Bilayer Tablet Dissolution Kinetics Based on a Degassing Cyclic Flow UV-Vis Spectroscopy with Chemometrics

Abstract

Dissolution kinetics of a bilayer direct compress tablet was evaluated by using degassing cyclic flow UV-visible (Vis) spectroscopy with chemometrics. The model bilayer nicotinamide (NA)–pyridoxine hydrochloride (PH) 100.0 mg tablets were prepared via the dual compress method. The fast diffusion layer of the bilayer tablet contained nicotinamide, microcrystal cellulose, beta-lactose, magnesium stearate, and croscarmellose sodium. The slow release layer contained pyridoxine hydrochloride and carnauba wax. The monolayer direct compress tablets were prepared as dual ingredient (API)s formulation tablets. The degassing cyclic flow UV-Vis spectroscopy dissolution test was carried out using the prepared tablets. The dissolution test conditions were as follows: time, 60 min; temperature, 37°C; paddle method, 50 rpm, and UV-Vis spectra measurement 1 time/min. The UV-Vis spectra of the flow solution were measured in the range of 240–380 nm. API concentration was predicted by partial least squares (PLS) regression models based on UV-Vis spectra. The dissolution kinetics of the bilayer and monolayer tablets were evaluated based on the UV-Vis spectra with the predicted API concentration profile. The degassing flow system could prevent air bubbles in the flow cell at 1800 min. Therefore, simultaneous determination of NA and PH concentration based on the PLS regression was suggested to have high accuracy. PLS regression has advantages over the conventional λ_max absorbance method of simultaneous determination. We found that the kinetics of the separated bilayer tablet can be evaluated by the same kinetic analysis method used for the single layer model tablet.

Introduction

Sustained drug delivery systems can reduce the frequency of therapeutic dosing.^1–3) Reducing the administration frequency increases the value of pharmaceutical products, which can thereby improve the quality of life of patients.^4–6) Dissolution testing of pharmaceutical tablets is one of the important in vitro tests for evaluating pharmaceutical preparations.⁷⁾ Bilayer compressed tablets composed of two matrix layers are widely employed for controlled drug release⁸⁾; in these, the dissolution kinetics can be easily controlled by adjusting the matrix ingredients.⁹⁾

UV-visible (Vis) spectroscopy or HPLC are widely used to evaluate dissolution kinetics via batch tests.^10–12) An HPLC batch test requires more sampling time, increased costs per sample, and a lot of organic solvents. Further, the time taken for batch analysis can be so long that it could limit their use in monitoring broth component concentrations during the process. The samples need to be analyzed directly. The dissolution testing system is desirable to be a circulating, high-speed, automatic, low-cost system without organic solvents. It is known that the flow system can contain air bubbles (see Supplementary Fig. S1). The presence of air bubbles in the flow cell will affect the optical path. The degassing cyclic flow dissolution tester with a wide-range UV-Vis spectra is suitable for pharmaceutical dissolution testing.

However, the UV-Vis spectra of over two kinds of active pharmaceutical ingredients (API) show overwrapped peaks.¹³⁾ It is difficult to determine API concentrations based on direct calculations, by using the absorbance integral method or absorbance values. The chemometrics calculations can quantitate such complex UV-Vis,¹⁴⁾ Raman,¹⁵⁾ near-IR,¹⁶⁾ terahertz,¹⁷⁾ or IR spectra¹⁸⁾ data from mathematic matrix calculations. Some common multivariate calibration methods are multiple linear regression,¹⁹⁾ principal component regression,²⁰⁾ and partial least square (PLS).²¹⁾ Andrade et al.²²⁾ reported the dissolution kinetics of a mebendazole–cambendazole double-content monolayer tablet using UV-Vis spectroscopy.

In a previous study,²³⁾ we predicted the pharmaceutical properties of a direct compress tablet using Near-IR spectroscopy with PLS regression. The constructed PLS regression models suggested pseudo polymorphic transformation of theophylline. The metastable calcium phosphate phase transformation and pseudo polymorphic transformation of carbamazepine were investigated based on IR spectra with multi-variate analysis.^24,25) The simultaneous determination in API formulation powders and co-crystals was carried out using PLS regression models based on powder X-ray diffractograms.^26,27)

In this study, we applied our newly developed degassing flow UV-Vis spectroscopy system for dissolution testing. We used a bilayer formulation tablet and chemometric approaches to evaluate the properties of bilayer tablet dissolution. The effect of a wax matrix on bilayer tablet dissolution kinetics and the dissolution mechanism were evaluated by using a flow UV-Vis spectroscopy dissolution test system.

