In situ artificial membrane permeation assay under hydrodynamic control: Correlation between drug in vitro permeability and fraction absorbed in humans

European Journal of Pharmaceutical Sciences 44 (2011) 299–309

Contents lists available at SciVerse ScienceDirect

European Journal of Pharmaceutical Sciences journal homepage: www.elsevier.com/locate/ejps

In situ artificial membrane permeation assay under hydrodynamic control: Correlation between drug in vitro permeability and fraction absorbed in humans Mateˇj Velicky´ a, Kin Y. Tam b,⇑, Robert A.W. Dryfe a,⇑ a b

School of Chemistry, University of Manchester, Oxford Road, Manchester M13 9PL, UK AstraZeneca, Mereside, Alderley Park, Macclesfield, Cheshire SK10 4TG, UK

a r t i c l e

i n f o

Article history: Received 1 April 2011 Received in revised form 28 July 2011 Accepted 10 August 2011 Available online 16 August 2011 Keywords: In situ permeation Stirring PAMPA Drug absorption Paracellular transport Unstirred water layer

a b s t r a c t The purpose of this study was to develop an in vitro permeation model that will predict the fraction of drugs absorbed in humans. A rotating-diffusion cell with two aqueous compartments, separated by a lipid-impregnated artificial membrane, was used to determine the permeability of drugs under conditions of controlled hydrodynamics. The measured effective permeability coefficient was modified to include the paracellular transport derived from a previously reported colorectal adenocarcinoma epithelial cell line (Caco-2) permeability study and the effects of unstirred water layer anticipated in vivo. Permeability data were collected for 31 different marketed drugs with known absolute oral bioavailability and human hepatic clearance data. Literature bioavailability values were corrected for the first pass hepatic clearance thus obtaining the fraction absorbed from intestinal lumen (fraction absorbed), Fa, while assuming that the fraction escaping intestinal extraction, Fg, was approximately 1. Permeability obtained under conditions of controlled hydrodynamics was compared with the permeability measured under unstirred conditions. It is shown that the optimized effective permeability correlates with the fraction absorbed. In contrast, permeability data obtained under unstirred conditions does not show a good correlation. The in vitro permeation model developed in this study predicts the fraction absorbed of the selected drugs in humans within experimental uncertainty. It has been demonstrated that the correlation with the fraction absorbed is greatly improved using the permeability data obtained under controlled hydrodynamics with paracellular transport included in the model. Ó 2011 Elsevier B.V. All rights reserved.

1. Introduction Since the introduction of the parallel artificial membrane permeability assay (PAMPA) by Kansy et al. (1998), the technique has received considerable attention in the pharmaceutical industry. PAMPA offers a simple physicochemical measure of permeability for research compounds. The throughput is higher than other cell-based permeability assays (e.g. Caco-2, Madin-Darby Canine Kidney cell line), and it is relatively cheap to run (Artursson and Karlsson, 1991; Fade, 1998; Irvine et al., 1999). PAMPA is a potentially useful tool in early phase discovery projects where permeability is seen to be an issue for the chemical structures of interest. For instance, in combination with other in vitro assays, such as metabolic assessment, aqueous solubility, plasma protein ⇑ Corresponding authors. Tel.: +44 (0) 161 306 4522; fax: +44 (0) 161 275 4734 (R.A.W. Dryfe), tel.: +44 (0) 1625 230338 (K.Y. Tam). E-mail addresses: [email protected] (K.Y. Tam), robert.dryfe@manchester. ac.uk (R.A.W. Dryfe). 0928-0987/$ - see front matter Ó 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.ejps.2011.08.007

binding, PAMPA could form part of the testing cascade to evaluate the suitability of research compounds. Over the past decade, several enhanced versions of PAMPA have been developed by different researchers, including variations in membrane composition, solution composition, and hydrodynamics/stirring. The organic phase immobilized within the membrane has been varied from the original use of egg lecithin in n-dodecane (egg-PAMPA) to modified composition as dioleoyl phosphatidylcholine in dodecane (DOPC-PAMPA) (Avdeef et al., 2001), the complex double-sink method (DS-PAMPA) (Avdeef, 2003a,b; Bermejo et al., 2004), n-hexadecane alone (HDM-PAMPA) (Wohnsland and Faller, 2001) or the mixture of lipids dissolved in 1,7-octadiene to form a ‘bio-mimetic’ PAMPA (BM-PAMPA) (Sugano et al., 2001). Avdeef et al. introduced a modification of DS-PAMPA that employed stirring of the donor compartment (Avdeef et al., 2004). On the other hand, some PAMPA practitioners still prefer to adopt the original experimental design developed by Kansy et al. (1998), with their choice of membrane compositions in setting up their own PAMPA assays. It has been show recently that

300

´ et al. / European Journal of Pharmaceutical Sciences 44 (2011) 299–309 Mateˇj Velicky

Abbreviations and symbols a permeability function A, B, C permeation constants BCS biopharmaceutics classification system BM-PAMPA bio-mimetic PAMPA cA(t) time-dependent acceptor concentration cB/cP total blood to total plasma drug concentration ratio Caco-2 colorectal adenocarcinoma epithelial cell line CLh first pass hepatic clearance Daq aqueous diffusion coefficient DOPC dioleyol phosphatidylcholine DOPC-PAMPA dioleoyl phosphatidylcholine PAMPA DS-PAMPA double-sink PAMPA e elementary charge egg-PAMPA egg lecithin PAMPA E(Du) electric potential drop function FR(rHYD/Rp) Renkin hydrodynamic sieving function F absolute bioavailability in humans Fa fraction absorbed from intestinal lumen fraction escaping intestinal extraction Fg Fh fraction escaping hepatic extraction %Fa (percent) fraction absorbed from intestinal lumen f(–), f(+) concentration fraction of an anionic, cationic form, respectively f(±/0) concentration fraction of neutral/zwitterionic form G, H fitting constants HDM-PAMPA hexadecane PAMPA HJP human jejunal permeability k concentration function from permeability measurement kB Boltzmann constant generic distribution coefficient Kd K Ad membrane/acceptor distribution coefficient KD membrane/donor distribution coefficient d

the original design of PAMPA has some limitations with regard to the understanding of permeability and the correspondence with conditions found in vivo (Avdeef, 2003a, 2005; Youdim et al., 2003; Avdeef et al., 2004, 2007; Korjamo et al., 2009; Velicky´ et al., 2010). In particular, the experiment is performed in a ‘static’ configuration where the permeating research compounds are allowed to diffuse through the PAMPA membrane, without any control of the hydrodynamics. For lipophilic molecules where the permeability is already close to the rate of diffusion across the unstirred water layer (UWL), the permeability data obtained from ‘static’ PAMPA conditions are of little use for ranking purposes. Another issue is that the PAMPA model does not account for any paracellular transport. This may potentially lead to an underestimation of permeability for some research compounds, especially at the early phase of drug discovery, where the size of the molecules tend to be smaller than the fully elaborated drug candidates. In our previous study (Velicky´ et al., 2010), we have developed a rotating-diffusion cell with two aqueous compartments, separated by a lipid-impregnated artificial membrane, for the determination of drug permeability under conditions of controlled hydrodynamics. With this novel experimental setup, we have addressed and resolved the three most neglected theoretical and experimental problems, namely: reproducible stirring (controlling the masstransport regime), in situ permeability measurements (accurate determination of the permeability coefficient) and use of the appropriate analytical model (considering lag-time, two-way flux, pH-gradient and membrane retention). The term ‘in situ’ in relation to the presented permeation assay reflects the fact that the detection of the drug molecule is carried out in real time inside the

