Phase behaviour analysis of solid dispersions of loperamide and two structurally related compounds with the polymers PVP-K30 and PVP-VA64

Article abstract of DOI:10.1016/j.ejps.2004.04.002

The purpose of the present study was to investigate the influence of the structure of a poorly water soluble model drug (loperamide) on the phase behaviour in solid dispersions with PVP-K30. Dispersions with PVP-VA64, a less hydrophilic polymer, were investigated as well in order to study the influence of differences in polymer structure and water content of the samples. The solid dispersions of PVP-K30 or PVP-VA64 with loperamide as well as with two fragments of this molecule were prepared by spray drying. The amount of residual solvents and water was determined with GC and thermogravimetric analysis (TGA). The drug loading of the dispersions was determined using high performance liquid chromatography (HPLC). The solid state properties were evaluated using powder-XRD, IR-spectroscopy and MT-DSC. All mixtures containing loperamide proved to be completely amorphous, whereas the dispersions containing the fragments are only amorphous in case the polymer content is high. The phase diagrams that were constructed clearly show that loperamide exhibits a different behaviour in the solid dispersions than its two building blocks. They also point to the presence of specific intermolecular compound-polymer interactions in the dispersions of one of the fragments with the two polymers. This was confirmed by the IR-results. Despite structural similarities, interactions in dispersions containing loperamide are far less important. In dispersions containing high concentrations of the other fragment, the DSC curves give indications for polymorphism whereas IR and XRD-spectra point towards inclusion of solvent in these samples.

Full text of DOI:10.1016/j.ejps.2004.04.002

European Journal of Pharmaceutical Sciences 22 (2004) 375–385
Phase behaviour analysis of solid dispersions of loperamide
and two structurally related compounds with the polymers
PVP-K30 and PVP-VA64
a
a
b
b
b
Ilse Weuts , Dieter Kempen , Annelies Decorte , Geert Verreck , Jef Peeters ,
b
a,∗
Marcus Brewster , Guy Van den Mooter
a
Laboratorium voor Farmacotechnologie en Biofarmacie, University of Leuven, Herestraat 49, B-3000 Leuven, Belgium
b
Johnson & Johnson Pharmaceutical Research & Development, Turnhoutseweg 30, B-2340 Beerse, Belgium
Received 3 December 2003; received in revised form 6 April 2004; accepted 14 April 2004
Available online 7 June 2004
Abstract
The purpose of the present study was to investigate the influence of the structure of a poorly water soluble model drug (loperamide) on the
phase behaviour in solid dispersions with PVP-K30. Dispersions with PVP-VA64, a less hydrophilic polymer, were investigated as well in order
to study the influence of differences in polymer structure and water content of the samples. The solid dispersions of PVP-K30 or PVP-VA64
with loperamide as well as with two fragments of this molecule were prepared by spray drying. The amount of residual solvents and water
was determined with GC and thermogravimetric analysis (TGA). The drug loading of the dispersions was determined using high performance
liquid chromatography (HPLC). The solid state properties were evaluated using powder-XRD, IR-spectroscopy and MT-DSC. All mixtures
containing loperamide proved to be completely amorphous, whereas the dispersions containing the fragments are only amorphous in case the
polymer content is high. The phase diagrams that were constructed clearly show that loperamide exhibits a different behaviour in the solid
dispersions than its two building blocks. They also point to the presence of specific intermolecular compound–polymer interactions in the
dispersions of one of the fragments with the two polymers. This was confirmed by the IR-results. Despite structural similarities, interactions
in dispersions containing loperamide are far less important. In dispersions containing high concentrations of the other fragment, the DSC
curves give indications for polymorphism whereas IR and XRD-spectra point towards inclusion of solvent in these samples.
©
2004 Elsevier B.V. All rights reserved.
Keywords: Differential scanning calorimetry; Solid dispersions; Phase behaviour; Crystallinity; Loperamide
1
. Introduction
Another approach to increase the solubility and dissolu-
tion rate is the use of solid dispersions (Leuner and Dress-
man, 2000). In these dispersions, the drug can be present in
the fully crystalline, fully amorphous or partially crystalline
and amorphous state. The enhanced dissolution properties of
eutectic mixtures (dispersions containing crystalline drug),
can be explained by both increased wettability and the cre-
ation of a microenvironment. The solubility of the drug par-
ticles is increased due to a high concentration of carrier in
the surrounding solution and the small size of the drug crys-
tals (Sekiguchi and Obi, 1961; Sekiguchi et al., 1964).
Often the amorphous state of the drug is preferred in solid
dispersions, since it shows a greater solubility and dissolu-
tion rate in comparison to the crystalline material. Glassy
solutions in which the drug is molecularly dispersed in the
carrier, represent the highest level of particle size reduction.
The enhancement of the oral bioavailability is currently
one of the greatest challenges in the development of poorly
water soluble drugs. One of the possible routes to overcome
this limitation is the use of salts. These salts may exhibit
a higher solubility in water and increased dissolution rate.
However, the use of salts is no guarantee for an increased
dissolution rate in the gastro-intestinal tract because of the
possible reconversion of salts into aggregates of their re-
spective acid or base forms (Serajuddin, 1999).
∗
Corresponding author. Tel.: +32 16 345829; fax: +32 16 345996.
E-mail address: guy.vandenmooter@pharm.kuleuven.ac.be
(G. Van den Mooter).
0
928-0987/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.ejps.2004.04.002

3
76
I. Weuts et al. / European Journal of Pharmaceutical Sciences 22 (2004) 375–385
After the carrier has dissolved, the drug is molecularly dis-
persed in the dissolution medium (Goldberg et al., 1966).
Furthermore, no energy is required to break up the crys-
talline structure of the drug (Taylor and Zografi, 1997).
Despite these benefits of the amorphous state with respect
to enhanced dissolution rate, it corresponds to a high energy
state and is physically metastable, especially when stored
relatively close to the glass transition temperature or Tg (less
than 40–50 K below Tg) (Hancock and Zografi, 1997). Often,
the addition of a polymer with a high Tg is sufficient to
prevent crystallization. The basis for this stabilization on a
molecular level is not yet clearly understood.
Some authors state that the stabilization of the amor-
phous drug is caused mainly by the presence of specific
drug–polymer interactions in dispersions of several com-
pounds with PVP (Khougaz and Clas, 2000; Matsumoto and
Zografi, 1999; Shamblin and Zografi, 1999; Taylor and
Zografi, 1997) and PVP-VA (Matsumoto and Zografi, 1999).
