International Journal of Pharmaceutics 167 (1998) 83 – 96

A controlled porosity drug delivery system

Padmaja Shivanand a, Omar L. Sprockel b,*

a
Di6ision of Pharmaceutics and Drug Deli6ery Systems, College of Pharmacy, Uni6ersity of Cincinnati Medical Center, Cincinnati,
OH 45267, USA
b
Bristol-Myers Squibb, Pharmaceutical Research Institute, One Squibb Dri6e, P.O. Box 191, New Brunswick, NJ 08903 -0191, USA

Received 27 January 1997; received in revised form 22 October 1997; accepted 13 January 1998

Abstract

The relationship between the drug release rate constant (Ko) and the physicochemical properties of porosity
modifiers incorporated in the polymer coat of a proposed compression coated drug delivery system was investigated.
The effects of particle size, hygroscopicity, solubility, absolute density and powder specific surface area on Ko were
related to their influence on the pore structure created. In general, porosity modifiers with larger particle sizes, smaller
specific surface areas, greater hygroscopicity coefficients or higher solubilities caused faster drug release, by creating
more conducting channels. The porosity modifier particle size and load were related to changes in the coat volume
(Vcoat), coat porosity (ofinal) and increased specific surface area (SSAd). The Vcoat decreased with an increased loading
or particle size, promoting faster release. Ko increased very slowly until ofinal and SSAd reached critical values
(approximately 38% and 0.1 m2 g − 1, respectively), after which Ko increased very rapidly. This information allows the
selection of a porosity modifier with the appropriate characteristics to provide a delivery system with the desired
release rate. Alternatively, one can specify the necessary coat characteristics after porosity modifier release that will
yield the desired release rate. © 1998 Elsevier Science B.V. All rights reserved.

Keywords: Compression coating; Porosity modifier; Release rate; Pore generation; Cluster; Particle size; Surface area;
Hygroscopicity; Coat volume

1. Introduction independent of their environment. Core tablets

surrounded by a polymeric coat offer a simple but
The focus of research efforts in the drug deliv- effective approach. The release rate can be altered
ery field has been on developing delivery systems easily by modifying the coat’s physical structure
that can deliver drug at a predictable release rate (porosity).
The coat can be applied by compressing the
coating material around the core (Mars, 1974;
* Corresponding author. Conte et al., 1983; Verhoeven et al., 1989). The

0378-5173/98/$19.00 © 1998 Elsevier Science B.V. All rights reserved.

PII S0378-5173(98)00047-7
84 P. Shi6anand, O.L. Sprockel / International Journal of Pharmaceutics 167 (1998) 83–96

release rate from these systems is controlled by release media, are the conduits by which drug is
several parameters, such as the amount of poly- released.
mer used, the surface characteristics of the poly- It is clear from the literature, that addition of
mer and the compressibility of the polymer a porosity modifiers to an insoluble polymer coat
(Shivanand and Sprockel, 1993). Because of the surrounding a core tablet increases the release
stresses encountered during dissolution, the poly- rate of the drug. However, the relationship be-
mer shell must have sufficient physical strength tween the physicochemical properties of the addi-
to remain intact during drug release. This means tive and the release rate constant have not been
that for a particular polymer the shell will need a studied. In this study we investigated the rela-
minimum thickness and density and changing the tionship between the properties of several soluble
release rate through the above mentioned meth- additives and their effect on the release rate con-
ods is limited. An alternative is to alter the per- stant of propranolol HCl from a compression
meability of the shell by dynamically modifying coated system, through changes in the coat.
the pore structure of the shell.
Fryklof et al. (1967) incorporated soluble addi-
tives (porosity modifiers) in their compressed 2. Materials and methods
shell, that upon dissolution increased the release
rate. They saw an initial lag time, after which the 2.1. Materials
release rate increased to a constant rate and
The following materials were used as received
finally declined to zero. The advantage of using
from the manufacturer without further purifica-
soluble additives in the coat material is that one
tion: cellulose acetate butyrate 500-1 (CAB) (a
can design a strong coat with a useful release
gift from FMC corporation); propranolol HCl (a
rate. Källstrand and Ekman (1983) spray coated
gift from Knoll Ag.); citric acid (CA); calcium
their core tablet with a suspension of polyvinyl
tartrate (CaT), calcium sulfate (CaSO4), manni-
chloride and micronized sucrose in a solvent sys-
tol, sodium chloride (Fisher Scientific); and su-
tem using conventional pan technology. They
crose (Kroger).
saw an increasing release rate with time followed
by the declining rate. Zentner et al. (1985) 2.2. Characterization of powders
demonstrated that, with increased loading of sor-
bitol, the increase in the release rate became 2.2.1. Absolute density
smaller. Therefore, increasing the release rate by The absolute densities of the CAB powder and
adding soluble additives had its limitations. the porosity modifiers were determined in tripli-
Thombre et al. (1989) showed that as sorbitol cate using a Micromeritics helium pycnometer.
dissolved out of the coat, the pores generated The powders were dried in an oven at 100°C for
formed networks through which water entered 12 h prior to analysis.
the system.
The percolation theory, in general, deals with 2.2.2. Particle size analysis
the number and property of pores and pore clus- To obtain the desired size fractions, samples of
ters (Stauffer, 1985). The percolation theory has CaT, NaCl and mannitol were placed on top of
been used to explain the relationship between a standard nest of sieves and shaken until there
pore generation and solute release from films was no detectable change in the weight of the
(Siegel, 1988). Dissolution of the porosity powder retained on each sieve. CA, CaT and
modifiers incorporated in the coat creates pores. sucrose were used as received. Samples of all the
As the number of pores increase, the pores com- powders were suspended in liquid paraffin and
bine to form a percolating pore cluster or con- the particle size distributions were determined
ducting channel, that spans the thickness of the microscopically. The geometric mean particle size
coat. These percolating pore clusters, filled with was obtained from log-probability plots.
 P. Shi6anand, O.L. Sprockel / International Journal of Pharmaceutics 167 (1998) 83–96 85

