Review Solubility and Dissolution Proˆle Assessment in Drug Discovery

Transcription

1 Drug Metab. Pharmacokinet. 22 (4): (2007). Review Solubility and Dissolution Proˆle Assessment in Drug Discovery Kiyohiko SUGANO*, Arimichi OKAZAKI, Shohei SUGIMOTO, Sumitra TAVORNVIPAS, Atsushi OMURA and Takashi MANO Global Research & Development, Nagoya Laboratories, Pharmaceutical R&D, Pˆzer Inc., Aichi, Japan Full text of this paper is available at Summary: The purposes of the review are to: a) Provide a comprehensible introduction of the-state-ofthe-art sciences of solubility and dissolution, b) introduce typical technologies to assess solubility and dissolution, and c) propose the best practice strategy. The theories of solubility and dissolution required in drug discovery were reviewed especially from the view point of oral absorption. The physiological conditions in the gastrointestinal uid in humans and animals were then brie y summarized. Technologies to assess solubility and dissolution in drug discovery were then introduced. Recently, these technologies have been improved by the laboratory automation and computational technologies. Finally, the strategies to apply these technologies for a drug discovery project were discussed. Key words: solubility; dissolution; drug discovery; high throughput; simulation; preformulation 1. Introduction 1.1 Background In the early 1990's, most innovative pharmaceutical companies started to use high-throughput screening and combinatorial chemistry in order to improve the productivity of drug discovery. As a result, the number of poorly soluble compounds has increased in drug discovery and development, and poor solubility has become an industry-wide concern (Fig. 1). 1) Oral administration is obviously the most preferable dosage route. The oral absorption of a drug is the tandem process of the dissolution and the intestinal membrane permeation of a drug in the gastrointestinal (GI) tract. Therefore, low solubility, a low dissolution rate, and low permeability, can all result in incomplete and variable oral absorption (Section 4). Intestinal membrane permeability is mostly governed by the chemical structure of a drug. In the current drug discovery and development paradigm, modiˆcations of a chemical structure are performed only during drug discovery. Therefore, a candidate compound must achieve an acceptable intestinal membrane permeability at some point during the drug discovery stage. In contrast, the dissolution proˆle can be improved by saltwsolid form selection and formulation, as well as through chemical structure modiˆcation. The saltwsolid form selection and the formulation studies start at the ˆnal stage of drug discovery, or in the early development stage. 2) Since the discovery-development transition is practically Received; May 11, Accepted; July 13, 2007 *To whom correspondence should be addressed: KiyohikoSUGANO, Global Research & Development, Nagoya Laboratories, Pharmaceutical R&D, Pfizer Inc., 5-2 Taketoyo, Aichi , Japan. Tel , Fax , Kiyohiko.Sugano@pˆzer.com Abbreviations used are: A GI, EŠective intestinal membrane surface area for absorption; A solid (t), Total solid surface area at time t; C GI (t), ConcentrationintheGItractattimet; C bile, Concentration of bile acid in biorelevant media; C bulk (t), Concentration in dissolution test bulk medium at time t; C s, bušer, Molar solubility of drug in simple bušer; C s, micelles, Molar solubility of compound in the micelles sphere; C water, Mol water per liter (55.55 molwl); D drug, DiŠusion coe cient of drug; D eš, EŠective dišusion coe cient; D i, DiŠusion coe cient of each compound state; Do, Dose number; f agg, Fraction of aggregate included drug molecule; f mono, Fraction of monomer drug molecule; h(t), Thickness of dišusion layer at time t; h j (t), DiŠusion layer thickness of spherical particle of particle size group; K a, Dissociation constant; k diss, Dissolution rate constant; K oct,octanol water partition coe cient; L j, Representative length of particle; mp, Melting point; N j, Number of particles in particle size group; P eš, EŠective intestinal membrane permeability; ph max,phofmaximumsolubility;re, Reynolds number; r j (0), Initial particle radius of particle size group; r j (t), Particle radius of particle size group; S GI, Solubility of drug in the GI uid; SR, Solubility ratio; S TCWPL, Solubility of compound in TCWPL media; S bile,medium, Solubility of compound in biorelevant media; S bušer, Solubility of compound in simple bušer; S bulk, Solubility of compound in dissolution test bulk medium; Sc, Schmidt number; S eq, Equilibrium solubility at the ph; S eq, int, Intrinsic equilibrium solubility (the solubility of free form); S eq, ion, Equilibrium solubility of dissociated species; DSf, Molar entropy of fusion; Sh, Sherwood number; S surface, Solubility of compounds at the solid surface; t,time;v bulk, Volume of bulk of dissolution test; X abs (t), Absorbed amount until time t; X j (0), Initial sample weight of particle size group; X solid (t), Amount of undissolved solid at time t; X solu (t), Dissolved amount in dissolution test bulk medium at time t; Z, Dimensionless solubility ratio parameter; Z?, Fitting parameter for initial slope to observed data; r, Density of drug. 225

2 226 Kiyohiko SUGANO, et al. Fig. 1. Number of publications containing the concept ``poor solubility drug'' as of December Carried out using SciFinder}. irreversible, it is necessary to map out a successful strategy to achieve an acceptable dissolution proˆle during the discovery stage. Therefore, a comprehensive assessment of solubility and dissolution is required in drug discovery whether or not it can be improved in the later stages. Various solubility assays are performed routinely in many pharmaceutical companies. 3 5) The science underlying solubility and dissolution is seemingly easy, but actually, very complicated and requires a breadth of knowledge beyond what one might imagine. 1.2 Purpose of this review The purposes of the review are to: a) Provide a comprehensible introduction of the-stateof-the-art sciences of solubility and dissolution. b) Introduce typical technologies to assess solubility and dissolution. c) Propose bests practice strategies. The expected readers of this review are drug discovery scientists, including those not familiar with this subject. This review focuses on drug discovery, not on drug development. Therefore, formulation development studies and bioequivalent studies etc. are beyond the scope of this review. 2. Theory 2.1 Deˆnition of terms in this article Misconceptions about solubility and dissolution may be caused by unclear terminology. Therefore, the terminology will ˆrst be clariˆed. The deˆnitions employed in this article are not necessarily the only ones, and may dišer among publications State of the molecule in a medium: After adding a solid compound into a blank medium, if it looks transparent to the eye, we often say it is ``dissolved'' and the medium is typically called a ``solution''. However, the molecule can exist in this transparent solution as [1] a monomer (a single molecule surrounded by solvent molecules), [2] a dimer or higher self-aggregate, [3] complexes with large molecules, [4] the micelle included state, or even [5] nano scale particles (Fig. 2). In the literature, with the exception of the last case, these are referred as ``solutions'' (the last example is often referred to as a nano-suspension). We will use this deˆnition of ``solution'' in this article unless otherwise noted Equilibrium solubility: A saturated solution is one in which the solute is in equilibrium with the solid phase (solute). Solubility is deˆned in quantitative terms as the concentration of solute in a saturated solution at a certain temperature. 6) [Martin's physical pharmacy and pharmaceutical science, 5th Ed. By P. K. Sinko.] Solubility is an equilibrium value per se. However, we use ``equilibrium solubility'' to avoid confusion when needed in this article (Table 1). The equilibrium solubility of a compound is deˆned as the concentration of the compound in a solution which is in contact with an excess amount of the solid compound when the concentration and the solid form do not change over time (at equilibrium). The conˆrmation of equilibrium by a time