However, Van den Mooter et al. (2001) clearly demonstrated
that the antiplasticizing effect of PVP-K25 in solid disper-
sions with ketoconazole was the only stabilizing factor.
Loperamide (4-(4-chlorophenyl)-4-hydroxy-N,N-dime-
thyl-alpha, alpha-diphenyl-1-piperidinebutanamide), one of
the compounds under investigation in this study, is present
Fig. 1. Molecular structures of the compounds and polymers.
in Fig. 1. Three low molecular weight compounds (LMWC)
were used: loperamide and two fragments of this molecule.
The molecular structures are also included in Fig. 1. Lop-
eramide and F2 were available from Janssen Pharmaceutica
®
as a salt (hydrochloride) in imodium , a commercially
available antidiarrhoea agent marketed by Janssen Cilag
(Berchem, Belgium). In this formulation, the salt is used
(
Beerse, Belgium). The purity of the product was assessed
with the aim to increase the oral bioavailability of lop-
eramide which is poorly soluble in water (established by
Janssen to be lower than 0.001 g in 100 ml).
by means of high performance liquid chromatography
(HPLC)-methods and was 99.8 (loperamide) and 99.9%
F2). Fragment 1 had to be synthesized. The following
(
In this study, solid dispersions of loperamide (the free
base) and two smaller molecules, representing fragments of
loperamide (fragment 1 (F1): 4-dimethylamino-N,N-dime-
thyl-2,2-diphenyl-butyramide and fragment 2 (F2): 4-(4-
chlorophenyl)-4-piperidinol), were prepared. The study of
these compounds as such in the amorphous state was per-
formed previously (Weuts et al., 2003). The polymers used
were PVP-K30 and PVP-VA64 and the dispersions were
prepared by spray drying.
The aim was to investigate the physical structure and
phase behaviour of the obtained dispersions and to look for
specific interactions between the small molecules and the
polymers. Attention was paid to the possible implications
of these interactions on the thermal properties. Emphasis
was made on the influence of the chemical structure of both
the compounds and the polymer on these interactions and
physicochemical properties.
procedure was used.
N-(Dihydro-3,3-diphenyl-2(3H)-furanylidene)-N-methyl
methanaminium bromide (precursor, structure represented
in Fig. 1), dimethylamine (DMA) and methylisobutylketon
◦
(MIK) were put in a reactor and heated to 70 C. Since a
heterogeneous reaction takes place (the precursor does not
dissolve in the solvent) vigorous stirring during 24 h was
required. After this reaction time, excess of unreacted pre-
cursor was removed from the MIK-layer by extraction with
water, while F1 remained dissolved in the organic layer. An
acid–base extraction was performed by addition of HCl to
the MIK-layer. The corresponding salt was formed which
was extracted into a water layer. The free base (F1) was
liberated by addition of ammonia to this aqueous layer. The
crystals were removed from the solution by centrifugation,
◦
washed with water and dried at 50 C. The purity was de-
termined with HPLC and proved to be more than 99.9%.
The structure of F1 was confirmed by NMR-spectroscopy.
2
. Materials and methods
2.1.1. Preparation of solid dispersions
The solid dispersions were prepared by spray drying of so-
lutions with an overall concentration (compound + polymer)
of 5% (w/v) in a Büchi Mini Spray Dryer B191 (Büchi
Labortechnik AG, Flawil, Switzerland). The compound and
polymer were dissolved in a 1/1 (v/v) mixture of methanol
and dichloromethane. The evaporation temperature was set
2
.1. Materials
The polymers used in this study were PVP-K30 and
PVP-VA64, which were obtained from BASF (Lud-
wigshafen, Germany). Their molecular structures are given

I. Weuts et al. / European Journal of Pharmaceutical Sciences 22 (2004) 375–385
377
◦
at 80 C. However, dispersions containing a high amount
The results indicated that the samples still contained a
considerable amount of water, especially the dispersions
containing a high amount of PVP-K30 (0.4–8.2%).
of F1 (≥60%) have a low Tg (Tg of the pure substance is
◦
−
10.5 C; Weuts et al., 2003) and hence, spray drying from
this solvent mixture led to sticky samples. The solubility
of the polymers and F1 in pure dichloromethane was high
enough to be able to prepare the dispersions from solutions
in this solvent. Due to the low boiling point of this sol-
vent (40 C) a lower evaporation temperature could be used
50 C) and the obtained samples were much less sticky.
All adjustable parameters of the equipment were kept con-
stant (feeding rate: 7.5 ml/min, air flow: 400 (arbitrary units),
power of the aspirator: maximum). Solutions containing 5,
0, 20, 30, 40, 50, 60 and 80% (w/w) of LMWC relative to
the total amount of LMWC and polymer were spray dried.
2.2.3. High performance liquid chromatography(HPLC)
The exact LMWC content of the dispersions was de-
termined using appropriate HPLC-methods. All measure-
ments were performed using a Merck Hitachi pump L7100,
a Merck Hitachi autosampler L7200 and a Merck Hitachi
UV-detector L7400. A Hypersil BDS-C18 (3 ␮m, 4.0 mm
× 100 mm) column (Agilent, USA) was used. All mea-
surements were performed at room temperature, the flow
rate was set at 1 ml/min, the injection volume was 10 ␮l
and the detection wavelength was 220 nm (except for F2:
210 nm).
◦
◦
(
1
◦
After spray drying, the dispersions were dried at 40 C
under vacuum until constant weight, after which they were
For the determination of the content of loperamide an
isocratic method was applied, using a 70/30 (v/v) mixture
of 1 mM tetrabutylammonium hydrogen sulfate in water
and acetonitrile. Methods using a linear gradient were used
for the determination of the content of the other LMWCs.
The mobile phase for the determination of F1 consisted of
0.1 mM tetrabutylammonium hydrogen sulfate in water (sol-
vent A) and acetonitrile (solvent B). The ratio of the sol-
vents changed as follows: from 90/10 (v/v, A/B) to 50/50
during the first 15 min, to 0/100 during the next 7 min, fol-
lowed by an isocratic period of 7 min and a change to 90/10
over the last 3 min of the method. For F2, acetonitrile (sol-
vent B) was used together with a 10 mmol Na2SO4 solution
in water which had been adjusted to pH 2 (solvent A) using
H2SO4. The ratio of the solvents changed as follows: from
100/0 (A/B) to 60/40 during the first 20 min, to 5/95 during
the next 12 min, followed by an isocratic period of 3 min
and a change to the starting composition of 100/0 over the
last 2 min.
