RESEARCH ARTICLES

Drug Microencapsulation by PLA/PLGA Coacervation in the Light of

Thermodynamics. 2. Parameters Determining Microsphere Formation
CLAUDIO THOMASIN, HANS P. MERKLE, AND BRUNO GANDER*

Contribution from the Deparment of Pharmacy, ETH, Winterthurerstr. 190, 8057 Zuerich, Switzerland.

Received February 4, 1997. Accepted for publication December 19, 1997.

ular weight distribution, conformation and hydrophobicity

Abstract 0 Phase separation (frequently called coacervation) of poly- of the polymer and by the amount and viscosity of the
(lactide) (PLA) and poly(lactide-co-glycolide) (PLGA) is a classical coacervating agent. Recently, the importance of basic
method for drug microencapsulation. Here, attempts have been made physicochemical parameters of PLGA coacervation and
to describe this process in the light of thermodynamics. Different coacervation kinetics for the formation of empty and drug-
PLA/PLGAs were dissolved in either dichloromethane or ethyl acetate, loaded microspheres was demonstrated.8 Besides, only
phase separated by addition of the coacervating agent silicone oil little information is available on the mechanisms of PLA/
(PDMS), and hardened in either octamethylcyclotetrasiloxane or PLGA phase separation under conditions relevant for drug
hexane. Various stages of phase separation were defined microscopi- microencapsulation, although fundamental aspects of poly-
cally, and the coacervate and continuous phases characterized with mer phase separation have been covered thoroughly some
respect to volume, composition, polymer molecular weight, and time ago (reviewed in part 1).9 Nonetheless, Hildebrand
rheological behavior. The optimal amount of PDMS was inversely and Hansen solubility parameters have been considered
proportional to the polymer molecular weight and hydrophilicity, and in the phase separation of ethylcellulose to form micro-
a coacervate viscosity of above 5−10 Pa s was required for stable capsules.10-12 Further, a new thermodynamic model has
coacervate droplets. The composition and, consequently, viscosity also been developed to quantify drug-polymer-solvent
of the coacervate and continuous phases depended on the polymer− interactions and to characterize and predict aqueous
protein encapsulation into PLA/PLGA microspheres by
solvent−PDMS interactions, as analyzed by the parameters χ (Flory),
spray-drying.13
δ (Hildebrand), and ∆intE (Hô). In general, the lattice model of Flory
In this study, the process of phase separation of various
and Scott describing polymer−polymer incompatibility best explained
types of PLA/PLGA was analyzed in some detail by
the results. The interaction parameters and viscosity of the phases establishing phase diagrams defining the various stages
were also helpful to explain microsphere characteristics such as of phase separation and by characterizing the coacervate
residual solvent and particle size. The data suggest that microsphere and continuous phases in terms of volume, composition,
formation by polyester coacervation is primarily driven by molecular polymer molecular weight, and rheological behavior. At-
interactions between polymer, solvents, and coacervating agent. tempts were made to interpret the results in thermody-
namic terms using three different thermodynamic models
based on χ (Flory), δ (Hildebrand) and ∆intE (Hô, 1994).
Introduction The thermodynamic models used have been described in
more detail elsewhere (part 1).9 Further, the importance
Biodegradable microspheres of poly(lactide) (PLA) or of phase separation conditions for the quality of biodegrad-
poly(lactide-co-gycolide) (PLGA) have received considerable able polyester microspheres was outlined. Most impor-
attention as delivery systems for peptide and protein drugs. tantly, the purpose of this study was to examine PLA/PLGA
Microencapsulation methods have been developed respect- phase separation under conditions relevant for drug mi-
ing the particular properties of this category of compounds. croencapsulation and to interpret the observations with the
Although most of the more recent microencapsulation aid of thermodynamic parameters. Clearly, the study was
studies focused on the process of solvent evaporation or not intended to analyze thermodynamically the PLA/PLGA
extraction,1 phase separation remains very attractive for phase separation.
highly water soluble compounds. Incidentally, the terms
“phase separation” and “coacervation” will be used inter-
changeably in this text, because the latter is a very Experimental Section
commonly used term to describe PLA/PLGA microsphere MaterialssMedical grade poly(D,L-lactide) (PLA) (Resomer-
formation by phase separation. Initial work on peptide R104, -202, -203, -206, and -209), poly(D,L-lactide-co-glycolide) 75:
microencapsulation in PLA/PLGA by phase separation was 25 (PLGA 75:25) (ResomersRG752, -755, and -756), poly(D,L-
performed in the early 1980s.2-5 However, these studies lactide-co-glycolide) 50:50 (PLGA 50:50) (ResomersRG502, -503,
did not especially explore the crucial properties of solvents, -504, and -505), and poly(L-lactide) (PLLA) (ResomersL206) were
coacervating agents, and polymers. More insight into the purchased from Boehringer Ingelheim, Ingelheim, FRG, and used
phase separation process was gained when the different as received. The molecular weights M h w/M
h n (kDa) of the polyesters,
consecutive steps of PLGA phase separation in a dichlo- as determined by GPC,14 were 14.6/7.7 (R202), 23.3/10.6 (R203),
129.7/60.3 (R206), 513.0/232.0 (R209), 17.0/10.5 (RG752), 68.6/38.3
romethane/silicone oil system were described.6,7 These (RG755), 88.2/36.3 (RG756), 13.7/7.4 (RG502), 35.0/15.7 (RG503),
different steps were shown to be influenced by the molec- 41.2/19.1 (RG504), 68.2/24.4 (RG505), and 69.5/26.9 (L206). For
the R104, a M h w of approximately 1.9 kDa was indicated by the
* Tel: +41 1 635 60 12. Fax: +41 1 635 68 81. E-mail: manufacturer. As certified by the manufacturer, residual mono-
bruno.gander@pharma.ethz.ch. mer content was below 0.5% and residual acetone was below 0.1%