3 Solubility and Dissolution Proˆle Assessment in Drug Discovery 227 Table 1. Deˆnition of solubiilty used in this review Name Incubation time Typical sample source Equilibrium Until no concentration W solid form change observed Crystalline powder Apparent A given time intented for equiliblium (hours to a day) Powder or DMSO stock solution Kinetic Immediately after the addition of a stock solution into an aqueous media (minutes to a few hours) DMSO stock solution Fig. 2. State of a molecule in a media. A molecule can exist in the transparent part as [1] monomer (a single molecule surrounded by solvent molecules), [2] dimer or higher self-aggregate, [3] complexes with large molecule, [4] micelles included state, or even [5] nano scale particles. course measurement is essential. At equilibrium, the chemical potential of the solid is equal to that of the solution. The origin of the solid in the solution is not necessary a crystal powder, but can be an amorphous solid or a concentrated sample solution in a rich solvent. It is often di cult to experimentally obtain the equilibrium solubility of a meta-stable, or an amorphous form, because they can transform into a more stable form Apparent solubility: In early drug discovery, it is not practical to conˆrm no change of the concentration and the solid form. However, a long incubation time with intention of reaching equilibrium can be set. 7 12) After a reasonably long incubation time (typically several hours to a day), the concentration of a compound in the solution which is in contact with the solid is deˆned as the apparent solubility. Even if the solid is crystalline, the solubility can not be ascribed as equilibrium solubility without the conˆrmation of any concentration change by a time course measurement. The origin of the sample is not necessarily a powder (Section 5.3.1), 10,11) but can be a concentrated sample solution in a rich solvent (such as DMSO) (Section 5.2.2). 7 9) Kinetic solubility: The kinetic solubility has often been referred to as the solubility of a compound measured starting oš with a sample stock solution in a rich solvent, regardless of the incubation time (Section 5.2). 13,14) In this article, the kinetic solubility is deˆned as the concentration immediately after the addition of the concentrated sample stock solution into an aqueous media (typically within several minutes to 1 2 hours) ) Therefore, the kinetic solubility represents the precipitation tendency. 19) DMSO is used as a rich solvent in most cases. Many reports have pointed out that kinetic solubility is signiˆcantly higher than equilibrium solubility. 9,10) Three reasons have been suggested for this discrepancy: 1 the solubilization ešect of DMSO, 2 the short incubation time, and 3 the ešect of the solid state. Sugaya et al. suggested that the discrepancy might be due to the dišerence in the solid form. 20) We demonstrated that a short incubation time could also be an additional reason. 9) Thermodynamic solubility: The thermodynamic solubility is often referred to as the solubility measured starting oš with a solid sample. 13,14,19) In this review, however, we do not use this term to avoid confusion Intrinsic solubility: Intrinsic solubility is the solubility of undissociated species. Therefore, we can have ``intrinsic equilibrium solubility'', ``intrinsic apparent solubility'' and ``intrinsic kinetic solubility''. Intrinsic solubility can be measured at a ph where the compound does not dissociate Supersaturation: Supersaturation represents a concentration which is higher than the equilibrium solubility. The supersaturation concentration will eventually settle down to match the equilibrium solubility. Supersaturation can occur, for example, in: Dissolution of salts, meta-stable form, amorphous, and solid dispersion (Section 2.3.4). ph shift from the high solubility region to low solubility region (Section 2.2.2). 21) Dilution of a sample solution in a rich solvent (e.g., DMSO) by an aqueous medium (Sections 2.1.4, 5.2.1) ph: Solubility is often measured by adding a bušer at a ph (``initial ph'') to an excess amount of compound to reach the maximum concentration. In the case of a dissociable compound, the ph can be shifted from the initial ph by the drug. In this article, ph at the concentration determination (``ˆnal ph'') is used unless otherwise noted Dissolution ratewintrinsic dissolution rate: The term ``good solubility'' often implies the tendency for fast and complete dissolution. However, in this arti-

4 228 Kiyohiko SUGANO, et al. cle, ``solubility'' does not imply any kinetic phenomena. ``Dissolution rate'' is used to represent the speed (the dimension has time unit in the denominator amountw time ). The intrinsic dissolution rate is the dissolution rate from a unit surface area of the solid drug (the dimension is amountwareawtime), but not the dissolution rate of an undissociated species. 2.2 Theories of solubility The equilibrium solubility in a biorelevant medium (Section 3.3) can be derived from the chemical structure in a stepwise fashion (Fig. 3). The intrinsic equilibrium solubility in water, pk a, and the micelles-water partition coe cient are the primary physicochemical parameters Intrinsic equilibrium solubility: The intrinsic equilibrium solubility of a compound is the primary solubility parameter directly relevant to the chemical structure. Based on the thermodynamic theory, the intrinsic equilibrium solubility can be further reduced to more fundamental chemical constants, such as enthalpy of melting, entropy of melting, enthalpy of mixing, entropy of mixing and the temperature of interest. 22) Yalkowsky and Valvani derived an equation for the intrinsic equilibrium solubility (the equilibrium solubility of undissociated form, S eq, int) ) log S eq, int =-log K oct DSf(mp-25) (1) 1364 where K oct is the octanol water partition coe cient, mp is the melting point, DSf is the molar entropy of fusion. Further discussion about the thermodynamics of the intrinsic equilibrium solubility is out of the scope of this review and has been extensively reviewed elsewhere. 22) The solid form of a compound has a large impact on its solubility (represented by the second term of Eq. (1)). It has been reported that, in most cases, the dišerence of the solubility was within 2 to 5 fold among crystal polymorphs, while the solubility from amorphous can be much higher (could beà1000 fold). 26,27) ph-solubility proˆle Theoretical ph-solubility curve: The phequilibrium solubility proˆle of a dissociable compound can be theoretically described with S eq, int, the equilibrium solubility of dissociated species (S eq, ion), and the dissociation constant (K a) ) When the drug molecule exists only in the monomer state, the equilibrium solubility at a ph (S eq) is described as: For mono acids (HA) S eq =S eq, intø Ka 1+ phºph [H ]» + max (2) S eq =S eq, ionø ] 1+[H+ K a» phàph max (3) For mono bases S eq =S eq, intø ] 1+[H+ K a» phàph max (4) S eq =S eq, ionø 1+ K a phºph [H ]» + max (5) where ph max is the ph of the maximum solubility at which the function changes. The typical ph equilibrium solubility proˆle of a mono basic compound is shown in Fig. 4. In the ph controlled region (acid phºph max, base phàph max), the slope of the logarithmic plot is 1. Therefore, a 1 unit shift of ph or pk a results in 10 fold Fig. 3. Cascade of solubility calculation. Once the intrinsic equilibrium solubility is obtained, the equilibrium solubility at each ph, the ešect of bile solubilization, and the dissolution rate can be theoretically predicted.

5 Solubility and Dissolution Proˆle Assessment in Drug Discovery 229 change of solubility. The maximum solubility of the ph-equilibrium solubility proˆle is limited by the solubility product. In the solubility product controlled region, the equilibrium solubility changes with the change in the concentration of counter ions (common ion ešect). Therefore, the species of the counter ion is an important factor when we measure the ph-equilibrium solubility proˆle. S eq, ion can be calculated from the solubility product (K sp) and the concentration of the counter ion in the bulk ([I]). 30) S eq, ion = -[I]+ [I]2 +4K sp 2 (6) Na + is most often used as a cation since it is the major cation in the physiological condition. Cl - is most often used in the low ph region, since it is the major anion in the gastrointestinal tract. However, phosphate, and citrate, etc., are also used in the neutral ph region. The theory of the ph-equilibrium solubility proˆle implies that, regardless of the initial solid form (free or salt), the ph-equilibrium solubility proˆle will become the same in the ph controlled region (Fig. 4). However, salt formation increases the dissolution rate (Section and 2.3.3), and might also increase the apparent solubility (Section 2.1.3, 2.1.4). When the ph titration method is used, the ph-apparent solubility curve can deviate from the theoretical curve, especially near the ph max. 21,32 34) When the drug forms aggregate, equilibrium to the aggregate state must additionally be taken into account. 35,36) Solubilization by bile: Bile micelles in the intestinal uid may increase the solubility of poor solubility drugs. Bile micelles mainly consist of bile acids and phospholipids (Section 3.2). The solubility ratio (SR) is deˆned as the ratio of solubility in bile micelles to that in water. 37) log SR=log (C S, micelleswc S, bušer) (7) C S, micelles = Sbile, medium -S bušer C bile (8) C S, bušer =S bušerwc water (9) where S bile, medium is the solubility of a compound in the biorelevant media with bile micelles, S bušer is the solubility of a compound in a simple bušer, C s, bušer is the molar solubility of the drug in a simple bušer, C s, micelles is the molar solubility of a compound in the micelles sphere, C bile is the concentration of bile acid in a biorelevant media, and C water is the mol water per liter (55.55 molwl). For the mixed micelles of taurocholic acid (TC) and lecithin (4:1), the solubility of an undissociable compound at any TC concentration can be calculated as (subscript ``bile'' is substituted to ``TCWPL''), S TCWLC =C TCWLC S eq, int C water 10 SR +S eq, int (10) SR=0.75 log K oct (11) For dissociable compounds, the partition of ionized species into the micelles has to be taken into account. The dišerence in the uncharged micelleswbušer distribution coe cients between ionized and unionized species of a drug were ca. 2 for acids and ca. 1 for base in log 10 scale. 38,39) Fig. 4 Typical ph-equilibrium solubility proˆle of a basic compound with pk a =8, S eq,int =0.001 mgwml, K sp (as a chloride salt)=100 mgwml, and molecular weight=400. Calculated based on ref. (30). ph was assumed to be adjusted only by HCl or NaOH. The activity coe cients of all species were assumed to be 1. In the ph controlled region (phàph max ), the slope of the logarithmic plot is 1. In the solubility product controlled region (phºph max ), the equilibrium solubility changes with the change in the concentration of counter ions (common ion ešect).