◦
stored in a dessicator over P2O in a freezer at −40 C to
5
avoid changes of the physical structure of the samples in
between preparation and the measurements.
2
.1.2. Preparation of physical mixtures
The physical mixtures were prepared by weighing the
appropriate amounts of crystalline LMWC (and amorphous
in the case of loperamide) and polymer and mixing them in
a mortar for 3 min.
2
2
.2. Methods
.2.1. Headspace gas chromatography (GCS)
Headspace gas chromatography was used to identify and
quantify the presence of residual organic solvents in the dis-
persions. The measurements were performed using a Var-
ian 3800 gas chromatograph (Varian gas chromatograph
CP3800, Walnut Creek, CA, USA) with splitter injector, FID
detector and a headspace sampler (Tekmar Dohrmann 7000,
Cincinnati, OH, USA). Approximately 100 mg of the sample
was dissolved in 2 ml 1,3-dimethyl-2-imidazolidinone in a
2.2.4. MT-DSC-measurements (modulated temperature
DSC)
2
2 ml headspace vial. Two 50 m silicagel columns (CP-SIL 5
CB: i.d. 0.32 mm, 5 mm and CP-WAX 52 CB: i.d. 0.32 mm,
.2 mm) were used. Nitrogen was used as a carrier gas. The
The MT-DSC-measurements (DSC-measurements with
temperature modulation) were performed on a Mettler-
e
1
Toledo DSC822 equiped with an intercooler and the re-
◦
◦
e
injector was held at 230 C and the FID detector at 270 C.
The oven temperature was changed from 40 to 220 C dur-
ing the analysis.
sults were analyzed with the STAR software version 6.10
◦
(Mettler-Toledo, Schwerzenbach, Switzerland). Mercury
and indium were used to calibrate the temperature scale and
enthalpic response. The calibration of temperature and en-
thalpy was validated at least every two days, using the same
standard materials. Deviation of the experimental value
from the theoretical one was less than 0.3 K for temperature
measurement and less than 2% for enthalpy measurement.
The samples (weight range 9–14 mg) were analyzed
in sealed 40 ␮l aluminum sample pans (Mettler-Toledo,
Schwerzenbach, Switzerland) with a pierced lid to allow
removal of water and other volatile substances. Samples
were weighed before and after measurement. The regis-
tered weight loss confirmed the TGA-results. The following
modulation parameters were used: underlying heating rate:
2 K/min, amplitude: 0.212 K and period: 40 s. The raw
The amount of residual solvent was for all samples
0.1%, indicating that the drying conditions after prepara-
≤
tion were adequate.
2
.2.2. Thermogravimetric analysis (TGA)
Thermogravimetric analysis was used to determine the to-
tal amount of volatile substances (organic solvents + water).
The measurements were performed with a TA Instrument
Hi-Res TGA 2950 (TA Instruments, New Castle, DE, USA)
and data station TA2100. Approximately 10 mg of sample
was weighed in an aluminum pan of 30 ␮l and heated from
◦
room temperature at a heating rate of 20 C/min. The end
point was set at 300 C.
◦

3
78
I. Weuts et al. / European Journal of Pharmaceutical Sciences 22 (2004) 375–385
data (modulated heat flow versus modulated temperature)
showed an undistorted sine wave indicating that the sam-
ples followed the modulation well. Measurements were
performed at least in duplo.
Table 1
Overview of the most important stretch vibrations of the pure substances
that can be affected by H-bonding
Substance
Wave number
Function
−
1
(cm
)
2
.2.5. X-ray diffraction measurements
PVP
1666
C=O (amide)
Powder X-ray diffraction was performed at room tem-
PVP-VA64
1677
1739
C=O (amide)
C=O (ester)
perature with a Philips PW Diffractometer (beam 173 mm).
Monochromatic Cu K␣-radiation (λ = 1.5406 Å) was ob-
tained with a Ni-filtration and a system of diverging, receiv-
Loperamide crystals
Loperamide glass
1614
3560 and 3390
C=O (amide)
OH (alcohol): two very
sharp peaks
◦
◦
ing and scattering slides of 1 , 0.2 mm and 1 , respectively.
The diffraction pattern was measured with a voltage of
1636
3411
C=O (amide)
OH (alcohol): broad band
◦
4
6
5 kV and a current of 20 mA in the region of 4 ≤ 2θ ≤
◦
◦
0 in a step scan mode of 0.02 per second.
F1
F2
1639
C=O (amide)
3281
NH (amino): very sharp
OH (alcohol): very broad
band
2
.2.6. Infra-red spectroscopy
IR-measurements were performed on a Perkin-Elmer
000 (Perkin-Elmer, Norwalk, CT, USA) at room temper-
3600–2800
2
ature using KBr-tablets containing approximately 0.35%
w/w) of dispersion. Analysis of the spectra was performed
(
g
In this equation Tgx stands for the T of the binary mixture
using the Spectrum v2.00 software (Perkin-Elmer, Norwalk,
CT, USA). Dispersions containing 10, 20, 30 and 80% of
LMWC were examined and compared with spectra obtained
from the pure substances and physical mixtures (containing
i gi i g
whereas w , T and ρ represent the weight fraction, T and
density of component i respectively.
This equation is valid in case of volume additivity, im-
plying that mixing of molecules of type 1 and 2, does not
lead to a volume contraction or expansion. This only holds
if the homomolecular interactions are of similar strength as
the heteromolecular intermolecular forces.
2
0 and 80% of LMWC). The pure substances of F1 and F2
were only available in the crystalline state. For loperamide,
both the crystalline and the amorphous form were at hand,
and were used to compare with the dispersions.
Hence, deviations between experimental T ’s and
g
these calculated ones indicate the presence of specific
compound–polymer interactions.
3
. Results and discussion
In order to study interactions between the polymer and the
LMWCs IR-measurements were performed. Table 1 sum-
marizes the peaks of some important stretch vibrations of
the pure components that can be affected by H-bonding.
H-bonding leads to a shift of the peak maxima towards lower
wave number (bathochromic shift) and very often the peak
width is increased.
