A Drug Release Study.pdf

A Drug Release Study From Hydroxypropylmethylcellulose(HPMC) Matrices Using QSPR Modeling

TARAVAT GAFOURIAN,1,2 AREZOO SAFARI,1 KHOSRO ADIBKIA,3 FATEMEH PARVIZ,1 ALI NOKHODCHI1,2

1Drug Applied Research Center and School of Pharmacy, Tabriz University of Medical Sciences,Daneshgah Street, Tabriz 51664, Iran

2Medway School of Pharmacy, Universities of Kent and Greenwich, Chatham, Maritimes, Kent ME4 4TB, England

3School of Pharmacy, Zanjan University of Medical Sciences, Zanjan, Iran

Received 20 June 2006; revised 15 February 2007; accepted 22 February 2007

Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.20990

Corresponde1634-883846; FaE-mail: t.ghafou

Journal of Pharm

� 2007 Wiley-Liss

3334 JOURN

ABSTRACT: This investigation is aimed at characterization of the mode of release fromtwo different substitution types of HPMC and the effect of chemical structure of drugsusing the QSPR (Quantitative - Structure–Property Relationship) technique. To thisend, release profiles of HPMC matrices of several drugs containing the same formulationand compressed at a constant pressure were studied. QSPR method was used toestablish statistically significant relationships between release parameters and thestructural descriptors. Structural descriptors consisted of molecular mechanical, quan-tum mechanical and graph-theoretical parameters, as well as the partition coefficientand the aqueous solubility of the drugs. The results showed that the most importantfactors determining the release profile from both HPMC K4M and HPMC E4M matriceswere the aqueous solubility of drugs (which could be substituted efficiently by dipolemoment) and the size of the drug molecules. Comparison of drug release from matricesprepared using the two grades of HPMC showed very distinct differences for some drugs,as evaluated by the similarity factor. The results indicated that the source of thedifference could be sought in the drug properties (as exemplified by the aqueoussolubility and surface area) as well as the rate of erosion (that depends mainly onthe polymer type). � 2007 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm

Sci 96:3334–3351, 2007

Keywords: matrix; release rate; HPMC
; QSAR; QSPR
INTRODUCTION

A method of obtaining a sustained-release productis to embed or disperse the solid medicinalcompound in an insoluble matrix by compressionof a physical mixture of the compound and apolymeric material.1 This has attracted consider-able attention and has been described by several

nce to: Taravat Gafourian (Telephone: 44-x: 44-1634-883927;[email protected]; [email protected])

aceutical Sciences, Vol. 96, 3334–3351 (2007)

, Inc. and the American Pharmacists Association

AL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DE

researchers.2–5 Matrix tablets have long beenused to obtain sustained drug delivery and it wasHiguchi who first presented a detailed mathema-tical analysis of this release.6

Hyroxypropylmethylcellulose (HPMC) is one ofthe most widely used polymers in the preparationof oral controlled drug delivery systems. Toachieve controlled release through the use of awater-soluble polymer such as HPMC, the poly-mer must quickly hydrate on the outer tablet skinto form a gelatinous layer. A rapid formation of agelatinous layer is critical to prevent wetting ofthe interior and disintegration of the tablet core.Once the protective gel layer is formed, it controls

CEMBER 2007

A DRUG RELEASE STUDY FROM HPMC MATRICES 3335

the penetration of additional water into the tablet.As the outer gel layer fully hydrates and dissolves,a new inner layer must replace it and be cohesiveand continuous enough to retard the influx ofwater and control drug diffusion. Hydroxypropylmethylcellulose (HPMC) products vary chemi-cally and physically. The major chemical dif-ferences are in degree of methoxyl substitution,moles of hydroxypropoxyl substitution (MS),and degree of polymerization. Varying ratios ofhydroxypropyl and methyl substitution in differ-ent products influence properties such as organicsolubility and the thermal gelation temperature ofaqueous solutions and swelling behavior.7

It has been shown that mechanism of release andthe release profiles from matrices depend not onlyon the type of the polymer but also on the propertiesof the drug.8–10 For example the release mecha-nisms of propranolol HCl and indomethacin fromHPMC matrices are significantly different.8,9 In astudy on the release of drugs from polyvinylalcoholmatrices, it was observed that the release rates ofpotassium chloride, phenylpropanolamine hydro-chloride and bovine serum albumin decrease as themolecular size of the drug increases.2 Although thephenomenon was attributed to decreased diffusiv-ity and molecular weight was taken as the criterionof molecular size, the study was not aimed atestablishing any explicit relationship. In anotherinvestigation where molecular weights of the drugsclassified in groups of roughly similar solubilitieswere compared, molecular weight and solubilitywere indicated as the possible factors affectingdrug release from matrices.11 Baveja et al. inves-tigated release characteristics of six water-soluble bronchodilators from HPMC K4M matricesin order to find correlation between releaserate and molecular geometry of the drugs.12 Theyshowed that despite almost identical aqueoussolubilities different drug molecules showed dif-ferent release rates from HPMC matrices, whichwas related to the accessible surface area of thedrugs. However, it must be stressed that as thedrugs used in the study consisted of only structu-rally related beta-blockers, the findings cannot beextrapolated to other drugs.

Quantitative Structure–Property Relationship(QSPR) is a valuable tool that employs specializedstatistical techniques to relate the propertyunder investigation to the molecular structureof the chemicals represented by physico - chemicalproperties or structural descriptors. The resultingQSPR models facilitate understanding of theproperty in terms of chemical structure, ultimately

DOI 10.1002/jps JOURNA

enabling the investigators to estimate the propertyforother similar compounds. The aimof thepresentinvestigation was to rationalize the release char-acteristics of different drugs from HPMC matricesin terms of the chemical structure of the drugs andthe properties of the polymer. To achieve this goal,separate QSPR models were established for drugrelease from HPMC K4M or HPMC E4M matricesand the resulting models were compared. Themodels were obtained using the release data of15 drugs (belonging to different chemical classes)from matrices prepared using HPMC K4M andE4M. Chemical structure of the drugs wererepresented by a wide range of molecular descrip-tors such as partition coefficient (logP), pKa, atomiccharges, orbital energies, dipole moment, length,surface area and electrostatic potentials on thesurface, molecular connectivity indexes and shapeindices, solubility and solubility parameter.

MATERIALS AND METHODS

Materials

Acetaminophen (Acros Organic, UK), diclofenacsodium (Sobhan Co., Iran), fluoxetine HCl (Pars-Daru, Iran), naproxen (Pars-Daru, Iran), pirox-icam (Sigma, USA), propranolol HCl (AcrosOrganic, UK), sulfamethoxazole (Logman, Iran),diltiazem HCl (Acros Organic, UK), ibuprofen(Acros Organic, UK), atenolol (Daru-Pakhsh,Iran), diphenhydramine HCl (Acros Organic,UK), imipramine HCl (Logman, Iran), theophyl-line monohydrate (Acros Organic, UK), trifluo-perazineHCl (Sobhan Co., Iran) and trimethoprim(Logman, Iran) were obtained. Two HPMC(Hypromellose) grades, HPMC (Methocel) K4MPremium CR and HPMC (Methocel) E4M Pre-mium CR were gifts from Colorcon, UK.

