`34 (2004) 945–956
`
`Penciclovir solubility in Eudragit films: a comparison of X-ray,
`thermal, microscopic and release rate techniques
`A. Ahmed a, B.W. Barry a, A.C. Williams a,∗
`
`, A.F. Davis b
`a Drug Delivery Group, School of Pharmacy, University of Bradford, Bradford, West Yorkshire BD7 1DP, UK
`b GlaxoSmithKline Consumer Healthcare Brands, Weybridge, Surrey, UK
`
`Received 27 July 2003; received in revised form 17 November 2003; accepted 21 November 2003
`
`Abstract
`
`The solubility of penciclovir (C10N5O3H17) in a novel film formulation designed for the treatment of cold sores was determined
`using X-ray, thermal, microscopic and release rate techniques. Solubilities of 0.15–0.23, 0.44, 0.53 and 0.42% (w/w) resulted
`for each procedure. Linear calibration lines were achieved for experimentally and theoretically determined differential scanning
`calorimetry (DSC) and X-ray powder diffractometry (XRPD) data. Intra- and inter-batch data precision values were determined;
`intra values were more precise. Microscopy was additionally useful for examining crystal shape, size distribution and homogeneity
`of drug distribution within the film. Whereas DSC also determined melting point, XRPD identified polymorphs and release data
`provided relevant kinetics.
`© 2003 Elsevier B.V. All rights reserved.
`
`Keywords: Eudragit films; Drug solubility; X-ray powder diffraction; Differential scanning calorimetry; Light microscopy; Release kinetics;
`Penciclovir
`
`1. Introduction
`
`[9-(4-hydroxy-3-hydroxymethylbut-1-
`Penciclovir
`yl)guanine], a synthetic nucleoside analogue,
`is a
`potent
`inhibitor of Herpes simplex virus (HSV1
`and 2). It has been marketed in topical preparations
`(Vectavir/Denavir) for the treatment of cold sores.
`A semi-solid polymer
`formulation of penciclovir
`(C10N5O3H17) has been developed, which upon ap-
`plication to the affected area, rapidly dries leaving a
`
`Corresponding author. Tel.: +44-1274-234756;
`∗
`fax: +44-1274-234769.
`E-mail address: a.c.williams@bradford.ac.uk (A.C. Williams).
`
`thin protective film. This layer is clear, dry to touch,
`substantive and aesthetically acceptable.
`It is important to characterise drug solubility within
`such a transdermal drug delivery system to understand
`and predict in vivo performance of the product [1]. The
`thermodynamic activity of the drug in the vehicle de-
`scribes the potential of the active ingredient to become
`available for its therapeutic purpose, i.e. the leaving
`potential. Higuchi [2] postulated that to achieve the
`maximum rate of drug penetration, the highest ther-
`modynamic potential should be utilised; this is usually
`a saturated system. The level of saturation depends on
`the amount and solubility of the drug in the vehicle
`and other factors such as the addition of solubility en-
`hancers (e.g. propylene glycol), which may result in
`
`0731-7085/$ – see front matter © 2003 Elsevier B.V. All rights reserved.
`doi:10.1016/j.jpba.2003.11.018
`
`Page 1
`
`Mylan v. MonoSol
`IPR2017-00200
`MonoSol Ex. 2023
`
`
`
`946
`
`A. Ahmed et al. / Journal of Pharmaceutical and Biomedical Analysis 34 (2004) 945–956
`
`a sub-saturated system and hence reduce the rate of
`drug delivery. Many formulations overcome problems
`caused through using solubility enhancers by adding
`excessive amounts of drug to the formulation, lead-
`ing to wastage of the active ingredient and poor ef-
`ficiency of the product. Increasing the drug loading
`within a preparation does decrease potential problems
`caused by depletion of the active ingredient. Contrary
`wise, a high solubility may reduce drug partitioning
`into the skin. Therefore, bases selected should balance
`optimum solubility and release properties [1]. Also,
`knowledge of the physical state of the drug (dissolved
`or suspended) in the vehicle is required to model ap-
`propriately its release kinetics [3].
`Hence, the solubility of a drug in its medium is
`an important determinant
`in formulation efficacy.
`However, it is difficult to measure such solubilities in
`semi-solids and films. Conventional methods such as
`filtration of a saturated drug solution and analysis [4]
`are inappropriate as it is difficult to remove excess
`crystals.