2.2.3. Intrinsic dissolution rates cm2 · per h. The solubility of NaCl was determined
The intrinsic dissolution rate of NaCl in 0.02% to be 0.35290.007 g − 1 cm3. Therefore, Kd for
v/v polyoxyethylene sorbitan monooleate in dis- NaCl under these conditions was 0.1539 0.024
tilled water was determined. A total of 50 mg of cm h − 1. This indicates that the release of NaCl
NaCl were compressed at 236 MPa on a hydraulic from the polymer coat was not dissolution rate
press (Carver press) using a 6 mm flat faced limited.
punch and die set. The NaCl core tablet was
placed in a 12 mm die containing the bottom flat 2.2.4. Hygroscopicity
faced punch. CAB (300 mg) was carefully poured The hygroscopicity coefficient for the porosity
on top of the core tablet. The powder bed was modifiers was determined at 89% relative humid-
compressed at 78 MPa. As a result, the NaCl core ity using saturated solutions of ZnSO4. The pow-
tablet was coated with the polymer on the annular ders were dried at 100°C for 24 h. Samples (1 g)
surface and on one planar surface. This limited of the porosity modifiers were placed in glass petri
dissolution to a single tablet face and maintained dishes of known weights. A saturated solution of
a constant surface area. The coated NaCl tablet ZnSO4 was placed at the bottom of the desiccator
was affixed to a metal substrate with the exposed and the system was allowed to stand for 24 h. The
surface facing upwards. The tablet on the sub- petri dishes containing the powder samples were
strate was placed at the bottom of a dissolution placed in the desiccator. At predetermined inter-
vessel containing 500 ml of the dissolution media vals, the petri dishes were removed, weighed, and
at 37°C. Media agitation was achieved with a replaced. The increase in weight was ascribed to
paddle rotating at 25 rpm. Dissolution fluid (5 ml) moisture uptake by the powder. The absorption
were removed at predetermined intervals and ana- of moisture can be described by the following
lyzed for NaCl content by polarography. equation (van Campen et al., 1980):
The mass transfer of drug from the solid sur-
Mt
face of the undissolved material to the solution is = (1− e − Kht) (3)
described by the Noyes – Whitney equation M
(Carstensen, 1974): where Mt and M are the amounts (g) of mois-
dC ture absorbed at time t and at time infinity,
= AKd(Cs −C) (1) respectively, and Kh is the hygroscopicity coeffi-
dt
cient (h − 1). The Kh for the porosity modifiers
where V is the volume of the dissolution media were calculated using linear least squares regres-
(cm3), C is the concentration of NaCl in the sion analysis.
solution (g − 1 cm3), t is time (h), A is the exposed
surface area of the NaCl core tablet (cm2), Cs is 2.2.5. Specific surface area
the solubility of NaCl (g − 1 cm3) and Kd, the The powder samples were dried in an oven at
intrinsic dissolution rate (cm h − 1), is: 100°C for 24 h. The specific surface area of the
powders was determined in triplicate by the BET
D
Kd = (2) method using a krypton/helium mixture with a
Vh
Micromeritics Flowsorb 2300 instrument.
where D is the diffusion coefficient of the solute in
the dissolution medium (or solvent), V is the 2.2.6. Solubility
volume of the dissolution medium (cm3) and h is The solubilities of the porosity modifiers were
the thickness of the boundary layer (cm). determined. Excess solid was placed in 10 ml test
The dissolution of 70% of the core tablet was tubes. A volume of 5 ml of 0.02% v/v poly-
linear with time (r 2 =0.997). Kd was calculated oxyethylene sorbitan monooleate in distilled water
from the slopes from these lines, which were was added. The caps were screwed on and sealed
determined using least squares regression. The with parafilm. The test tubes were agitated in a
average slope of these lines was 0.054 90.008 g − 1 shaker bath at 37°C for 3 days. The suspended
86 P. Shi6anand, O.L. Sprockel / International Journal of Pharmaceutics 167 (1998) 83–96

Table 1
Characteristics of the porosity modifiers

Porosity modifier Size (mm) K ah(h−1) SSAb (m2 g−1) r cabs (g ml−1) C ds (g ml−1)

NaCl 7 0.457 0.209 1.852 0.352

98 0.054 2.033
256 0.018 1.957
Calcium tartrate 18 0.004 0.504 2.037 0.030
22 0.437
98 0.402
Mannitol 40 0.227
83 0.012 0.186 1.554 0.269
98 0.167
Sucrose 21 0.016 0.454 1.593 0.701
Citric acid 270 0.603 0.186 0.844
CaSO4 33 0.207 4.080 2.711 0.002

a
Hygroscopicity coefficient.
b
Specific surface area.
c
Absolute density.
d
Solubility.