Experimental

Materials

Nicotinamide (NA) and Pyridoxine Hydrochloride (PH) were purchased from Wako Pure Chemical Industries, Ltd. (Japan). Polishing Wax-105 of carnauba wax (CWax) for tablet matrix was supplied by Freund Co. (Japan). KICCOLATE™ ND-200 of Croscarmellose sodium (CCS) and PH-102 of micro-crystalline cellulose (MCC) were supplied by Asahi Kasei Chemicals Co. (Japan). Lactose (LC) DCL-21 was obtained from DMV Fonterra excipients (Germany). Zartorius Arium 611 DI grade deionized water was used throughout.

Bilayer Direct Compress Tablet

The formulations of the tablets are listed in Table 1. Each bulk powder was ground using a mortar and pestle. The prepared powder was compressed using a 7.0-mm-diameter punch with a flat surface with a handy oil compressor (P-16B, RIKEN KIKI Co., Japan). Bilayer NA-PH 100.0 mg tablets of three types (named A, B and C group) were prepared by using the double compress method. During the first direct compress process, the mixture powder for the slow release wax layer was compressed at 10 kg/cm². After compressing, the mixture powder for fast release layer was inserted, and compressed to 30 kg/cm². The monolayer tablets (named D group) were prepared at 30 kg/cm².

Table 1. Formulation of the Bilayer and Monolayer 100.0 mg Tablets

Tablet group Layer Contents	Bilayer tablet						Monolayer tablet
	A		B		C		D
	Wax	FD	Wax	FD	Wax	FD	Mono
	Formulations (mg)
NA	0	10.0	0	10.0	0	10.0	10.0
PH	5.0	0	5.0	0	5.0	0	5.0
CWax	5.0	0	15	0	44.5	0	0
CSS	0	0.5	0	0.5	0	0.5	1.0
MCC	27.65	27.3	20.65	27.3	0	27.3	58.10
LC	11.85	11.7	8.85	11.7	0	11.7	24.90
MgST	0.5	0.5	0.5	0.5	0.5	0.5	1.0
Total	50.0	50.0	50.0	50.0	50.0	50.0	100.0

Wax; wax matrix layer, FD; fast dissolution layer; Total; total tablet weight of layer.

Degassing Cyclic Flow UV-Vis Spectroscopy Dissolution Test Systems

Figure 1 shows the degassing flow UV-Vis spectroscopy dissolution test system used in the present study. Dissolution behaviors were investigated using an automatic tablet dissolution tester (NTR-VS6P, Toyama Sangyo Co., Ltd., Japan). The dissolution test was performed by the paddle method at 50 rotations per minute in 900.0 mL of water at 37 ± 0.5°C with flow UV-Vis spectroscopy system.

Fig. 1. The Degassing Cyclic Flow UV-Vis Spectroscopy Dissolution Test System

DT; Automatic tablet dissolution tester, F; Metal filter, T; Degassing tube, P; Peristaltic pump, D; UV-Vis spectrophotometer, PC; Personal Computer, Flow rate; 1.25 mL/min.

The peristaltic pump (Rainin Dinamax RP-1, U.S.A.) was used for delivering solutions. PharMed tubing (0.5 mm i.d.) was used as pump tube and tube jointer. The clogging prevention metal filter was attached on suction port. The TB-0201 poreflon^® tube with 1 mm i.d. (Lot No. 100001, Sumitomo Electric Fine Polymer, Inc., Japan) of 5 cm was joined for degassing after metal filter.

