European Journal of Pharmaceutical Sciences 63 (2014) 124–131

Contents lists available at ScienceDirect

European Journal of Pharmaceutical Sciences

journal homepage: www.elsevier .com/ locate /e jps

The construction, prediction and measurement of co-crystal ternaryphase diagrams as a tool for solvent selection

http://dx.doi.org/10.1016/j.ejps.2014.06.0170928-0987/� 2014 Elsevier B.V. All rights reserved.

⇑ Corresponding author at: Department of Chemical Engineering, Institute ofChemical Technology Prague, Technická 5, 166 28 Prague 6, Czech Republic.Tel.: +420 220 443 236.

E-mail address: holanj@vscht.cz (J. Holan).

Jan Holan a,b,⇑, František Štepánek a, Pascal Billot c, Ludek Ridvan b

a Department of Chemical Engineering, Institute of Chemical Technology Prague, Technická 5, 166 28 Prague 6, Czech Republicb Zentiva k.s., U kabelovny 130/22, 102 00 Prague 10, Czech Republicc Sanofi SA, 13 quai Jules Guesde, Vitry-Sur-Seine, 94403 Cedex, France

a r t i c l e i n f o

Article history:Received 19 May 2014Received in revised form 25 June 2014Accepted 25 June 2014Available online 1 July 2014

Keywords:AgomelatineCo-crystalDITASolubilityTernary phase diagram

a b s t r a c t

Many active pharmaceutical ingredients (APIs) are poorly soluble and therefore poorly bioavailable. Inrecent years, the advances in crystal engineering have motivated the research in the design of pharma-ceutical co-crystals. This study examines the combination of solvent selection and co-crystal ternaryphase diagram prediction on the basis of solubility measurements into a single methodology that canbe integrated into the pharmaceutical process development workflow. Ternary diagrams constructedfrom agomelatine citric acid co-crystal solubility data were compared with those obtained by moderncalorimetric method called discontinuous isoperibolic thermal analysis (DITA). A suitable solvent forthe co-crystallization process has been chosen on the basis of agomelatine citric acid co-crystal solubility,which is connected to the yield of the crystallization process. Furthermore, the quality of final crystalsfrom crystallization experiments was evaluated.

� 2014 Elsevier B.V. All rights reserved.

1. Introduction

Enhancing the solubility of active pharmaceutical ingredients(APIs) is currently one of the main challenges for the pharmaceuti-cal industry (Good and Rodríguez-Hornedo, 2009). Many new mol-ecules discovered by high-throughput screening and combinatorialchemistry are increasingly larger and more lipophilic. These mole-cules are usually characterized by poor water solubility, which is akey issue for commercial production and introduction to the mar-ket (Aakeroy et al., 2009). Over the last decade, there has been agrowing interest in the design of co-crystals, which emerges as apotential method for improving bioavailability of drugs with lowaqueous solubility (Qiao et al., 2011). Co-crystals are defined ashomogenous crystalline materials containing two or more compo-nents as neutral molecules in the crystal lattice with defined stoi-chiometry at room temperature (Mohammad et al., 2011).According to this definition, the pharmaceutical co-crystals consistof API(s) and other components called co-formers (Qiao et al.,2011). Co-crystals can enhance other essential properties such asphysical and chemical stability, powder flowability, compressibility

and hygroscopicity (Lu and Rohani, 2009) and can be useful for dis-covering new co-crystal forms (Chadwick et al., 2009).

There are various approaches to obtaining a co-crystal. Solutioncrystallization (Chiarella et al., 2007), mechanical grinding (Deloriet al., 2012) or melt crystallization (Berry et al., 2008) are some ofthe general techniques used for preparing co-crystals in the pharma-ceutical industry. Solution crystallization is the most preferred pro-cess in terms of high volume production. However, this procedurebears a risk of crystallizing the single component phases (Thakuriaet al., 2013). Hence, solvent selection and ternary phase diagramconstruction is crucial for the crystallization process design. Severalmethods for obtaining ternary phase diagrams are known. Earliermethods involved analyzing both the liquid and the solid phase foreach composition. These analyses are extremely time consumingand generally incomplete (Ainouz et al., 2009). Other quantitativetechniques have been used on the basis of concentration measure-ment (spectroscopic methods, density, vapor pressure, conductiv-ity) (Hamedi et al., 2006; Grossjohann et al., 2012).