solids were allowed to settle and 3 ml of the A total of 50% (150 mg) of the coating material
clear supernatant were removed. The samples of was placed in a 12 mm die containing the bottom
solutions containing CaT or NaCl were filtered flat faced punch. The core tablet was carefully
and the first 1 ml of the filtrate was discarded. centered on the powder bed. The rest of the
The solute concentrations were determined by coating material was poured on top of the core
polarography after appropriate dilutions. The tablets and the system was compressed at 31 MPa
solubilities of CA, CaSO4, mannitol and sucrose on the hydraulic press. The coating material com-
were determined gravimetrically. A volume of 2 prised CAB only or mixtures of CAB and poros-
ml of the filtered solutions were pipetted into ity modifiers. CAB was used as the insoluble
polymer in the coating material. Several water
weighing pans of known weights. The pans con-
soluble materials were mixed with CAB in varying
taining the solutions were dried in an oven at
proportions by geometric dilution using a spatula;
50°C to constant weight.
these materials acted as porosity modifiers (Table
1). The compression coated tablets were stored at
2.3. Preparation of the coated tablet
24°C in closed HDPE containers for at least 1 day
before any testing occurred. This was done to
To make the core tablets, 50 mg of propra- allow post-compression elastic recovery of the
nolol HCl were compressed at 78 MPa on a tablets to occur.
hydraulic press (Carver Press) using a 6 mm flat
faced punch and die set. The punch faces and 2.4. Characterization of the coated tablet
the die wall were lightly dusted with magnesium
stearate before each compression. After ejection, 2.4.1. Tablet dimensions
the core tablets were carefully wiped clean to One day after compression the thickness and
remove the magnesium stearate. The core tablets the diameter of the core tablet and of the com-
were stored in closed HDPE bottles at 24°C for pression coated tablet were measured with a dig-
1 day before coating. ital micrometer.
 P. Shi6anand, O.L. Sprockel / International Journal of Pharmaceutics 167 (1998) 83–96 87

2.4.2. Drug release The specific surface area of the tablets after disso-
The release of propranolol HCl from the com- lution of the NaCl is defined as the SSAf (specific
pression coated tablets was determined in an surface area final). The difference between SSAi
aqueous solution of 0.02% v/v poyoxyethylene and SSAf is defined as SSAd (specific surface area
sorbitan monooleate at 37°C. Agitation was difference).
maintained at 100 rpm with a paddle (Hanson).
Samples of volume 4 ml were withdrawn periodi- 2.4.5. Porosity changes
cally and analyzed by spectrophotometry at 290 For tablets containing NaCl as the porosity
nm for the propranolol HCl concentration. The modifier, the increase in the coat’s porosity during
lost volume was replaced with fresh release media. NaCl release was calculated. The coat’s porosity
at time t (ot) is:
2.4.3. Porosity modifier release
In the NaCl release studies, the propranolol ot = o0 + oNaCl (4)
HCl core tablet was replaced with a CAB core where the tablet’s initial porosity (o0) is:
tablet to facilitate the polarographic analysis of
NaCl. The release of NaCl from the coat of the rapp(coat)
o0 = (5)
compression coated system was determined simi- rapp(coat)
lar to the propranolol HCl release studies, except and the porosity due to the release of NaCl from
that 5 ml of the media were removed periodically the coat (oNaCl) is:
and analyzed by polarography. The effective dif-
fusion coefficient for NaCl release was calculated MNaClVcoat
oNaCl = (6)
from Eq. (14) using non-linear least squares re- rNaCl
gression analysis.
where MNaCl is the amount of NaCl released and
rNaCl is the absolute density of NaCl.
2.4.4. Surface area changes
The volume of the coat (Vcoat) is:
To determine the changes in the surface area of
the coat at specified time points during NaCl Vcoat = pr 2t ht − p 2c hc (7)
release, 30 compression coated tablets containing
where rt is the radius of the tablet, ht is the
NaCl were made. The propranolol HCl core
tablet was replaced with a CAB core tablet to thickness of the tablet, rc is the radius of the core
facilitate data analysis. Three tablets were used as and hc is the thickness of the core. The overall
control and did not undergo any release study. apparent density of the coat (rapp(coat)) (g ml − 1)
The remaining 27 tablets were subjected to a is:
release study similar to that for propranolol HCl. rapp(coat)=(rcabXcab)− (rNaClXNaCl) (8)
At 0.25, 0.5, 0.75, 1, 1.5, 2, 4, 6 and 12 h, three
tablets were removed from the appropriate disso- where rcab is the absolute density of CAB and
lution vessels and dried in an oven at 100°C for at Xcab and XNaCl are the weight fractions of CAB
least 24 h. The surface area of the dried tablets and NaCl, respectively. For simplicity, we assume
were then determined using the BET method. A that there is no density gradient in the coat; such
krypton/helium mixture was used with a a density gradient may exist since the pressure
Micromeritics Flowsorb 2300 instrument. The transfer through the coat upon compression is not
surface area of each tablet was determined at least uniform. The absolute density of the coat
three times. The three control tablets were sub- (rabs(coat)) (g ml − 1) is:
jected to the same procedures. The surface area of Mtab − Mcore − MNaCl
the control tablets served as a reference point for rabs(coat)= (9)
Vcoat
the other tablets.
The specific surface area of the control tablets is where Mtab is the mass of the tablet and Mcore is
defined as the SSAi (specific surface area initial). the mass of the core (g).
88 P. Shi6anand, O.L. Sprockel / International Journal of Pharmaceutics 167 (1998) 83–96