© 1998, American Chemical Society and S0022-3549(97)00048-8 CCC: $15.00 Journal of Pharmaceutical Sciences / 269
American Pharmaceutical Association Published on Web 01/31/1998 Vol. 87, No. 3, March 1998
for all samples. Dichloromethane (DCM) and ethyl acetate (EtAc) 12, Haake, Karlsruhe, FRG). For this, the coacervate dispersions
(analytical grade, >99.5%) were from Merck, Darmstadt, FRG. were centrifuged at 3500 min-1 in sealed vials and cooled to 10
Hexane (Hex) and silicone oil DC-200 (PDMS) of different viscosity °C. Shear stress was recorded in a solvent saturated atmosphere
(110, 375, 1070 mPa s) were obtained from Fluka, Buchs, (glass chamber sealed over the sensor; Contraves, Oerlikon,
Switzerland. Octamethylcyclotetrasiloxane (OMCTS) was from Switzerland) for 1 min by increasing automatically the shear rate
Scheller, Zürich, Switzerland. PDMS of 3000 mPa s was prepared from 100 to 1000 min-1. The temperature was maintained at 15
by mixing aliquots of PDMS of 1070 and 10 700 mPa s (Plüss & ( 1 °C for all experiments.
Stauffer, Oftringen, Switzerland) according to a viscosity-nomo- The viscosity of the continuous liquid was determined in a
gram.15 DCM and EtAc served as polymer solvents, PDMS as cylinder-type viscometer (VT 501, Haake, Karlsruhe, FRG) at
coacervating agent, Hex and OMCTS as hardening agents to increasing and decreasing shear rates of 27-2700 s-1.
solidify the microspheres. Turbidity of Coacervation DispersionssThe stability of
Solubility ParameterssThe Hansen parameters of PLA were coacervate droplets against merging and sedimentation was
determined theoretically according to a group contribution method.16 characterized by monitoring light transmission at 600 nm (UV/
The theoretical values were compared to those estimated experi- VIS-200, Perkin-Elmer Überlingen, FRG) of coacervate dispersions
mentally through the solubility behavior of PLA in three solvent prepared at different solvent/nonsolvent ratios.
classes differing in their hydrogen-bonding capability.17 Typically, Preparation of MicrospheressEmpty and BSA-loaded mi-
200 mg of polymer was mixed with 4 mL of solvent, and the crospheres were prepared by dissolving the polymer (2 g) in DCM
mixture was shaken vigorously and examined for complete solubil- or EtAc at concentrations of 2.5, 5.0, or 10% (w/w). Half of the
ity at 20 ( 1 °C. All solvents that did not completely dissolve the volume of the respective solutions was transferred into a jacketed-
polymer were also tested at 50 °C. Solvents inducing only polymer vessel (250 mL) equipped with baffles and an anchor stirrer. The
swelling or causing turbidity were classified as nonsolvents. For protein solution (500 µL) was emulsified in the other half of the
the copolymers PLGA 75:25 and 50:50, the solubility parameter δ polymer solution by ultrasonication, and this dispersion was also
was estimated experimentally rather than calculated from group added to the vessel. Coacervation was induced by introducing a
contributions. predetermined amount of silicone oil at a rate of 4 g min-1 to obtain
Phase Diagrams of PLA/PLGA CoacervationsA 20 g stable coacervate droplets according to established phase dia-
portion of 2.5, 5.0, or 10.0% (w/w) polymer solutions in DCM, using grams. Unless specified otherwise, stirring was set at 1000 min-1
different PLAs and PLGAs 75:25 and 50:50, was mechanically and the temperature maintained at 10-15 °C. The coacervation
stirred at 250 rpm in an airtight two-neck round-bottom flask kept dispersion was slowly transferred into 1200 mL of hardening agent
at 15 ( 1 °C. Methylene blue (Fluka, CH-Buchs) was dissolved to solidify the microparticles. Stirring was continued for 30 min
in the organic solution to enhance optical contrasts in the whereafter the microspheres were collected on a sintered glass
coacervation mixtures. Aliquots of PDMS were added at 2 min filter and washed with 100 mL of hexane. The microparticles were
intervals to increase the amount of nonsolvent by 2-5% relative air-dried for 5 min and resuspended three times in water contain-
to the total weight. Before each new silicone oil addition, 0.2 mL ing 0.1% poloxamer F 68 (ICI, Cleveland, UK). After washing,
of the mixture were withdrawn, examined, and photographed the powder was dried for 12 h under laminar air flow followed by
under an optical microscope (Optiphot with Microflex UFX, Nikon, vacuum-drying at 0.5 mbar.
JPN-Tokyo). From the appearance, size, and stability of the Residual Solvents in the MicrospheressBSA-loaded mi-
droplets, phase diagrams were established. For PLA R202, the crospheres prepared at three different DCM/PDMS ratios were
experiment was done in both DCM (Lewis acid) and EtAc (Lewis analyzed by GC for their solvent and hardening agent content.
base) to study the influence of the solvent type. Silicone oil Typically, 100 mg of each sample were weighed into screw-cap
addition and withdrawal of coacervation mixture was done through vials and dissolved in 2.0 g of 1,4-dioxane. The dispersions were
an inserted Teflon tube under rigorous precautions to prevent loss centrifuged at 3000 rpm for 10 min. The solutions were cooled to
of the highly volatile solvents. 10 °C, and the clear supernatant subsequently transferred into 1
GPC of Coacervated PolymerssFor two different DCM/ mL HPLC vials and sealed. Analysis was performed on a GC 8500
PDMS ratios, the coacervate phases were separated by centrifuga- (Perkin-Elmer, Ueberlingen, FRG) using a 10 m wide bore fused
tion and collected with a syringe. DCM was evaporated to avoid silica column of 0.53 mm internal diameter, an additional retention
interference in the GPC chromatogram, and the polymer was gap of 1.0 m, a stationary phase of 5 µm methylsilicone (PS 255)
redissolved in tetrahydrofuran (THF) spiked with toluene. GPC cross-linked with 2% dicumyl peroxide (BGB, Adliswil, Switzer-
conditions were identical with those described elsewhere.14 land), and the carrier gas nitrogen with a flow rate equivalent to
Appearance, Volume, and Composition of Coacervate a retention time for methane of 35 s. The injected volume was
and Continuous PhasessThe influence of polymer type and 2.0 mL and the injector temperature set at 200 °C. The oven
molecular weight on the appearance and the volume of the temperature was set at 40 °C for 5 min and then increased to 200
coacervate phase was determined at 15 ( 1 °C for three different °C at a rate of 15 °C min-1. The total analysis time was 40 min.
solvent/nonsolvent ratios. For this, concentrations of 2.5, 5.0, and The solvents were detected by high-sensitivity FID at 250 °C.
10% (w/w) of poly(L-lactide) and poly(D,L-lactide) were dissolved Size Distribution of MicrospheressMicrospheres were char-
in DCM and supplemented by varying amounts of PDMS. The acterized for their particle size by laser light diffraction (Master-
dispersions were transferred into graduated vials and centrifuged sizer SX, Malvern, Malvern, UK), as described elsewhere.14
at 3500 min-1 for 15 min. The separated coacervate and continu- Calculations were based on Mie’s theory accounting for the optical
ous phases were examined for their appearance and volume properties of the polymer used. The refractive index of PLA and
fractions. The experiments were conducted in sealed containers PLGA was determined according to a group contribution method.16
to prevent loss of the volatile DCM.
The actual composition of the continuous and coacervate phases
was determined for the coacervation mixtures containing 1 g of Results
PLA R202, 19 g of DCM, and increasing amounts (5-46 g) of
PDMS. First, the theoretical solvent content of the continuous Solubility Behavior of PLA/PLGA in Different
liquid was calculated from the difference of volumes between the Solvent ClassessThe solubility range of PLA and PLGA
continuous phase and the added PDMS. This calculation was in solvents with different hydrogen-bonding capacity is
made under the assumptions that all PDMS remains in the presented in Table 1. All three polyester types were
continuous phase and no volume change occurs upon mixing DCM completely insoluble in solvents of strong hydrogen-bonding
and PDMS. The DCM content of the coacervate was calculated capacity. By contrast, the solubility range in solvents of
from the difference between the total amount of DCM used (19 g) poor or intermediate hydrogen-bonding depended on the
and the estimated amount in the continuous liquid. The calculated polymer type. Generally, moderate H-bonding solvents
results were verified gravimetrically with 5 mL of the continuous
phase, after evaporation of DCM at 150 °C.
showed the widest solubility range, which decreased in the
Rheological Behavior of Coacervate and Continuous order PLA > PLGA 75:25 > PLGA 50:50. Poor hydrogen-
PhasessThe rheological behavior of the coacervate phase, isolated bonding solvents with <17.8 MPa0.5 were classified here
at different solvent/nonsolvent ratios, was determined for poly(L- as nonsolvents.
lactide) and poly(D,L-lactide) at polymer concentrations of 2.5, 5, The thermodynamic parameters of the components used
and 10% (w/w) with a plate and cone sensor viscometer (Rheovisc in the coacervation process are compiled in Table 2. The