`techniques have been used in attempts
`Several
`to measure solubility in semi-solids and films. For
`oxybenzone, Kobayashi and Saitoh [5] collected the
`residual liquid separated from an ointment on stor-
`age and measured concentration. They confirmed the
`absence of crystals by microscopy, and the solubility
`determined by the residual liquid approach was in a
`range consistent with microscopic examination. Op-
`tical methods for solubility measurements have also
`provided accurate data. Gopferich and Lee [6] mea-
`sured clenbuterol solubility in polymer films; visible
`microscopy was the most sensitive of their techniques,
`with detection limit of 10% (w/w) compared to differ-
`ential scanning calorimetry (DSC) and release studies
`(limits were 12 and 13.5% (w/w), respectively).
`DSC has also been utilised to determine solubility of
`cholesterol in a silicone matrix [7] and for measuring
`propranolol [8], salicylic acid and chlorpheniramine
`[9] dispersed in polymer films. Plots of drug con-
`centration versus enthalpy of fusion and extrapolation
`to the intersect provided data for the drugs, although
`clearly these determinations provided solubilities at
`the melting point, not at room (or skin) temperature.
`Infra red attenuated total reflectance (IR-ATR)
`spectroscopy can determine solubility in acrylate ad-
`hesives [10,11]. The colorimetric determination of be-
`tamethasone in a topical vehicle by oxidation and then
`
`condensation of the 17 ␣-ketol group with phenylhy-
`drazine has also been successfully demonstrated [12].
`Chowhan and Pritchard [13] used partition data be-
`tween the vehicle and an aqueous phase, together with
`release data, to determine concentrations of corticoids
`in ointment bases.
`Interestingly, salicylic acid solubility in a hydrogel
`has been determined by X-ray powder diffractometry
`(XRPD). The intensities of salicylic acid peaks from
`its XRPD trace were linearly related to its weight per-
`cent in the formulation. The solubility of the acid in
`the hydrogel was taken as the intercept, determined to
`be 20% (w/w), but there was a large variance associ-
`ated with this measure [14].
`The objective of our work was to investigate the
`suitability of microscopy, DSC, XRPD and release
`experiments for determining penciclovir solubility in
`Eudragit NE30D films. Linear calibration lines were
`constructed for the DSC and XRPD data, intra- and
`inter-batch reproducibility of data was also deter-
`mined. The solubility values from each of the methods
`were compared and advantages and disadvantages of
`the techniques were considered.
`
`2. Experimental
`
`2.1. Materials
`
`Penciclovir (>99%, GSK, Weybridge, UK) was
`used as obtained, Eudragit NE30D (poly(ethyl acry-
`late methyl methacrylate)) was sourced from Rhom
`Pharma (Darmstad, Germany) and thickener Plasadone
`K90 (poly(vinylpyrolidine) (PVP)) was from ISP
`(Wayne, USA). HPLC grade methanol, buffer salt
`potassium dihydrogen orthophosphate and lithium
`fluoride standard (>99%) were supplied by Sigma
`(Dorset, UK).
`
`2.2. Formulation preparation
`
`The thickener PVP (0.5 g) was well stirred into
`the Eudragit NE30 dispersion (9.5 g). Penciclovir,
`0.025–10% (w/w) were mixed into the vehicle and
`equilibrated overnight; three batches for each drug
`loading were prepared. A film forming aid or plasti-
`cizer was not required since soft flexible films resulted
`◦
`after drying at 32
`C.
`
`Page 2
`
`
`
`A. Ahmed et al. / Journal of Pharmaceutical and Biomedical Analysis 34 (2004) 945–956
`
`947
`
`2.3. Film casting
`
`Films for microscopy, DSC and XRPD analysis
`were cast within a PVC template on a Teflon coated
`glass to obtain uniform sheets. The deposits were dried
`C for 24 h; thickness of dry films was 0.5 ±
`◦
`at 32
`0.1 mm (n = 30). Films were stored in a humidity
`◦
`cabinet at 32
`C at 38% r.h.
`
`2.4. Penciclovir loading in cast films
`
`Penciclovir content in the films were assessed by
`dissolving 100 mg samples in 10 ml ethanol before
`HPLC determination. Three samples at each drug con-
`centration from all three batches (n = 9), assessed
`drug homogeneity.
`
`2.5. HPLC analysis
`
`Penciclovir was analysed using a Hewlett-Packard
`−1,
`1100 HPLC instrument, with a flow rate of 1 ml min
`◦
`column temperature 30
`C, UV detection at λmax of
`254 nm and an injection volume of 100 l. The mobile
`phase was composed of methanol–potassium phos-
`phate (pH 7.0; 23 mM; 10:90 (v/v)), filtered and de-
`gassed. A guard column (Hypersil ODS C18 RP, 5 m,
`2.1 mm × 20 mm) cleaned the injected sample prior
`to separation on the main column (Hypersil ODS C18
`5 m, 150 mm). The method gave a linear response
`−1 with
`with concentration over the range 0–100 g ml
`r2 = 0.9999; limit of detection was 0.016 g ml
`−1
`−1.