From the MT-DSC-measurements, the transition temper-
atures of the different dispersions were derived. The total
heat flow signal was used to determine the melting tempera-
ture and heat of fusion. The glass transition temperature was
determined from the reversing heat flow signal. All transi-
tion temperatures were plotted as a function of composition
in order to obtain the phase diagrams. In the composition
axis the amount of polymer is expressed relative to the total
amount of dry substance (polymer + LMWC). The results
of the TGA- and GCS-measurements were used to correct
the contents obtained by the HPLC-experiments, in order to
obtain the concentration of LMWC relative to the amount
of dry product.
The difference in hygroscopic character of the polymers
is very clear in the IR-spectra. Not only the peak position
corresponding to the C
=
O band is shifted to a lower wave
number in the case of PVP-K30 indicating more H-bonding
(caused by the higher water content), also the shape of
the band is altered. The C
=
O band of the amide func-
tion is much broader in the PVP-K30 sample as in the
copolymer. The reduced amount of water leads to a smaller
extent of H-bonding in PVP-VA64. This is illustrated in
Fig. 2.
The phase diagrams also include a calculated Tg-curve
(points indicated as ᮀ) which was obtained using the
Gordon–Taylor/Kelley–Bueche equation (Gordon and
Taylor, 1952; Kelley and Bueche, 1961):
The amount of water of the different dispersions is listed
in Table 2. The values were obtained by subtraction of the
amount of organic solvent (determined with GCS) from
the total weight loss obtained by TGA. The table clearly
shows that the water content of the dispersions increases
as the amount of polymer increases. This effect is stronger
for PVP-K30 than for PVP-VA64 indicating the more hy-
drophilic character of the former. The data indicate that
w1Tg1 + Kw2Tg2
Tgx =
w1 + Kw2
with
ρ1Tg1
∼
K =
◦
ρ2Tg2
the drying procedure (40 C under vacuum until constant

I. Weuts et al. / European Journal of Pharmaceutical Sciences 22 (2004) 375–385
379
1
1
1
1
1
80
60
40
20
00
8
6
4
2
0
0
0
0
0
-20
0
0.2
0.4
0.6
0.8
1
(
a)
weight fraction PVP
1
1
1
1
80
60
40
20
Fig. 2. IR-spectra of PVP-K30 and PVP-VA64.
weight) is not sufficient to remove water that is being ab-
sorbed by the polymers.
100
8
6
4
2
0
0
0
0
0
3
.1. Solid dispersions with PVP-K30
The phase diagrams that were obtained from the disper-
sions of the three compounds with PVP-K30 are shown in
Fig. 3.
0
0.2
0.4
0.6
0.8
1
(
b)
weight fraction PVP
3
.1.1. Fragment 1/PVP-K30
Experimental glass transition temperatures could only be
1
1
80
60
determined in a small concentration range (weight fraction
of PVP (wPVP) ≥ 0.8). In the samples with a low content of
PVP (wPVP ≤ 0.4) the amount of amorphous material has
probably become too small to result in a detectable Tg. For
140
1
1
20
00
80
Table 2
Water content of the solid dispersions
6
0
Fragment 1/PVP
Fragment 2/PVP
Loperamide/PVP
0
0.2
0.4
0.6
0.8
1
(
c)
weight fraction PVP
LMWC Water
LMWC
(%)
Water
(%)
LMWC
(%)
Water
(%)
a
b
(%)
(%)
Fig. 3. Phase diagrams of PVP-K30 with the LMWCs: fragment 1 (a),
5
.3
4.2
4.3
3.6
3.3
2.3
2.0
2.1
1.1
5.0
10.2
20.4
30.9
40.9
51.4
63.4
84.8
6.8
5.7
4.4
5.0
3.5
3.2
2.5
1.3
5.2
8.9
8.2
7.7
6.3
5.1
4.0
2.5
1.8
1.5
fragment 2 (b) and loperamide (c) (᭡) Tm exp (᭜) Tg exp (ᮀ) Tg calc.
1
2
2
4
5
6
8
0.0
0.1
9.4
1.7
3.7
2.4
4.8
20.8
30.7
41.5
51.4
62.6
82.4
samples with an intermediate composition (0.4 < wPVP <
0.8) there is an overlap between the melting of the crystalline
drug and the glass transition. This also causes problems with
respect to Tg-determination.
The experimental Tg-values correspond very well to
the ones calculated by the Gordon−Taylor/Kelley–Bueche
equation, pointing to the absence of specific intermolecu-
lar forces in this binary system. This was confirmed with
IR-measurements as is being illustrated in Fig. 4. The peak
corresponding to the C=O stretch of the amide function
of PVP, is being shifted towards higher wave number as
the amount of F1 in the dispersion is being increased.
Fragment 1/PVP-VA
Fragment 2/PVP-VA
Loperamide/PVP-VA
LMWC
Water
(%)
LMWC
(%)
Water
(%)
LMWC
(%)
Water
(%)
(%)
5
9
.0
.9
2.8
2.1
2.2
2.0
2.1
2.0
1.4
0.4
4.9
9.9
2.2
2.2
3.1
1.9
2.3
0.9
2.4
0.5
4.5
10.2
20.1
30.6
40.5
51.3
61.6
82.3
2.7
2.0
1.8
1.7
1.3
1.2
0.7
0.6
1
2
3
4
5
7
9.3
8.0
6.6
5.6
7.4
7.3
20.1
30.3
40.0
50.3
59.2
79.0
−
1
It changes from 1666 cm for pure PVP, over 20% and
−
1
30% to 1683 cm for the dispersion containing 80% of
−1
the small molecule. The peak at around 1640 cm
that
a
is arising and increasing in intensity as the amount of F1
molecules increases, is being caused by F1. Furthermore,
%
%
(w/w) as determined with HPLC: average of two measurements.
(w/w): only one determination was performed.
b

3
80
I. Weuts et al. / European Journal of Pharmaceutical Sciences 22 (2004) 375–385
^
exo
Integral
-18.22 mJ
normalized -1.64 Jg^-1
Onset
Peak
91.44 ˚C
95.13 ˚C
5
mW
-
20
-0
20
40
60
80
100
120
140
160
˚C
1
0
20
30
40
50
60
70
80
90
100 min
e
K.U.Leuven Belgium: METTLER
METTLER TOLEDO STAR System
Fig. 4. IR-spectra of dispersions of PVP-K30 with fragment 1 molecules.
a) 0%, (b) 10%, (c) 20%, (d) 30% and (e) 80% of F1.