Regarding the specific determination of co-crystals ternary phasediagrams, Chiarella et al. (2007) described a ternary phase diagramof trans-cinnamic acid/nicotinamide co-crystal where the solidphase composition was analyzed by X-ray diffraction and the liquidphase by HPLC (high-performance liquid chromatography).Grossjohann et al. (2012) presented a phase diagram of benzamide/dibenzyl sulfoxide in water and investigated the solid-state character-istics, solubility and dissolution behavior by X-ray diffraction, DSC

J. Holan et al. / European Journal of Pharmaceutical Sciences 63 (2014) 124–131 125

(differential scanning calorimetry), TGA (thermogravimetric analy-sis), and ATR-FTIR (attenuated total reflection Fourier transforminfra-red spectroscopy). Ainouz et al. (2009) developed a calorimet-ric method called discontinuous isoperibolic thermal analysis (DITA)for the construction of the co-crystal phase diagram based on the sol-ubility of API and a co-former (glutaric acid).

The crystallization and co-crystallization process requires acareful selection of solvent. The solvent can have an influence onthe purity of the API, the size and shape of crystals, polymorphismand co-crystal formation itself. The most important factors for sol-vent selection are the yield and the physical quality of the crystal-lizing drug. Organic solvents are largely used in the syntheticpathway and crystallization (Mullin, 2001). For toxicological rea-sons, manufactures are increasingly required to minimize thenumber of solvents used in drug production. According to theQ3C guideline, solvents are divided into four classes (Grogowskaand Parzcewski, 2010). Class 1 contains human carcinogens andenvironmental hazards. Class 2 represents non-genotoxic animalcarcinogens or possible causative agents of irreversible toxicity,such as neurotoxicity or teratogenicity. Solvents in class 3 are com-monly used in the chemical and pharmaceutical industry and havea permissible daily exposure of up to 50 mg. For the fourth group,there are no adequate toxicological data enabling the formulationof an acceptable limit. Solvents used in this work were chosen fromthird class and are presented in the experimental section.

The aim of the present work was to combine solvent selectionand co-crystal ternary phase diagram prediction on the basis of sol-ubility measurements into a single methodology that can be inte-grated into the pharmaceutical process development workflow.Ternary diagrams constructed from solubility data, which is a sim-ple but not so accurate method, were compared with thoseobtained by a modern calorimetric method (DITA) (Ainouz et al.,2009) for the specific case of agomelatine/citric acid co-crystal(Zentiva, 2012). A suitable solvent for the co-crystallization processwas chosen on the basis of the co-crystal solubility, which is con-nected to the yield and solubility of agomelatine and citric acid.Furthermore, crystallization experiments were performed and thequality of the final crystals were evaluated.

2. Methodology

2.1. Chemicals

An active pharmaceutical ingredient (agomelatine) was pre-pared in Zentiva. Agomelatine is produced industrially in largequantities and it is used for the treatment of major depressive dis-order (Du et al., 2013). Citric acid was purchased from Alfa AesarGmbH & Co KG (Karlsruhe, Germany). The solvents methylethylketone (MEK), acetone, methanol, ethanol, isopropanol and tetra-hydrofuran (THF) were provided by Penta (Chrudim, Czech Repub-lic). Co-crystal (agomelatine/citric acid 1:1) was obtained fromsolution by cooling crystallization. 100 g (0.411 mol) of agomela-tine and 79 g (0.411 mol) of citric acid were dissolved in 200 mLof MEK at 80 �C in an agitated 1 L glass vessel. The solution wascooled down to 60 �C and seeded by co-crystal (3 g) prepared ina previous synthesis. The mixture was cooled to 0 �C during 3 hand stirred for a further half an hour at the final temperature.The resulting crystals were filtered, washed and dried overnight.The prepared co-crystal was confirmed by X-ray and NMR analysis.