270 / Journal of Pharmaceutical Sciences

Vol. 87, No. 3, March 1998
Table 1sSolubility Parameter Range of PLA and PLGA Polymers Table 3sCalculated Interaction Forces in Binary Mixtures Used for
PLA Coacervation
δ-range
H-bonding ∆δ ∆intEb
capacity of of solvents for PLA for PLGA75:25 for PLGA50:50 binary mixtures χa (eq 2) (eq 3)
solvents tested37 (MPa0.5) (MPa0.5) (MPa0.5) (MPa0.5) solvent−solute (eq 1) (MPa)0.5 (kJ mol-1)
poor 14.1−26.0 17.8−26.0 18.8−26.0 19.0−26.0 DCM−PLA 0.03 7.33 −50.04
moderate 15.1−30.1 17.4−30.1 17.4−30.1 18.2−30.1 EtAc−PLA 0.19 5.17 −73.22
strong 19.4−29.7 not soluble not soluble not soluble PDMS−PLA 0.86 10.42 −37.20
OMCTS−PLA 1.82 10.42 −37.20
hexane−PLA 1.85 14.22 −15.04
Table 2sThermodynamic Parameters for the Components Used in DCM−PDMS 0.48 7.62 −44.10
PLA Microencapsulation16,19,26,37-39 EtAc−PDMS 0.38 5.27 −67.44
Hansen Drago DCM−OMCTS 1.16 7.62 −44.10
molec molar Hildebranda EtAc−OMCTS 1.20 5.27 −67.44
compo- wt volume δd δp δh E C δ DCM−hexane 0.62 10.98 −34.66
nent (g mol ) (cm3 mol-1)
-1 (MPa0.5) (kJ0.5 mol-0.5) (MPa0.5) EtAc−hexane 0.54 9.12 −46.38
PLA 72.0b 56.5 15.8c 8.7c 11.1c 3.31d 2.00d 20.8 (m) OMCTS−PDMS 0.18 0.00c −58.17
DCM 84.9 63.9 18.2 6.3 6.1 1.76 0.23 19.8 (p) hexane-PDMS 0.02 4.03 −60.79
EtAc 88.1 98.5 15.8 5.3 7.2 3.31 2.00 18.6 (m) a
Calculated for T ) 298 °K. b ∆intE was calculated using V1 of the more
PDMS 74.1b 76.0 15.5 1.6 3.5 3.44e 3.07e 15.5 (p)
volatile solvent and the E and C parameters only for the mixtures where acid−
OMCTS 74.1b,f 76.0f 15.5f 1.6f 3.5f 3.44f 3.07f 13.1 (p)
base interactions can occur, i.e., those containing DCM as solvent. c Hansen
hexane 84.0 131.6 14.9 0 0 0 0 14.9 (p)
parameters for OMCTS were adopted from PDMS, hence ∆δ ) 0.
a p and m indicate poor and moderate H-bonding solvents. b Values for

the repeating unit of the polymers. c Values calculated from group contributions.
d
True values not available; values adopted from EtAc. e True values not
available; values adopted from CH3OCH3. f Values adopted from PDMS.