`and limit of quantification 0.055 g ml
`
`2.6. Microscopy
`
`Films were examined under a visible microscope
`(Nixon labophot 2A) at 20× magnification for the
`presence of penciclovir crystals and photographs taken
`with a Nikon C35 camera (Nikon, Japan).
`
`2.7. Differential scanning calorimetry (DSC)
`
`Temperature and enthalpy were calibrated with an
`indium standard and the thermal behaviour of the films
`was examined using DSC (Perkin-Elmer Series 7) us-
`◦
`−1 over 25–300
`◦
`ing 10
`C min
`C. Samples in triplicate
`(8–10 mg), sealed in aluminium pans, were scanned
`against an empty reference pan. Since penciclovir con-
`
`tent ranged from 10 to 0.025% (w/w), sample weights
`were so as to maintain constant drug amounts.
`The enthalpy of fusion of penciclovir was calcu-
`lated from the melting endotherm using Perkin-Elmer
`Pyris Software. The solubility at its melting point was
`determined from the intercept of a plot of enthalpy of
`−1) versus drug loading (% (w/w)).
`fusion (J g
`
`2.8. X-ray powder diffractometry (XRPD)
`
`A Siemens D5000 powder diffractometer (Siemens,
`Karlsruhe, Germany) equipped with a scintillation
`counter detector produced film diffractograms. Af-
`ter calibration with lithium fluoride, samples were
`exposed to Cu K␣ radiation, wavelength 1.5418 Å,
`◦
`through 2 nm slits from 2 to 60
`2θ with a step size
`◦
`of 0.05
`2θ and a count time of 1 s per step; the
`generator was set to 40 kV and 30 mA.
`Samples (area = 3 cm2) were weighed and placed
`in holders with triplicate determination of Batch 1 and
`one analysis for Batches 2 and 3, allowing calculation
`of intra- and inter-batch variation.
`Integrated peak intensities (peak areas) were cal-
`culated from the diffractograms using GRAMS 32
`version 5 software (Galactic Industries Corporation,
`USA). Integrated data were produced for five peaks in
`each diffractogram (2θ = 8, 11, 17, 18, 26
`◦
`), summed
`and adjusted for sample weight. A plot of I/I0 (I: sum
`of five peaks at particular weight fraction, I0: sum
`of five peaks for pure penciclovir powder) versus the
`weight fraction of drug yielded an intercept that pro-
`vided the solubility.
`
`2.9. Penciclovir release studies
`Films were cast into holders (area = 1 cm2) and
`◦
`placed in a oven for 24 h at 32
`C. A modified USP
`XXI rotating paddle method [8] determined the re-
`lease. The receptor was 250 ml of a 10 mM pH 7.4
`phosphate buffer maintained at 32 ± 1
`◦
`C (represent-
`ing surface skin temperature) agitated by paddles at
`50 rpm ensured sink conditions. Aliquots were re-
`moved at intervals, analysed using HPLC and replaced
`by fresh media. Formulations were tested in triplicate
`and release data were plotted according to Eq. (10).
`Solubility was determined from the differences in
`the rate of increase in the release rate constant as a
`function of drug loading.
`
`Page 3
`
`
`
`948
`
`A. Ahmed et al. / Journal of Pharmaceutical and Biomedical Analysis 34 (2004) 945–956
`
`2.10. Precision of data
`
`Precision was assessed using percentage relative
`standard deviation (%R.S.D.) calculated as:
`%R.S.D. = S.D.
`× 100
`mean
`The precision for each point on the calibration plots
`was calculated for intra- and inter-batch data (n = 3).
`
`(1)
`
`3. Results and discussion
`
`3.1. Penciclovir concentration in films
`
`Casting the semi-solid formulations and solvent
`loss upon drying concentrated the drug in the re-
`sulting films. The concentration of penciclovir was
`determined using an HPLC assay (see Table 1). Data
`precision was within 4% R.S.D. indicating that the
`drug was homogeneously distributed for the 100 mg
`sample size tested.
`
`3.2. Microscopy
`
`Microscopic examination provided direct visual ev-
`idence for the presence or absence of solid penci-
`clovir in the films with needle shaped crystals evident
`at high concentrations (Fig. 1A). As the penciclovir
`loading decreased, the number of crystals declined un-
`
`Table 1
`Penciclovir concentration in polymer films pre- and post-casting
`
`Cast film
`
`Polymer film
`
`Fig. 1. Photomicrographs of penciclovir polymer film at (A)
`14.66% (w/w) and (B) 1.5% (w/w) drug loading.