Preparation of Tablets

The matrices were prepared by mixing 20 g of thedrug (with particle size in the range of 45–125mm)with 20 g of HPMC K4M or HPMC E4M for 10 minusing a small double cone mixer. Magnesiumstearate was then added to the mixture and mixedfor a further 1 min. The final mixtures werecompressed on an 8-mm punch and die usingsingle punch machine (Erweka, Germany) at aconstant pressure of 10 kN. The weight of eachtablet was 202 mg which included 100 mg drug(49.5%), 100 mg HPMC K4M (49.5%) and 1%magnesium stearate.

L OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007

3336 NOKHODCHI ET AL.

Dissolution Studies

The united state pharmacopoeia (USP) basketmethod (Erweka, DT 6R, Heusenstamm, Ger-many) was used for all the in vitro dissolutionstudies. In this method, phosphate buffer (pH 6.8)without enzyme, was used as dissolution medium.The buffer was prepared according to UnitedStates Pharmacopoeia by adding 50 mL of 0.2 Mmonobasic potassium phosphate and 22.4 mL of0.2 M sodium hydroxide into a 200 mL volumetricflask and adding water to volume.13 Matrices wereplaced in 900 mL of the dissolution medium andmaintained at 37� 0.1 8C for 8 h at pH 6.8. Theamount of drug was 100 mg in all formulations.The rate of stirring was 100� 2 rpm. At appro-priate intervals (15, 30, 60, 90, 120, 180, 240, 300,

Figure 1. Chemical structures and the lma

study.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 200

360, 420, and 480 min), 5 mL of samples was takenand filtered through a 0.45 mm Millipore filter.The dissolution media was then replaced by 5 mlof fresh dissolution fluid to maintain a constantvolume. The samples were then analyzed byultraviolet/visible spectrophotometer at the max-imum absorption wavelengths (see Fig. 1). Themean of three determinations was used tocalculate the drug release rate from each of theformulations.

Dissolution Parameters

The parameters obtained from dissolution studieswere the time required for 50% of drug release(T50%), percentage of drug released after 8 h (Q8 h),

x wavelengths of the drugs used in this

7 DOI 10.1002/jps


release rates obtained using various kinetic models(k), and dissolution efficiency (DE8). Naproxen,piroxicam, trifluoperazine, and trimethoprimshowed less than 50% release during the 8 h ofdissolution studies. The slope and the intercept ofthe zero-order release were calculated for theseformulations (correlation coefficient above 0.99)and the results were used for the estimation of thetime required for the release of 50% drug.

The dissolution efficiency (DEt) of a pharma-ceutical dosage form is defined as the area underthe dissolution curve up to the time, t, expressed asthe percentage of the area of the rectangle (Eq. 1).14

DEt ¼

Rt0

ydt

y100t� 100% (1)

where y is the percent of drug dissolved at timet. Eq. (1) was used to calculate the dissolutionefficiency of the matrices where t was 8 h.

Kinetic Models

In order to obtain the rate of release, the releasedata from the matrices were fitted to the followingmathematical models: zero-order kinetic (Eq. 2),first-order kinetic (Eq. 3), square-root of timeequation (Higuchi equation, Eq. 4), and Peppasequation (Eq. 5).

Q ¼ kt (2)

ln ð100 �QÞ ¼ ln Q0 � kt (3)

Q ¼ kt1=2 (4)

Q ¼ ktn (5)

In Eqs. (2)–(5), Q is the percent of drug releasedat time t and k is the coefficient of the equations. InEq. (3)Q0 equals 100. In Eq. (5), k is the constantincorporating structural and geometric charac-teristics of the release device and n is the releaseexponent indicative of the mechanism of release.

Solubility Measurements

Solubility measurements were carried out inphosphate buffer pH 6.8 at 37� 18C. Saturatedsolutions were prepared by adding excessamounts of drugs to water and shaking on ashaker bath (Velp, Italy) for 48 h under constantvibration. After this period, the solutions werefiltered, diluted and analyzed by UV spectro-


photometer (Shimadzu, Japan). The mean of threedeterminations for each drug was reported.

Structural Parameters

A total of 75 structural descriptors for each drugwere obtained from various software packages.Table 1 gives a summary of the descriptorscalculated for the drugs and the software used.

The COSMIC force field was used for energyminimization prior to molecular mechanicalparameter calculations. For calculation of mole-cular orbital parameters, the three-dimensionalstructures of the drugs were minimized using theMNDO Hamiltonian in MOPAC version 7.0(QCPE, Department of Chemistry, Indiana Uni-versity, 800 East Kirkwood Ave., Bloomington, IN47405-7102). The structural descriptors obtainedfrom MOPAC consisted of Atomic charges, orbitalenergies, dipole moment, length, and principlemoments of inertia.

SMILES strings were entered into the MOL-CONN-Z software. MOLCONN-Z was used tocalculate graph theoretical descriptors. ACD/LogD Suite release 7.0 was used to obtain log P, log Dat pH 6.8, and pKa. The calculated values wereonly used when the experimental values werenot available in the database. The fractions ofdrugs that are ionized to anions (fiA) or cations(fiB), and the unionized fraction of drugs (fu) at pH6.8 were calculated using Eqs. (6)–(8). Note thatonly the first acidic pKa and the first basic pKa

were considered in the calculations.

fiB ¼ 1

1 þ anti logð6:8 � pKaÞ(6)

fiA ¼ 1

1 þ anti logðpKa � 6:8Þ (7)

fu ¼ ð1 � fiBÞ � ð1 � fiAÞ (8)

Solubility parameter and energy of vaporizationwas calculated using a group contributionmethod.15

Development of QSPRs

The QSPR endpoints were dissolution parametersdiscussed earlier under ‘dissolution parameters’,as well as the percentage released at differenttimes. Stepwise regression analysis was used todetermine statistically significant relationshipsbetween structural parameters discussed in latter


Table 1. The Physicochemical and Structural Descriptors Used in the Study

Method Parameter

Nemesis Solvent accessible surface area (SA), dipole moment calculated by the Charge-2method (m), the highest and the lowest electrostatic potentials on the solventaccessible surface (ESPþ and ESP�, respectively)

MOPAC 7.0(MNDOHamiltonian)

The energies of the highest occupied and the lowest unoccupied molecular orbitals(EHOMO and ELUMO, respectively), dipole moment (m), the highest and the lowestatomic charges in the molecule (Qþ and Q�), principle moments of inertia (IM),length of the molecule (L), molecular weight (MW)

MOLCONN-Z Simple and valence corrected molecular connectivity indexes including zero- through

fourth-order path (0xp–4xp and 0xvp � 4xv

p), fourth order path-cluster (4xpc and 4xvpc),

third- and fourth-order cluster (3xc,4xc,

3xvc and 4xv

c ), and fifth- and sixth-order chain(5xch and 6xch), the highest atomic electrotopological index (S(I)), molecular shapeindexes (0k� 3k and 0ka–3ka), and delta connectivity indexes (X1–X3)

ACD/log D Partition coefficient (log P), distribution coefficient at pH 6.8 (log D6.8),first acidic and first basic pKa values, the fraction unionized at pH 6.8 (fu),polarizability (a), molecular volume (V), molar refractivity (MR) and parachor (PA)