6 230 Kiyohiko SUGANO, et al. 2.3 Theories of dissolution Nernst-Brunner equation (for undissociable compounds): The dissolution of a drug is a kinetic process. Therefore, the dissolution process is basically described by a time dependent dišerential equation (d W dt). Two steps are involved in the dissolution from the solid surface. 40,41) The ˆrst step is the detachment of a molecule from the solid surface. The second step is the dišusion of the detached molecule across the dišusion layer adjacent to the solid surface. In most cases, rapid equilibrium (i.e., saturation) is achieved at the solid surface. Therefore, the second step determines the dissolution rate. The basic dišusion-controlled model was ˆrst described by Noyes and Whitney, 42) and later modiˆed by Nernst 43) and Brunner. 44) The Nernst-Brunner equation (NBE) postulates the existence of a dišusion layer (an unstirred layer) adhering to the solid surface (Fig. 5). The NBE is derived from the Fick's law of dišusion, which is based on the second law of thermodynamics. Asolid (t) Ddrug DR= =- (S bulk -C bulk (t)) dt h(t) Asolid (t) Ddrug Xsolu (t) =- h(t) Ø Sbulk - V bulk» (12) dxsolid (t) where X solid(t) is the amount of an undissolved solid at time t, A solid(t) is the total solid surface area at time t, h(t) is the thickness of the dišusion layer at time t, D drug is the dišusion coe cient of the drug, S bulk is the solubility of a compound in the bulk, C bulk (t) is the concentration in the bulk medium at time t, X solu(t) is the dissolved amount in the bulk solution at time t, and V bulk is the volume of the bulk Interpretation of Nernst-Brunner equation for prediction of oral absorption: In the NBE, the S bulk has two meanings and determines two characteristics of the dissolution proˆle (Fig. 6): (1) the solubility at the solid surface which ašects initial slope as A solid D drug S bulkwh, and (2) the solubility in the bulk medium which determines the plateau concentration (Fig. 6). C bulk(t) can not exceed S bulk, since the concentration gradient (S bulk -C bulk(t)) approaches zero as C bulk(t)increases(fig. 5). For oral absorption, these two characteristics of the dissolution proˆle, the initial slope and the plateau concentration, are of critical importance (Section 4). Eq. (12) is used in various oral absorption simulation programs ) However, the straight forward application of this equation is not appropriate in various cases. Eq. (12) is only valid when the compound is undissociable in the dissolution media and when the dissolution media do not contain any surfactant. In the following sections, the authors would like to introduce some clues on how to use Eq. (12) for describing oral absorption Dissolution of free acid and base: In the case of a free acid or a free base, the ph at the solid surface (ph 0) deviates from the bulk medium ph, because the molecule dissolving out of the solid surface reacts with the bušer species coming from the bulk medium (self bušering ešect) (A free acid lowers ph 0. A free base raises ph 0.) ) The solubility of the compounds at the solid surface (S surface) becomessmallerthanthatinthe bulk medium (S surface ºS bulk ). The shape of the dissolution proˆle is monotonic increase similar to that of an Fig. 5. Schematic presentation of dišusion-controlled model. The Nernst-Brunner equation (NBE) postulates the existence of a dišusion layer (an unstirred layer) adhering the solid surface. C bulk (t) can not exceed S bulk, since the concentration gradient (S bulk -C bulk (t)) approaches zero as C bulk (t) increases.

7 Solubility and Dissolution Proˆle Assessment in Drug Discovery 231 Fig. 6. Typical dissolution proˆles of a dissociable compound. Red line: dissolution at a bulk ph where the compound does not dissociate, green solid line: dissolution of free acid or base at a bulk ph where the compound dissociates (the surface ph is lower (acid) or higher (base) than the bulk ph), green dotted line: dissolution of free acid or base at a bulk ph where the compound dissociates when the surface ph is assumed to be same as the bulk ph, blue lines: dissolution of salts (solid line: without supersaturation, dotted line: with supersaturation and precipitation, dot-dash-dot line: with supersaturation and without precipitation). Fig. 7. Surface ph and Z value. Calculated based on ref. (49). Model compound: MW=350, D= cm 2 Wsec. BuŠer: HCO - 3, 6.7 mm. undissociable drug, and the plateau of the concentration is S bulk. However, the initial slope is determined by S surface (not by S bulk )asa solid D drug S surfacewh (green solid line in Fig. 6). Therefore, if we use the Eq, (12) with S bulk, the initial dissolution rate can be overestimated (green dotted line in Fig. 6). The ph 0 can be calculated from the intrinsic solubility of a compound, the concentration of the bušer species, and the dišusion coe cients of the compound and the bušer species. 48,49,55) As the intrinsic solubility of a compound increases andwor the bušer concentration decreases, the dišerence between ph 0 and the bulk ph increases (Fig. 7). Usually, the bušer capacity used for a dissolution test is higher than that of the physiological condition (Section 3.1). 56) Therefore, the self bušering ešect at the solid surface may often be underestimated by a dissolution test. 56) Once ph 0 is obtained, S surface can be calculated from the theoretical ph-solubility curve as described in Section The NBE can be modiˆed as DR=-Z SAsolid(t) Ddrug h(t) (S bulk -C bulk(t)) (13) Z= Ssurface S bulk (14) Z is the dimensionless solubility ratio parameter. The initial slope of the dissolution proˆle in sink condition (C bulk(t) 0) can be calculated as A solid D drug S surfacewh. The plateau concentration remains S bulk. Itisstressed that this treatment is an expedient to ˆt the initial slope and plateau. Numerial integration of the dišusion and and chemical reaction for all species is an expensive calculation. 57) Dissolution of salts: Inthecaseofsalts, 51 54) a typical dissolution proˆle is shown as the blue lines in Fig. 6. S surface can be calculated from the solubility product (K sp ) and the concentration of the counter ion in the bulk ([I]) as Eq. (6). However, it is resource intensive to measure K sp. Therefore, in drug discovery, it is practical to obtain the Z from the dissolution test (Sections 6 and 8.1). In addition, dissolution from a salt form can cause a supersaturation in the media (Section 6.3). If the supersaturation state can be maintained for longer than the transition time in the absorption part, it may be adequate to use the supersaturation concentration as S bulk. If not, we need to incorporate the precipitation equation into the oral absorption simulation. 58) Precipitation of free form can occur at the surface of the solid. 52) Dissolution into biorelevant media with bile acid: In a bulk medium with additives which interact