`
`til at 0.23% (w/w) only a few fragments were visible;
`at 0.15% (w/w) none were apparent (Fig. 1B). Based
`on these observations, penciclovir solubility was esti-
`mated to be between 0.23 and 0.15% (w/w). The ab-
`sence of any fine powder suggested that amorphous
`material or solid dispersions of penciclovir within film
`components were not formed.
`
`Penciclovir
`concentration
`(% (w/w))
`10.0
`7.5
`5.0
`2.5
`1.0
`0.75
`0.5
`0.25
`0.15
`0.1
`0.05
`0.025
`
`Penciclovir concentration
`(% (w/w) ± S.D.; n = 6)
`
`Relative standard
`deviation (%)
`
`3.3. Differential scanning calorimetry
`
`14.66 (0.32)
`10.94 (0.26)
`7.39 (0.17)
`3.81 (0.12)
`1.54 (0.05)
`1.14 (0.03)
`0.77 (0.03)
`0.39 (0.015)
`0.23 (0.003)
`0.15 (0.004)
`0.076 (0.002)
`0.039 (0.001)
`
`2.2
`2.4
`2.3
`3.1
`3.2
`2.6
`3.9
`4.0
`3.8
`2.7
`2.6
`2.6
`
`The penciclovir powder gave a single sharp en-
`dothermic peak with a melting point and enthalpy
`C and 140 ± 5 J g
`−1 (n = 3) in
`◦
`of fusion of 278
`agreement with product data sheet. Broad melting en-
`◦
`dotherms resulted at 276
`C for drug films (Fig. 2). As
`drug loading fell, the enthalpy of fusion correspond-
`ingly decreased up to 0.39% (w/w), beyond which
`no penciclovir melting events were recorded, imply-
`ing that drug solubility was below 0.39% (w/w). The
`amorphous nature of the drug free films was shown by
`the absence of melting events and by a raised baseline;
`
`Page 4
`
`
`
`A. Ahmed et al. / Journal of Pharmaceutical and Biomedical Analysis 34 (2004) 945–956
`
`949
`
`Fig. 3. (䉫) Experimentally determined and (×) theoretically cal-
`culated enthalpy of fusion as a function of penciclovir loading in
`polymer films (n = 3). Error bars represent standard deviation.
`
`could in principle be used to estimate drug solubility.
`However, in practice this approach was not possible
`since the experimental line overlapped that of the the-
`oretical determination at low penciclovir levels and
`was raised above it at high drug levels. Higher than
`expected experimental enthalpies of fusion may have
`resulted because of drug interaction with the polymer;
`the broad endothermic event presented difficulties for
`an accurate determination of the integrated area. Ad-
`ditionally, the drug solubility was relatively low and,
`hence, the difference between theoretical and experi-
`mental lines was marginal; this approach may be more
`appropriate for systems with higher solubilities. Fur-
`ther difficulties arose due to the high water content in
`films (up to 25% (w/w)), the loss of which may further
`concentrate samples during analysis.
`From the intercept of the experimental line a sol-
`ubility of 0.44 ± 0.12% (w/w) (n = 3) was deter-
`mined. This was close to <0.39% (w/w), which was
`the minimum drug concentration at which endotherms
`were observed on the thermograms. Thermal analysis
`results were higher than those estimated from visible
`microscopy (0.15–0.23% (w/w)) but used room tem-
`perature, whereas the DSC approach estimated solu-
`bility at the drug melting point. Thus, a higher value
`was expected for the thermal method.
`For enthalpy of fusion values of penciclovir loaded
`films, precision of data expressed as %R.S.D. was
`good intra-batch (<5% R.S.D.) except for the 7.39
`and 3.81% (w/w) samples where one outlying repli-
`cate caused a large %R.S.D. value for both intra-
`and inter-batch (Table 2). As expected, large R.S.D.
`values for inter-batch data resulted in comparison to
`
`C) of
`Fig. 2. Differential scanning calorimetry profiles (250–300
`penciclovir powder, penciclovir-loaded polymer film at decreasing
`drug loading, and drug-free polymer film.
`
`◦
`
`(2)
`
`there were no interfering peaks at the drug melting
`point.