Experimentaldetermination

Aqueous solubility (mg/mL) measured at 378C, in phosphate buffer pH 6.8

Otherparameters

The total number of oxygen and nitrogen atoms (NNþO), number of hydrogenatoms connected to oxygen or nitrogen (NH), number of double bonds (N––),number of rotatable bonds (Nrotat), total number of bonds (Nbond), numberof aromatic carbon atoms (Carom), number of aliphatic carbon atoms (Calip),logarithms of molecular surface area and weight (log SA and log MW), 1/3 and1/2 power of molar volume (V1/3 and V1/2), energy of vaporization and solubilityparameter (Ev and d)


section and the dissolution parameters. Distribu-tion of the dissolution parameters and thelogarithmically transferred dissolution para-meters were studied and those with the leastdeviation from normality were used in stepwiseregression analyses. The statistical analyses wereperformed using the MINITAB (release 13.1)statistical software. In order to avoid the risk ofchance correlations, loss of interpretability andpredictability, the number of parameters in themodels was kept as low as possible. Accordingly,only two parameters were allowed in stepwiseregression analysis. The following statisticaldetails of the models were noted: n, the numberof observations; R2, the squared correlationcoefficient; s, the standard deviation; F, the Fisherstatistic; and the p value. The figures in bracketswith the regression coefficients were standarderrors of coefficients.

Linear Discriminant Analysis (LDA)

The basic theory of LDA is to classify thedependent by dividing an n-dimensional des-criptor space into two regions that are separatedby a hyperplane defined by a linear discriminant


function as follows:

Y ¼ b0 þ b1x1 þ b2x2 þ . . .þ bnxn (9)

where, Y is discriminant score, that is, thedependent variable, x1–xn represent the specificdescriptors, and b are corresponding to weightsassociated with the respective descriptors. Step-wise LDA was performed using the softwareTSAR (Accelrys Software Inc.).

RESULTS AND DISCUSSION

Various factors could be accounted for the drugrelease mechanism from hydrophilic matrices.16

These include the geometry of the matrix,17

particle size of polymer and matrix swelling ratio(which depends on the HPMC type and controlswater and drug diffusion coefficients),18,19 poly-mer and drug concentrations,20,21 chain lengthand degree of substitution of the HPMC,19 as wellas the drug characteristics.22,23 An additionalfactor that might affect the drug release profile isthe constituents of the dissolution media. Forexample, it has been suggested that phosphatemitigates the hydration of HPMC compacts, aninteraction that is further amplified if the compact

7 DOI 10.1002/jps


incorporates diclofenac sodium.24 However, it wasshown that release of chlorpheniramine maleateand theophylline from HPMC matrices was notsignificantly different in phosphate buffer com-pared to water medium.25 In this study, thematrices consist of the same weight percentages ofthe drug and HPMC E4M or HPMC K4M,compressed under the similar pressure, with thesame die geometry. Furthermore, the samedissolution medium has been used throughoutthe investigation. Therefore, the release profilesfrom matrices prepared using the same polymertype are expected to be related to the intrinsic

Figure 2. Release profiles of 15 drug


properties of the drug and the interaction betweendrug and the polymer. Moreover, physicochemicalcharacteristics of drug alone should be able todetermine which of the release mechanisms,polymer swelling, dissolution/erosion at thematrix periphery and drug diffusion is thegoverning process in drug release. Figures 2and 3 show the dissolution characteristics ofHPMC matrices of these drugs. The figures showdifferent release profiles from HPMC matricescontaining the same polymer type for differentdrugs. Following the QSPR approach in thisstudy, the relationships were sought between

s from the HPMC K4M matrices.


Figure 3. Release profiles of 12 drugs from HPMC E4M matrices.


measured release characteristics and the struc-tural parameters of the drugs. The drugs belongedto various chemical classes covering a broad rangeof water solubility values, as well as acidic andbasic properties (Table 2).

QSPR Models for Release Parameters (RateConstants and Model Independent Parameters)

Various parameters have been employed in orderto evaluate and compare drug dissolution profilesfrom pharmaceutical formulations. Among the


most widely used parameters are release rateconstants from different kinetic models,26,27 andmodel independent parameters such as difference( f1) and similarity ( f2) factors,28,29 dissolutionefficiency,14 mean dissolution time,30 and percentreleased after a given time period.31 In this study,several informative parameters including kineticand model independent release parameters werecalculated from dissolution data. These includedrelease rates, dissolution efficiency, the amount ofdrug released after 8 h, and the time required forthe release of 50% of drug. Table 3 shows therelease rate constants calculated using different

7 DOI 10.1002/jps

Table 2. Water Solubility (mg/mL), pKa, Molecular Volume (V) and Molecular Weight (MW) Values of the DrugsUsed in the Study

Drug Acidic pKa Basic pKa

Water Solubility(mg/mL) Va MW

Acetaminophen 9.86a 1.72a 22.1 120.9 151.2Diclofenac sodium 4.01b �2.26a 17.76 206.8 296.1Fluoxetine HCl — 9.62b 50.0 266.7 309.3Naproxen 4.84a — 0.193 192.2 230.3Piroxicam 5.07b 2.33b 0.172 211.9 331.3Propranolol HCl 13.84a 9.47b 254 237.1 259.3Sulfamethoxazole 5.81a 1.39a 0.51 173.1 253.3Diltiazem HCl — 8.06b 1071 327.6 414.5Ibuprofen 4.41a — 0.23 200.3 206.3Atenolol 13.88a 9.16a 24.75 236.6 266.3Diphenhydramine HCL — 9.02b 1010 249.2 255.4Imipramine HCl — 9.4b 750 269.2 280.4Theophylline monohydrate 8.54b 1.7a 8.6 112 182.2Trifluoperazine HCl — 8.08b 610 328.7 407.5Trimethoprim — 7.12b 0.4 231.8 290.3

aCalculated by the ACD/pKa software.bObtained from ACD/pKa database.


kinetic models. Comparing the correlation coeffi-cients (r2 values) of different kinetic models inTable 3, it becomes evident that no single kineticmodel can be selected as the best model for all thedrugs. For example, the release of propranololhydrochloride best follows a Higuchian, Peppas orfirst order pattern whereas fluoxetine hydrochlor-ide follows a zero-order release pattern best.

As the kinetic constants from different modelscould not be compared with each other, the releaserates from each of the kinetic models wereseparately correlated against structural descrip-tors of the drugs using stepwise regressionanalysis and the QSPR models were developed.The results of these analyses have been presentedin Table 4 for HPMC K4M and in equation 14 forHPMC E4M. In stepwise regression analyses,water solubility was the first parameter thatentered the models. In step 2 of the analyses,molecular volume (V) entered the models forrelease from HPMC K4M. Therefore, it can beconcluded from stepwise regression analysis thatthe most important molecular property increasingthe release rate from both HPMC E4M and HPMCK4M matrices is water solubility (SW) of thedrugs. Furthermore, the larger molecular volume(V) of the drugs results in the smaller release ratesfrom HPMC K4M matrices. It must be notedthat in case of HPMC E4M matrices, the onlysignificant QSPR for release rates was obtainedfor the release rate constant of Peppas kinetic


model (Eq. 14).

log k ¼ �0:35 þ 0:10 logSW � 0:04L

n ¼ 11 s ¼ 0:101R2 ¼ 0:748

F ¼ 11:9 p ¼ 0:004

ð14Þ

In Eq. (14), length of the drug molecules (L) isthe second descriptor to be selected by stepwiseregression.