(
Fig. 5. DSC curves of dispersions of PVP-K30/fragment 1, containing 30
upper curve) and 80% (lower curve) of F1.
(
the broad band at 3464.2 cm 1 caused by H-bonding of
water present in the samples, decreases in intensity as the
amount of polymer decreases (not shown in the graph).
These observations indicate that the amount of H-bonding
of PVP lowers in the dispersions as the amount of LMWC
increases due to the fact that there is less water present in
these dispersions (confirmed by TGA; Table 2).
Crystalline F1 could be detected with DSC in dispersions
containing 30% of this substance or more. However, the
amount of crystalline material in this 30% dispersion is still
−
very small (column 4, Table 3a). Fig. 5 shows the DSC
curve (upper curve) of the 30% dispersion. Fig. 6 depicts
the XRD-spectra of several dispersions. From this graph,
crystallinity in the 30% dispersion is far from evident. Only
when the spectra are being enlarged, one can see minor
◦
traces of crystallinity (very shallow peaks at 2θ = 9.5 and
◦
1
3.8 ). From these two graphs it is clear that small amounts
of crystalline F1 (30% dispersion) are easier to trace with
DSC than with XRD.
The IR-data give evidence for the absence of crystalline
F1 in the dispersions with a low amount of F1 (10, 20 and
Table 3
Quantification of the heats of fusion of dispersions containing PVP-K30
30%). The absorption peaks corresponding to F1 are less
intense and sharp than in the spectra of the pure crystalline
compound and the physical mixtures. These sharp peaks
typical for the crystalline substance are only found in the
sample of the 80% dispersion. Probably the amount of crys-
talline material in the 30% dispersion is too low to detect
with IR-spectroscopy. In comparison to XRD and DSC, IR
is known to be the least sensitive technique to detect crys-
tallinity.
wPVP-K30
ꢀHm
J/g sample)
a) Fragment 1/PVP-K30
ꢀHm
(J/g F1)
F1c
(%)
PVP in
rest (%)
a
b
(
(
0
0
0
0
0
0
0
0
.703
.703
.583
.583
.463
.463
.376
.376
1.4
1.6
7.4
7.8
13.6
13.9
13.3
13.1
4.6
5.5
4.2
5.0
71.2
71.4
62.5
62.8
52.9
53.0
42.8
42.7
17.7
18.7
25.3
25.8
21.4
21.0
16.2
17.1
23.2
23.7
19.6
19.2
The DSC curve of the dispersion with composition
w
= 0.8 (Fig. 5, lower curve) shows signs of
fragment1
ꢀ
Hm
J/g sample)
b) Fragment 2/PVP-K30
ꢀHm
(J/g F2)
F2c
(%)
PVP in
rest (%)
a
b
(
(
800
0
0
0
0
0
0
0
0
.591
.591
.486
.486
.366
.366
.152
.152
1.8
1.3
10.0
8.5
16.4
15.2
72.5
71.1
4.4
3.2
3.4
2.4
60.0
59.7
52.6
52.0
41.8
41.4
34.0
33.2
6
4
2
00
00
00
0
19.5
16.5
25.8
23.9
85.5
83.8
14.9
12.6
19.7
18.2
65.3
64.0
a
%
Fic = 100 × mFi /mFi = 100 × mFi /(msample × wFi), mFi
c
c
c
◦
=
ꢀHm (J/g sample) × msample/ꢀH (J/g Fi).
m
b
%
PVP in rest = 100
×
mPVP/mrest = 100
×
msample
×
5
10
15
20
25
30
35
wPVP/(msample−mFi ).
2 theta
c
mFi , mFi, msample, mPVP: mass of crystalline Fi, Fi, sample and PVP,
c
respectively, wFi, wPVP: weight fraction of Fi and PVP in the sample
Fig. 6. XRD-spectra of dispersions of PVP-K30/fragment 1, containing
0, 30, 60, 80 and 100% of F1. The amount of F1 increases from the
lower to the upper curve.
◦
ꢀH
(J/g Fi): heat of fusion of pure compound Fi (=109 J/g for F1 and
m
1
31 J/g for F2).

I. Weuts et al. / European Journal of Pharmaceutical Sciences 22 (2004) 375–385
381
polymorphism. An overlap of several endo- and exothermic
events, seems to suggest the transformation of a less stable
into a more stable modification. However, this phenomenon
has also been investigated using other techniques (IR and
XRD) and the results of these experiments do not point to
the presence of polymorphs. The XRD- (see Fig. 6) and
IR-spectra look the same as for pure F1, not pointing to
polymorphism but to inclusion or entrapment of solvent in
the samples.
not only the result of the melting of F1-crystals. The energy
associated with the mixing of molten F1 and PVP is also
being recorded.
3.1.2. Fragment 2/PVP-K30
The phase diagram clearly shows a deviation between the
experimental and the calculated glass transition tempera-
tures. In dispersions with a low amount of F2, this deviation
is still quite small and increases suddenly when the amount
of F2 exceeds 20%. This sudden drop in the Tg-curve at
wPVP < 0.8 can be explained by the presence of water in
the samples. Due to the hydrophilic character of the poly-
mer, the samples contain a considerable amount of water (for
the exact values, see Table 2). The dispersions with a high
polymer content, have a relatively high Tg. By the time this
temperature is being reached, the water has evaporated from
the samples since pans with a pierced lid were used. Disper-
sions containing 80% of polymer or more have a Tg above
the boiling point of water. Hence the measured Tg corre-
sponds to a practically dry sample. Both the TGA and DSC
curves indicate that the evaporation of most of the water has
The melting points show a small melting point depres-
◦
sion upon addition of polymer (from 100.5 C for pure F1 to
9
◦
5.3 C for the dispersion containing 30% of this substance).
This is not unusual as melting point depression is a colliga-
tive property: it depends upon the number of molecules that
are being added. Since PVP-K30 is a polymer, addition of
a large amount by weight to F1 is required to have a con-
siderable amount of PVP-molecules and cause a clear melt-
ing point depression. A quantification of the measured heats
of fusion was made and the results are given in Table 3a.
Since the dispersion containing 80% of fragment 1 exhibits
an overlap of several endo- and exothermic events, no mean-
ingful heat of fusion could be calculated. Hence this disper-
sion is not included in the table.