`The theoretical enthalpy of fusion as a function of
`drug loading was calculated from that of pure drug:
` Ht = xp Hp
`where Ht and Hp are the theoretical and pure pen-
`ciclovir enthalpies of fusion, and xp is the weight frac-
`tion of penciclovir [7]. Theoretical and experimental
`enthalpies were plotted versus drug loading (Fig. 3)
`resulting in linear calibration lines (r2 = 0.9989 for
`experimental line); error bars on the theoretical lines
`were due to calculation from three values of Hp. Ex-
`perimental and theoretical lines agreed well. Since the
`theoretical line does not take into account the solubil-
`ity of penciclovir in the film, whereas the experimen-
`tal plot does, the difference between the two graphs
`
`Page 5
`
`
`
`950
`
`A. Ahmed et al. / Journal of Pharmaceutical and Biomedical Analysis 34 (2004) 945–956
`
`Table 2
`Precision of differential scanning calorimetry data, intra- and inter-batches
`
`Concentration of penciclovir in
`the polymer film (% (w/w))
`
`14.66
`10.94
`7.39
`3.81
`1.54
`1.14
`0.77
`0.39
`
`Intra-batch
`
`Enthalpy of fusion
`(mean ± S.D.,
`n = 3) (J g
`−1)
`25.9 (1.2)
`19.4 (0.6)
`13.2 (1.7)
`6.1 (1.4)
`1.9 (0.05)
`0.9 (0.05)
`0.8 (0.02)
`0.7 (0.01)
`
`Precision
`(%R.S.D.)
`
`4.6
`3.4
`13.0
`23.0
`3.0
`5.0
`3.0
`2.0
`
`Inter-batch
`
`Enthalpy of fusion
`(mean ± S.D.,
`n = 3) (J g
`−1)
`24.3 (1.8)
`19.8 (0.7)
`10.9 (1.5)
`4.8 (1.0)
`1.6 (0.1)
`1.3 (0.4)
`0.6 (0.4)
`0.24 (0.02)
`
`Precision
`(%R.S.D.)
`
`7.4
`3.5
`13.7
`20.8
`6.2
`30.0
`66.0
`8.3
`
`(w/w)), but not at xp of 0.023 (0.23% (w/w)). This in-
`dicated that the solubility of penciclovir in the films
`lies between these values or that the XRPD method
`was not sensitive enough to detect crystals below xp
`of 0.039.
`Theoretical predictions of intensity ratio (Iip/(Iip)0)
`as a function of weight fraction of penciclovir (xp)
`were made [14,16–18]. The intensity (I) of line i of
`the penciclovir component (p) in a film (f) is given as:
`Iip =
`
`ρp[xp(µ∗
`
`p
`
`Kxp
`− µ∗
`
`f
`
`) + µ∗
`f ]
`
`(3)
`
`where K is a constant, ρp the density of penciclovir, xp
`and µ∗
`p are the weight fraction and mass attenuation
`coefficient (MAC) of penciclovir, and µ∗
`f is the MAC
`of the film.
`The intensity of peak i of a sample containing only
`penciclovir is given by:
`(Iip)0 = K
`ρpµ∗
`
`(4)
`
`p
`
`intra-batch measurements. The use of <10 mg sam-
`ples in analysis (drug content ranged from 0.025 to
`10% (w/w)) especially as drug concentration reduced,
`increased the margin of error, indicating the heteroge-
`neous nature of the samples. However, larger sample
`sizes (100 mg), such as those used to determine drug
`concentration in films, indicated homogeneity. A re-
`duction in crystal size may improve the distribution of
`drug crystals and hence reduce errors experienced at
`small sample sizes.
`Another problem inherent in this system was the
`large sample sizes utilized to improve sensitivity of the
`method, thus slower equilibration hence lag present
`between programmed and the actual temperature.
`The calibration line was verified by preparing a film
`−1.
`sample at 9.2% (w/w), where Ht was 17.03 J g
`The difference between this sample and that predicted
`−1) was 5.4% similar
`from the calibration line (16.2 J g
`to the %R.S.D. shown for inter-sample variability.
`
`3.4. X-ray diffraction
`
`The XRPD diffraction pattern of penciclovir (Fig. 4)
`agreed with literature; a single polymorphic form (nee-
`dle crystals) was identified. The diffraction pattern
`of the drug-free film gave a raised baseline but no
`clear diffraction lines, characteristic for an amorphous
`material [15]. Therefore, peaks in the XRPD traces
`for drug-loaded films arose from penciclovir crystals.