2.2. Solubility measurement

The solubility of a crystallizing material (defined as the amountof a substance that dissolves in a given volume of solvent at aspecific temperature) is an indispensable requirement for the

successful development of a crystallization process (Grogowskaand Parzcewski, 2010). There are two basic types of solubility:kinetic and thermodynamic solubility. Kinetic solubility valuesare strongly time dependent and are not expected to be reproduc-ible between different kinetic methods. Kinetic solubility alsodepends on the physical properties of the dissolving material(polymorphism, particle size). On the other hand, thermodynamicsolubility represents the saturation solubility of a compound inequilibrium with an undissolved substance in the solvent at theend of the dissolution process. Thermodynamic solubility (equilib-rium solubility) is usually determined by a single measurement,generally after 24–48 h (Alsenz and Kansy, 2007).

In the present work, the kinetic solubility measurements of allcompounds were performed by Crystal 16 (Avantium, Nether-lands). Crystal 16 measures solubility on the basis of turbiditymeasurement in vials. The range of sample weight for agomelatine,citric acid and co-crystal solubility measurement by Crystal 16 wasapproximately 0.030–0.630 g per 1 ml of solvent depending on thetype of solid and solvent. Several solvents common for crystalliza-tion in the pharmaceutical industry were used (acetone, ethanol,methanol, isopropanol, THF and MEK). The temperature wasincreased from 0 to the boiling point of each solvent at the rate0.2 �C/min. When the transmission of light through the vialreached 100% of transmission corresponding to a pure solvent, allsolid material was considered dissolved. From the solubilitycurves, solubility data were used for ternary phase diagrams con-struction at two temperatures, namely 26 and 40 �C.

2.3. Construction of phase diagrams

2.3.1. TheoryTernary phase diagrams can be represented using a right trian-

gle or an equilateral triangle diagram. The equilateral triangle dia-gram is preferred due to its clarity. Each corner in the trianglerepresents a pure compound A, B and C. The triangle’s sides repre-sent the binary systems A + B, A + C and B + C. Triangular phase dia-grams are usually presented as isothermal. In the specific case ofco-crystal phase diagrams, the three components are the solvent,the API and the co-former.

The presence of a solvent in the crystallization process causes alowering the activation energy barrier for the rearrangement of thesolute molecules into a crystalline structure. Thus, the solvent hasa function of a ‘‘catalyst’’ with an influence on the kinetics but notthe thermodynamics of the transformation. The solvent as a thirdcomponent in the mixture minimizes the overall Gibbs free energyof the ternary system (Wouters, 2012). According to the Gibbs ruleof phases (Smirnov, 2006), the number of degrees of freedom (F)specifies how many parameters may be changed without changingthe phases and can be defined as:

F ¼ C � P þ 2 ð1Þ

where C is the number of independent components, P means thenumber of phases in thermodynamic equilibrium and 2 refers tothe external parameters pressure and temperature. To maintaintwo-dimensional representation in a ternary system, temperatureor pressure must be constant (Nyvlt, 1975). In the case of ternaryphase diagram of co-crystal measured at constant temperatureand pressure, the number of degrees of freedom can be reduce bytwo:

F 0 ¼ C � P ð2Þ

where F0 is the number of degrees of freedom with isobaric andisothermal conditions. A typical phase diagram for high a thermo-dynamic stability co-crystal 1:1 is presented in Fig. 1.

The diagram is divided into several areas with differentcomposition and phases as summarized in Fig. 1. The points I

126 J. Holan et al. / European Journal of Pharmaceutical Sciences 63 (2014) 124–131

and J represent solid API/co-former and solid co-crystal with afixed liquid composition. These points are called ‘‘invariants(eutectic)’’ and the number of degrees of freedom is 0. The pointsK and L correspond to API and co-former solubility. The API andco-former ratio in the co-crystal is given by the point M.