K OCT d M Mr NA P0 PAMPA Pe Pm Pp PTFE Pu PVDF Qh rHYD Rp S T t UWL VA, VD

octanol/water distribution coefficient membrane area molar mass Avogadro constant intrinsic permeability coefficient parallel artificial membrane permeation assay effective permeability coefficient membrane permeability coefficient paracellular permeability coefficient polytetrafluoroethylene unstirred water layer permeability coefficient polyvinylidene fluoride hepatic blood flow hydrodynamic radius intercellular junction pore radius transcellular/paracellular transport scaling factor absolute temperature time unstirred water layer acceptor, donor volume, respectively a hydrodynamic exponent b linear regression coefficient (hydrodynamic extrapolation) d path length ratio du unstirred water layer thickness Du electric potential drop at the channel surface Du electrical potential drop in intercellular junctions e/d porosity/path-length capacity factor g dynamic viscosity j = e/kBT constant m kinematic viscosity x angular velocity of stirring

permeation cell. In the present study, we apply this advanced assay to a diverse set of marketed drugs with the explicit aim of establishing a correlation with literature bioavailability data and developing a tool to predict the fraction absorbed in humans that could be used as a tool in early drug discovery. Primarily, PAMPA is designed to mimic transcellular transport across human intestine epithelial cells. As a result, the low permeability response of small hydrophilic molecules (Mr < 250 g mol1) often does not correlate with their fraction absorbed due to the simultaneous paracellular route for their in vivo transport. Inspired by recent work from the laboratories of Sugano et al. (2002, 2004)) and Avdeef (2010), Avdeef and Tam (2010) and Tam et al. (2010) we report herein the incorporation of a paracellular component in our model based on the Renkin function (Renkin, 1954). The paracellular permeability model derived from a recent detailed analysis of the Caco-2 permeability data of Adson et al. was employed for this purpose (Avdeef, 2010). The measured effective permeability coefficient is corrected for the paracellular transport occurring in vivo thus expanding the range of drugs that can be properly ranked by this permeation assay. As a part of this analysis, a novel approach to determine the optimal effective permeability as a function of UWL thickness is obtained, which can be converted to the corresponding stirring rate. The unstirred water layer, i.e. the aqueous layer adjacent to the membrane, where the flux of the solute is diffusion limited, is accurately controlled using the rotating-diffusion device (Molloy et al., 2008; Velicky´ et al., 2010). Firstly, the effective permeability is measured for at least two stirring rates. Then, using the known analytical solution relating solute flux to applied stirring rate

´ et al. / European Journal of Pharmaceutical Sciences 44 (2011) 299–309 Mateˇj Velicky

301

Fig. 1. Chemical structures of the 31 selected drug molecules.

(Levich, 1962), one can calculate the effective permeability at any given stirring rate. This approach, only applied to date on an empirical basis (Avdeef et al., 2004), allows us to match the permeation experiment to conditions anticipated in vivo (where UWL thickness is believed to be on the order of 10–1000 microns depending on the experimental methods and protocol used (Levitt et al., 1992; Adson et al., 1995; Fagerholm and Lennernäs, 1995; Avdeef et al., 2004; Lennernäs, 2007; Avdeef and Tam, 2010)).

In this work, 31 commercially available drug molecules (9 weak acids, 9 bases, 8 neutral molecules and 5 zwitterions, Fig. 1) were used as a training set for our hydrodynamic permeation assay. Drugs, where the first pass hepatic clearances are low to moderate, and with published human absolute bioavailability spanning low to high values, were deliberately selected. The human bioavailability data was corrected for the first pass hepatic clearance to avoid erroneous correlation with permeability. Drugs with no known

302

´ et al. / European Journal of Pharmaceutical Sciences 44 (2011) 299–309 Mateˇj Velicky

active transport were selected for the training set. Moreover, care was taken to include drugs with moderate to good aqueous solubility, to ensure the absorption process was not solubility-limited. In other words, the selected compounds generally fall within the Biopharmaceutics Classification Scheme (BCS) class I (high solubility and permeability) and III (high solubility and low permeability). These conditions (BCS class I and III, no known active transport issue, and moderate clearance) are deemed necessary to ensure that passive permeability is the major factor driving absorption. The goal was to find a correlation between the in vitro drug permeability coefficient and fraction absorbed determined in vivo, using the approach described above. The development of a predictive model for the absolute bioavailability with the capability to deal with solubility limited absorption, active transport and clearance issues is beyond the scope of this work. 2. Materials and methods 2.1. Materials Sodium phosphate (98.5%), buffer solutions for pH meter calibration (pH 4.00, 7.00, 10.0). 1,9-decadiene (96%), DOPC (1,2-Dioleoyl-sn-glycero-3-phosphocholine, approx 99%), stearic acid (grade I approx 99%), acetaminophen (98%), antipyrine, atenolol (98%), betamethasone (98%), cefixime trihydrate (98%), cephalothin sodium salt (99%), cetirizine dihydrochloride (98%), (±)-chlorpheniramine maleate salt (99%), colchicine (95%), diclofenac sodium salt, fexofenadine hydrochloride (>98%), midazolam hydrochloride, nafcillin sodium salt mohohydrate, naproxen (98%), norfloxacin, oxybutynin chloride (98%), pindolol (98%), (±)propranolol hydrochloride (98%), pyridoxine (98%), quinine, anhydrous (98%), salicylic acid (99%), theophylline, anhydrous (99%), tolbutamide, (±)-verapamil (min 99%), warfarin (min 98%) and zopiclone were purchased from Sigma–Aldrich, UK. Chlorthalidone, eprosartan, gatifloxacin, metolazone and risperidone were supplied by AstraZeneca, Alderley Park, UK. Sodium hydroxide (98.8%) and sulfuric acid (98%) were used for pH adjustment (Fisher Scientific UK Ltd). Sodium acetate (98%) was purchased from BDH Ltd. Precision ground glass donor tubes were obtained from Glass Precision Engineering Ltd (Leighton Buzzard, UK), ‘‘Durapore’’Ò PVDF (polyvinylidene fluoride) hydrophobic membrane filters (0.45 lm pore size, 125 lm thickness, 75% porosity, 13 mm diameter) were supplied by Millipore (cut to an apparent area 0.68 cm2). Water of 18.2 MX cm resistivity purified by a ‘‘PURELAB’’ Ultra-filtration unit (ELGA) was used for solution preparation. Flexible plastic foil ‘‘Parafilm’’ was obtained from Pechiney Plastic Packaging, USA. Solution pH was measured using a HI991300 pH meter (Hanna Instruments). UV spectra were acquired using a DH-2000-BAL spectrometer equipped with a DH-2000-BD deuterium bulb and fibre-optic cable (supplied by Ocean Optics, the Netherlands) and controlled using a USB2000 interface (Micropack GmbH). Rotation of the donor compartment was controlled using a Model 616 rotating-disc controller (EG&G Parc). The permeation cell, consisting of an acceptor compartment made of polytetrafluoroethylene (PTFE), was built in-house and has been described previously (Velicky´ et al., 2010).

originally present in the donor part and its flux across the membrane is detected in the acceptor part using in situ UV–visible spectrophotometry. The membrane consisted of 1,9-decadiene solution containing 1.5% weight DOPC and 0.5% weight stearic acid. Permeation was conducted at ambient temperature 22.1 ± 0.4 °C (mean ± standard deviation over 30 temperature measurements, measured during each experimental session). About 30 mM sodium phosphate buffer solutions were used to maintain pH, which was kept at 6.5 and 7.4 in donor and acceptor solutions, respectively. During the permeation experiments, the membrane was rotated at rates ranging from 60 to 600 rpm with respect to the donor/ acceptor compartment. The total duration of the permeation experiment was 20 min to ensure the sink conditions in the acceptor compartment were maintained. The organic phase(membrane)/ aqueous phase distribution coefficients were determined using the standard shake-flask method. The analytical model has been presented previously (Velicky´ et al., 2010). Briefly, two-way flux equations were applied, with the inclusion of pH-gradient, membrane retention and lag-time in the model. The concentration, cA(t), measured in situ in the acceptor compartment as a function of time was transformed to a function, k, whose logarithm shows a linear dependence on time:

k¼

A  VV DA cA ðtÞ

ð1Þ

B

lnðkÞ ¼ a  t þ C

ð2Þ

where A, B and C are permeation constants, VD and VA the volumes of the donor and acceptor, respectively. The effective permeability, Pe, was calculated from the slope of the logarithmic concentration function-time dependence, a:

MPe KA VD a¼ 1 þ Dd VD Kd VA

! ð3Þ

where M is the membrane area (corrected for the porosity), K Dd and K Ad are the solute distribution coefficients between the membranedonor and membrane-acceptor, respectively. Detailed description of the above equations can be found in (Velicky´ et al., 2010). Membrane permeability, Pm, purely representing transport through the membrane (no UWL contribution) was obtained by extrapolation from the hydrodynamic relationship between effective permeability and stirring rate (Levich, 1962; Albery et al., 1976; Amidon et al., 1982; Guy and Honda, 1984; Leahy and Wait, 1986):