`Qualitative analysis of the XRPD traces showed a
`decrease in the peak intensities as the weight frac-
`tion of penciclovir (xp) fell. Peaks due to drug crys-
`tals were still present at an xp value of 0.039 (0.39%
`
`Therefore, the ratio of intensities of line i in a pen-
`ciclovir mixture to the identical
`line in a sample
`containing only penciclovir can be determined by
`dividing Eq. (3) by Eq. (4) to give the final intensity
`equation:
`=
`
`) + µ∗
`
`f
`
`(5)
`
`Iip
`(Iip)0
`
`xp(µ∗
`
`p
`
`xpµ∗
`− µ∗
`
`p
`
`f
`
`The MAC of elements are available in the litera-
`ture [19] and can be used to calculate the MAC value
`
`Page 6
`
`
`
`A. Ahmed et al. / Journal of Pharmaceutical and Biomedical Analysis 34 (2004) 945–956
`
`951
`
`Fig. 5. (䉫) Experimentally determined and ((cid:8)) theoretically cal-
`culated mean intensity ratio as a function of weight fraction of
`penciclovir in polymer films (n = 3). Error bars represent stan-
`dard deviation. (Linear regression analysis and polynomial fit of
`the experimental data was performed.)
`
`depended on the drug content. At the highest weight
`fraction (penciclovir:film ratio of 0.15:0.85 of which
`0.45 was poly-EA/MMA, 0.1 PVP and 0.45 water) the
`µ∗
`−1. At the lowest
`f was calculated to be 7.98 cm2 g
`penciclovir weight fraction (ratio of 0.01:0.99 pen-
`ciclovir:film) the µ∗
`−1.
`f determined was 8.24 cm2 g
`Therefore, the MAC used was calculated for an in-
`termediate composition, 0.08:0.92 penciclovir:film, a
`−1 resulted and was used in all the
`value of 8.11 cm2 g
`calculations.
`The determination of µ∗
`f and µ∗
`p allowed theoreti-
`cal determination of intensity ratio as a function of xp,
`a plot of which (Fig. 5) was found to be linear (r2 =
`0.999). In simple powder mixes, for example, -and
`␦-mannitol [20], or mixtures of crystalline and amor-
`phous leukotriene biosynthesis inhibitor (MK-0591)
`[15], the MAC of the unknown and film components
`were the same. In a multicomponent system like ours
`with different MAC values for unknown and film, al-
`though a linear relationship, resulted the slope was not
`1 but of 0.83. In some cases a linear relationship does
`not result and the data has to be fitted to a modified
`form of Eq. (5) [17].
`To determine the (Iip)0 from the pure penciclovir
`trace and Iip as a function of xp, integrated intensities
`of five of the most intense peaks were summed for each
`trace. More than one peak was utilised to minimize the
`influence of preferred orientation on peak intensity.
`Peak area instead of peak height was selected since
`variations in particle size range affect peak shape and
`hence height, but does not influence peak area [18].
`The Iip/(Iip)0 ratio as a function of xp was calculated
`from the experimental data, and a plot was non-linear;
`
`wkµ∗
`
`k
`
`(6)
`
`Fig. 4. X-ray powder diffractometry profiles (0–50
`2θ) of pen-
`ciclovir powder, penciclovir loaded polymer film at decreasing
`weight fractions and drug free polymer film.
`
`◦
`
`of a compound from its elemental composition using
`Eq. (6):
`
`µ∗ = n(cid:1)
`k=1
`where wk is the weight fraction of elements and µ∗
`k
`are their respective MAC values. For penciclovir
`(C10N5O3H17),
`the weight
`fractions of elements
`(0.47, 0.27, 0.19, 0.07) and MAC of elements 4.6,
`7.5, 11.5, 0.43, respectively [19], gave a calculated
`−1. The MAC of the film (µf) was
`µp of 6.41 cm2 g
`the sum of the weight fraction of each compound in
`the film multiplied by the compounds MAC. Since
`the film consisted of poly-EA/MMA, PVP and water,
`the MAC was calculated for each compound first,
`−1. In
`resulting in values of 6.47, 5.50, 10.28 cm2 g
`the film, the weight fraction of the film components
`
`Page 7
`
`
`
`952
`
`A. Ahmed et al. / Journal of Pharmaceutical and Biomedical Analysis 34 (2004) 945–956
`
`regression analysis gave r2-value of 0.979 with a slope
`of 0.91 (Fig. 5).
`intensity ratio
`The theoretical and experimental
`plots agreed poorly, due to the dynamic and complex
`nature of the multicomponent system with different
`MAC values (Fig. 5). The film results depended on the
`ratio of components, which had been estimated from
`the weight of the dry film. Since the polymer film
`contained a high weight fraction of water (0.40–0.45),
`fluctuations in its content would affect the experimen-
`tally determined values and hence differ from the the-
`oretical. Other sources of error included the fraction
`of drug dissolved in the film and possible formation
`of amorphous solid dispersions, with drug precipitat-
`ing out on drying of cast films. Suryanarayanan et al.