Model independent dissolution parametersof DE8, T50%, and Q8 h for different drugs arelisted in Table 5. The relationships between theseparameters and physicochemical/structural des-criptors (QSPR models) obtained from stepwiseregression analysis have been presented inTable 6. The QSPR models in Table 6 confirmthe conclusion that drug release from HPMC K4Mand E4M matrices is primarily controlled by watersolubility. Moreover, in QSPRs for dissolution ofHPMC K4M matrices (Eqs. 15–17), the secondparameter is either V or 1xp, with the latter beingthe first order molecular connectivity index,mostly defining the size of the molecule.32 How-ever, in QSPRs for dissolution of HPMC E4Mmatrices (Eqs. 18–20), the second most significantparameter describing the dissolution is the lengthof the molecule.

The effect of drug solubility on release rate iswell documented.9,10,33,34 The theoretical basiscan be explained using Fick’s first law (Eq. 21).


Table

3.

Rel

ease

Rate

Con

stan

tsof

HP

MC

Matr

ices

Obta

ined

Fro

mD

iffe

ren

tK

inet

icM

odel

s;T

he

Cor

resp

ond

ing

Squ

are

dC

orre

lati

onC

oeffi

cien

tsare

Rep

orte

din

the

Pare

nth

eses

Dru

gHPMC

K4M

HPMC

E4M

ZeroOrd

erFirst

Ord

erPep

pas

Higuch

iZeroOrd

erFirst

Ord

erPep

pas

Higuch

i

Ace

tam

inop

hen

0.2

24

(0.9

98)

0.4

71

(0.9

92)

0.3

76

(0.9

96)

0.5

04

(0.9

93)

0.1

09

(0.8

07)

0.2

94

(0.9

87)

0.3

68

(0.9

78)

0.3

92

(0.9

97)

Dic

lofe

nac

sod

ium

0.1

02

(0.9

42)

0.1

81

(0.9

67)

0.2

13

(0.9

36)

0.3

11

(0.9

66)

0.1

15

(0.9

84)

0.1

88

(0.8

56)

0.2

42

(0.9

99)

0.3

94

(0.9

94)

Flu

oxet

ine

HC

l0.0

9(0

.999)

0.1

81

(0.9

64)

0.1

93

(0.9

66)

0.3

2(0

.965)

0.0

73

(0.8

04)

0.0

63

(0.8

90)

0.2

77

(0.9

49)

0.2

64

(0.9

18)

Nap

roxen

0.0

4(0

.937)

0.0

5(0

.924)

0.0

89

(0.9

71)

0.1

34

(0.9

26)

0.0

31

(0.9

59)

0.0

17

(0.9

70)

0.0

96

(0.9

91)

0.1

08

(0.9

87)

Pir

oxic

am

0.0

52

(0.9

84)

0.0

73

(0.9

94)

0.1

3(0

.989)

0.1

79

(0.9

9)

0.0

30

(0.9

92)

0.0

16

(0.9

91)

0.0

98

(0.9

7)

0.1

00

(0.9

65)

Pro

pra

nol

olH

Cl

0.1

23

(0.9

63)

0.2

76

(0.9

97)

0.3

6(0

.992)

0.3

49

(0.9

98)

0.1

04

(0.9

55)

0.1

41

(0.9

85)

0.2

97

(0.9

96)

0.3

61

(0.9

95)

Su

lfam

eth

oxazo

le0.0

73

(0.9

91)

0.1

15

(0.9

98)

0.1

4(0

.997)

0.2

47

(0.9

86)

——

——

Dil

tiaze

mH

Cl

0.1

6(0

.987)

0.3

6(0

.988)

0.3

66

(0.9

99)

0.4

06

(0.9

96)

0.1

04

(0.9

91)

0.1

35

(0.8

33)

0.2

44

(0.9

79)

0.3

50

(0.9

66)

Ibu

pro

fen

0.1

06

(0.8

96)

0.1

79

(0.7

97)

0.0

93

(0.9

51)

0.3

03

(0.8

09)

0.1

02

(0.9

92)

0.1

05

(0.9

56)

0.2

08

(0.9

89)

0.3

46

(0.9

76)

Ate

nol

ol0.0

71

(0.9

42)

0.1

23

(0.9

67)

0.2

05

(0.9

36)

0.2

47

(0.9

66)

0.0

47

(0.9

75)

0.0

28

(0.9

91)

0.1

40

(0.9

92)

0.1

61

(0.9

96)

Dip

hen

hyd

ram

ine

0.1

49

(0.9

48)

0.4

02

(0.9

93)

0.4

75

(0.9

94)

0.3

85

(0.9

94)

0.0

67

(0.9

44)

0.0

55

(0.9

81)

0.2

85

(0.9

80)

0.2

32

(0.9

88)

Imip

ram

ine

HC

l0.1

26

(0.9

96)

0.2

48

(0.9

66)

0.2

67

(0.9

89)

0.3

47

(0.9

71)

0.1

11

(0.9

93)

0.1

39

(0.9

17)

0.2

25

(0.9

76)

0.3

74

(0.9

82)

Th

eop

hyll

ine

0.1

64

(0.9

75)

0.3

43

(0.9

74)

0.3

22

(0.9

89)

0.4

17

(0.9

89)

0.0

999

(0.9

53)

0.1

11

(0.9

97)

0.2

56

(0.9

95)

0.3

48

(0.9

97)

Tri

flu

oper

azi

ne

HC

l0.0

56

(0.9

90)

0.0

75

(0.9

94)

0.0

98

(0.9

77)

0.1

88

(0.9

73)

——

——

Tri

met

hop

rim

0.0

38

(0.9

98)

0.0

46

(0.9

97)

0.0

57

(0.9

90)

0.1

28

(0.9

65)

——

——

Mea

nR

2valu

es0.9

70

0.9

67

0.9

78

0.9

66

0.9

46

0946

0.9

78

0.9

70


JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007 DOI 10.1002/jps

Table 4. QSPR Models for Release Rate Constants from HPMC K4M Matrices: a, b and c are Coefficients inthe Equation, log k¼aþ b. log SWþ cV, where k is the Release Rate Constant, Sw is Water Solubility, and V isthe Molecular Volume, n is the Number of Drugs, R2 is the Correlation Coefficient, s is Standard Deviation, and Fis the Fisher Statistic (Ratio of Variance Between Groups to the Overall Variance in Analysis of Variance)

Eq. Kinetic Model a b c n R2 s F p

10 Zero order �0.584 (�0.147) 0.160 (�0.031) �0.00286 (�0.0007) 15 0.704 0.137 14.3 0.00111 First order �0.210 (�0.202) 0.231 (�0.042) �0.00380 (�0.0010) 15 0.719 0.189 15.3 0.00012 Peppas �0.296 (�0.150) 0.207 (�0.031) �0.00301 (�0.0007) 15 0.784 0.140 21.8 0.00013 Higuchi �0.232 (�0.115) 0.123 (�0.024) �0.00212 (�0.0006) 15 0.693 0.108 13.5 0.001


Accordingly, when the drug is poorly water-soluble, dissolved and non-dissolved drug areboth present within the matrix, from which onlythe dissolved drug can diffuse into the dissolutionmedia. In other words, the concentration gradient,being smaller for such drugs, results in a reduceddiffusion according to Fick’s first law.35