1 1 1 ¼ þ 1=6 x0:5 Pe P m 0:62 D2=3 aq m

ð4Þ

where Daq is the aqueous diffusion coefficient of the solute, m is the aqueous kinematic viscosity and x is the angular velocity of stirring. In the case of an ionisable drug, the membrane permeability is a pH dependent parameter. For a drug with a known dissociation constant, pKa, one is able to calculate the intrinsic permeability, i.e. permeability of the neutral fraction, P0:

  P0 ¼ Pm 10ðpHpK a Þ þ 1

ð5Þ

2.2. Methods where the sign in the exponent is + for acids and  for bases. The permeability coefficients were determined using the method previously reported by this group (Velicky´ et al., 2010). Briefly, the permeation assay with a design similar to PAMPA, consisting of buffered aqueous donor and acceptor compartments separated by a lipophilic PVDF membrane, was employed. The solute (drug) is

2.3. Correction of bioavailability for first pass hepatic clearance The absolute drug bioavailability, F, is defined as a fraction of an administered dose of drug that reaches the systemic circulation.

´ et al. / European Journal of Pharmaceutical Sciences 44 (2011) 299–309 Mateˇj Velicky

Taking first pass hepatic metabolism and gut wall metabolism into consideration, F may be expressed as (Varma et al., 2010):

F ¼ Fa  Fh  Fg

ð6Þ

where Fa, Fh and Fg represent the fraction absorbed from the intestinal lumen, fraction escaping hepatic extraction, and fraction escaping intestinal extraction, respectively. Here, we make the assumptions that the gut metabolism of the drug in the training set is negligible (Fg  1) and the total blood to total plasma drug concentration ratio (cB/cP) is unity. In order to correlate the permeability directly with fraction absorbed from the intestinal lumen, the absolute bioavailability was corrected for the first pass hepatic clearance to obtain the fraction absorbed, using following equation (Obach et al., 1997; Yang et al., 2007):

Fa ¼

F ¼ F h  F g ð1  Q

F CLh Þ h ðC B =C P Þ

 Fg



F h ð1  CL Þ Q

ð7Þ

h

where CLh is the first pass hepatic clearance and Qh is the hepatic blood flow. CLh values were obtained from Obach et al., 2008al. (2008) and Qh = 23 ml min1 kg1 from Varma et al. (2010). The averaged absolute bioavailability values, F, were obtained from several sources (Sietsema, 1989; Tenero et al., 1998; Dollery, 1999; LaCreta et al., 2000; Goodman et al., 2006; Chen, 2008; Lappin et al., 2010; FDA, 2011). 2.4. Correction of permeability for paracellular transport The paracellular term proposed by Avdeef (2010), based on work of Adson et al. (1994), Adson et al. (1995), was used here to correct for the flux of small molecules through intercellular junctions in vivo. This term accounts for the paracellular transport, dependent on molar mass and the charge state of the drug. Paracellular permeability, Pp is based on three Caco-2 assay parameters: a pore capacity factor, e/d (porosity to pathlength ratio), pore radius, Rp, and electrical potential drop at the channel surface, Du. The equation used to calculate Pp is (Avdeef, 2010):

Pp ¼

e d

Daq F R

  r HYD EðDuÞ Rp

ð8Þ

where e/d is a porosity-pathlength capacity factor, Daq is the aqueous diffusion coefficient, calculated using an empirical formula based on solute molar mass, Mr (Avdeef, 2005):

log Daq ¼ 4:15  0:488 log Mr

ð9Þ

The term FR(rHYD/Rp) in Eq. (8) is the Renkin hydrodynamic sieving function for cylindrical water channels expressed as follows (Renkin, 1954):

FR

    2 rHYD rHYD ¼ 1 Rp Rp " #    3 r HYD r HYD rHYD 5 þ 2:09  1  2:104  0:95ð Þ Rp Rp Rp ð10Þ

where rHYD is the solute hydrodynamic radius and Rp is the pore radius. Molecular hydrodynamic radii were calculated using the Sutherland–Stokes–Einstein spherical-particle equation (Avdeef, 2010): 1

r HYD ¼

0:92 þ

21:8ðg mol Þ Mr

!

kB T 6pgDaq

ð11Þ

where kB is Boltzmann constant, T is absolute temperature, g is the dynamic viscosity of the solvent (0.00890 g cm1 s1, water, 25 °C, (Lide, 1995)).

303

The function E(Du) in Eq. (8) describes the electric field across the intercellular junctions (pores) due to negatively charged ions (Avdeef, 2010):

EðDuÞ ¼ fð=0Þ þ fðþÞ

jjDuj jjDuj þ fðÞ þjjDuj 1  ejjDuj e 1

ð12Þ

where f(±/0), f(+) and f() are the concentration fractions of neutral/ zwitterionic, cationic and anionic forms, respectively, j is a constant defined as j = e/kBT where e is the elementary charge and NA is the Avogadro constant. Du is the potential drop across the pore. In the present study, the parameters Rp = 12.9 Å, e/ d = 0.78 cm1, Du = 30 mV, were taken from Avdeef (14) who based these parameters on a re-analysis of the Caco-2 permeability data of several paracellular markers published by Adson et al. (1994). There were several reasons why Adson’s paracellular model was used. The model was derived from data obtained at 25 °C, therefore is applicable to the system presented here. Other paracellular models were also tested, Adson’s model, however, has proven to be the best choice as the fitting calculations provided a stable solution and the model implements the second largest pore size from paracellular models presented in Avdeef (2010). The twocomponent permeation model described earlier (Velicky´ et al., 2010) was corrected for the paracellular transport observed in the Caco-2 assay as follows:

1 1 1 ¼ þ Pe Pu S  Pm þ Pp

ð13Þ

where Pe, Pu, Pm and Pp are the effective, unstirred water layer, membrane and paracellular permeability coefficients, respectively. The variable, S, is a scaling factor. The choice of the scaling factor depends on the particular permeability assay/paracellular model used. In this report, the S value was varied and its optimum value was found from the best fit of the effective permeability and the fraction absorbed in the plug-flow model, as discussed later. The physical meaning of Eq. (13) is that the inverse of the effective permeability is directly proportional to the inverse of UWL permeability (independent of membrane or in vivo mimic properties) and the inverse of the membrane permeability corrected for paracellular transport across Caco-2 epithelial cells. The scaling factor, S, normalises membrane permeabilities to a level comparable to those generated from the paracellular model (see above). 2.5. Extrapolation of the effective permeability for the set UWL The hydrodynamic model reported previously allows extrapolation of the effective permeability to a given stirring rate mimicking hydrodynamic conditions found in the human small intestine (Velicky´ et al., 2010). The permeability can thus be optimised to mimic UWL properties in vivo. A given UWL thickness, du, can be transformed to a corresponding stirring rate using the following relationship based on the Levich equation (Levich, 1962): 1=6 1=2 du ¼ 1:61D1=3 x aq m

ð14Þ

where v is the kinematic viscosity of the solvent and x is the angular velocity of stirring. The measured effective permeability can be extrapolated to any given stirring rare providing the relationship between the two parameters is known. In practise, we measured Pe as a function of stirring rate and used the following equation to extrapolate Pe for a set x:

log Pe ¼ a  log x þ b

ð15Þ

where a and b are the linear regression coefficients determined by measuring Pe at two or more different stirring rates and a

304

´ et al. / European Journal of Pharmaceutical Sciences 44 (2011) 299–309 Mateˇj Velicky

Fig. 2. Block scheme of the experimental/analytical method.

represents the hydrodynamic exponent (0.5 in the diffusion limited case – Eq. (14)). The effective permeabilities of all 31 drug molecules were measured three times to find the average Pe. Cetirizine, diclofenac, naproxen, propranolol, theophylline, verapamil and warfarin were analysed at six different stirring rates (60, 80, 110, 160, 280 and 600 rpm). Chlorpheniramine, fexofenadine, norfloxacin, pindolol, pyridoxine, salicylic acid and tolbutamide were analysed at three different stirring rates (60, 160 and 600 rpm). The rest of the drug molecules were analysed at two different stirring rates (60 and 280 rpm). 2.6. Absorption data as a function of the effective permeability The fraction absorbed was fitted against the plug-flow absorption model (Amidon et al., 1988; Yu and Amidon, 1999):