2.5. Analysis through the media filled channels in the coat.

While the solution concentration in the pores was
Propranolol HCl was analyzed by spectropho- maintained constant by the solid drug present, a
tometry at 290 nm. The standard curve was linear constant release rate was seen. Upon depletion of
in the range of 1 – 50 mg l − 1 (r 2 =0.999). Sodium the solid drug source, the internal concentration
chloride was analyzed by differential pulse polar- declined causing a decrease in the release rate.
ography. A static mercury drop electrode was Similar profiles were seen by other investigators
used with a polarographic analyzer in the differ- (Fryklof et al., 1967; Conte et al., 1983; Käll-
ential pulse polarography mode (EG&G, Prince- strand and Ekman, 1983). Thus, the release of
ton, NJ). The analysis was carried out under a propranolol HCl from this system required sev-
nitrogen blanket (99.999% nitrogen). The NaCl eral rate processes to occur in parallel or in series.
aqueous (deionized water) solutions contained The constant release of propranolol HCl across
0.02% polyoxyethylene sorbitan monooleate and the compressed polymer barrier is described by
0.01 M tetraethyl ammonium hydroxide as the Eq. (10) (Shivanand and Sprockel, 1993):
supporting electrolyte. The following settings were
used: a scan range of 1.9 – 2.7 V; a scan rate of 5 dM DeA
= (Ci − Co) (10)
mV s − 1; a current of 100 mA; a pulse height of 50 dt h
mV; a drop time of 1 s; and a purge time of 4 where M is the amount of propranolol HCl re-
min. leased (mg); t is time (h); De is the effective
diffusion coefficient (cm2 h − 1); A is the effective
2.6. Statistical analysis
surface area (cm2); h is the coat thickness (cm); Ci
is the drug concentration inside the tablet (mg
A linear least squares method was used to
cm3); and Co is the drug concentration outside the
determine the release rate constant (Ko) for pro-
tablet (mg cm3). Zentner et al. (1985) pointed out
pranolol HCl and the dissolution rate constant for
that mass transfer from coated tablets is by
NaCl. A non-linear least squares method was
used to determine the effective diffusion coeffi-
cient for NaCl. One-way and two-way analysis of
variance were used to detect any significant differ-
ences between treatment means using the Minitab
statistical software. Tukey’s procedure for multi-
ple means comparison was used with a =0.05. A
multiple linear regression method was used to
study the relationship between Ko and the powder
properties.

3. Results and discussion

3.1. Drug release

The release profile from the compression coated

system was typically sigmoidal with a linear seg-
ment (Fig. 1). After penetrating the coat, the
release media dissolved the porosity modifier and Fig. 1. A typical release profile for propranolol HCl from the
created a saturated propranolol HCl solution in compression coated drug delivery system with mannitol as the
the core. Propranolol HCl then diffused outward porosity modifier.
 P. Shi6anand, O.L. Sprockel / International Journal of Pharmaceutics 167 (1998) 83–96 89

Table 2
The zero order release rate constants (Ko) and lag times (Tlag) from the various delivery systems

Modifier Load (%) Size (mm) Ko (mg h−1) Tlag (h)

None 30 0.999 9 0.145 29.648 9 5.871

NaCl 10 7 1.038 9 0.735 2.447 90.482
30 7 5.407 9 0.735 2.573 90.244
30 98 4.403 9 0.534 3.462 90.741
30 256 15.900 9 3.139 4.204 90.405
50 7 18.560 9 6.789 1.087 90.220
CaT 30 18 1.333 9 0.128 −0.3349 0.299
30 22 1.337 9 0.147 0.652 90.341
30 98 1.710 9 0.096 −0.9319 0.602
50 22 2.007 9 0.041 0.869 90.340
Mannitol 10 83 1.701 9 0.305 2.976 91.123
30 40 8.383 9 0.532 2.949 90.240
30 83 7.972 9 0.615 2.876 90.123
30 98 10.410 9 0.993 2.911 90.223
50 40 49.851 9 7.313 1.453 90.088
50 83 39.479 9 5.455 0.847 90.192
Sucrose 10 21 0.800 9 0.0158 4.883 90.864
30 21 5.775 9 0.541 2.862 90.404
50 21 84.968 9 12.353 0.888 90.200
Citric acid 30 270 8.969 9 1.345
CaSO4 30 33 2.106 9 0.149 1.058 90.417

diffusion and by osmotic pumping. The overall 3.2. Powder properties

release rate constant (Ko) is defined as:
A stepwise multiple regression method was used
DeACi
Ko = (11) to relate the physicochemical properties of the
h porosity modifiers listed in Table 1 to Ko. We
The release rate constants (Ko) and lag times found that a second degree polynomial adequately
(Tlag) were calculated from the slopes of the linear approximated the underlying relationship between
segments (see Table 2). The macroporous cellu- Ko and the properties (Fig. 2; r 2 = 0.996; slope
lose acetate butyrate membrane surrounding the = 0.998):
propranolol HCl core tablet was 2.7390.03 mm Ko = b0 + b1x1 + b2x2 + b4x4 + b5x5
thick. Because of the low porosity of the coat 2
(oo = 0.231), Ko was limited to 1.009 0.15 mg + b11x + b24x2x4 + b25x2x5 + b34x3x4 + b35x3x5
h − 1, with a lag time of 29.65 95.87 h. Addition (12)
of porosity modifiers to the coating material in- where bo is the intercept of the regression equa-
creased Ko tremendously. For example, at a 30% tion; bn are the coefficients of the regression equa-
level of porosity modifiers in the coat, Ko ranged tion; and xn are the properties. The coefficients of
from 1.3390.13 mg h − 1 to 15.9093.14 mg h − 1. the polynomial with their P-values are listed in
The relative increase in Ko between the various Table 3. We pooled all the non-significant (P]
porosity modifiers is due to their physicochemical 0.1) higher order terms with the error term. All
properties. The characteristics of the porosity the remaining terms in the statistical model are
modifiers used in this study are listed in Table 1. significant (PB 0.1), except the intercept (P=
90 P. Shi6anand, O.L. Sprockel / International Journal of Pharmaceutics 167 (1998) 83–96