J ¼ �DdC

dx(21)

In Eq. (21), J is the flux (the amount of materialflowing through a unit cross-section of a barrier inunit time), D is the diffusion coefficient, dC/dx isthe concentration gradient. With concentrationgradient being dependent on water solubility, thesecond parameter in Fick’s law (D) is expected tobe controlled by structural characteristics of thedrug. Diffusion coefficient has been theoretically

Table 5. Dissolution Efficiency (DE8), Time Required for 508 h (Q8 h) of HPMC Matrices

HPMC K4M

Drugs DE8 T50%

Acetaminophene 75.98 1.6Diclofenac sodium 57.31 3Fluoxetine HCl 44.12 4.8Naproxen 21.23 10.6Piroxicam 28.52 9.1Propranolol HCl 66.38 1.9Sulfamethoxazole 35.71 5.8Diltiazem HCl 70.10 1.8Ibuprofen 42.29 5Atenolol 51.38 3.6Diphenhydramine HCL 78.61 1.1Imipramine HCl 61.60 3.2Theophylline monohydrate 70.97 2.1Trifluoperazine 26.01 11Trimethoprim 16.60 12.8


related to the radius of the diffusing particle inStokes–Einstein equation for diffusion of aspherical particle within a continuous fluid.35

This has been used to justify the inverseproportionality between diffusion constant andcross sectional area of solute,36 or betweendiffusion constant and cube root of volume,V1/3 for diffusions through skin.37 Althoughdiffusion in polymers does not obey the Stokes–Einstein equation, it is still reasonable to expectthat the larger the diffusing molecule the moredifficult its movement within the HPMC gelmedium will be. This could explain the negativeeffect of molecular size (volume or length terms)observed in the QSPR models (Eqs. 10–20) on therelease rate or efficiency.

Although all of the equations (Eqs. 10–20) arestatistically significant (p< 0.05), the statisticalquality of the equations can be compared based on

% of Drug Release (T50%), and the amount Released after

HPMC E4M

Q8 h DE8 T50% Q8 h

99.42 61.91 1.0 100.1999.23 58.69 3 98.6479.79 54.71 1.9 71.5539.04 19.25 15.8 28.4447.51 17.01 14.1 30.2896.27 62.83 2.5 92.9561.16 — — —96.92 53.36 3.5 96.1392.65 50.47 4.35 86.0782.10 28.26 8.5 4497.57 48.53 3.5 68.6998.38 54.63 3 94.7298.18 58.31 2.8 87.0844.77 — — —31.42 — — —


Table

6.

QS

PR

Mod

els

for

Mod

elIn

dep

end

ent

Dis

solu

tion

Para

met

ers

ofH

PM

CM

atr

ices

Eq.

QSPR

mod

eln

R2

SF

pHPMC

Type

15

DE

8¼

701

(�88)þ

123

(�18)l

ogS

W�

2.0

1(�

0.4

3)V

15

0.7

92

82.1

22.8

0.0

00

K4M

16

log

T50%¼

0.0

03

(�0.2

05)�

0.2

39

(�0.0

43)

logS

Wþ

0.0

0391

(�0.0

010)V

15

0.7

25

0.1

91

15.8

0.0

00

K4M

17

Q8

h¼

127

(�19.5

)þ14.4

(�3.4

)lo

gS

W�

7.0

7(�

2.1

7)

1x

p15

0.6

26

16.8

10.1

0.0

03

K4M

18

DE

8¼

108

(�24.6

)þ8.7

4(�

3.3

0)l

ogS

W�

5.9

8(�

2.1

2)L

12

0.5

82

15.2

6.3

0.0

20

E4M

19

T50%¼�

5.1

3(�

4.7

8)�

2.6

7(�

0.6

4)

logS

Wþ

1.2

0(�

0.4

1)L

12

0.7

07

2.9

511.0

0.0

04

E4M

20

Q8

h¼

126

(�31.8

)þ12.3

(�4.2

7)l

ogS

W�

5.8

3(�

2.7

4)L

12

0.5

48

19.6

5.5

0.0

28

E4M

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEM


BER 200

the value of R2. R2 shows the percentage of thevariation in Y (dependent variable) that can beexplained by the independent variables (aqueoussolubility, volume or length). Other variations inY can be due to the factors other than thoseinvestigated here, or the experimental error. Inthe QSPR models (10)–(20), R2 values varybetween 0.548 and 0.792 (p values below 0.028).Generally, the QSPR models for Peppas releaseconstants (Eqs. 12 and 14) with R2 values of 0.784(p¼ 0.000) and 0.748 (p¼ 0.004) are statisticallysuperior to the models for other release rateconstants. A particularly good quality model isEq. (15) for the estimation of DE8 with R2 value of0.792 (p¼ 0.000).

Structure - Based QSPR Models forRelease Parameters

The QSPR models (10)–(20), are able to predictdifferentrelease endpointsusing watersolubilityofdrugsandthemolecularvolume/length.However itis most desirable to be able to estimate releasepropertiesusinganentirelystructure-basedmodel,without the need to obtain drug solubility. SuchQSPR models were obtained from stepwise regres-sion analysis between drug release parameters andstructural descriptors (Tables 7 and 8).

Table 7 shows that release rates are correlatedwith dipole moment (m) and a size term that iscube root of molar volume (V1/3), molecular weight(MW), or length of molecules (L). The QSPRs inTable 8 also contain dipole moment and a sizeparameter (mostly MW). It is possible that inthese equations dipole moment is merely describ-ing the drug solubility. In order to test thishypothesis, the structural descriptors that arerelated to solubility were explored using stepwiseregression analysis, from which Eq. (32) resulted.Eq. (32) shows that dipole moment is the mainstructural parameter describing the water solu-bility of this set of drugs. The second term in thisequation is fraction of drug molecules that are inanionic form at pH 6.8. The negative sign of fiAshows that acidic drugs (which ionize to anions)are less water soluble than the basic drugs in theset (see Fig. 4).

logSW ¼ 0:374ð�0:31Þ þ 0:127ð�0:021Þm

� 1:57 ð�0:32ÞfiAn ¼ 15 r2 ¼ 0:879 s ¼ 0:547

F ¼ 43:4 P ¼ 0:000

(32)

7 DOI 10.1002/jps

Table

7.

QS

PR

Mod

els

for

Dru

gR

elea

seR

ate

su

sin

gO

nly

the

Str

uct

ura

lP

ara

met

ers

Eq.

Kinetic

Mod

elQSPR

Mod

eln

R2

SF

pHPMC

Type

22

Zer

oor

der

logk¼

0.4

36

(�0.5

2)�

0.2

92

(�0.0

92)V

1/3þ

0.0

276

(�0.0

076)m

15

0.5

56

0.1

68

7.5

0.0

08

K4M

23

Fir

stor

der

logk¼

1.1

1(�

0.7

3)�

0.3

84

(�0.1

30)V

1/3þ

0.0

399

(�0.0

11)m

15

0.5

56

0.2

37

7.5

0.0

08

K4M

24

Pep

pas

logk¼

2.0

0(�

0.7

8)�

0.8

04

(�0.2

3)V

1/3þ

0.0

540

(�0.0

12)mþ

0.1

32

(�0.0

53)

L15

0.6

81

0.1

78

7.8

0.0

04

K4M

25

Hig

uch

ilo

gk¼

0.5

39

(�0.3

86)�

0.2

21

(�0.0

69)V

1/3þ

0.0

221

(�0.0

057)m

15

0.5

81

0.1

26

8.3

0.0

05

K4M

26

Pep

pas

logk¼�

0.5

66

(�0.1

56)þ

0.0

231

(�0.0

053)m

�0.0

0145

(�0.0

006)

MW

11

0.7

05

0.1

09

9.6

0.0

08

E4M

Table

8.