The ternary phase diagram can be constructed if the equilib-rium between the API or co-former, the co-crystal and solution(zones 1, 2, 3) is determined. The description of co-crystal/solutionequilibrium can be established by the apparent constant Kapp:

xaAxb

B ¼ Kapp ð3Þ

where x denotes molar or mass fraction of API (A) and co-former (B)regarding the stoichiometric coefficients a and b (Eq. (4)).

aAþ bB$ AaBb ð4Þ

If Eq. (3) is plotted on a triangular diagram for a 1:1 co-crystal,we obtain the theoretical phase boundary between zone 1 and 6 asshown in Fig. 1. The derivation of the apparent constant isdescribed more in detail in Ainouz et al. (2009) and Rodríguez-Hornedo et al. (2006).

2.3.2. InstrumentationThe discontinuous isoperibolic thermal analysis (DITA) is a calo-

rimetric method. A complete description of the DITA technique,instrumentation and principle of the method have already been pre-sented by Marchand et al. (2004) and Ainouz et al. (2009). In short,the method is based on diluting a suspension of one or two solidsby a pure solvent. The injection of the pure solvent initiates heatexchange involving temperature variation that is measured by anaccurate thermometer. From the temperature variation and the vol-ume of added solvent to the suspension, the borders between eachzones in the phase diagram can be determined.

The titration calorimeter used in this work was constructed inSanofi. The measuring cell was kept at a constant temperature by

Fig. 1. Typical 1:1 co-crystal phase diag

a surrounding jacket protected by an insulating foam. The accuratetemperature was ensured by a thermostat and temperature wasmeasured by a thermometer with a resolution ±10�3 K. Agitationwas carried out using a magnetic stirrer and the solvent was addedby a pump MCP CPF IP65 (Ismatec, Germany). The pump and thethermostat were controlled by a computer. The process was auto-mated by software Labworldsoft (IKA, Germany).

2.3.3. Measurement procedureTernary phase diagrams were first approximated on the basis of

solubility experiments for all considered solvents. Points K, L, Mand the point corresponding to co-crystal solubility located onthe curve separating zones 1 and 6 (Fig. 1) were traced using theProSim software. The intersection between the abscissas [KM],[ML] with the co-crystal solubility curve (point I and J) constructedusing Kapp gave the estimated ternary phase diagram. A more pre-cise ternary phase diagram of the co-crystal in the selected solvent(chosen after the solvent screening) was then measured by theDITA equipment. The measuring cell with a suspension of knownconcentration was immersed in water kept at a constant tempera-ture (26 and 40 �C). The amount of initial solid material (API or co-former and co-crystal) and solvent were calculated on the basis ofthe phase diagram predicted form solubility measurements. Asmall amount of pure solvent conditioned exactly at the same tem-perature as the suspension was then added to the slurry equili-brated at the set temperature. The entire experiment wascomposed of several injections of the solvent (typically 10–20).Each injection was carried out after equilibrium was established(30 min). A typical experiment started with a suspension corre-sponding to area 1, 2 and 3 (Fig. 2). When adding the solvent (dilut-ing), the composition of the system follows a straight line to thetop of the triangle as indicated in Fig. 2. The experiment finishedin zone 6. In total, at least eight points (K, L, M, N, O, P, Q, R)are needed for a valuable phase diagram. Therefore several

ram with description of the zones.

Fig. 2. Dilution trajectories during an experiment performed by DITA. The pointsdenoted as K, L, M, N, O, P, R correspond to phase transition. The pink points on thedilution trajectories represent added solvent into the suspension.

Fig. 3. Example of calorimetric data evaluation. The points N and O correspond todifferent zones of the diagram from Fig. 2.

J. Holan et al. / European Journal of Pharmaceutical Sciences 63 (2014) 124–131 127

temperature-versus-time curves and cumulated integration valuesversus the cumulated injection volumes were plotted. One exam-ple of such a plot is shown in Fig. 3. On such plot, changes in theslope are related to the intersections between the limits of the cor-responding phase diagram. The final ternary phase diagrams weretraced using the ProSim software.