cant may be surprising, except that the range of

values was relatively narrow (1.554–2.711 g ml −
1).
The intercept of the regression equation (bo)
represents Ko in the absence of any porosity
modifier. The predicted Ko with out porosity
modifier (1.039 0.05 mg h − 1) is very close to the
determined value (1.0090.145 mg h − 1). If one
looks at the main effects, the following conclusion
can be made. In general, porosity modifiers with
larger particle sizes, greater hygroscopicity coeffi-
cients, smaller specific surface areas or higher
solubilities caused a faster drug release.
The effect of porosity modifier particle size on
Ko is related to the ability of the pores created
after the dissolution of the solute particles to form
percolating pore clusters or conducting channels.
Fig. 2. Regression of the release rate constant (Ko) fitted from We assume that diffusion through the polymer
Eq. (12) on Ko determined experimentally. particles was negligible and that the release of
propranolol HCl occurred through media filled
0.265), particle size (P = 0.270) and absolute den- channels in the coat. The number and topology of
sity (rabs; r= 0.811). Particle size and absolute such channels directly affect the rate of drug
density were left in the model because the higher release from polymer matrices (Siegel and Langer,
order terms Size2, Kh · rabs, and SSA · rabs were 1990) and the pore topology is influenced by the
significant (P = 0.086, P= 0.035 and P =0.048, particle size (Desai et al., 1965).
respectively). That rabs by itself was not signifi- To study the effect of pore topology further,
three particle sizes each of calcium tartrate (CaT),
Table 3
mannitol and NaCl were studied at the 30% load-
Coefficients for the regression of the powder properties on Ko ing. The 30% loading was selected because it is
close to the percolation threshold and would,
Variable Property Coefficient P-value therefore, accentuate the effect of particle size on
Ko. The change in Ko with size was modest, unless
Constant Inherent Ko (mg 1.039 0.672 0.265
a relatively large particle size was used. For exam-
h−1)a
X1 Size (mm) −0.0319 0.021 0.270 ple, increasing the NaCl particle size from 7 to 98
X2 Kh (h−1)b 192.119 38.43 0.038 mm had a small impact on Ko (5.419 0.735 mg
X3 SSA (m2 g−1)c −99.10921.38 0.044 h − 1 compared to 4.409 0.534 mg h − 1) (Table 2).
X4 rabs (g ml−1)d 0.939 3.42 0.811 The NaCl 256 mm particle size increased Ko ap-
X5 Solubility (g 67.189 14.91 0.046
proximately three fold to 15.909 3.139 mg h − 1.
ml−1)
Xb1 Sizeb 0.00021 0.086 Increasing the CaT and mannitol particle sizes,
from 18 to 98 mm and 40 to 98 mm, resulted in
9 0.000066
X2X4 Kh · rabs −93.199 17.97 0.035
equally small increases in Ko.
X2X5 Kh · solubility −124.709 23.10 0.033 The effect of particle size on transmembrane
X3X4 SSA · rabs 47.189 4.38 0.048 transport is related to the formation of conduct-
X3X5 SSA · solubility −99.23932.71 0.094 ing channels across the membrane. In membranes
a
with relatively large pores transmembrane diffu-
Release rate constant.
b
Hygroscopicity coefficient. sion may occur through clusters of only a few
c
Specific surface area. pores, well below the percolation threshold
d
Absolute density. (Siegel, 1988). In other words, the larger the re-
 P. Shi6anand, O.L. Sprockel / International Journal of Pharmaceutics 167 (1998) 83–96 91

sulting pores the smaller the number needed to

form a conducting cluster. From our data, it is
apparent that below 100 mm, particle size has a
minor effect on Ko. Using a simple cubic lattice in
a computer simulation, Siegel et al. (1989) studied
the effect of solute particle size on the available
solute fraction (ASF). They reported that, below
the critical porosity, smaller particles had a
smaller ASF. With smaller particles there are
more layers per unit thickness. As the network of
pores required to connect an interior particle be-
comes longer and less probable, that particle is
less likely to be connected to the surface. There-
fore, formation of a conducting channel is more
probable with the NaCl 256 mm particle than with
the 7 mm particle at equal loading.
There is a general increase in Ko with a decrease Fig. 3. Effect of mannitol, NaCl and calcium tartrate load on
Ko.
in porosity modifier specific surface area (SSA).
To understand the inverse relationship between
SSA and Ko, one must envision the pores gener- artificial and that Cs is the controlling parameter.
ated after disssoution of the modifier particles. A This is logical since Ko depends on the number of
porosity modifier with a smaller SSA would gen- conducting channels rather than on the rate at
erate pores with smaller surface areas, on an which the channels are formed.
The modifier solubility determines the extent to
equivalent weight basis. To account for the
which the modifier will dissolve in the medium.
smaller surface area there would have to be fewer
This controls the pore structure and, thereby, the
and larger pores. These larger pores could easier
rate of propranolol HCl release (Ko). There is an
connect to form a conducting channel, which
increase in Ko with an increase in solubility. The
would explain the faster drug release.
increase in solubility caused a larger fraction of
There is a general decline in Ko with an increase the modifier to be released, creating more con-
in rabs. An increase in rabs means a decrease in the ducting channels. The dissolved solute may lower
modifier volume based on equal weight. Hence, the local solubility of the drug in the pores (Zent-
modifiers with a larger rabs will generate pores ner et al., 1991). The high concentration of dis-
with smaller volumes upon dissolution. This solved solute also may have increased the
means that there will be fewer conducting chan- viscosity of the fluid in the channels. Both factors
nels available for the diffusing propranolol HCl could reduce drug diffusion in the channels below
and hence, the slower release with a larger rabs. a theoretically calculated value.
A larger hygroscopicity coefficient (Kh) indi-
cates a greater affinity for water, and presumably 3.3. Pore cluster size
signify a faster dissolution of the modifier. The
dissolution rate controls the rate of pore genera- Three loading levels each of CaT, mannitol,
tion. At very low Kh values (50.1 h − 1), Ko is NaCl and sucrose were investigated to study the
independent of Kh. For example, the large differ- effect of cluster size on Ko. Fig. 3 shows the effect
ences in Ko between tablets containing CaT, man- of loading on Ko. A small increase in Ko was seen
nitol and sucrose are not reflect in their Kh values. at the 10% porosity modifier level compared to no
As Kh increases, it becomes a better predictor of porosity modifier. It was only at loading levels
Ko. Interestingly, the relationship between Kh and above 30% that the tremendous influence of the
Ko mirrors that between Kh and Cs. Therefore, it porosity modifier on Ko became apparent (Table
can be argued that the effect of Kh on Ko is 2).
92 P. Shi6anand, O.L. Sprockel / International Journal of Pharmaceutics 167 (1998) 83–96