8 232 Kiyohiko SUGANO, et al. with a compound, the compound can exist in various states (Section 2.1.1, Fig. 2). Each state has dišerent dišusion coe cient (D i) (Each state is represented as the subscript i). Thus, DR=-ØSf i D i i»ø Asolid(t) h(t)» (Sbulk -C bulk(t)) =-D eff Ø Asolid(t) h(t)» (Sbulk -C bulk(t)) (15) where D i is the dišusion coe cient of a compound in a state i, f i is the fraction of each states, and D eff is the ešective dišusion coe cient. In the case of dissolution into a biorelevant media with bile micelles (Section 3.2), at least two states should be taken into consideration; the aggregates included state (micelles or vesicles) and the free monomer state. 59) In this case, D eff can be simpliˆed to: D eff =D mono f mono +D agg f agg (16) f mono =1-f agg = SbuŠer S bile media (17) The subscripts ``mono'' and ``agg'' indicates the monomer drug molecule and the aggregates included drug molecule, respectively. For bile saltwlecithin micelles, S bile media can be predicted by Eq. (7) (9) from S bušer and K oct (Section 2.2.3). DiŠusion of aggregates is much slower than that of monomer drugs (Fig. 8) (Section 3.3) ) Treatment of poly-dispersed particle dissolution: To model the dissolution from particles, 58,65,66) estimation of the dišusion layer thickness of a particle is required. In addition, to model the poly-disperse particle size distribution, particles are divided into each particle size group (represented by the subscript j) (Fig. 9) DiŠusion layer thickness of spherical particles: According to the hydrodynamics theory, the apparent dišusion layer thickness of a spherical particle (h j(t)) with a particle radius (r j(t)) can be theoretically obtained by, 67) h L j j(t)= Sh =2r j(t) Sh (18) Sh=2+0.6Re 1W2 Sc 1W3 (19) where Sh is the Sherwood number, Re is the Reynolds number, Sc is the Schmidt number, and L j is the representative length of a particle. In the case of spherical particle, the representative length is the particle diameter. Eq. (19) is a semi-empirical equation called the Ranz-Marshall correction. In an aqueous medium with low agitation, the Sh for small particle (º30 mm) oatinginthemediais2,becausere is negligible. This theory suggests that, for small particles oating in a weakly agitated uid, the agitation strength dose not ašect the dissolution rate. Even though this theory is counterintuitive, it was conˆrmed experimentally. 68) As the particlesizeincreases,h becomes less than the radius. At low agitation, the h of a large particle is ca. 30 mm. 67) Consequently, for dissolution simulation, the h j(t) canbeset to be equal to the particle radius in r j (t)º30 mm range and h=30 mm inr j(t)à30 mm. h j(t)=r j(t) r j(t)º30 mm (20) h j(t)=30 mm r j(t)à30 mm (21) During the dissolution process, the h j(t) ofeachparticle decreases as the particle dissolves in r j(t)º30 mm range. Therefore, the decrease of particle radius results in the increase of the intrinsic dissolution rate (by ˆrst order) and the decrease in surface area (by second order). Fig. 8. Schematic presentation of dišusion of monomer and aggregates. DiŠusion of aggregates is much slower than that of monomer drugs. The real dišusion process is a random walk process.

9 Solubility and Dissolution Proˆle Assessment in Drug Discovery 233 Fig. 9. Particle size distribution and its time course change. Particles belonging to the same particle size group j are connected with arrows. The initial particle size distribution is semi-log distribution, d50=30 mm, with standard deviation of 2. S bulk =S surface =20 mmwml, dose=1 mg,media volume=50 ml, D eff = cm 2 Wsec. Calculated based on refs. (58) and (65) Calculation of surface area of each particle size group: When spherical particles are assumed, the surface area of a particle size group at time t (A j(t)) is calculated as, N j =X j(0) Ø 4prj(0)3-1 3 r» r j(t)=r j(0) Ø Xj(t) 1W3 X j(0)» (22) (23) A j(t)=n j 4pr j(t) 2 (24) where N j is the number of particles in a particle size group j, X j(0) is the initial sample weight of the particle size group, r j(0) is the initial particle radius of the particle size group, and r is the density of the drug. The r is ca. 1.3 gwcm 3 for many drugs. Eq. (23) describes the reduction of the particle radius astheparticledissolveintothemedium(fig. 9). The particle radius of a particle size group j reduces in proportion to the cube root of the sample weight ratio. By multiplying the surface area of each particle (4pr j(t)) by the number of particles (N j), the total surface area of a particle size group at time t is obtained (Eq. (24)). The particle size distribution can be measured by various methods. It has been pointed out that the results of each method could be inconsistent. In addition, in some cases, it is not appropriate to approximate the drug particles being spherical. 69) In such cases, the representative particle size distribution data can be obtained retrospectively from in vitro dissolution test data (Section 8.1.5). 70,71) Dissolution rate equation for oral absorption prediction: Combining Eq. (12) (24), 72) DR=-Z? Z D efføs j A j (t) h j(t)» (Sbulk -C bulk(t)) (25) Z? is additionally introduced to ˆt the initial slope to the observed data. Theoretically, for spherical particles, Z? =1. The deviation may be due to the ešect of particle shape, 69) inadequate measurement of particle size distribution, the micelles adhesion barrier, etc. 41) In the practical use of the commercially available oral absorption simulation programs, one of the coe cients in the Eq. (12) can be used as a descriptive lump constant of Z?, Z, D eff andwor SA j(t)wh j(t). 73,74) In conclusion, the input parameter for the dissolution simulation should be determined with meticulous care. As described above, the dissolution equations are based on various assumptions. In actual drug discovery, it is important to validate the theoretical equation by comparison with the experimental data (Section 8.1.5). 3. Physiology of gastrointestinal tract in relation to solubility and dissolution proˆle of a drug The primary objective of the solubility and dissolution rate assessment is to assess oral absorption. The physiological conditions in the GI tract, i.e., ph, bušer capacity, surfactant, etc., ašect the solubility and dissolution proˆle of a drug. Therefore, it is important to understand the physiological conditions in the GI tract. Physiological conditions in humans and animals dišer somewhat which could account for species dišerences. The anatomical, physiological, and biochemical characteristics of the GI tract in humans and animals have been extensively reviewed elsewhere ) 3.1 ph and bušer concentration The solubility and dissolution rate of dissociable com-

10 234 Kiyohiko SUGANO, et al. pounds are largely ašected by the ph as described in the previous sections (Section 2.2.2, 2.3.3). In humans, 80) in the fasted state, the ph in the stomach after the administration of 250 ml water is (Median 1.7). In the fed state, the ph in the stomach increases to about 6.4. The ph in duodenum and jejunum are and 6.8, respectively, in the fasted state. In dogs, stomach ph shows large individual dišerences. 81) The bušer concentration is also an important factor which determines the ph at the solid surface of the dissolving drug (Section 2.3.3). The main bušer species in the intestine is sodium-carbonate. The concentration of carbonate in the human intestine was reported to be 6.7 mm in the fasted state. 80) 3.2 Surfactants In humans, the average bile salt concentration in the jejunum is ca. 3 mm in the fasted state, and ca mm in the fed state. 80,82,83) However, the bile concentration shows a great deal of individual variation. The bile saltwphospholipids ratio is ca. 4:1. Rats lack the gall bladder and the bile continuously ows into the GI tract. 77) 3.3 Simulated gastro-intestinal uids Various simulated gastrointestinal uids have been proposed. The fasted stated simulated intestinal uids (FaSSIF) and the fed stated simulated intestinal uids (FeSSIF) are most widely investigated. 84,85) These uids contain taurocholic acid (TC) and phosphatidylcholine (PC) (Table 2). The bušer capacity of these media is higher than the physiological condition. 80,86) Variations of FaSSIF and FeSSIF to simulate the intestinal uid of dogs 68) or to apply for Caco-2 cell 87) have been reported. The grade of TC and PC was found to ašect the dissolution proˆle of a drug. 88) It is noteworthy that the preparation procedure ašects the state of the aggregates, especially in the case of human FaSSIF (TC=3 mm,pl =0.75 mm). When the powder bile acid and lecithin were directly diluted by the bušer, the media often became turbid. We prepare FaSSIF by diluting the concentrated TC WPC solution (TC=30 mm, PL=7.5 mm). The aggregation states of the bile acidwlecithin vary depending on the concentration ) Inthecaseof TCWPC=4:1 (same as FaSSIF and FeSSIF), at low concentration, the aggregates mainly form a vesicle (60-70 nm), while they form micelles (º10nm)athighconcentration (Fig. 10A). After the dilution of the concentrated TCWPL solution, the diameter of the aggregates increases (Fig. 10B). 94) The evaporation method 95) from dichloromethane is labor-intensive and environmentally unfavorable. 3.4 Gastrointestinal uid volume and ešective intestinal surface area The volume of the GI uid available for dissolution of a drug depends on the volume of co-administered uids, secretions and water ux across the gut wall. Yu Table 2. FaSSIF and FeSSIF a FaSSIF FeSSIF ph BuŠer capacity (mmolwlwdph) Osmolality (mosmolwkg) Taurocholic acid (mm) 3 15 Egg phosphatidylcholine (mm) a Ref. (85) suggested that the actual volume of the intestinal uid is around 600 ml and the ešective intestinal surface area for the oral absorption of a drug is about 800 cm 2. 96) In the biopharmaceutical classiˆcation system, 250 ml was used to represent the uid volume when a drug is administered with a cup of water. 97) 3.5 Transit time Since the oral absorption of a drug is a kinetic process, the transit time through the GI tract determines the absorbed amount of a drug. In humans in the fasted state, the average transit time in the stomach and the small intestine is ca. 12 minutes and 3 4 hours, respectively. 79) In dogs in the fasted state, the average transit time in the stomach and small intestine is ca minutes and 1 2 hours, respectively. 79) 3.6 Agitation strength Katori et al., reported that the agitation strength in the human intestine corresponded to rpm in the 900 ml- scale paddle method. 98) In dogs, the agitation strength was higher than in humans, and corresponded to rpm. 68,98,99) Agitation speed is important for the dissolution of coarse particles or disintegration of formulations and granules (Section ). The ešect of agitation strength on oral absorption was also observed in vivo. 100) 4. Solubility, dissolution, permeation and total oral absorption In this section, the interplay between solubility, dissolution and permeation is brie y discussed. A detailed discussion has been reported elsewhere. 96) 4.1 Compartment model Some compartment models have been used to simulate oral absorption, 73,101,102) e.g., the advanced compartment absorption transit model (ACAT model). 103) A simple compartment model is shown in Fig. 11. After the transit time has elapsed (Section 3.5), both the undissolved and dissolved drug move out of the compartment. The transfer of the undissolved particles can be simulated as the movement of particle groups created by further dividing a particle size group (Section 2.3.6). 4.2 DiŠerential equations There are two fundamental dišerential equations which describe absorption process in each GI position (=each GI compartment in the compartment model).