Hansen parameters of PLA were calculated according to a

group contribution method;16 for the copolymers, δ cannot
be calculated accurately by this method, because of the
nonrandom distribution of the repeating units. As ex-
pected, the solubility parameters of the coacervating
(PDMS) and hardening agents (OMCTS, hexane) are far
beyond the solubility range of PLA, whereas the δ values
of the solvents DCM and EtAc are in the same range as Figure 1sPhase diagram of poly(L-lactide) characterizing the phase separation
the δ of the polymer. process in a DCM/PDMS mixture. The dotted lines represent the boundaries
This solubility behavior was also in agreement with the between the stages 1−5, which can be microscopically distinguished during
estimated interaction parameters between various pairs of coacervation.
compounds used in the coacervation process. The values
in Table 3 were calculated according to three molecular process, the boundary conditions of PLA phase separation
interaction models, as presented in more detail in part 1 can be calculated, as detailed in part 1.9 Typically, in a
of this contribution:9 system PDMS/PLA/DCM, with a PDMS of 1070 mPa s (M hn
≈ 26 500 and x2a ≈ 350),18 and with a PLA of M h n of 7700
V1 (x2b ≈ 106), a volume fraction φ1 of 0.97 DCM was
χ≈ (δ - δ1)2 (1) calculated for (χc)2 according to:
RT 2

(x 1 1
)( ) 1
2
∆δ ) [4(δd1 - δd2)2 + (δp1 - δp2)2 + (δh1 - δh2)2]0.5 (2) (χc)2a,2b ) 0.5 + (4)
1 - φ1
x2a xx2b
∆intE ) -nV1δd1δd2 - nV1δp1δp2 - (E1E2 + C1C2) (3)
This equation was explained in more detail in part 1.9
Generally, intermediate to strong interactions, as rep- Therefore, the critical condition for polymer phase separa-
resented by low χ (Flory) and ∆δ (Hansen) values or high tion was reached when the total polymer concentration
interaction energies, ∆intE (Hô), were found between the (PDMS and PLA) was >3%. With the initial PLA concen-
solvents and PLA, on one side, and between the nonsol- tration of 5-10% used here, phase separation started
vents and solvents, on the other side. Moreover, the immediately upon PDMS addition as a result of polymer
interactions of solvent-PDMS and solvent-PLA are of repulsion.
similar order of magnitude. This may indicate that poly- If, for simplicity, the PLA/PDMS/DCM mixture is con-
mer desolvation by PDMS proceeds slowly, which should sidered as an apolar system, phase separation was also
facilitate the control of coacervation and reduce the risk of supported by the Van Oss theory (part 1).9 Taking into
polymer precipitation. When analyzing the data by a account the surface tensions of DCM (γ1 ) 27.2 mN m-1),20
nonparametric correlation test (Spearman), highly signifi- of PLA (γ2 ) 40 mN m-1, as estimated from group
cant correlations, rs, were found between the rank order contributions),16 and of PDMS (γ3 ) 21.5 mN m-1),19 the
of values ∆δ and ∆intE (rs ) 0.7253), between χ and ∆δ (rs corresponding equations (eqs 23 and 24 in part 1)9 pre-
) 0.7321), and between χ and ∆intE (rs ) 0.6882) (with rtab dicted polymer phase separation.
) 0.643 for R ) 0.01). PLA/PLGA Coacervation Phase DiagramssThe dif-
So far, coacervation was treated here primarily as a ferent stages of PLA phase separation are illustrated for
desolvation process, for which interaction parameters PLA L206 in Figure 1. Phase separation of this high
predict potential PLA/PLGA desolvation by PDMS ir- molecular weight PLA could be divided into five stages,
respective of the solvent used. When considering PLA/ each of them characterized by a typical microscopic ap-
PLGA coacervation as a polymer 1-polymer 2 repulsion pearance, rheological behavior, and phase volume. Gener-

Journal of Pharmaceutical Sciences / 271

Vol. 87, No. 3, March 1998
 Table 4sComposition, Volume, and Calculated Solubility Parameter
of Coacervate and Continuous Phases of 14.6 kDa PLA upon
Addition of Increasing Amounts of PDMS (stages 1−4 of Figure 1) to
a Solution of 1 g of PLA in 19 g of DCMc
coacervation stage
characteristics 1 2 3 4
composition (%, w/w)
PDMS 20 40 60 70
DCM 76 57 38 29
PLA 4 3 2 1