QS

PR

Mod

els

for

Mod

elIn

dep

end

ent

Dru

gR

elea

seP

ara

met

ers

usi

ng

On

lyth

eS

tru

ctu

ral

Para

met

ers

Eq.

QSPR

Mod

eln

R2

SF

pHPMC

Type

27

DE

8¼

631

(�129)þ

17.9

(�5.1

9)m

�1.5

3(�

0.5

2)

MW

15

0.5

38

122.4

7.0

0.0

10

K4M

28

logT

50%¼

0.2

63

(�0.1

98)�

0.0

261

(�0.0

09)m

þ0.3

98

(�0.1

20)

3x

c15

0.5

77

0.2

37

8.2

0.0

06

K4M

29

Q8

h¼

115

(�18.3

)þ2.8

9(�

0.7

3)m

�0.2

47

(�0.0

73)

MW

15

0.6

01

17.4

9.0

0.0

04

K4M

30

DE

8¼

96.5

(�20.1

)þ2.2

2(�

0.7

4)m

�7.8

4(�

2.5

2)

1x

p12

0.5

72

15.4

6.0

0.0

22

E4M

31

T50%¼�

0.1

1(�

3.9

)�

0.6

13

(�0.1

60)m

þ0.0

461

(�0.0

17)

MW

12

0.6

28

3.3

37.6

0.0

12

E4M

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 12, DECEMBER 2007



QSPR models for the percentage of drug released

The second approach in this investigation was theincorporation of all the release data in the form ofpercentage released (Q) at various release times inthe QSPR analysis. Table 9 is a list of the QSPRmodels obtained from stepwise regression ofQ against time and the structural descriptors ofdrugs. Different forms of Q – time relationships,adopted from Peppas, Higuchi, zero order, andfirst order kinetic models, were examined. Com-paring the statistics of the resulting QSPR models(Table 9), it can be seen that drug release fromboth HPMC K4M and HPMC E4M matrices bestfollows the Peppas kinetic model. This is of courseexpected due to the flexibility of Peppas modelthat can fit into the release profiles of bothdiffusion, and erosion controlled release mecha-nisms. Figures 5 and 6 show the plot between theobserved and calculated percentage released byEqs. (33) and (37), respectively. The QSPRs showthat apart from the time, drug aqueous solubilityis the most important parameter in determiningthe percentage released from matrices preparedfrom both grades of HPMC. However, it is notedthat coefficients of log SW are higher in QSPRs forrelease from HPMC K4M matrices in comparisonwith those in QSPRs for release from HPMC E4Mmatrices. In other words, solubility of drugs hasless effect on the release from HPMC E4M.Moreover, coefficients of time are also higher forHPMC K4M, indicating a generally higher drugrelease rates from matrices prepared with thisHPMC grade. In order to make a better com-parison of the coefficients of ‘time’ and ‘log SW’ inQSAR models for release from HPMC E4M andHPMC K4M, the same variables and the samedrug set were used for the development of the

Figure 4. Regression plot of log SW aga


models. Therefore, SA (surface area) rather thanV (volume) and the same set of drugs as those inEq. (37) for release from HPMC E4M was used forthe correlation of the amount released fromHPMC K4M. The resulting equation below forrelease from HPMC K4M still has a higher log tand log SW coefficients than Eq. (37), but a smaller(absolute) coefficient of SA.

logQ ¼ 1:57 ð�0:053Þ þ 0:622 ð�0:021Þ log t

þ 0:148 ð�0:008Þ logSW

� 0:00147 ð�0:0002ÞSAn ¼ 132 R2¼0:0907 s¼0:111 F¼418:6p ¼ 0:000

(41)

It should be emphasized that the findingpertains to average release profiles of all thedrugs and some drugs may show a higher releaserate from the HPMC E4M matrices. For example,according to Table 3, diclofenac sodium, fluoxetineHCl, naproxen and ibuprofen have higher Peppasrelease rates from HPMC E4M matrices thanfrom HPMC K4M matrices. It is generally thoughtthat HPMC K4M is a more efficient retardantthan HPMC E4M.38 However, in a comparativestudy on the release of propranolol HCl fromHPMC matrices, it has been shown that the orderof release rate from different grades of HPMCdepends dramatically on the drug: Polymer ratio.At low polymer: Drug ratio HPMC K4M has alower drug release rate. However, at higherpolymer: drug ratio of 1:1 HPMC E4M becomesmore retardant than HPMC K4M.39 A stepwiselinear discriminant analysis was performed inorder to find structural descriptors that are able toclassify drugs correctly into two groups of 0 for

inst the log SW calculated by Eq. 32.

7 DOI 10.1002/jps

Table

9.

QS

PR

Mod

els

for

the

Am

oun

tof

Dru

gR

elea

sed

at

Dif

fere

nt

Tim

eIn

terv

als

(Q)

from

Matr

ices

Pre

pare

du

sin

gtw

oG

rad

esof

HP

MC

;t

isti

me,

SW

isS

olu

bil

ity,V

isM

olar

Vol

um

e,an

dS

Ais

Sol

ven

tA

cces

sible

Su

rface

Are

a

Eq.

Kinetic

Mod

elQSPR

Mod

elN

R2

sF

pHPMC

Type

33

Pep

pas

logQ¼

1.6

8(�

0.0

45)þ

0.6

45

(�0.0

23)

logtþ

0.1

77

(�0.0

09)

logS

W�

0.0

0271

(�0.0

002)V

165

0.8

80

0.1

37

392.4

0.0

00

K4M

34

Hig

uch

iQ¼

26.3

(�4.4

3)þ

30.2

(�1.2

3)t

1/2

þ13.8

(�0.8

1)

logS

W�

0.2

22

(�0.0

2)V

165

0.8

47

12.0

3296.7

0.0

00

K4M

35

Zer

oor

der

Q¼

47.2

(�4.2

4)þ

8.7

8(�

0.3

7)tþ

13.8

(�0.8

4)

logS

W�

0.2

22

(�0.0

2)V

165

0.8

36

12.5

273.7

0.0

00

K4M

36

Fir

stor

der

ln(1

00�Q

)¼3.2

6(�

0.2

20)�

0.3

03

(�0.0

19)t�

0.4

57

(�0.0

44)

logS

Wþ

0.0

089

(�0.0

01)V

165

0.6

92

0.6

47

120.8

0.0

00

K4M

37

Pep

pas

logQ¼

1.8

2(�

0.0

57)þ

0.5

52

(�0.0

22)

logtþ

0.1

32

(�0.0

09)

logS

W�

0.0

0239

(�0.0

0022)