◦
been completed around 110 C. The deviation between the
experimental and calculated Tg’s in these samples must be
caused by the presence of specific F2−PVP interactions.
Dispersions containing less polymer, have a lower Tg.
Hence, the samples still contain a considerable amount of
water by the time the glass transition temperature is met. The
very large deviation that is registered in these samples (with
wPVP < 0.80) is caused by a combination of two phenomena:
intermolecular forces and the plasticizing effect of water.
The existence of H-bonds between F2 and polymer was
investigated with IR-spectroscopy. The peak position of the
C=O stretch vibration of PVP in the dispersions is the same
as for pure PVP and does not change with the polymer
concentration which is illustrated by the spectra shown in
Fig. 7a. This implies that the degree of H-bonding is the
same in all samples eventhough the amount of water de-
creases as the amount of F2 increases in the dispersions
(see Table 2). Hence, the PVP must form H-bonds with an-
other molecule, fragment 2. It can form H-bonds with the
amide function of the polymer by means of the hydroxyl
and amino group. An inhibition of crystallization of F2 due
to these interactions as has been mentioned in some other
papers (Taylor and Zografi, 1997; Matsumoto and Zografi,
1999), was not observed.
These data indicate that the degree of crystallinity of F1
increases as its amount in the solid dispersion increases.
Furthermore, F1 does not seem to dissolve in the polymer
up to a certain fixed extent, since the amount of PVP-K30 in
the residual, vitrifying phase (last column) is not a constant
value, but decreases as wPVP of the dispersion decreases.
This is not what one would expect and two possible causes
can be found for this peculiarity.
The mass of crystalline material might be underestimated
by the occurrence of undetectable melting. By this phe-
nomenon the following is understood. In a mixture exhibit-
ing miscibility in the liquid but not (or not completely) in
the solid state (as is the case here), the melting point of one
component (F1) is lowered by the presence of the second
component (PVP). This was already mentioned earlier in the
text. However, not only the melting point is influenced, the
melting takes place over a much broader temperature range.
From DSC curves it is often evident where the melting ends
but often it is far less clear where it starts. In the case the on-
set of melting is not clear, accurate integration of the melting
peak is impossible resulting in erroneous values for the heat
of fusion. However, this so-called “undetectable melting” is
probably not the cause of the variation in % PVP in rest (see
Table 3a) since this variation is very big. This would im-
ply that the heats of fusion that were determined are much
too low (even a factor 3–4 for some samples), which is very
unlikely.
Crystalline F2 could be detected both with XRD and DSC
in dispersions containing 40% of F2 or more. This is illus-
trated by Figs. 8 and 9. The melting point depression is much
◦
stronger than in the system F1/PVP-K30 (from 139.5 C for
◦
pure F2 to 109.4 C for the dispersion with 40% of this com-
A second and more probable cause is that the heat of fu-
sion of the pure compound (109 J/g) that is being used to
calculate mF1c (mass of crystalline F1-molecules), is not a
representative ꢀHm-value for the small and imperfect crys-
tals that might be formed in the dispersions. Furthermore,
the heat that is being measured by the DSC-equipment is
pound). It is not only caused by the addition of polymer but
also by thermal degradation of F2 (data not presented in this
paper).
A quantification of the measured heats of fusion is given
in Table 3b. The degree of crystallinity of F2 increases as the
amount of LMWC in the dispersion increases. Also here, the

3
82
I. Weuts et al. / European Journal of Pharmaceutical Sciences 22 (2004) 375–385
6
5
4
3
2
1
00
00
00
00
00
00
0
5
10
15
20
25
30
35
2
theta
Fig. 9. XRD-spectra of dispersions of PVP-K30/fragment 2, containing
, 30, 40 and 100% of F2. The amount of F2 increases from the lower
0
to the upper curve.
theoretical Tg’s practically coincide. There is no substan-
tial deviation between both, which leads to the conclusion
that heteromolecular interactions are of the same magnitude
than homomolecular interactions. In samples with a rela-
tively high drug load, (wPVP ≤ 0.5) the deviation between
theoretical and experimental Tg-values must be ascribed to
the presence of water.This hypothesis was verified using
IR-spectroscopy. The position of the C=O stretch vibration
of PVP-K30 is not changed, except for the dispersion con-
Fig. 7. IR-spectra of dispersions of fragment 2 with PVP-K30 (a) and
PVP-VA64 (b). (a) 10%, (b) 20%, (c) 30% and (d) 80% of F2.
−1
taining 80% of drug (1683 cm ). The absence of inter-
molecular H-bonds between the polymer and loperamide is
thus less clear as in the case PVP-K30/F1. However, an in-
dication for this absence is the fact that the C=O band of the
carrier is narrower in the dispersions with the drug than in
the pure polymer. This indicates that PVP-K30 forms less
H-bonds in the dispersions compared to the pure state (with
water).
compound does not seem to dissolve in the PVP up to a cer-
tain fixed level. The amount of polymer in the residual, vitri-
fying phase is not a constant value. Possible reasons for this
discrepancy were already mentioned before (PVP-K30/F1).
3
.1.3. Loperamide/PVP-K30
The phase diagram shows two distinct regions. In the
Even though F2 has an identical structure as a part of
the loperamide molecule, interactions are no longer im-
portant. This can be caused by several facts. In the drug,
samples with a low drug content, the experimental and the
^
exo
Integral
normalized -1.32 Jg^-1
Onset
Peak
-12.45 mJ
0
.5
mW
104.69 ˚C
109.49 ˚C
2
1
0
0
40
20
60
30
80
40
100
50
120
60
140
70
160
80
˚C
min
e
K.U.Leuven Belgiu m: METTLER
METTLER T OLEDO STAR System
Fig. 8. DSC curve of a dispersion of PVP-K30/fragment 2 and 40% of F2.

I. Weuts et al. / European Journal of Pharmaceutical Sciences 22 (2004) 375–385
383
one site for hydrogen bonding is lost: the amino function
3.2.1. Fragment 1/PVP-VA64
Dispersions containing more than 50% of carrier exhibit
a glass transition. This T follows the calculated curve quite
which carried a hydrogen atom in F2, has become a ternary
amine in loperamide (see molecular structures in Fig. 1).
Hence, the amino group in loperamide can no longer act as
a proton-donor. Furthermore, loperamide is a much larger
molecule than F2, which might cause steric hindrance.
g
well. Especially the highest ones (w
PVP-VA
≥ 0.9) are in
agreement with theory, whereas a small deviation occurs in
the lower ones. The deviation is however much smaller than
in solid dispersions of PVP-K30 with F2 and loperamide.