Coaca CLiqb Coaca CLiqb Coaca CLiqb Coaca Cliqb

volume (mL) 5.4 18.1 3.4 31.6 2.3 40.1 2.0 61.1
DCM content (%) 38.4 61.6 23.7 76.3 15.4 84.6 13.9 86.1
δ (MPa0.5) 19.9 18.2 20.0 17.4 20.1 16.7 20.2 16.4
Figure 2sAmounts of PDMS (1070 mPa s) defining the area of stable a Coac refers to the coacervate phase. b CLiq refers to the continuous liquid.
coacervate droplets for different polymer types and molecular weights. cThe solvent content of each phase is expressed as percentage of the total
amount of solvent used in the coacervation mixtures.
ally, the following stages were distinguishable: (1) Pseu-
doemulsion of PDMS droplets in DCM/PLA, (2) mixed
phase system of PDMS droplets in DCM/PLA besides DCM/ coacervating agent were the most prominent factors con-
PLA droplets in PDMS, (3) phase inversion and macro- trolling coacervation. For illustration, PLA coacervation
scopically detectable phase separation with the DCM/PLA required a higher percentage of PDMS when the solvent
phase starting to form unstable microdroplets merging EtAc (76% PDMS) was used instead of DCM (69% PDMS).
together, (4) stable dispersion of well-defined coacervate Obviously, partitioning of EtAc between coacervate and
droplets, and (5) aggregation and precipitation of polymer continuous liquid was lower, indicating a relatively weak
droplets (solid-liquid demixing). The phase borderlines interaction between this polymer solvent and PDMS.
greatly depended on the polymer type and molecular Composition of Coacervate and Continuous
weight and on the interaction between polymer-solvent LiquidsTable 4 shows some physicochemical character-
and solvent-nonsolvent. For microencapsulation, stages istics of coacervate and continuous liquids taken from
3 and 4 were of utmost importance, as the physicochemical typical stages during coacervation. The polymer selected
properties of these mixtures were found responsible for the for this analysis was a low M h w PLA because this polymer
wetting, spreading, and drug engulfment processes.21 type was found less subject to rapid desolvation and solid
The stability behavior of the coacervate droplets in stage state aggregation (stage 5). PLA phase separation was
4 was critical for the successful transfer of the coacervate induced by very low amounts of PDMS, because a single
dispersion into the hardening agent and the isolation of phase could only be observed at PDMS concentrations <2%.
discrete microparticles therefrom. If the borderline to stage Above this concentration, two distinct phases were detected
5 was crossed due to excessive PDMS addition, the coac- after centrifugation of the coacervation dispersion. The
ervate droplets aggregated irreversibly. Uncontrolled phase coacervate phase (Coac) primarily consisted of PLA and
separation was most critical for PLGA 50:50 and poly(L- DCM, whereas the continuous liquid (CLiq) contained
lactide), whereas PLGA 75:25 and PLA were more tolerant PDMS and DCM. We assume that most of the wall
to aggregation and readily formed single, partly gelled forming polymer had already separated at this stage and
droplets. that only low M h w lactide oligomers remained possibly in
The amount of PDMS defining the borderlines of stage the continuous phase. This assumption was supported by
4 and necessary to control the formation of well-defined the observation that the evaporation of DCM from the
coacervate droplets was determined for a whole range of continuous phase did not produce any turbidity. However,
polymers (Figure 2). While the homopolymers were char- the calculated solubility parameter value of CLiq in stage
acterized by a broad zone of stable coacervates, PLGA 75: 2 was very close to the solubility range of PLA (17.8-26
25 and even more PLGA 50:50 coacervates were stable only MPa0.5), indicating that this continuous liquid had enough
within a narrow range of PDMS concentration. The solvent power for PLA oligomers only. By raising the
process of polymer phase separation was fastest for PLGA amount of PDMS from 20 to 70%, the separated PLA phase
50:50, followed by PLGA 75:25 and PLA. Interestingly, the was continuously desolvated, as reflected by a decrease in
isomeric semicrystalline PLLA was coacervated at a sub- the coacervate volume and solvent content. At stage 3,
stantially lower PDMS concentration than the racemic polymer phase separation was complete because the solu-
amorphous PLA; this reflects another mechanism of phase bility parameter of the continuous liquid was below 17
separation, which is formation of crystalline domains in MPa0.5 and, hence, outside the δ-range of PLA.
the phase-separated droplets.40 Finally, an inverse pro- With increasing amount of PDMS, DCM was progres-
portionality was found when the midpoint of the necessary sively extracted from the coacervate into the continuous
PDMS amount was plotted versus M h w of the polymers; the phase. At stage 4, approximately 15% of the total DCM
linear correlation coefficients were r ) 0.987 for PLA, r ) content remained in the coacervate phase, which still
0.995 for PLGA 75:25, and r ) 0.987 for PLGA 50:50. represented more than 70% of the coacervate mass. Poly-
Although the polymer molecular weight distribution should mer solvation by DCM was still sufficient to prevent
also affect phase separation, this was not examined in this uncontrolled gelling or aggregation.
study, but has been addressed recently by other authors.6,7 Molecular Weight of Coacervated PolymerssThe
For PLA, linear correlation applied up to a M h w of 130 chromatographic analysis of PLA and PLGA coacervates
kDa. For the 513 kDa PLA (R209), the amount of PDMS did not reveal any fractionation of polymer molecular
(35%) necessary to stabilize the coacervate was higher than weight at the end of PDMS addition, irrespective of the
expected. In thermodynamic terms, the interactions be- polymer type and M h w. Although PLA oligomers could
tween polymer and solvent and between solvent and possibly remain in the continuous liquid, the latter could

272 / Journal of Pharmaceutical Sciences

Vol. 87, No. 3, March 1998
Table 5sCharacteristics of PLA and PLLA Coacervates
polymer polymer DCM PDMS coacervate coacervate light
type concn volume viscositya transmission
(%, w/w) (%, w/w) (%, w/w) (%, v/v) (mPa s) (∆% T/h)
L206 1.7 65.1 33.2 9.1 643 5.2
1.2 48.0 50.8 5.0 5932 19.9
1.1 41.7 57.2 2.4 ndb,c 2.1
3.3 63.4 33.3 15.7 1724 24.4
2.5 47.3 50.2 7.1 12011 7.6
2.1 40.4 57.5 5.8 ndb,c 1.1
6.6 59.4 34.0 23.8 1890 30.5
5.1 46.0 48.9 13.3 16239 6.1
R206 1.7 66.0 32.3 8.6 1385 32.3
1.3 48.8 49.9 2.8 9474 3.2
1.0 37.4 61.6 2.4 ndc 7.5 Figure 3sResidual processing liquids in 14.6 kDa D,L-PLA microspheres
obtained from different coacervation compositions: polymer solvent DCM
a Apparent viscosity at D ) 500 s-1. b Coacervates tended to gel. c nd:
(unfilled) and hardening agent OMCTS (filled).
not determined because viscosity was too high.