The coat of the control tablet (containing no 3.4. Mechanistic analysis

porosity modifiers) had a low porosity (0.2319
0.014). The probability of forming a conducting To fully understand how a porosity modifier
channel across this coat was low, which accounts influences the release rate constant (Ko), we inves-
for the slow release of propranolol HCl from the tigated coats containing NaCl as the solute. The
control tablet. Dissolution of porosity modifier changes in the coat before, during and after NaCl
particles created secondary pores, that expanded dissolution are used to explain the effect on Ko.
the pore cluster size and, thereby, increased Ko. A model for the three dimensional release of a
The effect of loading appears fairly straight solute from a polymer matrix was first developed
forward; increasing the modifier load increased Ko by Fu et al. (1976). Solute release from a cylinder
by increasing the number of conducting channels. of radius a and thickness 2l follows Fick’s second
The degree of increase, however, was very small

law:
for CaT. We attribute this to the low solubility of
CaT (Cs = 0.030 g ml − 1). A mass balance on the
dC
= De
d2C 1 dC d 2C
+ + 2
n
polymer shell after drug release, revealed that a dt dr 2 r dr d z
very small fraction of the incorporated CaT had t= 0 −l 5z 5l C= Co
dissolved. The percent of the incorporated CaT
released for the 10%, 30% and 50% CaT loadings 05 r5a
was 18.8, 11.9 and 14.2%, respectively. Therefore, t\ 0 z= 9l C= Cs (13)
it is not surprising that the CaT load did not
significantly increase Ko. This underscores the im- where Co is the initial NaCl concentration (g
portance of the available solute fraction (ASF) cm − 3) in the matrix; Cs is the concentration of
rather than the modifier loading as the controlling NaCl at the exposed surfaces (g cm − 3); a is the
factor. radius of the disk (cm); and 2l is the thickness of
To exemplify the relationship between ASF and the disk (cm). The solution to the diffusion equa-
Ko, NaCl (Cs =0.352 g ml − 1) was used. A higher tion with these initial and boundary conditions
ASF produced a larger Ko (Fig. 4). At a low ASF, was obtained by Carslaw and Jaeger (1959). Fu et
the mean cluster size was insufficient to form al. (1976) calculated the total mass transferred
many conducting channels, which restricted Ko. across all exposed surfaces and the fraction of
As the ASF increased the number of conducting solute released (F) at time t:
channels escalated, increasing Ko. Mt 8
F= = 1− 2 2 X. (2) c(2) (14)
M l a
m=
X. (k)= % exp(− Dea 2mt) (am) − k (15)
m=1

where Mt (g) is the amount of NaCl released at

time t; and M (g) is the amount of NaCl re-
leased at finite time.
n=
c(k)= % exp (− Deb 2nt) (bn) − k (16)
n=o

where De is the effective diffusion coefficient for

NaCl. b was calculated as:
(2n + 1)p
bn (17)
2l
and a was calculated from the positive roots of
Fig. 4. Effect of available solute fraction (ASF) on Ko. the zero order bessel function, Jo; i.e. Jo(aan) =0.
 P. Shi6anand, O.L. Sprockel / International Journal of Pharmaceutics 167 (1998) 83–96 93

Table 4
Characteristics of compression coated tablets containing NaCl

Size (mm) 7 7 7 98 256

Percent (%) 10 30 50 30 30
Ko (mg h−1) 1.038 5.407 18.560 4.403 15.900
Dea (cm2 h−1) 4.81×10−3 4.68×10−3 13.0×10−3 4.93×10−3 2.06×10−3
ASFb 0.563 0.701 0.974 0.860 1.019
Vccoat (cm3) 0.314 0.290 0.265 0.288 0.279
o dInitial 0.251 0.259 0.252 0.252 0.221
o eFinal 0.328 0.378 0.553 0.384 0.393
o fPeak 0.297 0.416 0.526 0.391 0.320
SSAgi (m2 g−1) 0.345 0.302 0.197 0.335 0.268
SSAhf (m2 g−1) 0.368 0.413 0.348 0.402 0.355
SSAid (m2 g−1) 0.023 0.111 0.151 0.067 0.087

a
Effective diffusion coefficient for NaCl.
b
Available solute fraction
c
Volume of the coat.
d
Initial coat porosity.
e
Final coat porosity.
f
Porosity at peak release rate.
g
Initial specific surface area.
h
Final specific surface area.
i
Difference in specific surface area.

