Combination of (M)DSC and Surface Analysis to Study the Phase Behaviour and Drug Distribution of Ternary Solid Dispersions

Miscibility of the different compounds that make up a solid dispersion based formulation play a crucial role in the drug release profile and physical stability of the solid dispersion as it defines the phase behaviour of the dispersion. The standard technique to obtain information on phase behaviour of a sample is (modulated) differential scanning calorimetry ((M)DSC). However, for ternary mixtures (M)DSC alone is not sufficient to characterize their phase behaviour and to gain insight into the distribution of the active pharmaceutical ingredient (API) in a two-phased polymeric matrix. MDSC was combined with complementary surface analysis techniques, specifically time-of-flight secondary ion mass spectrometry (ToF-SIMS) and atomic force microscopy (AFM). Three spray-dried model formulations with varying API/PLGA/PVP ratios were analyzed. MDSC, TOF-SIMS and AFM provided insights into differences in drug distribution via the observed surface coverage for 3 differently composed ternary solid dispersions. Combining MDSC and surface analysis rendered additional insights in the composition of mixed phases in complex systems, like ternary solid dispersions.


ABBREVIATIONS
. Phase behaviour of the different compounds plays a crucial role in this type of formulation as it will determine whether a given drug-polymer combination will be present as a glass solution (molecular dispersion) or a phase separated system (amorphous-amorphous or amorphous-crystalline phase separation).
Hence, the phase behaviour will have an influence on the release profile and stability performance of such a formulation. The technique of (Modulated) differential scanning calorimetry ((M)DSC) is generally used to study the phase behaviour and miscibility of a drug delivery system (DDS). Based upon the experimentally observed glass transition temperatures (T g s) the phase behaviour and composition of binary drug-polymer mixtures can be estimated by a number of expressions such as the Gordon-Taylor, Fox or Couchman-Karasz equation.
All of these equations are mathematical derivations on the same theme and were originally developed for ideal binary polymer systems (6). For binary pharmaceutical samples the Gordon-Taylor equation is the most frequently used expression to describe the relationship between the T g and sample composition (7). A similar expression for ternary mixtures has been developed but finds limited application due to the contradiction between the complexity of ternary mixtures and the assumptions made for this model equation (6). Therefore for ternary mixtures (M)DSC alone is not sufficient to characterize the composition of the different phases present, as for a mixed phase it is virtually impossible to determine the weight fraction of each of the three compounds. Knowledge of the composition of the different phases would inform on the distribution of a drug throughout the DDS.
Inhomogeneous drug distribution, such as surface enrichment, can only be detected by comparing the bulk characteristics of that formulation to its surface properties, indicating the need of surface analysis for ternary solid dispersions.
Information on both phase behaviour and spatial drug distribution is indispensable as these characteristics can be decisive for the release characteristics of a solid dispersion. For example, phase behaviour might influence the release, as particles containing a poorly soluble drug in the form of amorphous precipitates in the polymeric matrix are likely to have a slower release compared to ideal glass solutions where the drug is molecularly dispersed within a matrix. Spatial drug distribution might also influence the performance of a solid dispersion.
For instance, enrichment of drug at the surface results in an increased size or number of (amorphous) drug domains at the surface. This jeopardizes the increased solubility originating from the (molecular) dispersion of the poorly soluble drug. In that way surface-bulk distribution of the drug will determine the release kinetics of the sample. Additionally surface enrichment of an amorphous drug makes the solid dispersion more prone to surface crystallization and therefore might be detrimental to its stability. These issues highlight why surface characterization of solid dispersions is crucial in understanding how they perform with respect to dissolution and stability.
We have previously reported on the development of spray-dried polymeric microspheres for intramuscular injection for the long-term prophylaxis of infection with human immunodeficiency virus (HIV) (8,9). These shell structured microspheres contain two biocompatible polymers, poly(lactic-co-glycolic acid) (PLGA) and polyvinylpyrrolidone (PVP) and are made up of a PLGA-rich surface layer and an underlying PVP-rich phase (8).
For the present study a poorly soluble active pharmaceutical ingredient (API) was dispersed in these microspheres as a solid dispersion. The model drug used was an HIV protease inhibitor.
Three model formulations with varying ratios of API/PLGA/PVP were developed. As The main contribution of AFM for the analysis of solid dispersions lies in its potential to spatially resolve the phase separation at nanoscale resolution (9,(13)(14)(15)(16). Although this technique is not yet that frequently used for the characterization of pharmaceutical samples, it exhibits great potential as various mechanical properties (eg. stiffness, adhesion and friction) can be monitored to distinguish different compounds.
This study aims to clarify drug distribution in a multi-phased polymeric system. To do so, for the first time, (M)DSC was combined with ToF-SIMS and AFM to characterize the phase behaviour of ternary solid dispersions.