not be analyzed by GPC, as PDMS would interfere with

the chromatogram of oligomeric lactides.
Coacervate Volume and ViscositysThe addition of
PDMS to the PLA or PLGA solutions induced phase
separation, which was accompanied by volume and viscos-
ity changes in the coacervate and the continuous liquid.
Coacervate volume and viscosity were inversely propor-
tional and depended on polymer concentration and molec-
ular weight (Table 5). The viscosity of PLLA (L 206) and
PLA (R 206) coacervates showed a Newtonian behavior for
PDMS concentrations up to 35%. In this range (of unstable
coacervate droplets), the viscosity was only slightly affected
by increasing polymer concentration and attained values
of 0.6-1.8 Pa s, depending on the polymer type and Figure 4sEffect of the coacervating agent (PDMS) viscosity on the viscosity
molecular weight. Toward the midpoint of the stable of the continuous phase (dashed line) and the volume weighted particle size
coacervate range, the viscosity increased markedly, and the distribution of 68.6 kDa PLGA 75:25 microspheres. The bars represent the
behavior became pseudoplastic. At shear rates E500 range of 10−90% undersize diameters and the bold line inside the bar the
min-1, viscosities were 12-17 Pa s, depending on the average particle size.
polymer concentration. At higher polymer concentrations,
i.e., 10% initial polymer concentration, the coacervates With respect to the total amount of residuals, i.e., polymer
were too viscous for isolation and handling. solvent plus hardening compound, a low DCM/PDMS ratio
Coacervate droplets of stage 4 of the phase diagram were (1:4) produced microspheres with relatively low organic
stabilized by their own viscosity as well as by the increased liquid content. To minimize mainly residual DCM, a high
viscosity of the continuous liquid. Both factors counter- solvent:PDMS ratio (1:1.5) appeared more suitable. On the
acted merging and sedimentation, which was of great other side, EtAc residues were not analyzed here, but
importance for controlling microsphere size. Demixing similar findings may be expected, as shown recently in a
kinetics of the coacervate and continuous liquids was study on residual solvents in PLA/PLGA microspheres
followed by turbidimetric measurements. A pronounced prepared by coacervation.30 Interestingly, EtAc residuals
increase in the light transmission was found for coacerva- in microspheres were more than 10-fold higher than DCM
tion mixtures of relatively low PDMS content. When the residuals, which might be explained, at least partly, by the
PDMS concentration was increased from approximately 33 solubility parameters, but not by the Flory-Huggins model.
to 50%, or even to 60%, turbidity first increased markedly Effect of Coacervation Conditions on Micosphere
and then remained almost constant. An increase in light Particle SizesBesides hydrodynamic conditions, phase
transmission of 5-10%/h appeared perfectly acceptable for volumes and viscosities were considered important for
practical purposes, i.e., to prevent merging and sedimenta- coacervate droplet size and, hence, the particle size distri-
tion during the coacervation and hardening processes. bution of microspheres. Figure 4 illustrates that the
Effect of Coacervation Conditions on Solvent Resi- particle size distribution of PLGA 75:25 microspheres
dues in MicrospheressFor safety reasons, organic sol- became significantly narrower and the sizes smaller when
vent residues in microspheres must be minimized to an prepared with PDMS of increasing viscosity. This clearly
acceptable level. Figure 3 shows the impact of coacervation must be a direct consequence of the concomitant viscosity
conditions on the solvent residues. Increasing amounts of increase of the continuous phase, as indicated in the figure.
PDMS in the coacervation mixture diminished the residual This change in particle size distribution was mainly due
hardening agent OMCTS, but led to higher DCM residues. to the considerable reduction of the 90 and 50% undersize
Coacervate droplets produced at low PDMS concentration values, whereas the 10% undersize limit remained es-
allowed intimate interaction between the polymer solvent sentially unchanged. Thus, increased shear mediated
and the hardening agent. As a result, DCM was extracted primarily a breakdown of large coacervate droplets.
more efficiently under these conditions, but at the cost of Besides the viscosity of the coacervating agent, polymer
considerable OMCTS inclusion in the microspheres. When concentration and polymer M h w were important parameters
raising the amount of PDMS, the viscous polymer network influencing the microsphere size distribution. Raising the
counteracted diffusion and exchange of the components. polymer concentration increased the coacervate volume and
Hence, residual DCM in microspheres was slightly in- viscosity, which in turn produced larger coacervate droplets
creased, whereas the OMCTS level dropped considerably. (Figure 5). On the other side, small-sized microspheres