Tukey’s multiple means test (a =0.05) revealed In general, De is inversely related to Ko, which
no significant difference between the De at 10% may seem counter intuitive. The exception to this
and 30% NaCl loading, but there was a significant is the system containing 50% of 7 mm NaCl
increase in De at the 50% NaCl loading. The De particles. A closer examination of the coat dy-
for NaCl release from these systems reflects the namics explains this apparent paradox. NaCl re-
resistance encountered by the diffusing species. lease depends on a network of interconnected
The resistance encountered is a function of the NaCl particles regardless of the depth. Release of
number of conducting channels available for dif- propranolol HCl, however, depends on the num-
fusion and the tortuosity of those channels. NaCl ber of conducting channels across the coat. A
loading affected the number of channels available shallow, limited pore cluster would result in a fast
for diffusion. release of the accessible NaCl (large De), but a
The effect of modifier particle size on pore slow release of drug (small Ko). Conversely, a
structure was previously examined. Initially, there deep conducting pore cluster would result in a
is no change in De with particle size, with subse- slow release of NaCl (small De), but a fast release
quent increases in particle size resulting in a of drug (large Ko). An example of the former is
smaller De. A similar reduction in release rate the coat containing 10% of 7 mm NaCl particles
with increased particle size was reported previ- and of the latter is the coat containing 30% of 256
ously (Desai et al., 1965; Chandrasekaran and mm NaCl particles. At a 50% NaCl loading the
Paul, 1982; Foster and Parrott, 1990). Siegel pore cluster size (number of conducting channels)
(1988) proposed a pore structure inwhich solute is so large that the release of NaCl and propra-
release is through a central channel to which nolol HCl are fast.
pores are connected by narrow ducts. Solute Addition of NaCl to the coating material
transfer from these deadend pores is slow (Balazs modified the compression properties of the coat-
et al., 1985), resulting in a small De. Larger NaCl ing material. The apparent volume of the coat
particles would be more affected by this (Vcoat) decreased with an increase in NaCl loading
phenomenon. or particle size. This is reflected in the differences
94 P. Shi6anand, O.L. Sprockel / International Journal of Pharmaceutics 167 (1998) 83–96

from tablets containing the 256 mm particles. The

same arguments presented previously for the ef-
fect of size on Ko apply here.
To further study the interaction between poros-
ity and the release rate, the instantaneous release
rate for propranolol HCl (dM/dt) was correlated
with the coat’s porosity at time t (ot). ot was
calculated using Eqs. (4)–(7). The release data
was approximated adequately by a sigmoidal
function. To calculate dM/dt, we differentiated
the function with respect to time. Fig. 7 depicts a
typical relationship between dM/dt and ot. Ini-
tially, dM/dt increases very slowly with an in-
crease in ot. Close to the maximum porosity,
Fig. 5. Effect of porosity (ot) on release rate of propanolol HCl dM/dt escalates rapidly to a maximum, after
(at 30% loading of NaCl of 98 mm particle size). which dM/dt decreases sharply. In all cases, ex-
cept for the 30% 256 mm (i.e. 30% loading of
in Vcoat between the various formulations (Table NaCl of 256 mm particle size), the maximum
4). A smaller Vcoat denotes a thinner coat, which release rate occurs at a peak porosity, opeak. The
promotes faster drug release (Fig. 5). A thinner opeak increases with increased NaCl load (Table 4).
coat means a shorter conducting channel that This suggests that pore generation progresses as a
requires fewer open pores. retreating boundary from the surface of the coat
The initial porosity of the coat (o0) did not vary to the core as successive layers of NaCl particles
much with the NaCl loading, and only slightly dissolve. Appreciable drug release does not occur
with the particle size (Table 4). However, both the until this pore boundary reached the core tablet.
load and particle size of NaCl affected Ko (Table As the NaCl particle size increased, epeak de-
2). We, therefore, conclude that the oo did not creased (Table 4). This suggests that larger parti-
significantly affect Ko. cles can form conducting channels at lower
The final porosity (ofinal) is reached after release porosities, which we alluded to previously.
of the available solute fraction (ASF). A larger
ofinal represented a greater number of channels
available for the outward diffusion of propranolol
HCl. Using a scaling law, Hastedt and Wright
(1990) showed that the relative diffusivity of so-
lutes in a compressed matrix increased with
porosity. Fig. 6 depicts the relationship between
ofinal and Ko. Ko increased slowly until ofinal
reached a critical value (: 38%), after which Ko
increased very rapidly. Saltzman and Langer
(1989), in their study of bovine serum albumin
release from thin, solvent casted membranes, de-
termined a percolation threshold of 33%. The ofinal
increased significantly with the increase in NaCl
loading, but only slightly with the increase in
particle size (Table 4). The changes in ofinal explain
the increase in Ko with NaCl loading, but not Fig. 6. Effect of final coat porosity (ofinal) on Ko.
 P. Shi6anand, O.L. Sprockel / International Journal of Pharmaceutics 167 (1998) 83–96 95

Fig. 7. Effect of coat volume (Vcoat) on Ko.