Partial Least Squares Based on UV-Vis Spectra

The PLS regression models were applied to predict API concentration based on UV-Vis spectra of the sample solutions with the Unscrambler version 10.3. The details of the PLS regression model are described elsewhere.^28–31) Baseline optimization was adopted to prevent baseline shift of spectra. The software was run on TOSHIBA dynabook R734/W3K Intel core i7-4700 MQ, 2.4 GHz and 8 GB RAM laptop computer. The PLS regression models for NA and PH were constructed for predictive determination based on spectra of solutions for calibration. The root mean square error (RMSE) was calculated according to the error of each calibration model by using the following equation:

  

(1)

where ŷ_i is the predicted concentration of API, y_i the measured value, and n the number of samples. The negative values of the predicted API concentration were regarded as zero.

Results and Discussion

Calibration UV-Vis Spectra of Active Pharmaceutical Ingredients

The measured UV-Vis spectra of NA and PH solutions for the calibration models are shown in Fig. 2. The λ_max of both solutions showed concentration linear relationships obeying Beer’s law. The mixture solutions had absorbance in the range of 240–360 nm. The λ_max of the NA solutions was at 262 nm. The PH solutions had absorbance peaks at 290 nm. Machida and Nagai reported³²⁾ an in vitro dissolution test of PH for oral controlled release with hydroxypropyl cellulose. The dissolution kinetics of API was evaluated based on the time required for 50% dissolution of API at 292 nm. NA and PH mixture samples show overlapping of bands in the range of 240–290 nm; therefore, it is difficult to determine their API concentrations by using conventional λ_max calculations.

Fig. 2. UV-Vis Spectra of the NA and PH Solutions for Calibration Models

(Color figure can be accessed in the online version.)

Supplementary Fig. S2 shows the relationship between λ_max absorbance of PH and NA and their concentrations. For PH, there was a linear relationship between absorbance and concentration. In contrast, for NA, there was a non-linear relationship between absorbance and concentration.

UV-Vis Spectra of Prepared Tablet Dissolution Test of Bilayer NA-PH Tablets

Figures 3(a)–(d) show degassing dissolution test UV-Vis spectra of bilayer of A, B and C tablets and monolayer D tablet. The absorbance ranging from 240 to 360 nm increased with API dissolution. The increased spectral patterns agree a with the bulk material solution spectra shown in Fig. 2. The tested all bilayer tablets immediately separated each layer within 5 s. The disintegrated layer dispersed immediately in the solution within 60 s. This rapid dissolution was due to CCS of the disintegrated layer contents. CCS are widely used as super disintegrate for formulation tablet.^33–35) Ferero et al. reported³⁶⁾ CCS for solubilization in a direct compression formulation. PH absorbance increase speed of the tablet of dissolution of C group is slow. The tablet include sustained release matrix layer contain CWax 90%. The API-controlled release is considered to be attributed to the carnauba wax. Huang et al. reported³⁷⁾ a controlled release tablet with an acrylic polymer-wax matrix. They found that CWax contents affected the dissolution speed due to the cross linking of the matrix excipients. It was found that NA absorbance increase speed of disintegrate layer (240–280 nm) was faster than PH in all tablet groups.

Fig. 3. UV-Vis Spectra of the Bilayer and Monolayer Tablet Based on the Dissolution Test; Bilayer Group A Tablet (a), Group B Tablet (b), Group C (c), and Monolayer Tablet Group D (d)

The side color legend shows the dissolution times. (Color figure can be accessed in the online version.)

Machine Learning Prediction Accuracy and Constructed Model Analysis

PSL Model Construction to Predict API Concentrations for Dissolution Kinetics

The PLS regression models were constructed to predict both API concentration and determine the dissolution kinetics. The constructed model analysis is important to guarantee high prediction accuracy and predictive reliability properties. Table 2 listed cumulative percentage variances of the constructed PLS regression models of constructed PLS models. Principal component (PC) number is one of important factors to improve prediction accuracy. The variance in the UV-Vis spectra matrix X and the API concentration matrix Y at PC2 was 99.31 and 99.81, respectively. It is reasonable to select PC2 as an optimum PC number based on the cumulative percentage variance increase. In addition, the variance of absorbance seemed to relate to the chemical composition of NA and PH molecules. It was suggested that excipients (i.e., CCS, MCC, LC, CWax, and magnesium stearate) were present in very low concentrations to affect the UV-Vis spectra. Loadings are shown in Supplementary Fig. S3. PC1 and PC2 were found to be due to the NA and PH absorbance peaks, respectively.