%ðF a Þ ¼ G  ð1  expðH  Pe ÞÞ

ð16Þ

where G represents the Graetz number (a dimensionless number describes laminar flow in the intestine, treated as a tube, in the plug flow model). In the present study, G and H are regarded as fitting constants. %Fa is the fraction absorbed under the assumption that the drug is not a substrate for metabolic processes in the gut. The overview of the whole experimental/analytical method is shown in a block scheme in Fig. 2. 3. Results and discussion 3.1. Permeation hydrodynamics As shown earlier (Velicky´ et al., 2010), the hydrodynamic exponent a from Eq. (15) describes the permeation sensitivity to stirring. The lower the value of a, the more membrane limited

permeation becomes, conversely a higher a value means transport limited permeation which is sensitive to stirring. The a value depends on the drug lipophilicity, fraction of neutral species and diffusion coefficients in the aqueous and membrane phase. Theoretical limits are a = 0 (permeation controlled by transport through the membrane) and a = 0.5 (theoretical value according to Eq. (14) – permeation controlled by transport through the aqueous solution). Table 1 shows the physicochemical properties of 31 selected drugs. From this it is noted that a correlates with the lipophilicity of the molecule (membrane/buffer distribution coefficient at pH 6.5). Fig. 3 shows a as a function of lipophilicity for 31 studied drug molecules. The correlation can be described empirically by the following equation:

a ¼ 0:0125 ðlog K mem Þ3 þ 0:0424 ðlog K mem Þ2 þ 0:0732 log K mem þ 0:0905

ð17Þ

3.2. Prediction of fraction absorbed in humans The permeability data obtained for the thirty-one studied drug molecules are listed in Table 2. The effective permeability, Pe, membrane permeability, Pm and intrinsic permeability, P0 were determined using the method described earlier. The paracellular permeability, Pp, was calculated using Eq. (8). We first focused our attention on the permeability determined under unstirred conditions. The correlation between the fraction absorbed and the effective permeability under unstirred conditions for all thirty-one drug molecules is shown in Fig. 4. The colours of the symbols represent the charge state of the drugs at the physiological pH (red – acids, green – bases, blue – zwitterions, grey – neutral). It is clear that permeability determined under unstirred

´ et al. / European Journal of Pharmaceutical Sciences 44 (2011) 299–309 Mateˇj Velicky

305

Table 1 Physicochemical properties of 31 selected drug molecules. Drug

Mr/g mol1

charge statea

pKa

Daq/106cm2 s1b

c log K D d (6.5)

log K Ad (7.4)c

log K OCT (7.4)d d

Acetaminophen Antipyrine Atenolol Betamethasone Cefixime Cephalothin Cetirizine Chlorpheniramine Chlorthalidone Colchicine Diclofenac Eprosartan Fexofenadine Gatifloxacin Metolazone Midazolam Nafcillin Naproxen Norfloxacin Oxybutynin Pindolol Propranolol Pyridoxine Quinine Risperidone Salicylic acid Theophylline Tolbutamide Verapamil Warfarin Zopiclone

151.17 188.23 266.34 392.46 453.45 396.44 388.89 274.79 338.77 399.44 296.15 424.53 501.66 375.39 385.84 325.78 414.48 230.26 319.33 357.49 248.32 259.34 169.18 324.42 410.49 138.12 180.18 270.35 454.60 308.33 388.81

N N B N A A Z B N N A A Z Z N N A A Z B B B Z B B A N A B A B

9.63 (acid) Avdeef, (2003a) 1.44 (base) Avdeef (2003a) 9.54 Avdeef (2003a)  2.10, 3.73 Anacona and Estacio (2006) 2.35 Streng (1978) 2.90, 7.98 Tam and Quere (2001) 4.00, 9.20 Moyano et al. (2005) 9.40 (acid) Martin (1969) 1.70 (base) Windholz (1976) 3.99 Avdeef (2003a) 5.30 Caballero et al. (2001) 4.25, 9.53 Yasui-Furukori et al. (2005) 5.94, 9.21 Kalam et al. (2010) 9.70 (acid) Avdeef (2003a)  2.65 Hou and Poole (1969) 4.18 Avdeef (2003a) 6.25, 8.50 Takacs-Novak and Tam (2000) 6.90 Moyano et al. (2005) 9.54 Avdeef (2003a) 9.53 Avdeef (2003a) 4.90, 8.91 Takacs-Novak and Tam (2000) 8.55, 4.24 Avdeef (2003a) 3.10, 8.10 El-Barghouthi et al. (2005) 2.88, 13.55 (Avdeef, 2003a) 8.55 (acid) Avdeef (2003a) 5.40 Häußler (1958) 9.07 Avdeef (2003a) 4.82 Avdeef (2003a) 6.76 Avdeef (2003a)

6.12 5.49 4.64 3.84 3.58 3.82 3.86 4.57 4.12 3.81 4.40 3.69 3.41 3.92 3.87 4.20 3.74 4.98 4.25 4.02 4.80 4.70 5.79 4.21 3.76 6.39 5.61 4.60 3.57 4.32 3.86

0.70 ± 0.03 1.51 ± 0.00 2.00 ± 0.43 1.02 ± 0.02 1.21 ± 0.03 0.38 ± 0.01 0.27 ± 0.02 0.27 ± 0.01 0.54 ± 0.06 0.59 ± 0.02 0.95 ± 0.01 1.35 ± 0.26 0.62 ± 0.01 0.34 ± 0.01 0.27 ± 0.00 1.96 ± 0.01 0.76 ± 0.02 0.11 ± 0.02 0.29 ± 0.03 2.09 ± 0.00 0.07 ± 0.01 1.00 ± 0.02 1.96 ± 0.32 0.86 ± 0.00 0.01 ± 0.03 0.87 ± 0.23 0.97 ± 0.03 0.19 ± 0.04 1.43 ± 0.01 0.32 ± 0.02 0.22 ± 0.03

0.98 ± 0.22 e 1.10 ± 0.05 e 1.82 ± 0.31 0.85 ± 0.21 0.29 ± 0.02 0.98 ± 0.02 0.75 ± 0.01 e 0.48 ± 0.01 1.72 ± 0.37 0.89 ± 0.05 0.39 ± 0.03 0.66 ± 0.22 1.90 ± 0.02 2.10 ± 0.00 0.59 ± 0.06 0.38 ± 0.04 3.22 ± 0.01 0.52 ± 0.01 1.38 ± 0.02 1.92 ± 0.04 1.42 ± 0.01 0.83 ± 0.02 0.74 ± 0.02 e 0.79 ± 0.11 2.03 ± 0.00 0.75 ± 0.03 0.65 ± 0.00

0.26 0.11 1.70 1.90 1.71 2.21 1.24 1.13 0.98 1.12 1.11 0.90 0.60 0.82 1.64 3.40 0.72 0.23 1.04 3.65 0.20 1.25 1.10 2.16 2.05 1.62 0.12 0.45 2.42 0.86 1.10

a

A – acid, B – base, N – neutral, Z – zwitterion. Aqueous diffusion coefficients calculated using Eq. (4) in (Avdeef, 2005). Membrane/aqueous buffer drug distribution coefficients determined using shake-flask method at aqueous pH 6.5 or 7.4, respectively, temperature 22 °C (this report). Values are averages of three measurements, errors standard deviation of the average. d Octanol/water(pH 7.4) drug distribution coefficients determined using shake-flask method at 25 °C (AstraZeneca, Alderley Edge, UK). e log K Ad (7.4) was not measured for these neutral molecules and it was assumed to be equal to log K D d (6.5). b

c

Fig. 3. The dependence of a on the membrane/buffer distribution coefficient for the pH 6.5/7.4 permeation experiment. Distribution coefficients were determined by shake-flask method (pH 6.5, Table 1) and a calculated using Eq. (15). Horizontal axis error bars – standard deviation of three independent measurements, vertical axis error bars – goodness of fit used in a extrapolation (Eq. (15)).

conditions, and without correction for the paracellular transport component, does not provide a good predictive model. With the exception of cephalothin, nafcillin and eprosartan, all data points are clustered together with no apparent ranking (high permeability – high%Fa, low permeability – low%Fa). It is also important to notice that the permeability of the small molar mass drugs, (<200 g mol1, acetaminophen, antipyrine, pyridoxine, salicylic acid, theophylline) which generally show high fraction absorbed (>95), is systematically low.

