharmaceutics 475 (2014) 462–470

Wu, J.X., Berg, F.V.D., Sogaard, S.V., Rantanen, J., 2013. Fast track to a solid dispersion formulation using multiway analysis of complex interactions. J. Pharm. Sci. 102, 904–914. Yoo, S.U., Krill, S.L., Wang, Z., Telang, C., 2009. Miscibility/stability considerations in binary solid dispersion systems composed of functional excipients towards the design of multi-component amorphous systems. J. Pharm. Sci. 98, 4711–4723.

Yoshioka, S., Aso, Y., 2007. Correlations between molecular mobility and chemical stability during storage of amorphous pharmaceuticals. J. Pharm. Sci. 96, 960– 981. Zidan, A.S., Rahman, Z., Sayeed, V., Raw, A., Yu, L., Khan, M.A., 2012. Crystallinity evaluation of tacrolimus solid dispersions by chemometric analysis. Int. J. Pharm. 423, 341–350.