SA

132

0.8

70

0.1

21

284.7

0.0

00

E4M

38

Hig

uch

iQ¼

38.8

(�6.1

)þ27.5

(�1.3

8)t

1/2þ

9.8

3(�

0.8

7)

logS

W�

0.2

00

(�0.0

22)

SA

132

0.8

09

12.1

180.9

0.0

00

E4M

39

Zer

oor

der

Q¼

57.9

(�6.0

3)þ

7.9

9(�

0.4

2)tþ

9.8

4(�

0.9

0)

logS

W�

0.2

00

(�0.0

22)

SA

132

0.7

97

12.5

167.1

0.0

00

E4M

40

Fir

stor

der

ln(1

00�Q

)¼1.5

6(�

0.1

00)�

0.0

94

(�0.0

07)t�

0.0

95

(�0.0

148)

logS

Wþ

0.0

020

(�0.0

003)

SA

131

0.6

41

0.2

04

75.7

0.0

00

E4M



which release from HPMC E4M matrices is fasterthan release from HPMC K4M and 1 for whichrelease is faster from HPMC K4M matrices. Thisresulted in the following discriminant axis with83% correct classification:

Y ¼ 0:236SA þ 0:972 log fu þ 0:706confidenceestimate ¼ 0:833

(42)

The analysis shows that drugs with largersurface area and percentage unionized at pH7.4 are more likely to have faster release ratesfrom HPMC K4M matrices than from HPMC E4Mmatrices.

The two grades of hydroxypropylmethylcellu-lose, HPMC K4M and E4M share a similarmolecular weight but differ in degrees of sub-stitution, with the methoxyl percent being 22 forK4M and 29 for E4M. Hydroxypropyl percent forK4M and E4M grades is very similar (8.1 and 8.5,respectively). In order to further investigate theeffect of polymer type on the release kinetics, thesimilarity factor ( f2) was calculated using Eq. (43)and matrices of the same drug prepared withHPMC E4M or HPMC K4M were compared (ca.Table 10).

f2 ¼ 50 log

1 þ ð1=mÞXmj¼1

ðE4Mj � K4MjÞ2

" #�0:5

� 100

8<:

9=;

(43)

Average f2 for all the drugs was 38.04showing differences between matrices preparedfrom HPMCs E4M and K4M. Table 10 showsthat not all the drugs are liberated with adifferent profile from HPMC K4M and E4Mmatrices. The most different release profile isobserved for acetaminophen with f2 value of 20.On the other hand, the release profiles ofdiclofenac sodium from the two matrices arevery similar with f2 value of 76. Differencesbetween release profiles of the same drug frommatrices prepared using different grades ofHPMC could be determined by diffusion (ifsoluble) through the gel and by the rate oftablet erosion. Also listed in Table 10 is Peppascoefficient ‘n’ calculated for matrices of HPMCE4M and K4M. The table shows that average ‘n’ ishigher for matrices prepared from HPMC K4M,indicating higher contribution of erosion in therelease process from these matrices in comparisonwith the matrices prepared from HPMC E4M.


Figure 5. Plot of observed vs. calculated log Q (logarithm of percentage released) fromHPMC K4M matrices according to Eq. (33).


From the above discussion it can be deducedthat the overall lower release kinetics observed forHPMC E4M in comparison with HPMC K4Mshould be attributed to the lower erosion of thecorresponding matrices. The lower erosion ofthese matrices implies a higher dependence ofthe release process on diffusion, a process requir-ing small drug molecular size, which is confirmedby the larger coefficient of SA in Eq. (37) incomparison with that in Eq. (41). Figure 7 showsthat the higher the ‘n’ of the HPMC E4M matrices,the higher the similarities of the matrices preparedfrom the two HPMC grades. The deviations from

Figure 6. Plot of observed vs. calculated logHPMC E4M matrices according to Eq. (37).


this relationship (error) observed in Figure 7 islikely to be due to the varying contribution ofdiffusion and the effect of drug solubility that is notaccounted for in the graph.

CONCLUSION

The process of drug release from HPMC matricesinvolves the two routes of erosion of the matrix,and diffusion of the dissolved drug through theHPMC gel, each to a varying extent depending onthe properties of the drug as well as the polymer

Q (logarithm of percentage released) from

7 DOI 10.1002/jps

Table 10. Similarity Factor between MatricesPrepared Using HPMC E4M and K4M and Peppas ‘n’Coefficients for the Matrices

Drug f2 n (E4M) n (K4M)

Acetaminophen 20.36 0.560 0.644Diclofenac sodium 76.03 0.683 0.676Fluoxetine HCl 43.57 0.532 0.579Naproxen 69.73 0.536 0.657Piroxicam 48.61 0.469 0.604Propranolol HCl 67.38 0.595 0.496Sulfamethoxazole 0.703Diltiazem HCl 39.85 0.560 0.505Ibuprofen 48.95 0.643 0.908Atenolol 32.10 0.549 0.777Diphenhydramine HCL 29.25 0.424 0.406Imipramine HCl 58.68 0.630 0.546Theophylline monohydrate 47.04 0.659 0.623Trifluoperazine HCl 0.714Trimethoprim 0.783

Average 38.04 0.570 0.647


grade. QSPR method employed in this investiga-tion was able to outline the major drug propertiesresponsible for the control of release from HPMCmatrices. These included drug solubility and themolecular size, with enhancing and reducingeffects on the drug release, respectively. The effectof these properties can be explained on the basis ofthe effect of solubility and molecular size on thediffusion of a drug within a fluid medium. Thesimilarity factor calculated in order for comparison

Figure 7. Plot of similarity factor ( f2) betweK4M or HPMC E4M vs. Peppas ‘n’ coefficient


of HPMC K4M and HPMC E4M matrices revealedthat the matrices perform differently in most cases.Comparison of the QSPRs for release from twotypes of HPMC matrices led to the conclusion thatHPMC E4M matrices have a lower tendency toerode resulting in overall inferior drug releasekinetics for some drugs from this grade of HPMC incomparison with HPMC K4M. Furthermore, phy-sicochemical properties of the drugs (surface areaand percentage unionized at pH 6.8) determinewhether the release is faster from HPMC K4M orfrom HPMC E4M.

REFERENCES

1. Christenson GL, Dale LB. 1962. Sustained releasetablets. U.S. Pat. 3065143.

2. Korsmeyer RW, Gurny R, Doelker E, Buri P, Pep-pas NA. 1983. Mechanism of solute release fromporous hydrophilic polymers. Int J Pharm 15:25–35.

3. Lapidus H, Lordi NG. 1966. Some factors affectingthe release of a water-soluble drug from com-pressed hydrophilic matrices. J Pharm Sci 55:840–843.

4. Huber HE, Christenson GL. 1968. Utilization ofhydrophilic gums for the control of drug substancerelease from tablet formulations II. Influence oftablet hardness and density on dissolution beha-viour. J Pharm Sci 57:164–166.

5. Buri P, Doelker E. 1980. Formulation des com-primes a liberation prolongee. II. Matrices hydro-philies. Pharm Act Helv 55:189–197.

en matrices prepared either with HPMCfor HPMC E4M matrices.



6. Higuchi T. 1963. Mechanism of sustained-actionmedication: Theoretical analysis of rate of releaseof solid drugs dispersed in solid matrices. J PharmSci 52:145–1149.