Table 2. The Cumulative Percent Variance of the Calibration Model Based on the Constructed PLS Models

PC	Cumulative percent variance
	X: UV spectra	Y: Concentration of APIs
PC1	62.49	96.07
PC2	99.31	99.81
PC3	99.99	99.88
PC4	100.00	99.98

The predictive parameters of the constructed PLS regression models for NA and PH concentration are listed in Table 3. The RMSE for prediction (RMSEP), SEP, tendency, and relative values of the model properties at PC2 are listed. The constructed PLS regression models at PC2 for NA and PH have a high ability of prediction.

Table 3. The Predictive Parameters of the Constructed Model for NA and PH Concentration

	The parameters of constructed PLS models
	RMSEP	SEP	a	b	Coefficience
NA PC1	0.00683	0.00691	0.0936	3.4E-03	0.254
NA PC2	0.00017	0.00017	0.9996	−8.2E-08	1.000
PH PC1	0.00223	0.00224	0.9923	4.5E-04	0.998
PH PC2	0.00156	0.00158	0.9996	1.6E-05	0.999

Regression Vector of the PLS Regression Models to Predict NA and PH Concentrations

Figure 4 shows the regression vector (RV) for NA concentration (black line) and PH concentration (gray line). The RV show correlation points between X (spectra data) and Y (predictive concentration) on the constructed PLS regression model. RVs for NA and PH have a pronounced peak at 264 nm and 292 nm, respectively. The UV-Vis spectra of the PH solution agrees with the other experimental spectra data.³²⁾

Fig. 4. Regression Vectors of Constructed PLS Regression Models for NA and PH

Predicted API Concentration and the Bilayer Tablet Dissolution Kinetic Analysis

Figures 5(a) and (b) shows predicted dissolution profiles of direct compressed bilayer tablet based on UV-Vis spectra with previous constructed PLS regression models on PC2. The UV-Vis spectra of dissolution test was applied to the PLS regression models for NA (Fig. 5(a)). The dispersion of the tablet component due to the high disintegration property of CCS was found in the dissolution profiles. The predicted NA and PH concentrations were plotted in the figure. NA concentrations of the bilayer tablet (group A, B and C) immediately increased by 0.01 mM within 8 min; the NA concentrations of the monolayer tablet (group D) required 20 min to exceeded by 0.01 mM. Figure 5(b) shows the predicted PH concentration based on the constructed PLS regression models for PH concentration on PC2. The faster release was observed in monolayer tablet (group D). The containing carnauba raw bilayer tablet (group A, B and C) had slower dissolution ratios than monolayer tablet. It was under 0.003 mM concentration at 60 min in the 90% containing CWax tablet (group C tablet). The dissolution profiles of group C tablet suggested that containing wax matrix construct the crosslink structure in the layer. Miyagawa et al. investigated³⁸⁾ the mechanical strength and dissolution profiles of the diclofenac with wax matrix granules tablet. They discussed three types of dissolution mechanisms: swelling-insoluble, swelling-soluble, and non-swelling-soluble model. In our dissolution profile, the dissolution mechanism of the tablet group A and B can be assigned to the swelling soluble model. The tablet group C was classified in non-swelling-soluble model.

Fig. 5. The Predicted Concentrations in the Degassing Cyclic Flow UV-Vis Spectroscopy Dissolution Test System

(Color figure can be accessed in the online version.)

Table 4 shows the NA dissolutions profiles of a bilayer and monolayer tablet. The dissolution profile correlation coefficients were calculated based on zero order, 1st order, and Higuchi models. The Higuchi model³⁹⁾ was fitted at a concentration range of 5% < C < 80%, because the model is a sink model. The highest coefficient values of tablet group A, B, and C of fast dissolution layer of NA containing CCS were obtained in zero order, zero order, and Higuchi model. The most suitable model for the monolayer D tablet was the Higuchi model. The monolayer D group tablet of PH dissolution kinetics were listed in Table 5. The slow release layer containing PH and monolayer followed Higuchi model.