11 Solubility and Dissolution Proˆle Assessment in Drug Discovery 235 Fig. 10 Diameter of the aggregates of bile acidwlecithin measured by dynamic light scattering (DLS). (A) EŠect of the concentration of bile acid W lecithin. Phosphate bušer (ph 6.5) containing 30 mm bile acidw7.5 mm lecithin was diluted to each concentration by the blank phosphate bušer. Molar ratio of bile acidwlecithin was ˆxed to 4. (B) Diameter change of the aggregates of bile acid Wlecithin in FaSSIF (3 mm bile acidw0.75 mm lecithin) prepared by dilution of concentrated FaSSIF (30 mm bile acidw7.5 mm lecithin) by the blank bušer. - dxsolid(t) = dxsolu(t) =k diss (S GI -C GI(t)) dt dt =k diss Ø SGI - Xsolu(t) V GI» (26) dx abs(t) =- dxsolu(t) =P eff A GI C GI(t) dt dt (27) The ˆrst equation is the Noyes-Whitney equation. k diss is the dissolution rate constant. A more sophisticated expression of k diss is described in section S GI is the solubility of a drug in the GI uid (the state of a solute is deˆnedasinsection2.1.1).c GI(t) is the concentration in the GI tract at time t (same state as S GI). V GI is the volume of the GI uid. The second equation represents the intestinal membrane permeation rate. X abs (t) isthe absorbed amount, P eff is the ešective intestinal membrane permeability, 104) and A GI is the ešective intestinal membrane surface area for permeation. S GI, V GI, P eff and A GI can dišer at each GI position. 4.3 Permeability data P eff is usually expressed as 10-4 cmwsec scale, which is 30 to 100 fold higher than that of planer in vitro assays such as Caco-2 and PAMPA, due to the expansion of the surface area by villous structure in the intestine (Calculation of P eff assumes the intestine as a smooth tube). (Fig. 12). 104,105) The maximum value of P eff is around cmwsec due to the existence of the unstirred water layer (UWL), which is ca mm, adjacent on the intestinal membrane. It is noteworthy that, the thicknessoftheuwlin vitro planer membrane could reach mm without stirring. 106,107) In addition, in vitro, the membrane retention could be extensive, leading to artifact underestimation of permeability. 108,109) In vivo, the blood ow washes away the drug from the Fig. 11. Compartment model to represent the oral absorption from a solid dosage. After the transit time has elapsed (Section 3.5), both the undissolved and dissolved drug move out of the compartment. The transfer of the undissolved particles can be simulated as the movement of particle groups created by further dividing a particle size group. intestinal membrane. 110) In the case of UWL limited permeation, apparent permeability is not ašected by ph, because the dišusion coe cient of both dissociated and undissociated species are similar (Fig. 12). 4.4 EŠective concentration In a biorelevant media, drugs can exist in various states, e.g., the molecularly dispersed free state (monomer), the bile micelles bound state, and the excipient bound state (Section 2.1.1). Therefore, ``Which state is ešective for membrane permeation?'' is a question of interest. According to the free fraction theory, the concentration of the free monomer state is ešective

12 236 Kiyohiko SUGANO, et al. for membrane permeation. Therefore, even if the bile micelles increase the solubility of a drug, oral absorption may not be increased in the case of solubility limited absorption (Section 4.5.3), because the concentration of the free monomer state is consistent (Section 2.2.3). Some reports support this theory. 111) However, this contradicts some other experimental observations. 112) The oral absorption of poorly soluble drugs is often increased in the fed state compared to the fasted state. 113) In addition, it has been suggested that the apparent concentration in a biorelevant media, such as FaSSIF or FeSSIF, can predict the oral absorption more accurately than that in a simple bušer (Section 6.1.2). 73,74,88) Compounds with poor solubility tend to have high lipophilicity and high membrane permeability. In such cases, the ešective intestinal membrane permeability is limited by the unstirred water layer (UWL) which is adjacent to the intestinal membrane (Fig. 12). The UWL superimposes the mucus layer on the intestinal membrane. The luminal agitation does not reach the UWL. Previously, by using an artiˆcial membrane without a mucin layer, Amidon et al. demonstrated that, when the membrane permeability is limited by the UWL, not only free monomer drug molecules, but also micelle incorporated drug molecules are ešective for permeation across the rate limiting layer, because the micelles can dišuse the UWL. 114) In ešect, the micelles can act as carriers of solute across the UWL to the membrane surface. The free fraction theory can not be applied to the UWL limited permeation. In addition, it was suggested that the micelles can penetrate the mucus gel layer. 115,116) After passing through the rate-limiting layer, the drug can be released from the bile micelles and quickly penetrate the epithelial cellular membrane. Therefore, the micelle incorporated drug molecule may be bioavailable to some extent. However, the dišusivity of micelles in the mucus layer is slower than that of free monomer drug molecules. Further investigation may be necessary to theoretically deˆne the bioavailable solubility. In addition to solubilization by bile, solubilization by excipients also requires similar consideration. 117) 4.5 Categorization of oral absorption proˆle The causes of poor oral absorption can be categorized into three types based on the balance between the solubility, dissolution rate and permeability of a drug. 103) Dissolution rate limited oral absorption: If the permeation rate is much larger than the dissolution rate, the dissolved drug instantly disappears from the intestinal uid (C GI(t) is negligible at any time). In this case, the oral absorption is ``dissolution rate limited'' (Fig. 13A and Fig. 14A). The absorbed amount increases proportionally as the dose amount increase, because the total solid surface area is in proportion to the dose amount (and the number of particles) (Section ). Particle size reduction increases the solid surface area leading to an increase in the absorbed amount Permeability limited oral absorption: If permeation is slow and dissolution is fast, oral absorption is ``permeability limited'' (Fig. 13B and Fig. 14B). The dissolved amount accumulates in the intestinal uid. The administered dose (X solid(t=0)) is completely dissolved in the intestinal uid, but the concentration in the intestine remains lower than S GI. The absorbed amount increases proportionally as the dose amount increase, because the dose amount increases the concentration in the intestinal uid. Particle size reduction would not Fig. 12. Membrane permeation of a drug which can complex with additives. In a biorelevant media, drugs can exist in various states, e.g., the molecularly dispersed free state (free monomer), the protonated state, the bile micelles bound state, and the excipient bound state. In the case of high membrane permeability compound, after passing through the UWL, the drug can be released as a free monomer and quickly penetrate the epithelial cellular membrane.

13 Solubility and Dissolution Proˆle Assessment in Drug Discovery 237 Fig. 13. Bucket presentation of oral absorption. (A) Dissolution rate limited absorption. (B) Permeability limited absorption. (C) Solubility limited absorption. Fig. 14. EŠect of dose amount and particle radius on the absorbed amount. ašect the absorbed amount Solubility limited oral absorption: If the concentration of a drug in the intestine reaches S GI,the solid drug can no longer dissolve into the intestinal uid. In this case, the oral absorption is ``solubility limited'' (Fig. 13C and Fig. 14C). Usually, a poor solubility compound is lipophilic and may have high membrane permeability. However, since the permeability value has an upper limit due to the existence of UWL (Section 4.3), if the dissolution rate is fast, the drug may accumulate in the intestinal uid to reach S GI. Theabsorbed amount does not increase proportionally as the dose amount increase. Particle size reduction might not increase the absorbed amount.