7. Rajabi-Siahboomi AR, Bowtell RW, Mansfield P,Henderson A, Davies MC, Melia CD. 1994. Struc-ture and behavior in hydrophilic matrix sustained-release dosage forms. 2. NMR-imaging studies ofdimensional changes in the gel layer and core ofHPMC tablets undergoing hydration. J ControlRelease 31:121–128.

8. Ford JL, Rubinstein MH, Hogan JE. 1985. Propra-nolol hydrochloride and aminophylline release frommatrix tablets containing hydroxypropylmethylcel-lulose. Int J Pharm 24:339–350.

9. Ford JL, Rubinstein MH, Hogan JE. 1985. Dis-solution of a poorly water soluble drug, indometha-cin, from hydroxypropylmethylcellulose controlledrelease tablets. J Pharm Pharmacol 37:33P.

10. Tahara K, Yamamoto K, Nishihata T. 1996. Appli-cation of model-independent and model analysis forthe investigation of effect of drug solubility on itsrelease rate from hydroxypropyl methylcellulosesustained release tablets. Int J Pharm 133: 17–27.

11. Rao KVR, Devi KP, Buri P. 1990. Influence ofmolecular-size and water solubility of the soluteon its release from swelling and erosion controlledpolymeric matrices. J Control Release 12:133–141.

12. Baveja SK, Ranga Rao KV, Singh A, Gombar VK.1988. Release characteristics of some bronchodila-tors from compressed hydrophilic polymericmatrices and their correlation with moleculargeometry. Int J Pharm 41:55–62.

13. United States Pharmacopoeia, 2003. vol. 26, p.2524–2525.

14. Khan KA. 1975. Concept of dissolution efficiency.J Pharm Pharmacol 271:48–49.

15. Fedors RF. 1974. A method for estimating both thesolubility parameters and polar volumes of liquids.Poly Eng Sci 14:147–154.

16. Siepmann J, Peppas NA. 2001. Modeling of drugrelease from delivery systems based on hydroxy-propyl methylcellulose (HPMC). Adv Drug DelivRev 48:139–157.

17. Ford JL, Rubinstein MH, McCaul F, Hogan JE,Edgar PJ. 1987. Importance of drug type, tabletshape and added diluents on drug release kineticsfrom hydroxypropylmethylcellulose matrix tablets.Int J Pharm 40:223–234.

18. Velasco MV, Ford JL, Rowe P, Rajabi-Siahboomi AR.1999. Influence of drug:hydroxypropylmethylcellu-lose ratio, drug and polymer particle size and com-pression force on the release of diclofenac sodiumfrom HPMC tablets. J Control Release 57:75–85.

19. Campos-Aldrete ME, Villafuerte-Robles L. 1997.Influence of the viscosity grade and the particlesize of HPMC on metronidazole release from matrixtablets. Eur J Pharm Biopharm 43:173–178.


20. Mitchell K, Ford JL, Armstrong DJ, Elliott PNC,Rostron C, Hogan JE. 1993. The influence of con-centration on the release of drugs from gels andmatrices containing Methocel. Int J Pharm 100:155–163.

21. Fu XC, Wang GP, Liang WQ, Chow MSS. 2004.Prediction of drug release from HPMC matrices:effect of physicochemical properties of drug andpolymer concentration. J Control Release 95:209–216.

22. Ford JL, Mitchell K, Rowe P, Armstrong DJ, ElliottPNC, Rostron C, Hogan JE. 1991. Mathematicalmodelling of drug release from hydroxypropyl-methylcellulose matrices: Effect of temperature.Int J Pharm 71:95–104.

23. Mitchell K, Ford JL, Armstrong DJ, Elliott PNC,Hogan JE, Rostron C. 1993. The influence of drugson the properties of gels and swelling characteris-tics of matrices containing methylcellulose orhydroxypropylmethylcellulose. Int J Pharm 100:165–173.

24. Rajabi-Siahboomi AR, Adler J, Davies MC, MeliaCD. 1994. Particle swelling and the mechanism offailure of HPMC matrices. Proc 3rd Assoc: 21.

25. Levina M, Rajabi-Siahboomi AR. 2004. The influ-ence of excipients on drug release from hydroxy-propyl methylcellulose matrices. J Pharm Sci 93:2746–2754.

26. Vueba ML, Batista de Carvalho LA, Veiga F, SousaJJ, Pina ME. 2004. Influence of cellulose etherpolymers on ketoprofen release from hydrophilicmatrix tablets. Eur J Pharm Biopharm 58:51–59.

27. Kim H, Fasihi R. 1997. Application of a binarypolymer system in drug release rate modulation.1. Characterization of release mechanism. J PharmSci 86:316–322.

28. Moore JW, Flanner HH. 1996. Mathematical com-parison of dissolution profiles. Pharm Technol 20:64–74.

29. Koester LS, Ortega GG, Mayorga P, Bassani VL.2004. Mathematical evaluation of in vitro releaseprofiles of hydroxypropylmethylcellulose matrixtablets containing carbamazepine associated to b-cyclodextrin. Eur J Pharm Biopharm 58:177–179.

30. Tanigawara Y, Yamaoka K, Nakagawa T, Uno T.1982. New method for the valuation of in vitrodissolution time and disintegration time. ChemPharm Bull 30:1088–1090.

31. Huang YB, Tsai YH, Lee SH, Chang JS, Wu PC.2005. Optimization of pH-independent release ofnicardipine hydrochloride extended-release matrixtablets using response surface methodology. Int JPharm 289:87–95.

32. Hall LH, Kier LB, Murray WJ. 1975. MolecularConnectivity. 2. Relationship to water solubilityand boiling-point. J Pharm Sci 64:1974–1977.

33. Tahara K, Yamamoto K, Nishihata T. 1995. Overallmechanism behind matrix sustained release (SR)

7 DOI 10.1002/jps


tablets prepared with hydroxypropyl methylcellu-lose 2910. J Control Release 35:59–66.

34. Akbari J, Adrangui M, Farid J, Siahi-Shadbad MR,Nokhodchi A. 2000. Factors affecting the releaserate of a poorly soluble drug (carbamazepine)from HPMC matrices. STP Pharm Sci 10:473–478.

35. Martin A, Bustamante P, Chun AHC. 1993. Physi-cal Pharmacy. Baltimore, USA: Lippincott Williams& Wilkins. p. 400–401.

36. Mitragotri S. 2002. A theoretical analysis of per-meation of small hydrophobic solutes across thestratum corneum based on scaled particle theory.J Pharm Sci 91:744–752.


37. Buchwald P, Bodor N. 2001. A simple, predictive,structure based skin permeability model. J PharmPharmacol 53:1087–1098.

38. Rajabi-Siahboomi AR, Bowtell RW, Mansfield P,Davies MC, Melia CD. 1996. Structure and behaviorin hydrophilic matrix sustained release dosageforms. 4. Studies of water mobility and diffusioncoefficients in the gel layer of HPMC tablets usingNMR imaging. Pharm Res 13:376–380.

39. Mitchell K, Ford JL, Armstrong DJ, ElliottPNC, Hogan JE, Rostron C. 1993. The influence ofsubstitution type on the performance of methylcel-lulose and hydroxypropylmethylcellulose in gels andmatrices. Int J Pharm 100:143– 154.


Documents

A Drug Release Study.pdf