Table 4. Model Fitting of NA Dissolution Profiles for Kinetic Evaluation

Tablet group	Model fittings for NA dissolution profile
	Equations
	Zero	1st	Higuchi
A	0.9361	0.7093	0.9298
B	0.973	0.7587	0.951
C	0.9244	0.7479	0.9875
D	0.9423	0.8361	0.9915

Table 5. Dissolution Kinetics Evaluation Based on the Predicted PH Concentration Profile

Tablet group	Higuchi model fitting on PH dissolution profiles
	A	B	C	D
a	0.5264	0.2566	0.2764	N/A
b	−0.9874	−0.448	−0.5305	N/A
R2	0.9868	0.9946	0.9984	N/A
t (5% < C)	4	3	4	3
t (C < 80%)	12	25	24	N/A

Khaled et al.⁴⁰⁾ reported a 3D bilayer tablet containing an immediate release layer and a sustained release layer. A 3D lamination method that could control tablet dissolution was also reported. Kulkarni and Bhatia⁴¹⁾ reported a bilayer floating tablet comprised of atenolol and lovastatin with a hydroxypropyl methyl cellulose matrix. The optimized bilayer tablet had a flow of 8 h in the stomach. The bilayer tablet system could be a very useful controlled release system. This system can be developed to obtain better and efficacious tablets in the future.

Conclusion

In this study, we demonstrated that bilayer tablets can elicit dual functions in a single tablet and predicted API concentrations based on degassing flow UV-Vis spectra and PLS regression models. The regression vector of the constructed PLS regression models showed API-specific UV-Vis spectral peaks. The dissolution kinetics were evaluated based on the predicted NA and PH concentrations. The system can eliminate cumbersome manual effort and gassing cell window problem. The degassing cyclic flow UV-Vis spectroscopy could be applied to reduce dissolution test cost, increase sample measurement time-resolution, and prevent air contamination in the flow tube.

Acknowledgments

The authors thank Prof. Purnendu K. Dasgupta; Chemistry and Biochemistry, University of Texas at Arlington, Mr. Shimizu Shota; Faculty of Pharmaceutical Sciences, Tokyo University of Science, and Prof. Kenichi Hamada; Institute of Biomedical Sciences, Tokushima University for their help in interpreting the significance of the results of this study. This study was partly supported by JSPS KAKENHI, Grant number JP18H0615.

Conflict of Interest

The authors declare no conflict of interest.

Supplementary Materials

The online version of this article contains supplementary materials.