2.4. Crystallization experiments

Cooling crystallization experiments were performed in theEasymax equipment (Mettler Toledo, USA), which includes two100 mL vessels monitored by a Focused Beam Reflectance (FBRM)probe. The vessels have been equipped with four-blade 45� angleimpellers (d/D = 0.5, h/H = 0.1) which were set to 200 rpm. Thecrystallization process was carried out as unseeded crystallization.The initial concentration of co-crystal in the solvent was 0.5 g/ml.The solution was cooled down to 0 �C at the rate of 20 �C/h.

3. Result and discussion

The solubility of agomelatine, citric acid and co-crystal, as wellas the calculated Kapp constants at 26 and 40 �C were obtained from

Table 1Measured solubility of agomelatine, citric acid and co-crystal with Kapp constants calculatioand isopropanol based on 3 repeat measurements. It is assumed that the relative errors fofollowed.

Solvent Agomelatine [g/ml] Citric acid [

Solubility and Kapp constants at the temperature 26 �CAcetone 0.186 0.195Isopropanol 0.109 ± 0.004 0.239 ± 0.00Methanol 0.369 0.947Ethanol 0.225 0.451MEK 0.185 ± 0.004 0.040 ± 0.00THF 0.555 0.493

Solubility and Kapp constants at the temperature 40 �CAcetone 0.354 0.253Isopropanol 0.234 ± 0.004 0.316 ± 0.00Methanol 0.831 1.096Ethanol 0.404 0.560MEK 0.322 ± 0.004 0.055 ± 0.00THF 0.850 0.521

solubility curves measured in acetone, isopropanol, methanol, eth-anol, MEK and THF. The solubilities are summarized in Table 1.

The co-crystal mass fraction was calculated regarding to the co-crystal composition, which is 55.9% of agomelatine and 44.1% ofcitric acid on a mass basis (1:1 on a molar basis). The Kapp constantwas obtained from Eq. (3) where the stoichiometric coefficients ofagomelatine and citric acid are 1. From the obtained data, the ter-nary phase diagrams in each solvent were constructed.

Fig. 4 shows the phase diagrams of agomelatine–citric acid co-crystal at 40 �C in the six investigated solvents. On the basis of thediagram appearance, it is possible to make a first estimate of theexpected crystallization behavior. All diagrams are sufficientlysymmetric to suggest a stable co-crystal and a crystallization pro-cess with a low probability of the simultaneous crystallization ofthe pure API or the pure co-former. The phase diagrams are sym-metric due to a similar solubility of the API (agomelatine) andthe co-former (citric acid), which is crucial for the prevention ofcrystallizing the single component phases. From the perspectiveof API and co-former solubility, all the six suggested solventsshould in principle be suitable for the crystallization process.

In order to select the most appropriate solvent from the candi-dates, other physical and chemical properties should therefore alsobe taken into account. One of the major requirements for a crystal-lization process is high yield and material purity, which representeconomic aspects of drug manufacture. The measured solubilitycurves of agomelatine–citric acid co-crystal in the six consideredsolvents are shown in Fig. 5 and the calculated theoretical yieldis summarized in Table 2.

n for the temperatures 26 and 40 �C. The standard deviation was determined for MEKr other solvents are comparable, since identical measurement methodology has been

g/ml] Co-crystal [g/ml] Kapp

0.133 0.00502 0.082 ± 0.004 0.0022 ± 0.0002

0.673 0.05250.264 0.0154

2 0.091 ± 0.004 0.0025 ± 0.00020.547 0.0358

0.249 0.01413 0.143 ± 0.004 0.0058 ± 0.0002

1.006 0.07710.432 0.0309

3 0.152 ± 0.004 0.0062 ± 0.00020.730 0.0501

MEK Methanol THF

Isopropanol Ethanol Acetone

Fig. 4. Ternary phase diagrams of agomelatine citric acid co-crystal at 40 �C, constructed from the solubility data.

Table 2Calculated theoretical yield of co-crystal from solubility measurement.