Journal of Pharmaceutical Sciences / 273

Vol. 87, No. 3, March 1998
 mers PLA were less subject to uncontrolled aggregation of
coacervate droplets or precipitation than the more hydro-
philic PLGA copolymers. Consequently, the range of stable
coacervate droplets, also referred to as “stability window”,6,7
was clearly enlarged with the homopolymers. This is
consistent with the findings that increased polymer M h w and
hydrophilicity, e.g. due to a larger proportion of low M hw
compounds as reflected by larger polydispersity values,
lower the polymer solvation by DCM and, consequently,
reduce the amount of PDMS needed for phase separation.6,7
The particular solvation behavior of PLGA 50:50 copoly-
mers has also been revealed by laser light scattering.23,24
Therefore, it is not surprising that PLA and PLGA 50:50
Figure 5sEffect of polymer type, molecular weight, and initial concentration differ in their phase separation behavior, although their
in the solvent on the particle size distribution of microspheres. The 1070 apparent solubility parameters are similar.25,26 The find-
mPa s PDMS was used for phase separation. The bars represent the 10− ings emphasize the importance of detailed polymer char-
90% undersize diameters and the bold line inside the bar the average particle acterization to interpret phase separation phenomena.27
size.
Phase separation induced by low molecular weight
nonsolvents often occurs near the solubility borderline of
were obtained with low M h w polymers. This is ascribed to
polymers.10 Beyond this limit, the increasing difference in
the low coacervate viscosities obtained with these polymers
solubility parameter between the coacervate phase and the
and to the necessary large volume and high viscosity of
continuous liquid increases the risk of polymer precipita-
the continuous phase.
tion. This process became critical for PLGA 50:50 and the
When considering particle size distributions, interfacial
semicristalline PLLA, while for the amorphous PLA phase
properties of the coacervate and continuous phases have
separation occurred well within the solubility range, with
also to be taken into account. Interfacial tension acts as a
a lower risk of precipitation. Polyester coacervation in-
driving force to minimize the surface free energy of the
duced by PDMS did not result in polymer fractionation,
dispersed phase and, hence, promotes coalescence of coac-
as the Mh w distribution of the coacervates was comparable
ervate droplets. The coacervated systems generally showed
to that of the polymer powder. This observation strongly
low interfacial tension, as measured by spinning drop
supports the mechanism of polymer 1-polymer 2 incom-
tensiometry.21 This may be explained by the increased
patibility. While the onset of polymer phase separation
adhesion forces generated by solvent partitioning between
was primarily determined by the polymer properties, the
the coacervate and continuous phases. The interfacial
characteristics of the coacervate were strongly governed
tension of DCM/PLA/PDMS mixtures at two different
by continuous desolvation and partitioning of solvent and
PDMS concentrations was in the range of 0.9-1.5 mN m-1.
PDMS between the two separated phases.
Slightly lower values (0.6-0.8 mN m-1) were obtained in
the EtAc/PLA/PDMS dispersions. Thanks to these low The added amount of PDMS influenced the solvent
interfacial tension values, moderate stirring should be content and viscosity of the coacervate. At the beginning
sufficient in the manufacturing process to reduce and of the “stability window”, the colloid rich phase exhibited
control the coacervate droplet size to a desired level and a nearly Newtonian behavior but became pseudoplastic at
to prevent fast coalescence. Typically, a stirring rate of higher PDMS concentration. Comparable results were
350 min-1 was found adequate to obtain PLGA 75:25 (68.8 reported for an ethylcellulose-chloroform-ethanediol mix-
kDa) microspheres in the size range of 10-90 µm. Increas- ture.28 In the case of hydroxypropylmethylcellulose coac-
ing the stirring rate from 350 to 1250 min-1 did not greatly ervates, the Newtonian flow was ascribed to minor polymer
affect the mean particle size. Therefore, in the systems chain interactions, whereas pseudoplastic behavior was
studied here, stirring rate and interfacial tension were not associated with three-dimensional structure.11 Irrespective
critical factors for microsphere size. of the type of rheological behavior, the concomitant viscos-
ity effect of the continuous and coacervate phases greatly
contributes to the stabilization of the coacervate droplets
Discussion and counteracts agglomeration or uncontrolled precipita-
The results presented in this study show that the organic tion of the colloid.
phase separation process is primarily driven by molecular The physicochemical properties discussed so far also bear
interaction forces between the compounds present. We consequences on solvent residues in the microspheres. For
consider these molecular interactions as key parameters this type of delivery system, residual solvents are probably
determining the composition, solvent content, viscosity, a key issue in safety considerations. Typically, highly
interfacial tension, and wetting properties of the coacervate viscous coacervates are more difficult to desolvate com-
and continuous phases. They are likely to affect the surface pletely by a hardening agent. Correspondingly, such
characteristics, porosity, particle size, and amount of microspheres contain a higher amount of residual polymer
residual solvent of the final microspheres. solvent and a lower amount of residual hardening agent.
In coacervation, the actual physicochemical properties This observation is consistent with previous findings
of a given polymer type greatly influence the phase showing the importance of a high coacervating agent/
separation process. Typically, high M h w polymers phase- solvent ratio for preparing microspheres with low harden-
separate at lower nonsolvent concentration than low M hw ing agent residues.29 Furthermore, the question arises to
polymers. This observation is in agreement with the what extent the coacervating agent is entrapped in micro-
fundamental prediction that higher M h w polymers are less spheres. From a thermodynamic point of view, inclusion
miscible and, therefore, have a lower concentration limit of nonsolvents, i.e., coacervating and hardening agents,
of phase separation.22 Similarly, according to Scott’s might be most effectively reduced by minimizing the
theory, phase-separation of high M h w polymers is expected interaction energy between polymer and nonsolvents (∆intE;
at lower concentrations of the phase-separating compounds, see part 19). However, the actual intermediate interaction
because the critical polymer 1-polymer 2 interaction between the hardening compounds and the polymer sol-
parameter is lowered (part 1).9 Moreover, the homopoly- vents suggests that solvent extraction from the coacervate