Again the presence of water is at the origin of this phe-
nomenon. Since however PVP-VA64 takes up less water
than PVP (see Table 2), it is evident that the effect is smaller
than with the latter polymer.
−
1
The absence of the sharp peaks at 3560 and 3390 cm
typical for the crystalline drug, supports the amorphous
state of loperamide in all the solid dispersions. XRD-
and DSC-results also point towards fully amorphous
samples.
To exclude the fact that specific interactions might be
at the origin of this deviation, IR-measurements were per-
formed. The peak position of the C=O stretch vibration of
the ester function of the polymer is practically unchanged.
3
.2. Solid dispersions with PVP-VA64
The phase diagrams that were obtained from the disper-
sions of the three compounds with PVP-VA64 are shown in
Fig. 10.
=
The band corresponding to the C O stretch of the amide
function shows a shift towards higher wave number (going
−1
from 1677 for pure PVP-VA64 to 1684 cm ) as the amount
of F1 in the dispersion increases. Such a hypsochromic shift
is caused by a reduction in H-bond formation by the amide
function of the polymer. The shift is smaller than in the case
of PVP-K30 which is quite reasonable since PVP-VA in the
pure state forms less H-bonds (with water) than PVP.
Both DSC-measurements and XRD-data give evidence
for the presence of crystalline F1 in solid dispersions with
1
1
20
00
8
6
4
2
0
0
0
0
0
4
=
0% of this substance or more (very shallow peaks at 2θ
9.5 and 13.8 ). The XRD-spectra from Fig. 11 illustrate
◦ ◦
-20
the appearance and growth of typical crystalline peaks from
a drug load of 40% on. The IR-measurements confirm the
presence of crystalline material in the 80% dispersion. The
dispersions containing 60 and 80% of F1 exhibit again melt-
ing behaviour pointing to inclusion of solvent.
0
0.2
0.4
0.6
0.8
1
(
a)
weight fraction PVP-VA
160
140
120
100
In Table 4, a quantification of the integration of the melt-
ing peaks has been made. The number of data in the table is
limited since a quantification of the samples exhibiting mul-
tiple melting peaks is useless. Based upon the two disper-
sions in the table one might conclude that incorporation of
F1 molecules in the carrier does not take place to a certain
fixed extent, because in this case the amount of PVP-VA64
80
60
40
20
0
0
0.2
0.4
0.6
0.8
1
(
b)
weight fraction PVP-VA
1
20
10
00
1
1
9
8
7
6
0
0
0
0
0
0.2
0.4
0.6
0.8
1
(
c)
weight fraction PVP-VA
Fig. 11. XRD-spectra of dispersions of PVP-VA64/fragment 1, containing
0, 30, 40, 80 and 100% of F1. The amount of F1 increases from the
lower to the upper curve.
Fig. 10. Phase diagrams of PVP-VA64 with the LMWCs: fragment 1 (a),
fragment 2 (b) and loperamide (c) (᭡) Tm exp (᭜) Tg exp (ᮀ) Tg calc.

3
84
I. Weuts et al. / European Journal of Pharmaceutical Sciences 22 (2004) 375–385
Table 4
Quantification of the heats of fusion of dispersions fragment 1/PVP-VA64
Fig. 7b shows the spectra of the different dispersions of
F2 with PVP-VA64. The peak position resulting from the
absorption of the ester function (of VA) is not altered in the
wPVP-VA
ꢀHm
J/g sample)
ꢀHm
(J/g F1)
F1c
(%)
PVP-VA in
rest (%)
a
b
(
=
dispersions. The C O stretch vibration of the amide function
decreases as the amount of F2 in the dispersion increases
0
0
0
0
.634
.634
.545
.545
3.1
3.1
6.4
5.8
8.4
8.5
14.1
12.6
7.7
7.8
13.0
11.6
65.3
65.3
57.9
57.5
−1
−1
(
going from 1677 cm in pure PVP-VA, over 1676 cm
−1
in the 10 and 20% dispersion, 1668 cm in the 30% dis-
−
1
persion to 1663 cm in the 80% dispersion). This is a clear
indication for an increase of the amount of H-bonds that are
being formed by the amide function of the polymer upon
addition of F2.
a
%
Fic = 100 × mFi /mFi = 100 × mFi /(msample × wFi), mFi
c
c
c
◦
=
ꢀHm (J/g sample) × msample/ꢀH (J/g Fi).
m
b
%
PVP in rest = 100
×
mPVP/mrest = 100
×
msample
×
wPVP/(msample−mFi ).
c
mFi , mFi, msample, mPVP: mass of crystalline Fi, Fi, sample and PVP
This is where dispersions of PVP and PVP-VA with F2
deviate. With PVP the extent of H-bonding of the amide
function was not altered, resulting in a constant wave num-
ber, regardless the composition of the solid dispersion: the
H-bonds that were present in PVP in the pure state, are just
being replaced by H-bonds with F2 (see Fig. 7a). In PVP-VA
the amount of H-bonds in the pure state is much lower as
in PVP. Hence, the interactions between the amide function
of PVP-VA and F2, lead to an increase in the degree of
H-bonding and a lowering of the wave number.
c
respectively wFi, wPVP: weight fraction of Fi and PVP in the sample
◦
ꢀ
H
(J/g Fi): heat of fusion of pure compound Fi (=109 J/g for F1 and
m
1
31 J/g for F2).
in the residual phase would be a constant value. The same
remarks that were previously made when discussing the re-
sults of PVP-K30 with F1 and F2 regarding this topic can
be made for these dispersions as well.
3
.2.2. Fragment 2/PVP-VA64
The only dispersion in which crystalline LMWC was de-
Unlike in the dispersions of F2 with PVP-K30, the
compound−polymer interactions might be at the origin of
the lack of crystallinity of the compound (except for the
sample containing 80% of LMWC) (Taylor and Zografi,
1997; Matsumoto and Zografi, 1999).
For these dispersions again, the presence of crystalline
material in the 80% dispersion could be confirmed by the
IR-data.
tected is the one containing 80% of this compound. The DSC
curve shows a broad endothermic peak which is being pre-
ceded by a small exothermic signal. Hence crystallization
takes place upon heating. However, the unheated sample al-
ready contains an important amount of crystalline material
since the exothermic signal is much smaller than the melt-
ing peak. The presence of crystals at room temperature was
clearly confirmed by XRD-data (see Fig. 12).