Solubility [g/ml] Theoretical yield [%]

20 �C Boiling point

THF 0.479 1.260 62.0Methanol 0.566 2.000 71.7Acetone 0.110 0.520 78.8Ethanol 0.200 1.700 88.2MEK 0.065 ± 0.004 0.710 ± 0.004 90.8Isopropanol 0.087 ± 0.004 1.400 ± 0.004 93.8

Fig. 5. Solubility curves of agomelatine–citric acid co-crystal in selected solvents.

128 J. Holan et al. / European Journal of Pharmaceutical Sciences 63 (2014) 124–131

In order to maximize the yield of the cooling crystallization pro-cess, the solubility differences at low and high temperature shouldbe prominent, whereby the high temperature is limited by the boil-ing point of the solvent and the low temperature is convenientlychosen to be the room temperature (20 �C). The theoretical yieldof the crystallization process is calculated as the differencebetween the co-crystal solubility at the boiling point and at roomtemperature, although lower temperatures are sometimes chosenfor the final manufacturing process in order to further improvethe yield.

On the basis of solubility measurements, two solvents with atheoretical yield of over 90% were identified. As the next step ofthe solvent selection methodology, crystallization experimentswere performed using these solvents. The initial concentration ofco-crystal in the solvent was 0.5 g/ml, which corresponds approx-imately to a saturated solution at temperature 70 �C in MEK and60 �C in isopropanol. The solution was cooled down to 0 �C at therate of 20 �C/h. The crystallization process was carried as describedin Section 2.4, without seeds and nucleation was observed by a

FBRM probe. There are significant differences in the viscosity ofMEK and isopropanol. Isopropanol has a dynamic viscosityg = 1.304�10�3 Pa s at 40 �C, while MEK has a dynamic viscosityg = 0.334�10�3 Pa s at the same temperature. Viscosity affects bothnucleation and crystal growth, as well as the ease of subsequentcrystal isolation from the mother liquor. Generally, higher viscosityhas a negative effect on mixing and primary nucleation occurs at ahigher supersaturation. This event results in extremely fine parti-cles with poor filterability and long drying time. In some casesthe nucleation from a high viscosity solution at high supersatura-tion may lead to amorphisation. Fig. 6 shows the morphology ofobtained crystals from MEK and isopropanol.

The crystals crystallized from MEK are noticeably larger thanthose obtained from isopropanol, which is consistent with the dif-ference in viscosity of the two solvents. The crystal size and mor-phology also influence the filterability of the final material. Whilethe larger crystals obtained from MEK were easy to filter, the smal-ler crystals obtained from isopropanol had the consistency of amud, difficult to even remove from the crystallization vessel. Afterdrying, the material crystallized from isopropanol had to be groundup because of strong agglomeration caused by capillary forcesbetween the small particles during drying. With respect to crystal-linity of the product, the crystallization from both solvents

Fig. 6. Morphology of agomelatine–citric acid co-crystal obtained by cooling crystallization from MEK (a) and isopropanol (b). The scale-bar represents 200 lm.

Fig. 7. Solubility data of agomelatine, citric acid and co-crystal in MEK.

Table 3Solubility of each component expressed as mass fraction at 26 and 40 �C.

Points(Fig. 6)

Agomelatine Citric acid MEK Kapp

Solubility at 26 �CK 0.186 ± 0.004 – 0.813 ± 0.004L – 0.048 ± 0.002 0.952 ± 0.004P 0.057 ± 0.004 0.045 ± 0.002 0.898 ± 0.004 0.0025 ± 0.002M 0.559 0.441 –

Solubility at 40 �CK’ 0.286 ± 0.004 – 0.714 ± 0.004L’ – 0.064 ± 0.003 0.936 ± 0.004P’ 0.088 ± 0.004 0.070 ± 0.003 0.841 ± 0.004 0.0062 ± 0.002M 0.559 0.441 –

J. Holan et al. / European Journal of Pharmaceutical Sciences 63 (2014) 124–131 129

provided crystalline agomelatine citric acid co-crystal that wasanalyzed by XRPD.