14 238 Kiyohiko SUGANO, et al. 5. Solubility assessment technologies in drug discovery In the above sections, the sciences of solubility and dissolution were brie y reviewed. In the following section, technologies to assess the solubility and dissolution proˆle will be discussed. The technologies should be practical in drug discovery, as well as scientiˆcally correct. Since the available compound amounts and allowed turnaround time are dišerent at each discovery stage, the technologies suitable for each discovery stage are dišerent (Section 8). The technologies of the solubility assessment in drug discovery can be classiˆed into three categories based of the available resources: Chemical structural information (in silico prediction). DMSO sample stock solution. Powder material. The last method is also referred as the thermodynamic solubility method or shake- ask method. 5.1 In silico prediction The in silico prediction of the intrinsic equilibrium solubility of a drug from its chemical structure has been extensively investigated. 118,119) There are many publications about the fragment based method 120,121) and the descriptor based method 122) with various statistical techniques. Introduction of these methods is out of the scope of this article and has been thoroughly reviewed elsewhere. 118,119) The in silico prediction of solubility is more di cult than that of log K ow, because the enthalpy of melting (or melting point) is di cult to predict (Section 2.2.1). 118, ) It was reported that the standard error of in silico prediction of the intrinsic equilibrium solubility could reach log unit (2 10 fold in normal scale). The prediction error might be larger for new structure compounds in drug discovery. 126,127) For the prediction of the solubility at a ph, the error is multiplied by the prediction error of pk a (Section 2.2.2). 5.2 Solubility measurement using DMSO stock solution (DMSO solution precipitation (DMSO-SP) method) In early drug discovery, samples are usually supplied as a DMSO stock solution. Therefore, the DMSO-SP method is the method of choice in this stage. A concentrated sample stock solution in DMSO ( mm) is used as the sample source. As the DMSO sample solution is diluted by an aqueous medium, precipitation occurs. Both sequence order, adding the DMSO stock into an aqueous bušer and vice versa, have been reported in the literature (the former sequence was more common). The concentration can be determined by turbidity (Section 5.2.1), 15,16,128) UV, 7,8,12,17,18) HPLC, 9,10) or LCWMS. In the UV, HPLC, or LCWMS methods, the precipitant is removed by ˆltration andwor centrifugation before analysis. To remove the remaining DMSO, freeze dry or co-evaporation can be applied. 17) At the 1z level or lower, the solubilization ešect of DMSO is expected to be less than 2 fold increase. 129) Kinetic solubility: The kinetic solubility measurement 15) employs a short incubation time (Section 2.1.4). A concentrated sample stock solution is added to an aqueous bušer in a stepwise manner. The sample concentration, at which precipitation occurred immediately after the addition, is taken as the solubility value. Several modiˆcations of this method by using a 96 well plate were reported. A nephelometer or ow cytometer was used for the detection of precipitation. Caution must be exercised that the data from the kinetic solubility should be used only for the intended application (Section 8.2.1). 19) Apparent solubility (DMSO-SP method): The apparent solubility measurement employs a long incubation time (Section 2.1.3). It seems that the precipitant of the DMSO-SP method had been blindly believed to be amorphous. However, the precipitant can be crystal. We demonstrated that the precipitant of more than half of the model compounds were observed as crystalline by polarized light microscopy (PLM) analysis. 9) When the incubation time was 20 hours and the precipitant was crystalline, DMSO-SP solubility was similar to equilibrium solubility. However, when the incubation time was 10 min andwor the precipitant was not crystalline, DMSO-SP solubility was higher than the equilibrium solubility (Fig. 15). These results suggest that the information regarding the solid form of the precipitant is important when interpreting the solubility data. In addition, we developed an automated birefringence diagnostic system for drug discovery usage. 9) This method (long incubation time+plm observation) is called ``Nagoya solubility'' in tribute to the founding site. 5.3 Solubility measurement using powder material (PWD method) Apparent solubility (PWD method): In the PWD experimental procedure, 10,11,130) powder material is weighed into a vial and an appropriate bušer is added. The weighing process is usually done manually and requires several milligrams of compounds. After a given incubation period intended to reach equilibrium (Sections 2.1.3), the residual solid is removed by ˆltration or centrifugation. Sample quantiˆcation of the ˆltrate is performed by means of UV, HPLC or LCWMS.Itis recommended to observe the solid state of the residual solid by polarized microscope andwor powder X ray. 11) ph-solubility proˆle: In the presence of an excess amount of a drug powder in the media, the ph of the bušer can be changed by the self bušering ešect of the drug (Section 2.1.8). Therefore, the ph after the incubation period must be measured. A time course meas-

15 Solubility and Dissolution Proˆle Assessment in Drug Discovery 239 Fig. 15. DMSO-SP solubility versus equilibrium solubility. Incubation time: (A) 10 min incubation, (B) 20 h incubation. The solid form of precipitant, diamond: crystalline, open circle: partially crystalline, closed circle: no crystal observed. The dotted line indicates the case when the DMSO-SP solubility was equals to the equilibrium solubility. Ref. (9). urement should be performed to assure equilibrium. The solid form of a drug can be identiˆed before and after incubation by powder X-ray andwor dišerential scanning calorimetry. Hydrate formation can sometimes be observed. The potentiometric method can be used to assess the ph-solubility proˆle. 21,34) 6. Dissolution measurements in drug discovery 6.1 Mini-scale dissolution test Procedure: Takano et al. demonstrated that the mini-scale dissolution test can quantitatively predict the fraction of a dose absorbed in humans for a dozen of structurally diverse model drugs. 74) They used FaS- SIF, a small medium volume (50 ml), and low agitation speed (50 rpm). FaSSIF and the low agitation speed were intentionally used to represent the in vivo condition (Section 3.3 and 3.6). Agitation strength in the mini-scale dissolution test (50 rpm, mini-paddle) corresponds to rpm in the 900 ml-scale paddle method. 131) Miniaturization reduced the sample consumption. The mini-scale dissolution test requires only a few milligrams of materials In vivo predictability of mini-scale dissolution test: An adequate in vivo predictability for structurally diverse compounds is required in drug discovery. Dozens of human Faz (fraction dose absorbed) data are required to evaluate the in vivo predictability. However, only a few Faz have been published for poor solubility drugs. Takano et al. employed several assumptions to increase the number of model drugs. The following three conditions were employed to neglect the dissolution in the stomach and re ect the dissolution in the intestine: Poor soluble neutral drugs. Poor soluble free acidic drugs. Poor soluble free weak basic drugs (achlorhydria patients and the patients receiving gastric acid blockers). Otherwise, a two chamber dissolution test apparatus (stomach and intestine) is required (Section 6.3). In addition to the above conditions, in order to increase the number of model drugs, the followings were considered to approximate Faz: Faz calculated from their oral bioavailability (BA z) andfh assuming that hepatic clearance is the only elimination pathway. Relative BAz of a solid dosage vs. a solution dosage. Relative BAz of a solid dosage in the fasted vs. fed state. Intestinal bile levels are much higher in the fed than in the fasted state. Because the bile salts and lecithin can enhance the solubility of lipophilic drugs and almost complete absorption may occur in the fed state. Relative BAz of basic compound with high gastric ph vs. low gastric ph. Finally, twelve poor solubility model drugs with approximate human Faz data were collected from the literature. A mini-scale dissolution test was then performed using a commercial tablet. Since the particle size distribution data of the active ingredient in the marketed tablet was not available, they lumped the initial particle size term, dišusion coe cient term, and dišusion layer thickness term into one ˆtting parameter (z). Reduction of the surface area associated with the dissolution of particles was taken into account as the 2W3 power of the remaining amountwinitial dose ratio. Permeation of the model drugs was assumed to be limited by the unstirred water layer. Faz was predicted as described in Section 4 and By using FaSSIF, human Faz was adequately predicted for structurally diverse compounds. Simple bušer without bile and lecithin underestimated the Faz