`[14] associated the large variance in the estimation
`of integrated peak intensity to amorphous scattering
`of X-rays by the non-crystalline ingredients. The dif-
`ference in the theoretical and experimental line could
`possibly be exploited to give a solubility value. How-
`ever, the experimental line was above theoretical and,
`since large errors were associated with it, the assess-
`ment was not possible. Use of this difference could
`potentially be of value in a system were the drug had
`a high solubility.
`The calibration curve was verified by preparing an
`additional film at 9.2% (w/w), which provided an in-
`tensity ratio of 0.112; the difference between this value
`and that from the calibration plot 0.105 was 6.7%.
`The precision of the X-ray data expressed as
`%R.S.D. was again lower intra-batch than inter-batch
`(see Table 3). The maximum R.S.D. of 3.8% intra-
`batch gave an indication of the instrumental errors,
`whereas the inter-batch maximum value of 16.2%
`R.S.D. resulted from a combination of instrumen-
`
`tal errors and between-batch sample preparation
`differences.
`Solubility calculated from the x intercept of the in-
`tensity ratio versus xp gave −1.68±0.43% (w/w) from
`linear analysis, thus the relationship is non-linear. A
`curve-fit of the experimental data gave a correlation
`coefficient of 0.9976, with an intercept xp = 0.053
`(0.53% (w/w); Fig. 5). Suryanarayanan et al. [14] suc-
`cessfully estimated the solubility of salicylic acid in a
`hydrogel at 20% (w/w) using XRPD. In our system,
`the solubility was <0.5% (w/w). Due to the relatively
`large instrumental and experimental errors associated
`with the method, it was deemed unsuitable for solu-
`bility determinations at very low concentrations.
`Numerous sources of error have been identified
`in quantitative XRPD analysis (e.g. [14,17]). Hurst
`et al. [21] separated errors associated with the XRPD
`method into three groups;
`instrumental,
`inherent
`properties of the compound and parameters related to
`preparation and mounting of samples. Some of these
`errors may have had a significant impact in our X-ray
`analysis.
`Preferred orientation is the non-random crystal
`packing in X-ray holders and is especially problem-
`atic with powder mixtures; it can give an error in peak
`intensity of up to 100%. Grinding samples and filling
`holders from the side usually combat this. In our sys-
`tem, this effect was minimised since the drug crystals
`were randomly oriented in a viscous semi-solid vehi-
`cle. Upon casting the film and drying, the crystals re-
`mained randomly oriented throughout the matrix (see
`Fig. 1). This was reflected in the diffraction pattern,
`where reproducible intensities resulted (Fig. 4).
`Another possible error arises from microabsorption
`effects [16]. When two substances of different MAC
`
`Table 3
`Precision of X-ray powder diffractometry data, intra- and inter-batches
`
`Concentration of penciclovir in
`the polymer film (% (w/w))
`
`14.66
`10.94
`7.39
`3.81
`1.54
`
`Intra-batch
`
`Inter-batch
`
`Sum of integrated intensities
`(mean ± S.D., n = 3; AU × 2θ)
`2530 (43)
`2109 (32)
`1479 (32)
`1044 (40)
`440 (13)
`
`Precision
`(%R.S.D.)
`1.7
`1.5
`2.1
`3.8
`3.0
`
`Sum of integrated intensities
`(mean ± S.D., n = 3; AU × 2θ)
`2434 (168)
`2082 (189)
`1527 (156)
`987 (160)
`421 (40)
`
`Precision
`(%R.S.D.)
`7.0
`9.1
`10.3
`16.2
`9.3
`
`Page 8
`
`
`
`A. Ahmed et al. / Journal of Pharmaceutical and Biomedical Analysis 34 (2004) 945–956
`
`953
`
`and for a homogeneous solution:
`
`(cid:3)1/2
`
`(12)
`
`D π
`
`(cid:2)
`
`kH = 2A
`
`where kH is the release rate constant, the slope of a
`plot Q versus t1/2.
`Several assumptions apply for Eqs. (8) and (9) in-
`cluding that the drug is homogeneously distributed
`throughout the vehicle, that only the drug diffuses out
`and that sink conditions prevail. Providing these con-
`ditions are met, then a plot of Q versus t1/2 should be
`linear for at least 30% of loaded drug released [2] as
`verified by Bodomeier and Paeratakul [8,23].
`Drug release from the films followed Higuchi ki-
`netics so a plot of cumulative amount released versus
`square root of time was linear. An initial ‘burst’ re-
`lease preceded a constant release rate phase (Fig. 6A).
`The initial rapid release can be attributed to dissolved
`drug and crystals at the film surface [8,23]. The sec-
`ond phase had a lower release rate than the first due to
`the receding boundary layer; drug must dissolve prior
`to diffusion through the film for suspensions, thus re-
`ducing release rate.