Materials
Poly   The inlet temperature was set to 115°C and the feed rate was 6 mL/min. The co-current drying air had a flow rate of 0.2 m³/min and the atomizing air was supplied with a pressure of 1.25 bar.

Modulated differential scanning calorimetry
The bulk miscibility behaviour of the spray-dried microspheres was determined by MDSC (Q2000, TA Instruments, Leatherhead, UK). Thermal Analysis Software (Version 4.4A) was used to analyze the obtained data. Crimped aluminum pans (TA Instruments, Brussels, Belgium) were selected for the analysis of the samples. An empty pan was used as a reference and the masses of the reference pan and of the sample pans were taken into account. The DSC cell was purged with a nitrogen flow rate of 50 mL/min.
Indium and octadecane were chosen for temperature calibration. The enthalpic response was calibrated with indium. The modulation parameters used were a heating rate of 1°C/min, a period of 40 s and an amplitude of 1°C. Calibration of the heat capacity was done using sapphire. Samples were analyzed from -20°C to 220°C. Glass transitions were analyzed in the reversing heat flow signals.

Scanning electron microscopy
SEM was used to study the morphology and particle size of the samples which were prepared by fixing an amount of powder on an aluminum stub using double-sided carbon tape. The samples were coated with a gold-palladium mixture by sputtering for 45 s at 20 mA. Field emission gun scanning electron micrographs (FEG-SEM) were taken by using a Philips XL30 ESEM-FEG instrument (Philips, Eindhoven, The Netherlands) at an acceleration voltage of 10 kV.

Time-of-flight secondary ion mass spectrometry
Spray-dried samples were adhered to double-sided adhesive tape in order to produce an immobile surface suitable for ToF-SIMS analysis. The data were acquired using a ToF-SIMS For any given sample, the measured secondary ion intensity for each polymer and API marker peak was normalized to the total intensity count to enable a semi-quantitative comparison of the different samples.

Atomic force microscopy
The spray-dried powders were applied onto a fresh mica surface using a Gilson pipette tip, and slightly blown with pressured nitrogen gas.

Morphological characterization of microspheres
The particle size and surface morphology of the microspheres of the different formulations were compared via SEM. Although a statistically significant number of particles were not analyzed, SEM could be used as an indicator for both particle size and morphology. Figure 2 shows that there are no striking differences regarding particle size and morphology when comparing the three model formulations. Particles are spherical, with a smooth, intact surface and an estimated diameter approximately between 1 µm and 7 µm.

Phase behaviour
MDSC was used to determine the T g s of the pure compounds, which were 38°C for PLGA, 56°C for the API and 174°C for PVP under the given experimental conditions. The phase behaviour of the model formulations was examined and the resulting thermograms are displayed in Figure 3. For each sample two mixing T g s were observed, the first one approximating to the T g of PLGA and the second one shifting towards the T g of pure PVP (Fig. 3). The known T g of the API (around 56°C) was not observed in the thermograms.

Surface chemical analysis
ToF-SIMS was used to chemically analyze the surfaces of the microspheres and hence solid dispersions. The spatial distribution of the API and PLGA at the sample surface is represented in Figure 4. There is a negligible amount of PVP observed at the particle surface which is consistent with previous studies (8,9) and depicted in Figure 2 of the Supplementary Information. At the micron-scale resolution of the ToF-SIMS data the drug seems to be homogeneously distributed at the surface of the three model formulations. For the formulation containing 20 wt% API a significantly higher intensity for the marker of the API is observed. Figure 5

Surface topographical analysis
Simultaneous mapping of topography and surface properties using QNM in PeakForce mode revealed the surface structure of the three formulations. observed with a smooth homogeneous surface evident (Fig. 6A-B, Fig. 8A-B). The 20 wt% API sample shows significant heterogeneity in surface topography (Fig. 7A-B). Similarly the corresponding stiffness maps only show significant heterogeneity in the 20 wt% API sample ( Fig. 6C, 7C, 8C).