16 240 Kiyohiko SUGANO, et al. Fig. 16. Human Faz prediction for structurally diverse poor solubility compounds by mini-scale dissolution test and oral absorption simulation. Produced based on the data in ref. (74). Fig. 17. Typical DP-system apparatus. The DWP system consists of two half-chambers and a Caco-2 monolayer mounted between them. Both sides of the monolayer were ˆlled with transport medium and are constantly mixed by magnetic stirrers. Compounds were applied to the apical side as a solid, suspension, or solution. (Fig. 16). 6.2 Dissolution-permeation method Simultaneous assessment of dissolution and permeation is required to evaluate special formulations such as nano particles, solid dispersion and self emulsifying drug delivery systems (SEDDS), since the state of the drug molecule from these formulations may not be the same as the ``bioavailable'' form (Section 4.4) Chamber based methods: Several types of in vitro systems in which the dissolution and permeation process were combined were reported ) Kataoka et al. 87,135) developed a dissolutionwpermeation system (DW Psystem,Fig. 17), designed especially for poorly watersoluble compounds. This will be a useful tool for predicting in vivo absorption at the discovery stage because it: (1) consumes a small amount of compounds with a simple procedure, (2) can evaluate the food ešect, 87) and (3) has potential to be applied to the evaluation of prototype formulations as well. 87) The DWP system 87,135) consists of two half-chambers and a Caco-2 monolayer (1.77 cm 2 ) mounted between them. Both sides of the monolayer were ˆlled with transport medium (apical side; ph 6.5, volume 8 ml, basal side; ph 7.4, volume=5.5 ml) and were constantly mixed by magnetic stirrers. Compounds were applied to the apical side as a solid, suspension, or solution. In this DWP system, the physiological conditions of the GI tract and the clinical dose of drugs were taken into account: the apical volume was set to 8 ml, which corresponds to about 1W100 of the in vivo human volume and thus 1z of the clinical dose was applied to the system. They also reported that when modiˆed FaS- SIF or modiˆed FeSSIF with isotonic osmolality was used as an apical media, there were no damage to the Caco-2 monolayer and the ešect of food intake was evaluated correctly by the DWP system. 87) From 13 compounds experiments, the standard curves between the permeated amount in the DWP systemandthein vivo human Faz (IVIVC) for both fasted and fed state were obtained. Given the standard curves, Faz for an unknown compound can be predicted. The applicability of the IVIVC for lipid based formulation was conˆrmed by albendazole and danazol formulations. 87) well plate based methods: Sugano and Sakai suggested the possibility of constructing a dissolution permeation system in a 96 well format. 136) Their method is an application of the parallel artiˆcial membrane permeation system (PAMPA) ) One disadvantage of PAMPA for the formulation study is the lack of the mucus layer. Without the mucus layer, the formulations (particles, micelles etc.) can directly interact with the membrane. To attach the mucus layer onto the artiˆcial membrane which consisted of phospholipids and an organic solvent, a hot melt agarosew mucin gel (1z(wWv)) was impregnated to the hydrophilic ˆber scašold which is physically attached on the membrane. The hydrophilic scašold enables the formation of a thin mucus layer. Without the hydrophilic scašold, the agarosewmucin mixture makes a droplet on the hydrophobic PAMPA membrane and does not form a uniform thin layer. They demonstrate that the ešect of food can be adequately assessed by the mucus layer adhered PAMPA. 6.3 Multi-compartment dissolution system

17 Solubility and Dissolution Proˆle Assessment in Drug Discovery 241 Several multi-compartment dissolution systems have been developed to represent the ešect of gastric ph on oral absorption ) In these systems, the gastric compartment was tandemly connected to the intestinal compartment. Furthermore, a multi-compartment dissolution permeation system was also reported. 145) These systems are especially useful when the compound is poorly soluble base and their salts, e.g., ketoconazole, albendazole, cinnarizine and dipyridamol. These compounds have high solubility at low ph and low solubility at neutral ph. The bioavailabilities of these drugs were largely ašected by the gastric ph ) The dissolved amount during the stomach transition and the supersaturation retention in the intestine are important features which can be investigated by multicompartment dissolution systems. To date, it is impossible to theoretically predict the supersaturation retention potential (=inverse of the precipitation tendency) based on the chemical structure or physicochemical parameters such as the intrinsic equilibrium solubility. Since there are large individual dišerences of stomach ph, and since the stomach ph can be altered by the coadministration of antacid, H2blockers or proton pump inhibitors, the oral absorption of a drug product should not be ašected by the stomach ph. 7 In vivo studies To assess the ešect of solubility and dissolution by in vivo experiments, we must be careful when preparing the formulation. A test compound is dosed to animals assuspensionsinvehiclesorascapsules.thevehicle could be composed of inactive polymers like cellulose derivatives and small amounts of wetting agent like Tween 80. Conventional preparation of the suspension is to use a mortar and pestle. In the preparation process, the solid compound is subjected to destructive forces by sliding, crushing, andwor compression. This may change the solid characteristics of the compound such as particle size, crystal form and crystal defect. It is important to evaluate the crystal form and particle size in the suspension since these are critical factors ašecting the dissolution proˆle (Section ). In addition, homogeneity of the suspension should be conˆrmed, since a poorly soluble compound often has low wettability and will not be dispersed evenly in the vehicles. If the formulation is not homogeneous, the dose amount will vary among animals, leading more variability in the PK proˆle. Furthermore, the above characteristics might change during storage and it is preferable to conˆrm the stability of the formulation for experimental duration. Selection of animal species is also important. The dišerence in the physiology of animals and humans should be taken into consideration (Section 3). Dogs are preferentially used to assess the ešect of poor solubility on oral absorption. The use of gastric acidity-controlled beagle dogs may reveal the ešect of stomach ph on oral absorption. 81,150,151) 8 Strategies of solubilitywdissolution assessment in drug discovery There is no versatile technology for every stage in drug discovery (Fig. 18). Each technology has its pros and cons, which may include throughput, dynamic range, accuracy, sample amount, solvent, etc.. A project proceeds in accordance with its screening strategy, which can be expressed with a screening cascade (Fig. 19). 152) In this section, we will discuss the strategies for the solubility and dissolution proˆle assessment in drug discovery. The strategy should be aligned with that of whole project and other screenings. 8.1 Oral absorption aspect Library design: The quality of a compound library directly ašects the quality of a lead compound, and may impact the quality of a clinical candidate compound and the success rate of drug development. Therefore, a lead compound with reasonable solubility had better be discovered from the compound library. Generally, during the lead optimization stage, the average solubility of a compound series decreases because the average of molecular weight (MW) and lipophilicity would increase to achieve a high pharmacological potency and selectivity. 153) The large number of hydrogen bonds and high lipophilicity might cause high crystalline energy and high hydrophobicity (low solubility of super cooled liquid drug in water), leading to low solubility (Section 2.2.1). In order to ˆnd a high quality lead compound, ``druglikeness'' 15,154) as well as structural diversity should be considered in library design. ``Drug-likeness'' is related to the molecular weight, the number of hydrogen bond donors and acceptors, and lipophilicity. These factors also ašect the solubility. Therefore, the concept of ``drug-likeness'' includes the solubility. High MW might cause either or both of ADME property issues and synthetic complexity, and might lead to development candidates with lack of ``drug-likeness''. Therefore, it is preferred that the MW of the library compounds is as low as possible (e.g., less than 400). 155) For library design, an in silico approach is ordinary (Section 5.1). 129) However, at present, the prediction accuracy of in silico tools is not completely satisfactory. If one keeps in remind its limitations, an in silico tool can be used for the purpose of library design Lead selection and lead optimization: In this stage, the compound amount available for solubility assessment is less than a few milligrams, and the report due date is about a week. Therefore, a small-scale highthroughput method is essential for solubility determination. At the cascade level where in vitro ADME studies are performed, an apparent solubility screening with