In order to improve the ranking for these drugs, a paracellular transport component is introduced whose contribution increases with decreasing molar mass. Controlled hydrodynamics are also used to mimic stirred conditions in vivo. The stirred conditions were set to be 0.5 rad s1 of angular velocity, which corresponds to a UWL thickness of 200 lm and 160 lm for the smallest and largest molecules, respectively (salicylic acid, Mr = 138.12 g mol1 and fexofenadine, Mr = 501.66 g mol1) based on Eq. (14). The UWL thickness adjacent to the human intestinal epithelium has been reported to be in the range of 35–274 lm (Levitt et al., 1990, 1992; Fagerholm and Lennernäs, 1995; Lennernäs, 2007). The chosen UWL thickness of 200 lm in the present study is within the range found in the literature and corresponds well with a median value of 188 lm for antipyrine as reported by Fagerholm and Lennernäs (1995). Fig. 5 shows the fraction absorbed of the 31 drug molecules as a function of the effective permeability (log Pe) with the solid curve representing the best fit to Eq. (16). The optimized parameters are: H = 100 cm1 s1, G = 2.14  106, S = 0.014. The dotted curves are derived from estimated errors in the log Pe values with an average value of 0.35, which is based on the lab-to-lab variability of the Caco-2 paracellular parameters as estimated by Avdeef (2010). For the upper dashed curve, parameter H is 100 cm1 s1, while a value of 85 cm1 s1 is used for the lower dashed curve (in part, to reflect the estimated errors in the %Fa data). The estimated errors in the %Fa data were determined as a range of values spread over different published %Fa values. For those molecules where errors in the published %Fa values are not available, the average error value of 9% was used. It can be seen that bioavailabilities of the molecules

´ et al. / European Journal of Pharmaceutical Sciences 44 (2011) 299–309 Mateˇj Velicky

306

Table 2 Bioavailability and permeability data of 31 selected drug molecules. a

Drug

%Fa

Acetaminophen Antipyrine Atenolol Betamethasone Cefixime Cephalothin Cetirizine Chlorpheniramine Chlorthalidone Colchicine Diclofenac Eprosartan Fexofenadine Gatifloxacin Metolazone Midazolam Nafcillin Naproxen Norfloxacin Oxybutynin Pindolol Propranolol Pyridoxine Quinine Risperidone Salicylic acid Theophylline Tolbutamide Verapamil Warfarin Zopiclone

89 ± 11 100 ± 9 65 ± 18 82 ± 9 49 ± 16 0±9 86 ± 14 58 ± 30 67 ± 10 55 ± 9 82 ± 18 14 ± 9 58 ± 9 100 ± 9 68 ± 9 68 ± 32 35 ± 9 87 ± 13 58 ± 23 100 ± 0 98 ± 2 100 ± 9 95 ± 5 85 ± 15 75 ± 25 99 ± 9 84 ± 16 88 ± 8 84 ± 4 93 ± 7 100 ± 9

log Pe (unstirred)b

log Pmc

log P0c

log Ppd

log Pe⁄(200 lm)

5.40 ± 0.03 5.31 ± 0.02 5.23 ± 0.16 4.66 ± 0.02 5.69 ± 0.02 4.37 ± 0.02 4.82 ± 0.02 5.01 ± 0.01 5.06 ± 0.03 4.78 ± 0.02 4.32 ± 0.02 5.78 ± 0.02 4.91 ± 0.17 5.63 ± 0.01 5.30 ± 0.01 4.35 ± 0.02 5.73 ± 0.03 4.55 ± 0.04 5.72 ± 0.15 4.53 ± 0.01 5.59 ± 0.07 5.07 ± 0.01 6.10 ± 0.09 5.09 ± 0.01 4.93 ± 0.03 5.03 ± 0.03 5.63 ± 0.11 4.59 ± 0.02 5.04 ± 0.03 4.49 ± 0.01 4.65 ± 0.01

5.50 ± 0.03 5.09 ± 0.03 4.64 ± 0.03 3.98 ± 0.01 5.11 ± 0.08 5.67 ± 0.04 4.25 ± 0.01 3.64 ± 0.01 4.70 ± 0.08 5.48 ± 0.02 3.37 ± 0.01 4.95 ± 0.02 4.50 ± 0.04 5.34 ± 0.05 4.85 ± 0.03 3.03 ± 0.03 4.85 ± 0.02 4.14 ± 0.01 5.39 ± 0.02 2.97 ± 0.00 4.85 ± 0.16 3.65 ± 0.01 5.23 ± 0.03 3.60 ± 0.00 3.81 ± 0.07 4.82 ± 0.10 5.74 ± 0.10 4.09 ± 0.02 3.34 ± 0.04 3.95 ± 0.02 3.89 ± 0.02

5.50 ± 0.03 5.09 ± 0.03 2.43 ± 0.03 3.98 ± 0.01 0.71 ± 0.08 1.52 ± 0.04 4.12 ± 0.01 0.94 ± 0.01 4.70 ± 0.08 5.48 ± 0.02 0.90 ± 0.01 3.73 ± 0.02 4.60 ± 0.04 5.23 ± 0.05 4.85 ± 0.03 3.03 ± 0.03 1.00 ± 0.02 1.69 ± 0.01 5.24 ± 0.02 2.43 ± 0.00 1.98 ± 0.16 0.21 ± 0.01 5.10 ± 0.03 1.55 ± 0.00 2.19 ± 0.07 0.66 ± 0.10 5.76 ± 0.10 2.71 ± 0.02 0.89 ± 0.04 2.21 ± 0.02 3.44 ± 0.02

6.09 6.22 6.24 6.80 7.23 7.09 6.80 6.26 6.67 6.82 6.83 7.14 7.06 6.70 6.79 6.63 7.13 6.63 6.52 6.54 6.18 6.22 6.16 6.40 6.62 6.32 6.19 6.73 6.73 6.79 6.63

6.07 6.15 6.04 5.78 6.77 6.95 6.02 5.46 6.30 6.71 5.21 6.63 6.27 6.58 6.43 5.03 6.58 5.90 6.46 5.07 6.07 5.45 6.12 5.44 5.67 6.17 6.18 5.87 5.27 5.75 5.69

e

a Fraction absorbed obtained as described in Section 2.2.1. The errors are determined as a range of values spread over different literature sources. For those molecules where errors were not available, the average error value of 9% (in italic) was used in the analysis. b Effective permeability coefficient obtained under unstirred conditions, not corrected for paracellular transport. c Membrane (Pm) and intrinsic (P0) permeability coefficients obtained from hydrodynamic extrapolation described in text. d Paracellular permeability coefficient calculated from Eq. (8). e Optimised effective permeability coefficient, calculated from Eq. (13) using scaling factor S = 0.014, stirred conditions x = 0.5 rad s1 (UWL  200 lm). All permeability coefficients are in units of cm s1, effective permeability coefficients are averages of three separate measurements, errors are standard deviation of the average.