3.2.3. Loperamide/PVP-VA64
In all solid dispersions a Tg could be registered. This
shows a clear deviation from the calculated values, indicating
the presence of important intermolecular interactions. This
was already mentioned when the dispersions of PVP-K30
with this substance were discussed. In these latter disper-
sions the results were a bit distorted by the effect of water.
In the mixtures of PVP-VA64 with F2, this water becomes
less important and the presence of interactions becomes very
clear.
In the solid dispersions of loperamide and PVP-VA64, no
traces of crystalline material are found. All samples show
a single glass transition. The deviation between the experi-
mental points and the calculated ones is much smaller than
in the analogue samples with PVP-K30. This is caused by
the fact that the dispersions with PVP-VA64 contain less
water due to the less hydrophilic character of the polymer.
Hence, the phase diagram points to the absence of specific
drug–carrier interactions.
The IR-spectra indicate that both C=O bands of the
polymer are unaffected. As was stated in the discussion
of the pure substances, H-bonding is quite unimportant in
PVP-VA64. The fact that the addition of loperamide in the
solid dispersions has no effect on either of the carbonyl
groups of the polymer, points to the absence of H-bonds
between drug and PVP-VA64.
4. Conclusions
The IR-data showed that F1 does not form H-bonds with
the polymers which is evident from its molecular structure:
it has no functional groups that can act as a proton donor.
Hence, the experimental and calculated Tg-values practically
coincide.
Fig. 12. XRD-spectrum of a dispersion of PVP-VA64/fragment 2, con-
taining 80% of F2.

I. Weuts et al. / European Journal of Pharmaceutical Sciences 22 (2004) 375–385
385
Specific fragment–carrier interactions were established in
the dispersions with F2. In the dispersions with PVP-VA64
they might be at the origin of the absence of crystalline ma-
terial (except in the 80% sample) whereas this is clearly not
the case in the dispersions with PVP-K30. Here, crystalline
compound could be traced in samples containing 40% of F2
or more.
Despite the similar molecular structure, loperamide does
not form H-bonds with the polymers. This might be caused
by the presence of the tertiary amino function (contrary to
the secondary amino function in F2), leaving only the hy-
droxyl group as a possible H-bonding site on the drug. How-
ever, steric hindrance might also prohibit H-bonding with
the polymers.
Gordon, S., Taylor, J.S., 1952. Ideal copolymers and the second-order
transitions of synthetic rubbers I. Non-crystalline copolymers. J. Appl.
Chem. 2, 493–500.
Hancock, B.C., Zografi, G., 1997. Characteristics and significance of the
amorphous state in pharmaceutical systems. J. Pharm. Sci. 86, 1–12.
Kelley, F.N., Bueche, F., 1961. Viscosity and glass temperature relations
for polymer–diluent systems. J. Pol. Sci. 50, 549–556.
Khougaz, K., Clas, S.-D., 2000. Crystallization inhibition in solid dis-
persions of MK-0591 and poly(vinylpyrrolidone) polymers. J. Pharm.
Sci. 89, 1325–1334.
Leuner, C., Dressman, J., 2000. Improving drug solubility for oral de-
livery using solid dispersions. Eur. J. Pharm. Biopharm. 50, 47–
60.
Matsumoto, T., Zografi, G., 1999. Physical properties of solid molec-
ular dispersions of indomethacin with poly(vinylpyrrolidone) and
poly(vinylpyrrolidone-co-vinylacetate) in relation to indomethacin
crystallization. Pharm. Res. 16, 1722–1728.
The performed study showed that the existence of spe-
cific interactions between polymer and LMWC is not a
requirement to obtain completely amorphous samples. Dis-
persions of loperamide with the carriers were all amorphous,
whereas IR-spectroscopy and the phase diagrams pointed
towards the absence of such interactions. This was already
shown for other dispersions as well (Van den Mooter et al.,
Sekiguchi, K., Obi, N., 1961. Studies on absorption of eutectic mixtures I.
A comparison of the behaviour of eutectic mixtures of sulphathiazole
and that of ordinarry sulphathiazole in man. Chem. Pharm. Bull. 9,
866–872.
Sekiguchi, K., Obi, N., Ueda, Y., 1964. Studies on absorption of eutectic
mixtures II. Absorption of fused conglomerates of chloramphenicol
and urea in rabbits. Chem. Pharm. Bull. 12, 134–144.
Serajuddin, A.T.M., 1999. Solid dispersion of poorly water-soluble drugs:
early promises, subsequent problems, and recent breakthroughs. J.
Pharm. Sci. 88, 1058–1066.
Shamblin, S.L., Zografi, G., 1999. The effects of absorbed water on the
properties of amorphous mixtures containing sucrose. Pharm. Res. 16,
1119–1124.
Taylor, L.S., Zografi, G., 1997. Spectroscopic characterization of inter-
actions between PVP and indomethacin in amorphous molecular dis-
persions. Pharm. Res. 14, 1691–1698.
2
001). In this respect, our data point to the importance of
the ease of vitrification and stability of the amorphous state
of the pure compounds. It has previously been shown that
from the three molecules that were studied, loperamide was
most easily obtained in the amorphous state (even upon
◦
very slow cooling: 0.2 C/min) and once in the amorphous
state, was the most stable glass (Weuts et al., 2003).
Van den Mooter, G., Wuyts, M., Blaton, N., Busson, R., Grobet, P.,
Augustijns, P., Kinget, R., 2001. Physical stabilization of amorphous
ketoconazole in solid dispersions with polyvinylpyrrolidone K25. Eur.
J. Pharm. Sci. 12, 261–269.
References
Weuts, I., Kempen, D., Six, K., Peeters, J., Verreck, G., Brewster,
M., Van den Mooter, G., 2003. Evaluation of different calorimetric
methods to determine the glass transition temperature and molecu-
lar mobility below Tg for amorphous drugs. Int. J. Pharm. 259, 17–
Goldberg, A.H., Gibaldi, M., Kanig, J.L., Mayersohn, M., 1966. Increasing
dissolution rates and gastrointestinal absorption of drugs via solid
solutions and eutectic mixtures IV. Chloramphenicol-urea system. J.
Pharm. Sci. 55, 581–583.
25.