18 242 Kiyohiko SUGANO, et al. Fig. 18. Stage gates in drug discovery and development. There is no versatile method for every stages of drug discovery. Fig. 19. Typical screening cascade in drug discovery. HTS: high throughput screening, uhts: ultra HTS, PADMET: physicochemical propertywabsorption WdistributionWmetabolismWexcretionWtoxicology. crystallinity assessment (Section 5.2.2) 9) can be performed in parallel with a membrane permeability assay such as PAMPA ) Since most of the other assays use a DMSO sample stock solution, it would be e cient to share the same DMSO solution (Section 8.2.1). Compounds which passed the in vitro assays would then be evaluated by in vivo assays. Apparent solubility assay with powder material accompanied with crystallinity evaluation (Section 5.3.1) should take place at this stage. This data can be useful to select an appropriate formulation and to interpret the in vivo results (Section 7). It is noteworthy that the solid form of a compound in drug discovery is not always the same as that of a ˆnal drug product, because the solid form of a drug candidate is usually ˆnalized during the later discovery or development stage. In our laboratory, ca. 60z of the submitted powder materials for the apparent PWD solubility measurement was crystal Candidate selection and active pharmaceutical ingredient form selection: A detailed solubility proˆle (the ph-solubility proˆle and the ešect of bile) should be studied at this stage as a part of the preformulation study. 156,157) The ešect of bile should be evaluated to assess the ešect of food on oral absorption (Sections 3.3, 3.2). If the candidate compounds (may be 1 3 compounds from a project) have low solubility, salt formation, co-crystal formation, 158) andwor particle size reduction would be investigated to improve the dissolution proˆle. SaltsWco-crystalWsolid form selection (active pharmaceutical ingredient (API) form selection) is performed in late drug discovery or early drug development. 2, ) Nowadays, high throughput technologies for API form ˆndings are widespread, ) e.g., automatic polarized light microscope, 9,166) powder X-ray dišraction 167) and Raman spectroscopy. 166,168) It is reported that, to make a stable salt suitable for development, dišerence of pk a between the compound and counter ion must be larger than ,162) In addition, in the case of very weak acid or base (may be pk a Àca. 7 or pk aºca. 5, respectively), it is di cult to ˆnd a successful development case. At this stage, the available compound amount is about mg. 156,157) The dissolution rate is one of the key factors when we select the saltswsolid form, especially when the drug has low solubility. (Section 8). As shown in Section 2.3, the dissolution proˆle can be predicted theoretically from static parameters, i.e., intrinsic solubility, pk a,andlogk oct. However, in our experience, the prediction of a dissolu-

19 Solubility and Dissolution Proˆle Assessment in Drug Discovery 243 tion proˆle often fails. For saltwsolid form selection, the intrinsic dissolution rate measured by the rotate disk method may be the most desirable data, since it neglects the ešect of particle size which can be modiˆed in the later stages. However, at this stage the available compound amount is limited. In addition, since the surface area of a disk is much smaller than that of the particles, the concentration of the drug in the medium can be low and di cult for quantiˆcation. Therefore, the rotate disk method is not practical in drug discovery. To select an API form at this stage, the mini-scale dissolution test can be performed (Sections 6.1). In addition, in vivo pharmacokinetic study dosing the suspension of the API is also performed at this stage (Section 7). This pharmacokinetic proˆle is the basic data for the dose setting and formulation choice in subsequent pharmacologicalwtoxicological studies and ˆrst in human (FIH) formulation. However, in vivo studies are resource intensive. In addition, from the viewpoint of animal ethics, the number of animal experiments should be reduced. Therefore, the experimental protocol should be carefully planned to fulˆll the purpose of the study Exploring formulations Particle size reduction: If API form selection was not successful to improve the dissolution pro- ˆle, particle size reduction would be the next measure. In the case of the dissolution rate limited absorption (Section 4.5.1), particle size reduction may be ešective to increase Faz. The ešectiveness of standard size reduction can be assessed by using a theoretical calculation (Section 2.3). Usually, the initial dissolution rate is the reciprocal of the particle size or the reciprocal of square of the particle size. In drug discovery, the conventional preparation method is to use a mortar and pestle (Section 7). It is worth mentioning that particle size reduction can change the solid form, especially to an amorphous state. Therefore, a milling feasibility study should be simultaneously performed Special formulations: If API form selection and standard particle size reduction (ca mm) was not successful to improve the dissolution proˆle, special formulations such as nano-particle, solid dispersion and SEDDS could be the next trial to achieve a target product value (Section 8.1.6). In the case of the solubility limited absorption (Section 4.5.3), formulation technologies to solubilize compound or to produce supersaturation state may both be ešective. Highthroughput formulation screening may facilitate this approach ) Nano-particle, ) solid dispersion and SEDDS 169,175) can be prepared on a small scale. The DWP system may also be useful to investigate these formulations (Section 6.2). However, the successful development of special formulations is not always guaranteed. Therefore, we suppose that the primary solution to poor solubility would be to ˆx it in chemistry (compound structure or API form). 176) If poor solubility is inevitable, we should then challenge special formulations. To reduce the risk to challenge the special formulations while exploring the possibilities, it is preferable to experimentally examine these special formulations as early as possible in the drug discovery-development process. At the same time as the assessment of oral absorption, develop-ability (stability, production suitability, production facility, ease of quality control, etc.) and market competitiveness (compliance, cost of goods, development speed, etc.) should be addressed. Special formulations are always expensive in time and man power. The maximum loading dose of special formulations is often smaller than that of the standard formulations. To avoid over and under expectation on special formulations, su cient accountability for the project team is necessary (Section 8.4). An experienced formulation scientist with balanced perspective is required. It is important to have a decision tree before the in vivo oral absorption study, because the discovery project team is liable to be fascinated with the signiˆcant enhancement of oral absorption and put aside develop-ability and market competitiveness Strategy for human Faz prediction by computational simulations: It has been suggested that computational simulation can increase the e ciency of API form selection and special formulation investigation. 45) Today, several computational simulation programs are commercially available. 46) However, computational simulation is not always accurate since it is based on many assumptions as described in the previous sections. Most of the computer simulation programs calculate the numeric integration of (a kind of) Nernst- Brunner equation (Eq. (26)) and (a kind of) membrane permeation equation (Eq. (27)) (Section 4). As shown in Section 2, the dissolution process can be complicated depending on the drug substance (dissociable or non dissociable, free or salt, particle size and shape) and the dissolution medium (ph, surfactant, agitation strength). It was suggested that the dissolution rate in a bulk media should not be estimated from the equilibrium solubility in the bulk media based on the assumption of a linear relation. 177) D eš and Z should be taken into the calculation (Sections , and 2.3.4). The users should understand the theoretical assumption employed in the simulation and should input scientiˆcally appropriate parameters for the intended use of simulation. It is important to validate the computational simulation for each API form by comparison with the real experimental data. We propose a step by step validation strategy as shown in Fig. 20. The dišerential equation of dissolution can be validated by dissolution test. If the

20 244 Kiyohiko SUGANO, et al. theoretical curve deviates from the observation, the Z? parameter (Eq. (26)) or k diss (Eq. (27)) can be obtained from the dissolution test. 73,74) The simulation is then validated with the in vivo animal data. Finally, human Faz is predicted by the simulation. These validation processes may also provide information about the conˆdence in prediction. In addition, the computer simulation can be used to investigate the variation of oral absorption by using Monte Carlo procedure Criteria: It is not practical to set standardized and clear cut criteria for solubility, because (1) a candidate compound for development is selected by considering the balance of pharmacology, toxicology, pharmacokinetics and formulation (production) suitability, and (2) criteria should dišer among each project. For example, drugs for acute diseases should be absorbed immediately after dosing, while this is not necessary for a chronic disease drugs. However, for guidance, it is beneˆcial to set a target solubility value. The target solubility value can be derived by several approaches: Theory based approach (deductive approach) Data based approach (inductive approach) Knowledge based approach In the theory based approach, the solubility criteria can be derived from target Faz, permeability, dose strength, and particle radius (Section 4). If we have a target Faz and an applicable particle radius (5 10 mm), we can draw a solubility-permeability-dose criterion diagram. 178,179) The diagram used for the biopharmaceutical classiˆcation system (BCS) 97,180) canalsobeusedindrug discovery. 8,181,182) Dose amount and solubility can be compiled to the dose number (Do). 178) Table 3 shows the criteria proposed by Lipinski with modiˆcation. 183) Dose amount Do= Solubility 250 ml (28) By the data based approach, the solubility of marketed drugs which are known to have a poor oral absorption in humans due to its low solubility, can be used as a reference (Table 4). It is preferable to measure the solubility of these drugs using an in house method. In addition to the above two methods, in our opinion, pearls of wisdom of highly experienced experts are very ešective (Table 5). 184) We can turn the tacit knowledge of experts into formal knowledge (criteria) by administering questionnaires to experts. In the case of solubility, the criteria derived by the above three approaches showed strong Fig. 20. Step by step strategy to predict the oral absorption in humans. (I) The dišerential equation of dissolution can be validated by dissolution test. (II) If the theoretical curve deviates from the observation, the Z? parameter (Eq. 26) or k diss (Eq. 27) can be obtained from the dissolution test. (III) The simulation is then validated with the in vivo animaldata.(iv)finally,humanfaz is predicted by the simulation.