Fig. 4. Correlation between the fraction absorbed and effective permeability under unstirred conditions (donor/acceptor pH = 6.5/7.4). The permeability data are averages of three measurements, horizontal axis error bars standard deviation of the three, vertical axis error bars taken from Table 2. The horizontal axis error bars are smaller then data points where not shown. Colour scheme: red – acids, green – bases, blue – zwitterions, grey – neutral. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

are generally well predicted, within experimental uncertainty, by the approach developed in this study. The above treatment including a paracellular component in the permeability term has been extended to different hydrodynamic conditions, namely membrane limited (infinite stirring) and static (no stirring) cases. The comparison of the plug-flow model sum

Fig. 5. Correlation between the fraction absorbed and optimised effective permeability (including paracellular components calculated using Eq. (13) for the set hydrodynamic conditions – angular velocity 0.5 rad s1 (ca 200 lm of UWL thickness)). The horizontal axis error bars (not shown in the graph) are considered to be 0.35 log units. The solid curve is the best fit of the function represented by Eq. (16). The dashed curves are based on estimated errors in the log Pe values (0.35). Colour scheme: red – acids, green – bases, blue – zwitterions, grey – neutral. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

errors for the membrane (x ? 1 rad s1), unstirred case (x = 0 rad s1) and stirring at in vivo transport rate (x = 0.5 rad s1) is shown in Fig. 6. The optimised effective permeability at x = 0.5 rad s1 (corresponds to UWL thickness of 200 lm) exhibits the lowest error value, thus providing the best correlation with the

´ et al. / European Journal of Pharmaceutical Sciences 44 (2011) 299–309 Mateˇj Velicky

Fig. 6. Comparison of the sum errors for the correlation of the optimized effective permeability with fraction absorbed. Results are shown for the membrane (x ? 1 rad s1), unstirred (x = 0 rad s1) and stirred (x = 0.5 rad s1) conditions. Sum error = R(%F(literature) – %F(model))2 + 1000  Rn, where n represents a count of the point (including the error bar) falling outside the dotted lines in Fig. 5. The graph leads to an optimised UWL thickness value of 200 lm having the best effective permeability – fraction absorbed correlation.

Table 3 Contribution of unstirred water layer, paracellular and transcelullar components to the optimised effective permeability coefficient. Drug

Mr/g mol1

%UWL

%para

%trans

Acetaminophen Antipyrine Atenolol Betamethasone Cefixime Cephalothin Cetirizine Chlorpheniramine Chlorthalidone Colchicine Diclofenac Eprosartan Fexofenadine Gatifloxacin Metolazone Midazolam Nafcillin Naproxen Norfloxacin Oxybutynin Pindolol Propranolol pyridoxine Quinine Risperidone Salicylic acid Theophylline Tolbutamide Verapamil Warfarin Zopiclone

151.17 188.23 266.34 392.46 453.45 396.44 388.89 274.79 338.77 399.44 296.15 424.53 501.66 375.39 385.84 325.78 414.48 230.26 319.33 357.49 248.32 259.34 169.18 324.42 410.49 138.12 180.18 270.35 454.60 308.33 388.81

0 0 0 1 0 0 2 9 0 3 3 1 1 0 0 27 6 1 4 43 1 7 3 9 12 2 0 1 17 1 2

95 84 64 9 35 73 16 13 43 74 2 31 16 76 45 1 25 18 80 1 76 15 86 9 9 67 96 14 2 9 11

5 16 36 89 65 27 82 78 57 23 95 68 83 24 55 71 69 81 15 56 23 78 10 82 79 30 4 85 80 90 87

fraction absorbed. Note the large difference between the stirred and static case as opposed to only a small difference between the stirred and membrane cases. This suggests that PAMPA permeability with proper control of hydrodynamics, and the incorporation of a paracellular component are useful for the prediction of the fraction absorbed in humans (Avdeef et al., 2004, 2007; Velicky´ et al., 2010). The contributions of the three different transport mechanisms; namely UWL limited, transcellular, and paracellular for the 31 molecules are listed in Table 3. The percentages are calculated using Eq. (13), with the optimized S value and an UWL thickness of 200 lm. Our permeation model has offered some insights into

307

the transport mechanisms. As shown in Table 3, for molecules with molar mass < 250 g mol1, the transport is predominantly governed by the paracellular route. As the mass decreases it is more likely that the molecule would be transported via a paracellular route. For molecules with molar mass > 250 g mol1, the transcellular pathway becomes the major route of transport, but the paracellular route appears to be available for molecules even with molar mass > 300 g mol1. The absorption of zwitterions (e.g. norfloxacin, gatifloxacin) or anion (cefixime) may be enhanced by active processes via organic anion transporter proteins (Dobson and Kell, 2008). Colchicine is another surprising molecule which shows a significant percentage of paracellular transport despite its molar mass approaching 400 g mol1. It has been reported that multiple efflux processes are involved in the absorption of colchicines in small intestine, which contributes to its low and variable bioavailability (Dahan et al., 2009). Colchicine absorption in humans may be complicated by the fact that this drug is known to cause diarrhoea by the decrease of intestinal water transport (Rachmilewitz et al., 1978). Nevertheless, the measured Pm value of colchicine is very low (see Table 2), which clearly signals potential issues in its intestinal absorption. It should be noted that the active processes could be difficult to delineate from the passive paracellular transport, and are beyond the scope of the present study. This issue would, however, merit further investigation to unravel the transport mechanisms. In the present study, Fg is assumed to be close to unity in the data analysis. While this assumption may be valid for most of the drugs of low hepatic clearance, the molecules, which are likely to be cytochrome P450 substrates, could be vulnerable to gut wall metabolism (Gertz et al., 2010). In such a case, it would be desirable to take Fg into consideration. Midazolam and verapamil are typical examples, with reported Fg of 0.51 and 0.65, respectively (Gertz et al., 2010). Taking these Fg values into consideration, the %Fa values of these two molecules would be 100%, which are in good agreement with our model predictions (Fig. 5). It has been suggested that Fg could also depend on the dose strength, and saturable first pass metabolism is more likely to occur in the small intestine, because of the substantial drug concentration gradient during absorption (Lin et al., 1999). It is plausible that Fg value could vary in normal healthy human subjects, as in the case of midazolam (Paine et al., 1996). The variability in gut metabolism is partly reflected in the large Y-error bar of midazolam (Fig. 5). As far as we are aware, little Fg data, if any, has been published on other drugs studied in this work. It is noted that the total blood to total plasma drug concentration ratio (CB/CP) is assumed to be unity. For drug-like molecules the ratio may vary from 0.5 to 2, which could introduce error in the estimation of Fh (see Eq. (7)). It would be beneficial to carry out independent measurements to determine CB/CP for the molecules of interest. However the majority of molecules used in this work are of low hepatic clearance and the error is not expected to be significant. Indeed, a simulation study by Yang et al. (2007) revealed that the error in Fh is likely within ±0.2 units (for CB/CP spans from 0.7 to 2), provided that the hepatic clearance values are not greater than 25% of the the liver blood flow. Despite the selection of molecules were restricted to drugs with low to moderate first pass hepatic clearances, there are only 3 molecules, namely pindolol, propranolol and verapamil, which show hepatic clearance values greater than 25% of the liver blood flow (<10% of the total number of drugs in the set). We envisage that for majority of drugs (>90%), oral absorption was not hampered by extensive gut wall and/or hepatic metabolisms. 3.3. Applicability of the predictive model We note that Avdeef and Tam (2010) have developed a similar in-combo model (based on cell-based permeability combined with

308

´ et al. / European Journal of Pharmaceutical Sciences 44 (2011) 299–309 Mateˇj Velicky