References

• 1) Arora S., Ali J., Ahuja A., Khar R. K., Baboota S., AAPS PharmSciTech, 6, E372–E390 (2005).
• 2) Singh B., Chakkal S. K., Ahuja N., AAPS PharmSciTech, 7, E19–E28 (2006).
• 3) Tahara K., Yamamoto K., Nishihata T., J. Control. Release, 35, 59–66 (1995).
• 4) Tiwari S. B., Murthy T. K., Rayeendra Pai M., Mehta P. R., Chowdary P. B., AAPS PharmSciTech, 4, 18–23 (2003).
• 5) Jagdale S. C., Agavekar A. J., Pandya S. V., Kuchekar B. S., Chabukswar A. R., AAPS PharmSciTech, 10, 1071–1079 (2009).
• 6) Boyapally H., Nukala R. K., Douroumis D., Drug Deliv., 16, 67–74 (2009).
• 7) Wang L., Asgharnejad M., J. Pharm. Biomed. Anal., 21, 1243–1248 (2000).
• 8) Singh S. D., J. Appl. Pharm. Sci., 1, 43–47 (2011).
• 9) Nirmal J., Saisivam S., Peddanna C., Muralidharan S., Godwinkumar S., Nagarajan M., Chem. Pharm. Bull., 56, 1455–1458 (2008).
• 10) Van Eerdenbrugh B., Froyen L., Van Humbeeck J., Martens J. A., Augustijns P., Van den Mooter G., Eur. J. Pharm. Sci., 35, 127–135 (2008).
• 11) Shekunov B. Y., Chattopadhyay P., Seitzinger J., Huff R., Pharm. Res., 23, 196–204 (2006).
• 12) Arima H., Yunomae K., Miyake K., Irie T., Hirayama F., Uekama K., J. Pharm. Sci., 90, 690–701 (2001).
• 13) Azzouz T., Tauler R., Talanta, 74, 1201–1210 (2008).
• 14) Dinç E., Serin C., Tuğcu-Demiröz F., Doğanay T., Int. J. Pharm., 250, 339–350 (2003).
• 15) Szostak R., Mazurek S., Analyst, 127, 144–148 (2002).
• 16) Blanco M., Eustaquio A., Gonzalez J. M., Serrano D., J. Pharm. Biomed. Anal., 22, 139–148 (2000).
• 17) Takeuchi I., Otsuki M., Kuroda H., Makino K., J. Drug Deliv. Sci. Technol., 45, 70–75 (2018).
• 18) Calvo N. L., Kaufman T. S., Maggio R. M., J. Pharm. Biomed. Anal., 107, 419–425 (2015).
• 19) García C., Arana G., Etxebarria N., Fernández L. A., Microchim. Acta, 136, 15–21 (2001).
• 20) Sivakesava S., Irudayaraj J., Ali D., Process Biochem., 37, 371–378 (2001).
• 21) Armenta S., Garrigues S., De la Guardia M., Anal. Chim. Acta, 596, 330–337 (2007).
• 22) Andrade S. R., Scarminio I. S., Nery M. M., de Oliveira A. C., J. Pharm. Biomed. Anal., 33, 655–665 (2003).
• 23) Otsuka Y., Yamamoto M., Tanaka H., Otsuka M., Biomed. Mater. Eng., 25, 223–236 (2015).
• 24) Otsuka Y., Takeuchi M., Otsuka M., Ben-Nissan B., Grossin D., Tanaka H., Colloid Polym. Sci., 293, 2781–2788 (2015).
• 25) Otsuka Y., Ito A., Matsumura S., Takeuchi M., Tanaka H., Colloid Polym. Sci., 293, 3471–3478 (2015).
• 26) Otsuka Y., Ito A., Takeuchi M., Tanaka H., J. Pharm. Sci., 106, 3458–3464 (2017).
• 27) Otsuka Y., Ito A., Matsumura S., Takeuchi M., Pal S., Tanaka H., Chem. Pharm. Bull., 64, 1129–1135 (2016).
• 28) Tsuchikawa S., Kobori H., J. Wood Sci., 61, 213–220 (2015).
• 29) Nascimento P. A. M., de Carvalho L. C., Júnior L. C. C., Pereira F. M. V., de Almeida Teixeira G. H., Postharvest Biol. Technol., 111, 345–351 (2016).
• 30) Quintelas C., Mesquita D. P., Ferreira E. C., Amaral A. L., Talanta, 194, 507–513 (2019).
• 31) Kirchler C. G., Pezzei C. K., Beć K. B., Mayr S., Ishigaki M., Ozaki Y., Huck C. W., Analyst, 142, 455–464 (2017).
• 32) Machida Y., Nagai T., Chem. Pharm. Bull., 28, 1082–1089 (1980).
• 33) Matji A., Carvajal L., Conde F., Peña M. A., Donato N., Serrano D. R., Torrado J. J., J. Drug Deliv. Sci. Technol., 42, 237–244 (2017).
• 34) Bindra D. S., Stein D., Pandey P., Barbour N., Pharm. Dev. Technol., 19, 285–289 (2014).
• 35) Battu S. K., Repka M. A., Majumdar S., Rao Y., M. Drug Dev. Ind. Pharm., 33, 1225–1232 (2007).
• 36) Ferrero C., Munoz N., Velasco M. V., Muñoz-Ruiz A., Jiménez-Castellanos R., Int. J. Pharm., 147, 11–21 (1997).
• 37) Huang H. P., Mehta S. C., Radebaugh G. W., Fawzi M. B., J. Pharm. Sci., 83, 795–797 (1994).
• 38) Miyagawa Y., Okabe T., Yamaguchi Y., Miyajima M., Sato H., Sunada H., Int. J. Pharm., 138, 215–224 (1996).
• 39) Higuchi T., J. Pharm. Sci., 50, 874–875 (1961).
• 40) Khaled S. A., Burley J. C., Alexander M. R., Roberts C. J., Int. J. Pharm., 461, 105–111 (2014).
• 41) Kulkarni A., Bhatia M., Iran J. Pharm. Res., 8, 15–25 (2010).