`The release rate constant obtained from the slope of
`Higuchi plots increased as drug concentration in the
`films rise. This elevation in release rate constant was
`−2 min
`−1/2) when drug was
`greater (0.0041 mcg cm
`dissolved because the concentration gradient between
`film and sink increased as drug loading approached
`saturation. Beyond saturation, when excess drug was
`suspended, the increase in the release rate constant
`as a function of drug loading was at a slower rate
`−2 min
`−1/2) since the concentration
`(0.0005 mcg cm
`gradient was at its maximum and drug has to disso-
`lute to maintain the concentration gradient. Hence,
`a plot of rate constant for each film as a function
`of drug loading revealed two slopes. Linear regres-
`sion analysis of data when drug was dissolved in the
`film gave a straight line, r2 = 0.999 and slope =
`−2 min
`−1/2. When drug was in excess
`0.0041 mcg cm
`(suspension), gave a straight line, r2 = 0.999 with a
`slope = 0.0005 mcg cm
`−2 min
`−1/2. Extrapolation of
`both lines (dotted lines in Fig. 6) to intersection es-
`timated drug solubility. The solubility of penciclovir
`in the polymer film determined from this method was
`0.42% (w/w), close to that estimated from visible
`microscopy studies.
`
`mix, this affects the accuracy of the intensity mea-
`−1)
`surements. Since MAC of penciclovir (6.41 cm2 g
`−1) are different, errors due to
`and the film (8.11 cm2 g
`microabsorption need to be considered.
`Sample thickness also requires consideration and
`should be adequate to prevent loss in intensity. The
`required sample thickness can be estimated from [17]:
`l ≥ 3.2 sin θ
`µ∗ρ(cid:10)
`
`(7)
`
`where l is the sample thickness, θ the incident angle
`◦
`), µ∗
`of the X-rays (8–26
`the mass attenuation co-
`efficient of the sample and ρ(cid:10)
`the density. The den-
`sity was calculated from sample area (7.1 cm2), thick-
`ness (0.05 cm ± 0.01 cm) and weight (0.2 ± 0.03 g),
`−3. The value of µ∗
`the density value was 0.63 g cm
`−1 (xp, xf = 0.15, 0.85) to
`ranged from 7.8 cm2 g
`−1 (xp, xf = 0.01, 0.99). Therefore, an in-
`8.3 cm2 g
`−1 was used in the cal-
`termediate value of 8.1 cm2 g
`◦
`culation. The l value resulting for 2θ values of 8
`is
`◦
`l > 0.08 and for 26
`was l > 0.27 cm. Since the aver-
`age sample thickness (0.05 cm) was less than that cal-
`culated for maximum diffracted intensity, results may
`have been erroneous through intensity loss due to in-
`adequate sample thickness.
`
`3.5. Release kinetics
`
`Higuchi [2,22] stated that release from a planar sys-
`tem having dispersed drug (suspension) or dissolved
`drug (solution) in a homogeneous film should follow
`the relationship:
`suspension Q = [D(2A − CS)CSt]1/2
`(cid:2)
`(cid:3)1/2
`solution Q = 2A
`Dt
`π
`
`(8)
`
`(9)
`
`where Q is the amount of drug released after time t
`per unit exposed area, D the diffusivity of the drug in
`the film, A the initial total drug concentration, and CS
`the drug solubility in the matrix.
`Both equations describe drug release as being linear
`with the square root of time:
`Q = kHt1/2
`For a homogeneous suspension:
`kH = [D(2A − CS)CS]1/2
`
`(10)
`
`(11)
`
`Page 9
`
`
`
`954
`
`A. Ahmed et al. / Journal of Pharmaceutical and Biomedical Analysis 34 (2004) 945–956
`
`Fig. 6. (A) Higuchi square root of time plots for penciclovir release from polymer films. ((䊊) 0.039, ((cid:8) ) 0.076, (䉫) 0.15, (×) 0.77, (䊐)
`) 3.81% (w/w)). (B) Release rate constant determined from Higuchi plots as a function of drug loading. ((䊐) drug dissolved in
`1.54, (
`the film, (䉫) drug dissolved and suspended in the film, (. . . ) extrapolation of best fit lines).
`
`3.6. Comparison of techniques
`
`A summary of the advantages and disadvantages
`of these solubility techniques is shown in Table 4.
`The visible microscopy (VM) method was simple,
`rapid and, together with the XRPD technique, was
`non-destructive and, thus, could be conducted prior
`to DSC and drug release studies, where samples
`were destroyed. Besides VM, the other approaches
`were more complex, time consuming, data analysis
`was complex and expensive instrume