On the basis of the previous results, methylethyl ketone hasbeen chosen as a suitable solvent for the crystallization process.Hence, a ternary phase diagram with MEK was determined bythe DITA method. The purpose of the DITA measurement was toconstruct a more accurate ternary phase diagram than that

Fig. 8. Ternary phase diagram at 26 (blue lines) and 40 �C (red lines) constructed on the binterpretation of the references to color in this figure legend, the reader is referred to th

constructed from the solubility data. However, the ternary phasediagram constructed from the solubility data is a pre-requisitefor the planning of the DITA measurements, i.e. for the expectedcross-over points on the dilution curves (cf. Fig. 3 in Section 2.3.3).The solubility curves of agomelatine, citric acid and the co-crystalin MEK measured by Crystal 16 are shown in Fig. 7.

The key points of the phase diagrams constructed from the sol-ubility data are summarized in Table 3 for two temperatures, 26and 40 �C. The corresponding ternary phase diagram is presented

asis of agomelatine, citric acid and co-crystal solubility measured by Crystal 16. (Fore web version of this article.)

Fig. 9. Phase diagram measured by DITA at 26 �C. The gray curve shows the phase boundary estimated from the pure component solubility data.

Fig. 10. Phase diagram measured by DITA at 40 �C. The gray curve shows the phase boundary estimated from the pure component solubility data.

130 J. Holan et al. / European Journal of Pharmaceutical Sciences 63 (2014) 124–131

in Fig. 8 and clearly shows the differences of the zone boundaries atthe two different temperatures.

The phase diagram for agomelatine citric acid co-crystal mea-sured by DITA is shown in Figs. 9 and 10 for the temperatures of26 and 40 �C, respectively.

The points in Fig. 9 represent measured values by the DITA cal-orimeter and the lines correspond to a model by Eq. (3). For com-parison, the gray line represents the original diagram constructedon the basis of solubility measurements performed in Crystal 16.The most interesting part of the phase diagram is the apex of thetriangle, magnified in both Figs. 9 and 10. Although the higher tem-perature is connected with a higher occurrence of noise, the lines(Eq. (4)) fit the measured points well. Although there is a small dif-

ference between the diagrams based on solubility measurementsand those obtained by calorimetry, both methods provide consis-tent sets of data, which means that the simpler solubility-basedmethod can be used with a relatively high degree of confidencefor making process-related decisions such as solvent selection.

4. Conclusion

The selection of solvent for a co-crystallization process and theconstruction of ternary phase diagrams on the basis of solubilitydata and measured by a calorimetric (DITA) technique werepresented. Several solvents were considered (ethanol, methanol,

J. Holan et al. / European Journal of Pharmaceutical Sciences 63 (2014) 124–131 131

isopropanol, methylethyl ketone, THF and acetone) and used forthe measurement of agomelatine, citric acid and co-crystal solubil-ity. On the basis of theoretical yield, isopropanol and MEK were ini-tially selected as the most promising solvents and unseededcooling crystallization experiments were carried out in these sol-vents to evaluate process characteristics and product quality. Dueto higher viscosity of the solution with isopropanol, the obtainedcrystals were significantly smaller than those obtained fromMEK. Filtration and drying of small crystals is inefficient andtime-consuming at both laboratory and process scale. The crystal-line material prepared from MEK was well filterable because ofbigger, non-agglomerated crystals. On the basis of better productquality and processability, MEK has been chosen as a suitable sol-vent for agomelatine citric acid co-crystallization process.

Ternary phase diagrams were constructed using two methods –a simple solubility-based method and a more sophisticated calori-metric DITA technique. After comparing the differences betweenthe phase diagrams obtained by each method, it can be concludedthat the measurement of solubility in Crystal 16 – which has to bedone in any case as part of the solvent screening for yield calcula-tion – is sufficient for the construction of a ternary phase diagram.Thus constructed phase diagrams may serve for further optimiza-tion of the crystallization process conditions. The ternary phasediagram of agomelatine co-crystal will be used in the processscale-up and it has also been used in the registration procedureas a new form of API in the market.