a paracellular transport component) for predicting human jejunal permeability (HJP). Obviously HJP is the most direct parameter for developing models based on any in vitro permeability measures. In reality, jejunal permeability is often unavailable and is not routinely evaluated in pharmaceutical industry. The work by Avdeef and Tam (2010) reviewed, to the best of our knowledge, the most extensive collection of HJP data (53 compounds) to date. On selection of the training compounds employed here, it was found that bioavailability data were readily available on a wider range of marketed drugs, which allowed a degree of chemical diversity to be introduced into the training data set. Generally, the oral route is the preferred way of administration for pharmaceuticals because of patient convenience. However, other routes of administration could be developed if the treatment is likely to offer clinical benefits to patients. Therefore, it is of great interest for the discovery project to identify research compounds that are unlikely to show good oral exposure at an early stage. Poor oral bioavailability could arise for a number of reasons, including metabolic liability (high clearance), efflux transport or other active processes, gut wall metabolism, suboptimal physical properties such as poor solubility, or poor permeability. As the lead chemical series progress along the discovery pipeline, a cascade of in vitro tests would be performed to identify the aforementioned issues that could potentially lead to poor oral exposure. Instead of testing the research compounds on live animals, this simplistic approach would allow the discovery project to spot the issues more quickly, and trigger the design and synthesis of the next round analogues with the goal to moderate any risk. The model developed in this study would help to predict the fraction absorbed in the absence of other detrimental factors such as clearance, efflux, solubility. Together with other in vitro assessments in the testing cascade, this model would be particularly useful to help the prioritisation of the most promising research compounds for pharmacokinetics studies, with the goal of reducing the usage of live animals on those compounds that are unlikely to show a reasonable level of exposure. While the rotating-diffusion cell described in this work studies only one compound at a time, it is crucial to establish the rigorous computation approach to derive the permeability parameter, and develop the predictive model. We are currently in the process of developing a parallelized/high throughput version of this permeability assay, and result will be reported in due course. 4. Conclusions A hydrodynamic permeation method is presented herein and shown to be a good tool to predict fraction absorbed of orally administered drugs in humans. The optimised effective permeability coefficient is calculated at stirred conditions corresponding to the unstirred water layer thickness of 200 lm and includes the paracellular transport component to mimic transport through the intercellular junctions found in vivo. The optimised effective permeability coefficient and the effective permeability coefficient obtained under unstirred conditions are correlated with the fraction absorbed. Comparison of the two correlations shows that the optimisation method presented here outperforms the unstirred PAMPA method for the prediction of fraction absorbed in humans. Acknowledgment We thank our industrial collaborator (AstraZeneca) and EPSRC for funding. References Adson, A., Burton, P.S., Raub, T.J., Barsuhn, C.L., Audus, K.L., Ho, N.F.H., 1995. Passive diffusion of weak organic electrolytes across Caco-2 cell monolayers:

uncoupling the contributions of hydrodynamic, transcellular, and paracellular barriers. Journal of Pharmaceutical Sciences 84, 1197–1204. Adson, A., Raub, T.J., Burton, P.S., Barsuhn, C.L., Hilgers, A.R., Audus, K.L., Ho, N.F.H., 1994. Quantitative approaches to delineate paracellular diffusion in cultured epithelial cell monolayers. Journal of Pharmaceutical Sciences 83, 1529–1536. Albery, W.J., Burke, J.F., Leffler, E.B., Hadgraft, J., 1976. Interfacial transfer studied with a rotating diffusion cell. Journal of the Chemical Society-Faraday Transactions I 72, 1618–1626. Amidon, G.E., Higuchi, W.I., Ho, N.F.H., 1982. Theoretical and experimental studies of transport of micelle-solubilized solutes. Journal of Pharmaceutical Sciences 71, 77–84. Amidon, G.L., Sinko, P.J., Fleisher, D., 1988. Estimating human oral fraction dose absorbed: a correlation using rat intestinal membrane permeability for passive and carrier-mediated compounds. Pharmaceutical Research 5, 651–654. Anacona, J.R., Estacio, J., 2006. Synthesis and antibacterial activity of cefixime metal complexes. Transition Metal Chemistry 31, 227–231. Artursson, P., Karlsson, J., 1991. Correlation between oral drug absorption in humans and apparent drug permeability coefficients in human intestinal epithelial (CACO-2) cells. Biochemical and Biophysical Research Communications 175, 880–885. Avdeef, A., 2003a. Absorption and Drug Development. Solubility, Permeability, and Charge State. Wiley, Interscience. Avdeef, A. (2003b). High-throughput Measurement of Permeability Profiles, In: Drug Bioavailability, Wiley-VCH, Weinheim. Avdeef, A., 2005. The rise of PAMPA. Expert Opinion on Drug Metabolism & Toxicology 1, 325–342. Avdeef, A., 2010. Leakiness and size exclusion of paracellular channels in cultured epithelial cell monolayers–interlaboratory comparison. Pharmaceutical Research 27, 480–489. Avdeef, A., Bendels, S., Di, L., Faller, B., Kansy, M., Sugano, K., Yamauchi, Y., 2007. PAMPA – critical factors for better predictions of absorption. Journal of Pharmaceutical Sciences 96, 2893–2909. Avdeef, A., Nielsen, P.E., Tsinman, O., 2004. PAMPA – a drug absorption in vitro model: 11. matching the in vivo unstirred water layer thickness by individualwell stirring in microtitre plates. European Journal of Pharmaceutical Sciences 22, 365–374. Avdeef, A., Strafford, M., Block, E., Balogh, M.P., Chambliss, W., Khan, I., 2001. Drug absorption in vitro model: filter-immobilized artificial membranes: 2. Studies of the permeability properties of lactones in Piper methysticum Forst. European Journal of Pharmaceutical Sciences 14, 271–280. Avdeef, A., Tam, K.Y., 2010. How well can the caco-2/madin-darby canine kidney models predict effective human jejunal permeability? Journal of Medicinal Chemistry 53, 3566–3584. Bermejo, M., Avdeef, A., Ruiz, A., Nalda, R., Ruell, J.A., Tsinman, O., Gonzalez, I., Fernandez, C., Sanchez, G., Garrigues, T.M., Merino, V., 2004. PAMPA – a drug absorption in vitro model 7. comparing rat in situ, Caco-2, and PAMPA permeability of fluoroquinolones. European Journal of Pharmaceutical Sciences 21, 429–441. Caballero, R., Delpón, E., Valenzuela, C., Longobardo, M., González, T., Tamargo, J., 2001. Direct effects of candesartan and eprosartan on human cloned potassium channels involved in cardiac repolarization. Molecular Pharmacology 59, 825– 836. Chen, C., 2008. Physicochemical, pharmacological and pharmacokinetic properties of the Zwitterionic antihistamines cetirizine and levocetirizine. Current Medicinal Chemistry 15, 2173–2191. Dahan, A., Sabit, H., Amidon, G.L., 2009. Multiple efflux pumps are involved in the transepithelial transport of colchicine: combined effect of P-glycoprotein and multidrug resistance-associated protein 2 leads to decreased intestinal absorption throughout the entire small intestine. Drug Metabolism and Disposition 37, 2028–2036. Dobson, P.D., Kell, D.B., 2008. Carrier-mediated cellular uptake of pharmaceutical drugs: an exception or the rule? Nature Reviews Drug Discovery 7, 205–220. Dollery, C.T. (1999). Therapeutic Drugs. Churchill Livingstone. El-Barghouthi, M.I., Masoud, N.A., Al-Kafawein, J.K., Zughul, M.B., Badwan, A.A., 2005. Host–guest interactions of risperidone with natural and modified cyclodextrins: phase solubility, thermodynamics and molecular modeling studies. Journal of Inclusion Phenomena 53, 15–22. Fade, V., 1998. Link between drug absorption solubility and permeability measurements in Caco-2 cells. Journal of Pharmaceutical Sciences 87, 1604– 1607. Fagerholm, U., Lennernäs, H., 1995. Experimental estimation of the effective unstirred water layer thickness in the human jejunum, and its importance in oral drug absorption. European Journal of Pharmaceutical Sciences 3, 247–253. FDA (2011). Drug Information Online. Available from : . (accessed 5-04-11). Gertz, M., Harrison, A., Houston, J.B., Galetin, A., 2010. Prediction of human intestinal first-pass metabolism of 25 CYP3A substrates from in vitro clearance and permeability data. Drug Metabolism and Disposition 38, 1147–1158. Goodman, L.S., Gilman, A.G., Hardman, J.G., Limbird, L.E., 2006. Goodman & Gilman’s the Pharmacological Basis of Therapeutics. McGraw-Hill Health Professions Division, New York. Guy, R.H., Honda, D.H., 1984. Solute transport resistance at the octanol – water interface. International Journal of Pharmaceutics 19, 129–137. Häußler, V.A. and Hajdú, P. (1958). Mitteilung über die Dissoziationskonstante und Löslichkeit von Rastinon ‘‘Hoechst’’ Ò Archiv der Pharmazie 291, 531–536.