274 / Journal of Pharmaceutical Sciences

Vol. 87, No. 3, March 1998
by OMCTS or hexane is probably a lengthy and incomplete 5. Lapka, G. G.; Mason, N. S.; Thies, C. Process for preparation
process. Therefore, nonsolvents with both a low potential of microcapsules. U.S. Patent No. 4,622,244, 1986.
6. Ruiz, J. M.; Tissier, B.; Benoit, J. P. Int. J. Pharm. 1989,
for interaction with the polymer and a large molecular 49, 66-77.
volume may be preferable. These properties should coun- 7. Ruiz, J. M.; Busnel, J. P.; Benoit, J. P. Pharm. Res. 1990, 7,
teract the inclusion of the nonsolvents in the free space of 928-934.
the coacervated entangled polymer matrix. Consequently, 8. Nihant, N.; Grandfils, C.; Jérôme, R.; Teyssié, Ph. J. Con-
trolled Release 1995, 35, 117-125.
phase separation driven by polymer 1-polymer 2 incom- 9. Thomasin, C.; Merkle, H. P.; Gander, B. J. Pharm. Sci. 1998,
patibility must be advantageous over processes induced by 87, 259 (preceding paper in this issue).
low molecular weight nonsolvents such as petroleum ether 10. Robinson, D. H. Drug Dev. Ind. Pharm. 1989, 15, 2597-2620.
or mineral or vegetable oils. In the former process, the two 11. Weiss, G. Mikroverkapselung von Ibuprofen mit magen-
saftresistanten Polymeren durch einfache Koazervation;
phases are formed as a consequence of polymer 1-polymer Ph.D. Thesis, University of Regensburg, D.-Regensburg,
2 repulsion and each of them is composed of one polymer 1991.
type. In the latter process, phase separation is induced 12. Moldenhauer, M. G.; Nairn, J. G. J. Controlled Release 1992,
by polymer desolvation, thereby enhancing the risk of 17, 205-218.
13. Gander, B.; Merkle, H. P.; Nguyen, V. P.; Hô Nam-Trân J.
inclusion of nonsolvent within the coacervate droplets. A Phys. Chem. 1995, 99, 16144-16148.
more detailed study focusing on the aspect of residual 14. Thomasin, C.; Corradin, G.; Men, Y.; Merkle, H. P.; Gander,
solvents in biodegradable microspheres prepared by coac- B. J. Controlled Release 1996, 41, 131-145.
ervation has recently been published.30 15. Technical information from Dow Corning, Midland, MI.
16. Van Krevelen, D. W. Properties of polymers, 3rd ed.; Elsevi-
Besides influencing the amount of residual solvents, the er: Amsterdam, 1990; pp 189-225; 287-320.
viscosity of the coacervation mixture greatly affects the 17. Burrell, H. In Polymer Handbook; Brandrup, J., Immergut,
morphology and particle size distribution of microspheres. E. H., Eds.; Wiley: New York, 1966; p 341.
Predictions of the coacervate droplet size are difficult to 18. Klimisch, H. M. In The Analytical Chemistry of Silicones;
Lee Smith, A., Ed., Wiley: New York, 1991; pp 117-121.
make because of the continuous change of phase volumes, 19. Baney, R., In Structure-Solubility Relationships in Polymers;
viscosity, and interfacial tension. These physicochemical Harris, F. W., Seymour, R. B., Eds.; Academic Press: New
parameters together with the hydrodynamic process condi- York, 1977; pp 225-231.
tions all contribute to the particle size distribution.31-33 Key 20. Lide, D. R. CRC Handbook of Chemistry and Physics, 75th
ed.; CRC Press: Boca Raton, FL, 1994; p 6-151.
parameters determining the mean droplet size in a dis- 21. Thomasin, C.; Merkle, H. P.; Gander, B. Int. J. Pharm. 1997,
persed system were found empirically and compiled in an 147, 173-186.
equation.34 From this equation, the mean droplet or 22. Dobry, A.; Boyer-Kawenoki, F. J. Polymer Sci. 1947, 2, 90-
particle size (d) is directly proportional to the ratios of 100.
23. Bendix, D. Proc. Int. Symp. Controlled Release Bioact. Mater.
coacervate to continuous phase viscosity (ηcoac/ηcont) and 1990, 17, 248-249.
volume (Vcoac/Vcont), to the polymer concentration (c), and 24. Rafler, G.; Jobmann, M. Pharm. Ind. 1994, 56, 565-570.
the interfacial tension (γ) between both phases, and 25. Siemann, U. Proc. Int. Symp. Controlled Release Bioact.
inversely proportional to the stirring rate (N) and the ratio Mater. 1985, 12, 53-54.
26. Siemann, U. Eur. Polym. J. 1992, 28, 293-297.
of stirrer to vessel diameter (D). The effect of viscosity and 27. Donbrow, M. In Microcapsules and Nanoparticles in Medicine
interfacial tension on particle size was investigated more and Pharmacy; Donbrow, M., Ed.; CRC Press: Boca Raton,
closely in other systems.35,36 The obtained results in those FL, 1992; pp 20-23.
studies agree well with our findings, in as much as the 28. Nixon, J. R.; Nimmanit, U. J. Microencapsulation 1985, 2,
103-110.
minor effect of interfacial tension and the major effect of 29. Lewis, D. H.; Sherman, J. D. Low residual solvent micro-
the coacervate viscosity on particle size is concerned. spheres and microencapsulation process. PCT patent ap-
Different concepts have been presented here to describe plication, No. WO 89/03678, 1989.
30. Thomasin, C.; Johansen, P.; Alder, R.; Bemsel, R.; Hottinger,
the silicone oil induced coacervation process of biodegrad- G.; Altorfer, H.; Wright, A. D.; Wehrli, E.; Merkle, H. P.;
able polyesters. It has become evident that thermodynamic Gander, B. Eur. J. Pharm. Biopharm. 1996, 42, 16-24.
and other physicochemical considerations may be useful 31. Kipke, K. D. Verfahrenstechnik 1981, 15, 563-566.
but not sufficient to describe the behavior of the coacer- 32. Sweeney, E. T. An Introduction and Literature Guide to
Mixing; BHRA Fluid Engineering Series; Bedford, 1978; Vol.
vated polymer phase, which in turn will be crucial for the 5, pp 5-40.
interaction with the drug material to be coated. Finally, 33. Oldshue, J. Y. In Fluid Mixing Technology, Chemical Engi-
we would like to emphasize that this study was not neering; Oldshue, J. Y., Ed.; McGraw-Hill Publications: New
intended to analyze thermodynamically PLA/PLGA phase York, 1983; pp 125-140.
34. Arshady, R. Polym. Eng. Sci. 1989, 29, 1746-1757.
separation. Clearly, the values given for χ, ∆δ, and ∆intE 35. Arai, K.; Konno, M.; Matunaga, Y.; Saito, S. J. Chem. Eng.
(Table 3) were not determined by thermodynamic mea- Jpn. 1977, 10, 325-329.
surements but represent calculated rough estimates with 36. Sanghvi, S. P.; Nairn, J. G. J. Microencapsulation 1992, 9,
limited validity. Thus, the manufacturing scientist per- 215-227.
37. Barton, A. F. M. Handbook of Solubility Parameters and
forming coacervation studies for microencapsulation should Other Cohesion Parameters, 2nd ed.; CRC Press: Boca Raton,
not rely exclusively on such calculated parameters, but use FL, 1991; pp 95-111.
them as approximate indication for optimizing systems. 38. Barton, A. F. M. Handbook of Polymer-Liquid Interaction
Clearly, PLA/PLGA phase separation for microsphere Parameters and Solubility Parameters; CRC Press: Boca
Raton, FL, 1990; pp 125-147; 293-296.
preparation is a most complex process which cannot be 39. Drago, R. S.; Dadmun, A. P.; Vogel, G. C. Inorg. Chem. 1993,
described and understood sufficiently well by considering 32, 2473-2479.
only thermodynamic principles. 40. Flory, P. J. Principles of Polymer Chemistry, Cornell Uni-
versity Press: Ithaca, NY, 1953; pp 541-594.

References and Notes

1. Benoit, J.-P.; Marchais, H.; Rolland, H.; Vande Velde, V. In Acknowledgments
Microencapsulation, Methods and Industrial Applications;
Benita, S., Ed.; Marcel Dekker: New York, 1996; pp 35-72. The authors thank Dr. Ernst Wehrli, Laboratory for Electron
2. Fong, J. W. Processes for preparation of microspheres. U.S. Microscopy I, ETH (Zurich, Switzerland), for making the SEM-
Patent, No. 4,166,800, 1979. micrographs and Dr. Hô Nam-Trân, School of Pharmacy, Univer-
3. Sanders, L. M.; Kent, J. S.; McRae, G. I.; Vickery, B. H.; Tice, sity of Lausanne (Lausanne, Switzerland), for helpful discussion
T. R.; Lewis, D. H. J. Pharm. Sci. 1984, 73, 1294-1297. on thermodynamic aspects.
4. Sanders, L. M.; Vitale, K. M.; Mc Rae, G. I.; Mishky, P. B. J.
Controlled Release 1985, 2, 187-195. JS970048J

Journal of Pharmaceutical Sciences / 275

Vol. 87, No. 3, March 1998