Acknowledgements

Financial support from the Specific University Research (MSMT20/2014) is gratefully acknowledged.

References

Aakeroy, C.B., Forbes, S., Desper, J., 2009. Using cocrystals to systematicallymodulateaqueous solubility and melting behavior of an anticancer drug. J.Am. Chem. Soc. 131, 17048–17049.

Ainouz, A., Authelin, J.R., et al., 2009. Modeling and prediction of cocrystal phasediagram. Int. J. Pharm. 374, 82–89.

Alsenz, J., Kansy, M., 2007. High throughput solubility measurement in drugdiscovery and development. Adv. Drug Deliv. Rev. 59, 546–567.

Berry, D.J., Seaton, C.C., Clegg, W., et al., 2008. Applying hot-stage microscopy to co-crystal screening: a study of nicotinamide with seven active pharmaceuticalingredients. Cryst. Growth Des. 8, 1697–1712.

Chadwick, K., Davey, R., et al., 2009. The utility of a ternary phase diagram in thediscovery of new co-crystal forms. CrysEngComm. 11, 412–414.

Chiarella, R.A., Davey, R.J., Peterson, M.L., 2007. Making co-crystals – the utility ofternary phase diagrams. Cryst. Growth Des. 7, 1223–1226.

Delori, A., Friscic, T., Jones, W., 2012. The role of mechanochemistry andsupramolecular design in the development of pharmaceutical materials.CrystEngComm. 14, 2350–2362.

Du, W., Zhou, Y., et al., 2013. Investigation of physicochemical properties and in-vitro in-vivo evaluation of agomelatine polymorphs. Asian J. Pharm. Sci. 8, 181–190.

Good, J.D., Rodríguez-Hornedo, N., 2009. Solubility advantage of pharmaceuticalcocrystals. Cryst. Growth Des. 9, 2252–2264.

Grogowska, K., Parzcewski, A., 2010. Organic solvents in the pharmaceuticalindustry. Acta Pol. Pharm. 67, 3–12.

Grossjohann, Ch., Eccles, S.K., et al., 2012. Characterisation, solubility and intrinsicdissolution behaviour of benzamide: dibenzyl sulfoxide cocrystal. Int. J. Pharm.422, 24–32.

Hamedi, H.M., Laurent, B., et al., 2006. Construction of solid–liquid phase diagramsin ternary systems by titration calorimetry. Thermochim. Acta. 445, 70–74.

Lu, J., Rohani, S., 2009. Preparation and characterization of theophylline–nicotinamide cocrystal. Org. Process Res. Develop. 13, 1269–1275.

Marchand, P., Lefebvre, L., et al., 2004. Diastereomeric resolution rationalized byphase diagrams under the actual conditions of the experimental process.Tetrahedron 15, 2455–2465.

Mohammad, M.A., Alhalaweh, A., Velaga, S.P., 2011. Hansen solubility parameter asa tool to predict cocrystal formation. Int. J. Pharm. 407, 63–71.

Mullin, J.W., 2001. Crystallization, fourth ed. Butterworth-Heinemann, Oxford.Nyvlt, J., 1975. Rovnováhy tuhá faze-kapalina, first ed. Academia, Praha.Qiao, N., Li, M., et al., 2011. Pharmaceutical cocrystals: an overview. Int. J. Pharm.

419, 1–11.Rodríguez-Hornedo, N., Nehm, S.J., et al., 2006. Phase solubility of cocrystals are

explained by solubility product and solution complexation diagrams. Cryst.Growth Des. 6, 592–600.

Smirnov, B.M., 2006. Principles of Statistical Physics, first ed. Wiley-VCH,Weinheim.

Thakuria, R., Delori, A., et al., 2013. Pharmaceutical cocrystals and poorly solubledrugs. Int. J. Pharm. 453, 101–125.

Wouters, J., 2012. Pharmaceutical Salts and Co-crystals, first ed. RSC Publishing,Cambridge.

Zentiva, K.S., 2012. WO2012146371, December 01, 2012.