The initial specific surface area (SSAi) of the 4. Conclusion

tablet refers to the void spaces present before
dissolution of any NaCl. The SSA increased To accurately control the rate of drug release
rapidly on dissolution of the incorporated NaCl, from the proposed delivery system a suitable
to reach a plateau in approximately 1.5 h (SSAf: , porosity modifier must be chosen, based on its
final specific surface area). The effect of including physicochemical properties and incorporated at
NaCl in the coat on SSAi is derived from the
effect on coat volume (Vcoat). For example, an
increase in NaCl loading causes a reduction in
Vcoat, which in turn decreases the SSAi. Both SSAi
and SSAf are inversly related to Ko. The same
reasons proposed previously for the inverse rela-
tionship between Vcoat and Ko would explain this
inverse relationship. The difference between the
specific surface areas (SSAd =SSAf − SSAi) corre-
lated well with Ko, with 30% 256 mm (i.e. 30%
loading of NaCl of 256 mm particle size) being the
exception. SSAd quantifies the newly created
space available for difffusion of propranolol HCl.
The increase in Ko with SSAd was small until
SSAd reached a critical value (approximately 0.1
m2 g − 1) above which Ko increased dramatically Fig. 8. Effect of increased coat specific surface area (SSAd) on
(Fig. 8). Ko.
96 P. Shi6anand, O.L. Sprockel / International Journal of Pharmaceutics 167 (1998) 83–96

the appropriate loading level. In general, a poros- Fryklof, L.E., Sandell, E., Ostholm, G.I.V., 1967. Medicinal
tablet and a method for its preparation. US Patent
ity modifier with a high hygroscopicity coefficient, 3317394.
a moderately high solubility, a low absolute den- Fu, J.C., Hagemeier, C., Moyer, D.L., 1976. A unified mathe-
sity, a relatively large particle size and a low matical model for diffusion from drug-polymer composite
specific surface area will increase the drug release tablets. J. Biomed. Mater. Res. 10, 743 – 758.
Hastedt, J.E., Wright, J.L., 1990. Diffusion in porous materi-
rate. Alternatively, one can specify the necessary als above the percolation threshold. Pharm. Res. 7, 893–
coat characteristics after porosity modifier release 901.
that will yield the desired release rate. In general, Källstrand, G., Ekman, B., 1983. Membrane-coated tablets: a
if the coat provides for a large available solute system for the controlled release of drugs. J. Pharm. Sci.
72, 772 – 775.
fraction, a small coat volume, a large final poros- Mars, P., 1974. Compositie met vertraagde afgifte. Dutch
ity and a large increase in specific surface area, the Patent 7313696.
rate of drug release will increase. Saltzman, W.M., Langer, R., 1989. Transport rates of proteins
in porous materials with known microgeometry. Biophys.
J. 55, 163 – 171.
Shivanand, P., Sprockel, O.L., 1993. Release of propranolol
Acknowledgements HCl from a tablet coated with a macroporous membrane.
Int. J. Pharm. 92, 35 – 45.
The authors are grateful to the NIH for their Siegel, R.A., 1988. Modeling of drug release from porous
polymers. In: Rosoff, M. (Ed.), Controlled Release of
support, in part, with this project. Drugs: Polymers and Aggregate Systems. VCH, New
York.
Siegel, R.A., Kost, J., Langer, R., 1989. Mechanistic studies of
References macromolecular drug release from macroporous polymers
I. Experiments and preliminary theory concerning com-
pleteness of drug release. J. Control. Release 8, 223 –236.
Balazs, A.C., Calef, D.F., Deutch, J.M., Siegel, R.A., Langer, Siegel, R.A., Langer, R., 1990. Mechanistic studies of macro-
R., 1985. The role of polymer matrix structure and inter- molecular drug release from macroporous polymers II.
particle interactions in diffusion-limited drug release. Bio- Models for the slow kinetics of drug release. J. Control.
phys. J. 47, 97–104. Release 14, 153 – 167.
Carslaw, H.S., Jaeger, J.C., 1959. The Conduction of Heat in Stauffer, D., 1985. Introduction to Percolation Theory. Taylor
Solids. Clarendon Press, Oxford. and Francis, Philadelphia.
Carstensen, J.T., 1974. Theories of dissolution-single particu- van Campen, L., Zografi, G., Carstensen, J.T., 1980. An
late systems. In: Leeson, L.J., Carstensen, J.Y. (Eds.) Dis- approach to the evaluation of hygroscopicity for pharma-
solution technology. The Industrial Technology Section of ceutical solids. Int. J. Pharm. 5, 1 – 18.
the Academy of Pharmaceutical Sciences, Washington DC. Verhoeven, J., Schutte, S.C., Peschier, L.J.C., Danhof, M.,
Chandrasekaran, S.K., Paul, D.R., 1982. Dissolution-con- Junginger, H.E., 1989. The design of a dry-coated con-
trolled transport from dispersed matrixes. J. Pharm. Sci. trolled-release tablet for oxprenolol with microporous
71, 1399 – 1402. polypropylene powder. J. Control. Release 10, 205 – 217.
Thombre, A.G., Zentner, G.M., Himmelstein, K.J., 1989.
Conte, U., Colombo, P., Caramella, C., La Manna, A., 1983.
Mechanism of water transport in controlled porosity os-
Press-coated systems for drug release control. Polymers in
motic devices. J. Membr. Sci. 40, 279 – 310.
Medicine I: Biomedical and Pharmaceutical Applications.
Zentner, G.M., Rork, G.S., Himmelstein, K.J., 1985. Osmotic
Plenum, New York. flow through controlled porosity films: an approach to
Desai, S.J., Simonelli, A.P., Higuchi, W.I., 1965. Investigation delivery of water soluble compounds. J. Control. Release
of factors influencing release of solid drug dispersed in 2, 217 – 229.
inert matrices. J. Pharm. Sci. 54, 1459–1464. Zentner, G.M., McCelland, G.A., Sutton, S.C., 1991. Con-
Foster, T.P., Parrott, E.L., 1990. Release of highly water-solu- trolled porosity solubility- and resin modulated osmotic
ble medicinal compounds from inert, heterogenous ma- drug delivery systems for release of diltiazem hydrochlo-
trixes. I: Physical mixture. J. Pharm. Sci. 79, 806 – 810. ride. J. Control. Release 16, 237 – 244.