´ et al. / European Journal of Pharmaceutical Sciences 44 (2011) 299–309 Mateˇj Velicky Hou, J.P., Poole, J.W., 1969. The amino acid nature of ampicillin and related penicillins. Journal of Pharmaceutical Sciences 58, 1510–1515. Irvine, J.D., Takahashi, L., Lockhart, K., Cheong, J., Tolan, J.W., Selick, H.E., Grove, J.R., 1999. MDCK (Madin–Darby canine kidney) cells: a tool for membrane permeability screening. Journal of Pharmaceutical Sciences 88, 28–33. Kalam, M.A., Sultana, Y., Ali, A., Aqil, M., Mishra, A.K., Chuttani, K., 2010. Preparation, characterization, and evaluation of gatifloxacin loaded solid lipid nanoparticles as colloidal ocular drug delivery system. Journal of Drug Targeting 18, 191–204. Kansy, M., Senner, F., Gubernator, K., 1998. Physicochemical high throughput screening: parallel artificial membrane permeation assay in the description of passive absorption processes. Journal of Medicinal Chemistry 41, 1007–1010. Korjamo, T., Heikkinen, A.T., Mönkkönen, J., 2009. Analysis of unstirred water layer in in vitro permeability experiments. Journal of Pharmaceutical Sciences 98, 4469–4479. LaCreta, F.P., Kaul, S., Kollia, G.D., Duncan, G., Randall, D.M., Grasela, D.M., 2000. Interchangeability of 400-mg intravenous and oral gatifloxacin in healthy adults. Pharmacotherapy 20, 59S–66S. Lappin, G., Shishikura, Y., Jochemsen, R., Weaver, R.J., Gesson, C., Houston, B., Oosterhuis, B., Bjerrum, O.J., Rowland, M., Garner, C., 2010. Pharmacokinetics of fexofenadine: evaluation of a microdose and assessment of absolute oral bioavailability. European Journal of Pharmaceutical Sciences 40, 125–131. Leahy, D.E., Wait, A.R., 1986. Solute transport resistance at water–oil interfaces. Journal of Pharmaceutical Sciences 75, 1157–1161. Lennernäs, H., 2007. Intestinal permeability and its relevance for absorption and elimination. Xenobiotica 37, 1015–1051. Levich, V.G., 1962. Physicochemical Hydrodynamics. Englewood Cliffs, London. Levitt, M.D., Furne, J.K., Strocchi, A., Anderson, B.W., Levitt, D.G., 1990. Physiological measurement of luminal stirring in the dog and human small bowel. Journal of Clinical Investigation 86, 1540–1547. Levitt, M.D., Strocchi, A., Levitt, D.G., 1992. Human jejunal unstirred layer: evidence for extremely efficient luminal stirring. American Journal of Physiology – Gastrointestinal and Liver Physiology 262, G593–G596. Lide DR (1995). Handbook of Chemistry and Physics, 76th ed., CRC. Lin, J.H., Chiba, M., Baillie, T.A., 1999. Is the role of the small intestine in first-pass metabolism overemphasized? Pharmacological Reviews 51, 135–157. Martin, A.N., 1969. Physical Pharmacy. Lea & Febiger, Philadelphia. Molloy, B.J., Tam, K.Y., Matthew Wood, J., Dryfe, R.A.W., 2008. A hydrodynamic approach to the measurement of the permeability of small molecules across artificial membranes. Analyst 133, 655–659. Moyano, M.A., Rosasco, M.A., Pizzorno, M.T., Segall, A.I., 2005. Simultaneous determination of chlorpheniramine maleate and dexamethasone in a tablet dosage form by liquid chromatography. Journal of AOAC International 88, 1677–1683. Obach, R.S., Baxter, J.G., Liston, T.E., Silber, B.M., Jones, B.C., Macintyre, F., Rance, D.J., Wastall, P., 1997. The prediction of human pharmacokinetic parameters from preclinical and in vitro metabolism data. Journal of Pharmacology and Experimental Therapeutics 283, 46–58. Obach, R.S., Lombardo, F., Waters, N.J., 2008. Trend analysis of a database of intravenous pharmacokinetic parameters in human for 670 drug compounds. Drug Metabolism and Disposition 36, 1385–1405. Paine, M.F., Shen, D.D., Kunze, K.L., Perkins, J.D., Marsh, C.L., McVicar, J.P., Barr, D.M., Gillies, B.S., Thummel, K.E., 1996. First-pass metabolism of midazolam by the human intestine. Clinical Pharmacology and Therapeutics 60, 14–24. Rachmilewitz, D., Fogel, R., Karmeli, F., 1978. Effect of colchicine and vinblastine on rat intestinal water transport and Na–K–ATPase activity. Gut 19, 759–764.

309

Renkin, E.M., 1954. Filtration, diffusion, and molecular sieving through porous cellulose membranes. The Journal of General Physiology 38, 225–243. Saitoh, R., Sugano, K., Takata, N., Tachibana, T., Higashida, A., Nabuchi, Y., Aso, Y., 2004. Correction of permeability with pore radius of tight junctions in Caco-2 monolayers improves the prediction of the dose fraction of hydrophilic drugs absorbed by humans. Pharmaceutical Research 21, 749–755. Sietsema, W.K., 1989. The absolute oral bioavailability of selected drugs. International Journal of Clinical Pharmacology Therapy and Toxicology 27, 179–211. Streng, W.H., 1978. Microionization constants of commercial cephalosporins. Journal of Pharmaceutical Sciences 67, 666–669. Sugano, K., Hamada, H., Machida, M., Ushio, H., Saitoh, K., Terada, K., 2001. Optimized conditions of bio-mimetic artificial membrane permeation assay. International Journal of Pharmaceutics 228, 181–188. Sugano, K., Takata, N., Machida, M., Saitoh, K., Terada, K., 2002. Prediction of passive intestinal absorption using bio-mimetic artificial membrane permeation assay and the paracellular pathway model. International Journal of Pharmaceutics 241, 241–251. Takacs-Novak, K., Tam, K.Y., 2000. Multiwavelength spectrophotometric determination of acid dissociation constants. Part V: microconstants and tautomeric ratios of diprotic amphoteric drugs. Journal of Pharmaceutical and Biomedical Analysis 21, 1171–1182. Tam, K.Y., Avdeef, A., Tsinman, O., Sun, N., 2010. The permeation of amphoteric drugs through artificial membranes – an in combo absorption model based on paracellular and transmembrane permeability. Journal of Medicinal Chemistry 53, 392–401. Tam, K.Y., Quere, L., 2001. Multiwavelength spectrophotometric resolution of the micro-equilibria of cetirizine. Analytical Sciences 17, 1203–1208. Tenero, D., Martin, D., Ilson, B., Jushchyshyn, J., Boike, S., Lundberg, D., Zariffa, N., Boyle, D., Jorkasky, D., 1998. Pharmacokinetics of intravenously and orally administered eprosartan in healthy males: absolute bioavailability and effect of food. Biopharmaceutics and Drug Disposition 19, 351–356. Varma, M.V.S., Obach, R.S., Rotter, C., Miller, H.R., Chang, G., Steyn, S.J., El-Kattan, A., Troutman, M.D., 2010. Physicochemical space for optimum oral bioavailability: contribution of human intestinal absorption and first-pass elimination. Journal of Medicinal Chemistry 53, 1098–1108. Velicky´, M., Bradley, D.F., Tam, K.Y., Dryfe, R.A.W., 2010. In situ artificial membrane permeation assay under hydrodynamic control: permeability-pH profiles of warfarin and verapamil. Pharmaceutical Research 27, 1644–1658. Windholz, M. (1976) (Ed.). The Merck Index. Merck and Co., Inc., Rahway, NJ. Wohnsland, F., Faller, B., 2001. High-throughput permeability pH profile and highthroughput alkane/water log P with artificial membranes. Journal of Medicinal Chemistry 44, 923–930. Yang, J., Jamei, J., Yeo, K.R., Rostami-Hodjegan, A., Tucher, G.T., 2007. Misuse of the well-stirred model of hepatic drug clearance. Drug Metabolism and Disposition 35, 501. Yasui-Furukori, N., Uno, T., Sugawara, K., Tateishi, T., 2005. Different effects of three transporting inhibitors, verapamil, cimetidine, and probenecid, on fexofenadine pharmacokinetics. Clinical Pharmacology and Therapeutics 77, 17–23. Youdim, K.A., Avdeef, A., Abbott, N.J., 2003. In vitro trans-monolayer permeability calculations: Often forgotten assumptions. Drug Discovery Today 8, 997–1003. Yu, L.X., Amidon, G.L., 1999. A compartmental absorption and transit model for estimating oral drug absorption. International Journal of Pharmaceutics 186, 119–125.