DISCUSSION
MDSC was used to thermally study the phase behaviour of the three model formulations. For each sample two mixing T g s were observed, the first one approximating to the T g of PLGA and the second one shifting towards the T g of pure PVP (Fig. 3). In addition, the absence of a T g around 56°C (T g of the API) indicates that the API is molecularly dispersed in a phase separated system made up of a PLGA-rich phase and a PVP-rich phase. Preceding work has already revealed that hollow spheres are formed when PLGA and PVP are spray-dried together with a PLGA surface layer and an underlying PVP layer (8,9). The presence of PLGA at the particle surface is due to its surface activity and a consequence of restructuring of the methyl side groups of PLGA towards the droplet-air surface (8,19). Moreover these two polymeric phases display a small degree of miscibility which is determined by the PLGA to PVP ratio (8). The additional presence of an API will influence this phase behaviour.
Phase behaviour of the polymeric PLGA/PVP matrix was studied by estimating the polymer miscibility in these binary systems based upon the observed T g s using the Gordon-Taylor equation for binary mixtures (equation 1) (7).
In this equation, the weight fraction of each compound is represented by w, the glass transition temperature by T g and the subscripts 1 and 2 represent the compounds with the lowest and the highest glass transition temperatures respectively. The constant K can be assessed by using the Simha-Boyer rule (equation 2), with ρ being the density of the amorphous compounds.
Consequently the amount of PVP present in the PLGA layer and the amount of PLGA present in the PVP layer could be calculated. The PVP content in the PLGA phase was thus estimated to be 15% and the PLGA content in the PVP phase 3% (9).
The complex phase behaviour of ternary, or even quaternary (when taking residual solvent into account), systems is however difficult, if not impossible to assess using this equation. An extended form of the Gordon- Taylor equation for ternary mixtures has been developed   (equation 3).
With w representing the weight fraction of each compound, T g the glass transition where the subscripts 1, 2 and 3 represent the compounds with the lowest, middle and the highest glass transition temperatures, respectively. The constants K 1 and K 2 are analogously obtained by This extended formula was originally used to account for residual (plasticizing) solvent (20)(21)(22) and was expanded to the analysis of ternary solid dispersions (23). However, this extended form finds limited application due to the contradiction between the complexity of ternary mixtures and the assumptions made for this model equation. Generally the Gordon-Taylor equation assumes additivity of the specific volumes of the compounds involved as well as a linear change in volume with temperature (6). Obviously the likelihood that those assumptions apply decreases with increasing complexity of the system.  Figure 9.
The observed differences between the formulations studied can be explained by the fact that a change in formulation parameters (here a change in the ratio API/PLGA/PVP and thus composition of the feed solution) can influence particle formation during spray drying. A differently composed feed solution can result in different feed viscosity, evaporation rate or solidification point of the compounds. These parameters might affect the distribution of the API in the formed microspheres and therefore the degree of API surface coverage. This is illustrated by the Peclet number (P e ) which is used to predict compound distribution of the API during the particle formation process (equation 4).
In this equation the evaporation rate is represented by ĸ and D i stands for the diffusion coefficient of solute i. This formula clearly indicates how a change in evaporation rate might influence API distribution. Evaporation rate can be influenced by a varying feed composition as the affinity between solvent-polymer and solvent-drug molecules influences the evaporation rate. The extent of this influence will depend on the amount of polymer or drug molecules (with affinity for the solvent) present in solution (24). The potential influence of feed viscosity on API distribution and (surface coverage) is illustrated by the Stokes-Einstein equation which pinpoints the different parameters that influence the diffusion coefficient of a compound (equation 5).
In this equation D is the diffusion coefficient, r the globular radius, T the absolute temperature and η the viscosity of the solution. k B is the Boltzmann constant.
The higher the concentration of a compound in the feed solution, the closer it will be to its solubility limit and the faster it will reach the solidification point during the particle formation process. This possible difference of solute deposition time might also influence the spatial distribution of the API (25).
The lack of substantial differences in microspheres size and morphology (Fig. 2) when comparing the three model formulations indicates that differences in formulation parameters did not influence particle size and morphology.
In addition to influencing the particle formation process, changing the ratio of the three components might result in different mixed phases each displaying different solubility of the API and again a different distribution of the API in the spray-dried microspheres.

CONCLUSIONS
This study demonstrated how a combination of (M)DSC and the surface analysing techniques ToF-SIMS and AFM offer synergistic benefits for the characterisation of the phase behaviour and drug distribution of ternary solid dispersions (API/PLGA/PVP).
MDSC showed that the investigated formulations consisted of two mixed phases, a PLGArich phase and a PVP-rich phase, in which the API is present as a glass solution. ToF-SIMS informed on the spatial composition of the ternary solid dispersions: the PLGA-rich phase covers the surface of the microspheres, the PVP-rich phase is situated underneath. The degree of API surface coverage varies for the different formulations but is not in agreement with the bulk API concentration. AFM imaging and mechanical mapping coupled nanoscale spatial information about the microsphere surface to these findings, indicating structural and compositional heterogeneity in the 20 wt% API samples (compared to 10 wt% and 30 wt% API samples), which was consistent with their higher than expected surface coverage of API.
Interestingly, the distribution of the API in the ternary solid dispersions depended on formulation parameters. The extent of API surface coverage and therefore the distribution of the API over both polymeric phases differed significantly for the three formulations. As the location of the API in the microspheres might significantly influence the performance of the formulation (release behaviour and physical stability) insight into how formulation and process parameters influence the spatial distribution of the drug in these ternary solid dispersions would allow rational design of control release profiles and stability performance.